ASphalt

Volume 82
ASPHALT PAVING
TECHNOLOGY
2013
JOURNAL OF THE ASSOCIATION
OF ASPHALT PAVING TECHNOLOGISTS
Denver, Colorado
April 710, 2013
DEStech Publications, Inc.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Copyright DEStech Publications, Inc.

Provided by IHS under license with DESTECH Licensee=Singapore-Jurong/3219500185, User=T, Kumanan
No reproduction or networking permitted without license from IHS Not for Resale, 05/01/2017 18:27:10 MDT
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Asphalt Paving Technology 2013Volume 82
Produced by:
DEStech Publications, Inc.
439 North Duke Street
Lancaster, Pennsylvania 17602 U.S.A.
Copyright 2013 by the Association of Asphalt Paving Technologists

A l l r i g h t s r es er v e d
No part of this publication may be reproduced, stored in a

retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of the Association of
Asphalt Paving Technologists.
Printed in the United States of America

10 9 8 7 6 5 4 3 2 1
M ai n e n t r y u n d er t i t l e:
Asphalt Paving Technology 2013Volume 82
ISSN No. 0270-2932

Journals of the Association were printed for the meetings listed below
and may be obtained from the Secretary-Treasurer, 6776 Lake Drive,
Suite 215, Lino Lakes MN 55014
1974 Williamsburg, Vol. 43 15.00 1999 Chicago, Vol. 68 75.00

1975 Phoenix, Vol. 44 20.00 1999 75th Ann, Volume 68A 25.00
1977 San Antonio, Vol. 46 25.00 2000 Reno, Vol. 69 75.00
1982 Kansas City, Vol. 51 40.00 2001 Clearwater, Vol. 70 80.00
1983 Atlanta, Vol. 52 45.00 2002 Colorado Springs, Vol. 71 80.00
1984 Scottsdale, Vol. 53 50.00 2003 Lexington, Vol. 72 80.00
1986 Clearwater, Vol. 55 50.00 2004 Baton Rouge, Vol. 73 90.00
1988 Williamsburg, Vol. 57 55.00 2005 Long Beach, Vol.74 90.00
1989 Nashville, Vol. 58 55.00 2006 Savannah, Vol.75 100.00
1990 Albuquerque, Vol. 59 55.00 2007 San Antonio, Vol. 76 100.00
1991 Seattle, Vol. 60 60.00 2008 Philadelphia, Vol. 77 125.00
1992 Charleston, Vol. 61 60.00 2009 Minneapolis, Vol. 78 150.00
1993 Austin, Vol. 62 60.00 2010 Sacramento, Vol. 79 150.00
1994 St. Louis, Vol. 63 60.00 2011 Tampa, Vol. 80 180.00
1995 Portland, Vol. 64 70.00 2012 Austin, Vol. 81 180.00
1996 Baltimore, Vol. 65 70.00 2013 Denver, Vol. 82 200.00
1997 Salt Lake City, Vol. 66 75.00
1998 Boston, Vol. 67 75.00
Prices to AAPT members for extra copies of the Journal are $5 less than
those listed.
"The Association of Asphalt Paving Technologists is not responsible for the

statements and opinions advanced in its publications."
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Copyright 2013
The Association of Asphalt Paving Technologists
ISSN 0270-2932
I.S.I. Certified
Reprints of Articles included in Volumes which are out-of-print may be

obtained from the AAPT Office.
Printed in the United States of America by DEStech, Lancaster, PA, US

ASSOCIATION OF ASPHALT PAVING TECHNOLOGISTS
2013
W.J. Emmons Annual Award
The Board of Directors announces the recipients of the W.J. Emmons

Annual Award of the Association of Asphalt Paving Technologists for the
best paper presented at the 2012 Meeting in Austin, TX, held April 14,
2012, to be:
Winner:
Effects of Interface Condition Characteristics on Open-Graded

Friction Course Top-Down Cracking Performance
Yu Chen, Gabriele Tebaldi, Reynaldo Roque, George Lopp and

Yumin Su; University of Florida
Runner-up:
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Laboratory Evaluation of Asphalt Mixtures Containing Dry Added

Ground Tire Rubber and a Processing Aid
Gaylon L. Baumgardner, J. Michael Hemsley, Walter Jordan III,

Paragon Technical Services; Isaac L. Howard, Mississippi State
University
AAPT Scholarship Fund
The Board of Directors announces the 2013 winners of the AAPT

Scholarship to be:
Matthew Gersch, Auburn University
Mirella Villani, Delft University of Technology

AAPT Board of Directors 2012-2013
From left:
1st Row: Maria Bonaquist, Jeyna Rowe, Marcia Prowell, Rebecca McDaniel,
Emma McDaniel Burley, Betty Skok, Amy Epps-Martin
2nd Row: Ramon Bonaquist, Geoffrey Rowe, Brian Prowell, Eileen Soler,
Mike Anderson, Rey Roque, David Newcomb, Gene Skok
2013-2014 Board of Directors
President: Ramon Bonaquist

First Vice-President: Brian Prowell
Second Vice-President: Geoffrey Rowe
Director-at-Large: Bill Buttlar
Director-at-Large: Randy West
Past President: Rebecca McDaniel
Past President: David Newcomb
--`,`,,,``,,```,,`,,,,,`,``

Introduction of Mr. Ken Kandhal
as 2012 Honorary Member of AAPT
by E. Ray Brown
It is my pleasure to recognize Mr. Prithvi (Ken) Kandhal who has

been elected as honorary member in the Association of Asphalt
Paving Technologists (AAPT). Ken has been actively involved in
materials, mix design, and construction of asphalt pavements for over
40 years and is clearly deserving of this award. Ken has been very
active in all activities of AAPT since he joined the organization in
1972. As a result of his many contributions and leadership, he has
helped to advance AAPTs reputation to the preeminent organization
that it is today.
Kens interest in asphalt pavements began in India where he was first

introduced to asphalt as a highway agency employee. After working
there a few years, Ken moved to the US and began work as a graduate
student at Iowa State University where he received his Masters
Degree in Civil Engineering in 1969. After graduation he went to
work for the Pennsylvania Department of Transportation where he
served as the Chief Asphalt Engineer from 1970-1988. In 1988, he
began work at the National Center for Asphalt Technology where he
first served as Assistant Director and then Associate Director for many
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

years. When he retired in 2001, Ken was awarded the title of
Associate Director Emeritus of NCAT. Since his retirement Ken has
continued to work as a consultant on various asphalt projects
worldwide.
While with the Pennsylvania DOT and NCAT, Ken was very
successful in performing research and in publishing the results. Some
of his more significant accomplishments have involved work related
to: Superpave restricted zone, stripping of asphalt pavements,
performance testing of HMA mixtures, construction guidance for
longitudinal joints, improved tests for determining bulk specific
gravity of fine aggregate, testing and characterization of mineral
fillers, and aggregate tests related to HMA performance. He is also
one of the key authors of the textbook Hot Mix Asphalt Materials, Mix
Design, and Construction.
During his career, Ken has published more than 120 technical journal
articles as well as many other articles in magazines, conference
proceedings, and technical reports. Ken has published more than 20
AAPT papers during a 30-year period with his first paper being
published in 1973. During this time he published the AAPT papers in
four different decades: five in the 1970s, seven in the 1980s, six in
the 1990s and four in the 2000s. This shows a record of continued
active participation in AAPT activities.
Ken received the Emmons Award for best AAPT paper in 1989 for his
work on stripping of asphalt mixtures. He has been a member of
AAPT for over 40 years and served as President from 1999-2000.
During his year as President, Ken wrote personnel letters to all visitors
attending the annual AAPT conference personally thanking them for
their attendance and requesting that they consider joining AAPT. As a
result of his efforts there was a significant increase in AAPT
membership.
Ken has been very active in a number of other technical societies and
organizations. He is a registered engineer in the state of Pennsylvania.
He is a Fellow in ASTM and served as the Chairman of Committee
D04 on Road and Paving Materials (19981999). As a result of his
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

support to ASTM, he received the Award of Merit in 2003. He has
served in TRB where he was Chairman of Committee A2D02 from
1982-1988. He is now an emeritus member of A2D02. He is an
ASCE Fellow and served as Chairman of the Subcommittee on
Bituminous Materials from 1988-1992.
Ken is very worthy of this recognition and it is now my pleasure to

present to you, my good friend, Mr. Ken Kandhal.
Prithvi S. Kandhal Remarks
I am humbled by this honor bestowed upon me especially when it

comes from my peers in all continents of the world.
It is time to reflect back on circumstances and people who helped

increase my knowledge of asphalt paving technology during the last
40 years. It is not possible to name all the individuals; I will mention
only those with whom I worked closely. Over the years several people
assisted me in discharging my duty as President of the AAPT;
Chairman of the ASTM Main Committee D04 on Road and Paving
Materials; Chairman of the TRB Committee A2D02 on Asphalt
Mixtures; and Chairman of the ASCE Subcommittee on Bituminous
Materials.
As a young highway engineer in a desert region of India, which did

not have any stone quarry, I had the opportunity to construct about one
mile of road using dune sand and asphalt binder. Fortunately, I had a
copy of the Highway Engineering Handbook edited by K.B. Woods of
Purdue University, which I used for guidance. I used both Hubbard
Field and Marshall testing equipment at that time.
Having developed interest in asphalt, I came to Iowa State University

in Ames, Iowa, in 1968 for pursuing graduate study. Both Prof. Ladis
Csanyi, who invented the foamed asphalt process and Prof. Dah-yinn
Lee were engaged in asphalt research there. I even got the opportunity
to make a batch of foamed asphalt using steam. I completed my
graduate study under Prof. Lee researching asphalt absorption by
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

aggregates and development of an innovative method of determining
bulk specific gravity of aggregates using dyes.
Then in 1970 I joined the Pennsylvania Department of Transportation

(Penn DOT) as State Bituminous Engineer where both Jim Moulthrop
and Ron Cominsky were my colleagues. Later they played key roles in
administering SHRPs Asphalt Program. I was successful in
implementing up-to-date asphalt paving technology in Pennsylvania
with the able assistance of Carl Lubold of the Pennsylvania Asphalt
Pavement Association and Carlos Rosenberger of the Asphalt
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Institute. I am obliged to Bill Koehler who as Engineer of Tests gave
me the necessary environment and the freedom in my advancement as
an asphalt paving technologist.
Fortunately, in Penn DOT I was in charge of almost all asphalt-related

activities such as routine testing, research, development of
specifications and test methods, and troubleshooting in the field. This
helped me to get a well rounded experience in asphalt technology. One
thing really helped me in authoring a number of papers for AAPT,
TRB and ASTM while at Penn DOT. During the winter time, asphalt
technicians in my laboratory had little work to do. When visitors used
to tour the laboratory, it was a chronic complaint that technicians
(state employees!) are sitting idle. To address that problem, I started to
collect all types of aggregates from various quarries across the state
and also collect samples of AC-20 asphalt cement from various
refineries. Then I would draw up a research work plan. That would
keep all the technicians busy during the winter time and a lot of test
data were generated. Obviously, I had to sit down, analyze the test
data and write a research paper. We had just changed from penetration
grading to viscosity grading of asphalt cements in the US. There was a
lot to be researched to fully understand the viscosity graded asphalts.
There was another motivation to write research papers. Every time a

new governor came to office in Pennsylvania, the first thing which
was done was to restrict travel of state employees outside the state.
Only those who had a paper to present were allowed travel outside
Pennsylvania. By writing papers, I could attend AAPT and TRB
meetings regularly. During my tenure at Penn DOT of about 17 years,

I had the opportunity to author about 35 refereed papers. Very few
people in state DOTs publish technical papers. However, I was
inspired by Chuck Hughes and Gale Page, both working for state
DOTs, to do so. Gene Skok of AAPT, Bob Nady of Iowa, Bob
Dunning, and Byron Ruth of University of Florida also encouraged me
by appreciating my practical papers.
In 1987, I got a call from Dr. Freddy Roberts, the then Director of the
National Center for Asphalt Technology (NCAT) to consider joining
the newly established center at Auburn University, Alabama. I joined
NCAT in March 1988 because it was a challenge to establish a new
center with a skeleton staff of three engineers: Freddy Roberts as
Director and Ray Brown and myself as two Assistant Directors.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Fortunately, SHRP came at the right time. NCAT was awarded the
SHRP A-003 Project on Asphalt-Aggregate Interaction: Adhesion and
Absorption. I had the privilege of working with Bob Dunning, who
moved to Auburn to work on that project. Ray Brown and I were also
successful in obtaining several major NCHRP projects to sustain the
Center in the initial formative years. Both of us feel proud when we
see that NCAT has come a long way now to be recognized as a
premier asphalt research center in the world.
As Associate Director of NCAT and a member of the graduate faculty

of Civil Engineering at Auburn University, I had the opportunity of
advising and working with several bright Masters and Ph.D. students.
They were in alphabetical order: Shane Buchanan, Sanjoy
Chakraborty, Allen Cooley, Steve Cross, Kee Foo, John Haddock,
Mike Huner, Maqbool Khatri, Cynthia Lynn, Todd Lynn, Rajib
Mallick, S.S. Rao, Jay Winford, Yiping Wu, and Jingna Zhang. Rajib
Mallick and Allen Cooley, who are talented researchers, worked with
me on several research projects.
While at NCAT, I was also fortunate to work with Freddy Roberts,

Ray Brown, Doug Hanson. Frazier Parker, Mary Stroup-Gardiner,
Randy West, Don Watson, Brian Prowell, Buzz Powell, and Dave
Timm. Being a key instructor for the NCAT Professor Training
Course, I had the opportunity to meet many asphalt technologists in
the academia across the US.

Although I retired from NCAT in 2001, it is difficult to get the asphalt
out of my system. Therefore, whenever my services are needed in any
country for troubleshooting prematurely distressed asphalt pavements
or training in asphalt technology I take up the challenge. All what I
have learned from colleagues like you has been helpful to me in
meeting such challenges.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
I am also thankful to my wife, Uma, who had to put up with my

obsession with asphalt all these years.
In closing, I again thank the AAPT Board of Directors and all

members for giving me this honor which I will cherish. Thanks.

Key Note Address
Review and Implications of IARC 103 for the Asphalt Pavement

Industry
Anthony J. Kriech, Director of Research, Heritage Research

Group
Bitumen (asphalt in the U.S.) was reviewed by the International

Agency for Research on Cancer (IARC) in October of 2011. This
eight day comprehensive review of the literature resulted in IARC
classifying Occupational Exposure to straight-run bitumens and their
emissions during road paving as possibly carcinogenic to humans
(Group 2B). IARC determined this classification by reviewing the
relevant peer-reviewed journals related to cancer and bitumen. IARC
conducts these reviews through scientific working groups composed
of four areas of expertise. These include exposure, human, animal and
mechanistic subgroups. These subgroups evaluate the studies within
each area of expertise before taking it to the overall group for
discussion.
The classifications are assigned based on epidemiological evidence in
humans, experimental evidence in animals, mechanistic and other
relevant studies. IARC classifies agents as follows: Category 1 -
there is evidence of cancer in humans. Category 2A is probably
carcinogenic to humans, includes agents with limited evidence in
humans and sufficient evidence in experimental animals. Category 2B
- is used for agents with limited evidence of carcinogenicity in humans
and less than sufficient evidence in animals. Category 3 - is used for
agents where there is inadequate evidence in humans and experimental
animals. Category 4 - is for agents with evidence suggesting lack of
carcinogenicity in humans and experimental animals. In some cases,
such as bitumen, classification can be upgraded based on strong
evidence from mechanistic and other relevant data.
Human Data
By far the most extensive epidemiological study of bitumen workers, a
study of over 80,000 European bitumen workers in eight countries, did
not find any link to occupational activities and excess cancers (Olsson
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

et al., 2010). The authors investigated the contribution of exposure to
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
bitumen, other occupational agents, and tobacco smoking to lung
cancer risk. We found no consistent evidence of an association
between indicators of either inhalation or dermal exposure to bitumen
and lung cancer risk. A sizable proportion of the excess mortality
from lung cancer relative to the general population observed in the
earlier cohort phase is likely attributable to high tobacco
consumption, and possibly to coal tar exposure, while other
occupational agents do not appear to play an important role.
Animal
A significant 24-month inhalation study was conducted by Fuhst et al.
(2007) to evaluate possible carcinogenic effects of paving bitumen
emissions on rats. The fume condensate used for this study was
matched to paving industry field results. Data from this study showed
no increase in the number of tumor-bearing animals in any of the
bitumen-exposed groups as compared to the clean air control group at
concentrations up to 172.5 mg/m3. Typical asphalt paving exposure is
<1 mg/m3. Field-matched bitumen fume condensates collected from
the headspace above paving tanks (Kriech et al., 2007) were evaluated
in two-year dermal carcinogenicity assays in mice (Clark et al., 2011).
The authors concluded that the paving fume condensate was not
carcinogenic under the test conditions.
Mechanistic
The IARC subgroup on Mechanistics and Other Relevant Studies
reviewed a number of mechanistic studies involving paving asphalt.
The focus of many of these studies related to the presence of
Polynuclear Aromatic Hydrocarbons (PAHs) and Compounds (PACs).
There was evidence that some of these compounds were mutagenic
and carcinogens producing genotoxic activities. IARC determined
that bitumen fume caused cellular stress and disrupted cellular defense
systems.
Other mechanistic studies showed genotoxic effects in both animals
and humans. This was the first time that IARCs mechanistic subgroup
upgraded an agent which was a complex mixture (asphalt) rather than
a simple agent (chemical), especially when human and animal
evidence was inadequate (negative).

Implications
IARC determines classification of hazards, not risk. Hazard and risk
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
distinctions are important. Hazard is the potential to cause harm; risk
is the likelihood of harm considering the degree of exposure. It is
possible to have a cancer hazard but not have a significant risk.
Normally risk is determined by regulatory agencies or established
scientific bodies.
The paving industry has been active for many years in reducing
exposure to workers. These include engineering controls on pavers as
well as warm mix technologies that reduce exposure through reducing
temperatures during paving operations. Recently published studies
also found that replacing diesel oil with biodiesel (B100) for cleaning
equipment and tools also greatly reduced workers exposure as well as
simply wearing gloves.
The industry should continue to look for ways to further reduce
exposure through reduction in temperatures. Improved methods of
measuring exposure focused on reduction in PAH exposures will help
reduce exposure in the workplace of paving operations.
References
Olsson, A., Kromhout, H., Agostini, M., Hansen, J., Lassen, C.F., Johansen, C.,
Kjaerheim, K., Langrd, S., Stckerm, I., Ahrens, W., Behrens, T., Lindbohl,
M.L., Heikkil, P., Heederik, D., Portengen, L., Shaham, J., Ferro, G., de Vocht,
F., Burstyn, I., Boffetta, P., A case-control study of lung cancer nested in a
cohort of European asphalt workers, Environ. Health Persp. 118 (2010) 1418-
1424.
Fuhst, R., Creutzenbeg, O., Ernst, H., Hansen, T., Pohlmann, G., Preiss, A. and
Rittinghausen, S., 24 Months Inhalation Carcinogenicity Study of Bitumen
Fumes in Wistar (WU) Rats, J. Occup. Environ. Hyg. 4(1) (2007) 20-43.
Kriech, A.J., Osborn, L.V., Wissel, H.L., Redman, A.P., Smith, LA. and Dobbs,
T.E., Generation of Bitumen Fumes Using Two Fume Generation Protocols
and Comparison to Worker Industrial Hygiene Exposures, J. Occup. Environ.
Hyg. 4(1) (2007) 6-19.
Clark, C.R., Burnett, D.M., Parker, C.M., Arp, E.W., Swanson, M.S., Minsavage,
G.D., Kriech, A.J., Osborn, L.V., Freeman, J.J., Barter, R.A., Newton, P.E., and
Stewart, C.W., Asphalt Fume Dermal Carcinogenicity Potential: I. Dermal
Carcinogenicity Evaluation of Asphalt (Bitumen) Fume Condensates, Regul.
Toxicol. Pharmicol. 61(1) (2011) 9-16.

Table of Contents
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
TECHNICAL SESSIONS
Development of a Failure Criterion for Asphalt Mixtures under
Fatigue Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
JUN ZHANG, MOHAMMADREZA SABOURI, MURTHY N. GUDDATI
and Y. RICHARD KIM
Effect of Particle Mobility on Aggregate Structure Formation
in Asphalt Mixtures . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
NIMA ROOHI SEFIDMAZGI, POUYA TEYMOURPOUR
and HUSSAIN U. BAHIA
Hot Mix Asphalt Pavement Frictional Resistance as a Function
of Aggregate Physical Properties . . . . . . . . . . . . . . . . . . . . 59
JOHN E. HADDOCK and JOAN P. O'BRIEN
IlliTCLow Temperature Cracking Model for Asphalt
Pavements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
ESHAN V. DAVE, WILLIAM G. BUTTLAR, SOFIE E. LEON, BEHZAD BEHNIA
and GLAUCIO H. PAULINO
Effect of Short-Term Conditioning and Long-Term Aging on
Permanent Deformation Characteristics of Asphalt Mixtures . . . 127
HALEH AZARI and ALAEDDIN MOHSENI
Evaluating Photocatalytic Asphalt Pavement Effectiveness in
Real World Environments through Developing Models:
A Statistical and Kinetic Study . . . . . . . . . . . . . . . . . . . . . 151
HEATHER DYLLA, SOMAYEH A SADI, MARWA HA SSAN
and LOUAY N. MOHAMMAD

Laboratory Conditioning Protocols for Warm-Mix Asphalt . . . . 177
FAN YIN, LORENA GARCIA CUCALON, AMY EPPS MARTIN,
EDITH ARAMBULA, ARIF CHOWDHURY and EUN SUG PARK
Merits of RAP Dominated Warm Mixed Flexible Pavement
Base Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
ISAAC L. HOWARD, JESSE D. DOYLE and BEN C. COX
Evaluation of High RAP-WMA Asphalt Rubber Mixtures . . . . . 253
WALAA S. MOGAWER, ALEXANDER J. AUSTERMAN, LOUAY MOHAMMAD
and M. EMIN KUTAY
Rutting and Moisture Damage Resistance of High RAP Warm
Mixed Asphalt: Loaded Wheel Tracking vs. Conventional
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
JESSE D. DOYLE and ISAAC L. HOWARD
Effect of Long-Term Aging on RAP Mixtures: Laboratory
Evaluation of Plant Produced Mixtures . . . . . . . . . . . . . . . . 327
JO SIA S DANIEL, NELSON GIBSON, SEAN TARBOX, AUDREY COPELAND
and ADRIAN ANDRIESCU
Evaluation of Use of Recycled Asphalt Shingles in HMA . . . . . . 367
FUJIE ZHOU, HONGSHENG LI, SHENG HU, ROBERT LEE, TOM SCULLION,
GERMAN CLAROS, JON EPPS and JOE BUTTON
Evaluating the Effect of Rejuvenators on the Degree of Blending
and Performance of High RAP, RAS, RAP/RAS Mixtures . . . . . 403
WALAA S. MOGAWER, ABBA S BOOSHEHRIAN, SIAVA SH VAHIDI
and ALEXANDER AUSTERMAN
Effect of Rubber Characteristics on Asphalt Binder
Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
J. RICHARD WILLIS, PAMELA TURNER, CLAYTON PLEMMONS,
CAROLINA RODEZNO, TOM ROSENMAYER, CODRIN DARANGA
and DOUG CARLSON
Understanding Mechanisms Leading to Asphalt Binder Fatigue
in the Dynamic Shear Rheometer (DSR) . . . . . . . . . . . . . . . 465
C A S S I E H I N T Z a n d H US S A I N B A H I A
Fatigue Endurance Limit for HMA Based on Healing . . . . . . . . 503
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
MENA I. SOULIMAN, WALEED ZEIADA, MICHAEL MAMLOUK

and KAMIL KALOUSH
Investigation of Fracture Properties of California Asphalt
Mixtures Using Semi-circular Bending and Beam
Fatigue Tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
HAMED HAKIMELAHI, SHADI SAADEH and JOHN HARVEY

Development and Validation of a Rutting Model for Asphalt
Mixtures Based on the Flow Number Test . . . . . . . . . . . . . . 557
MARIA C. RODEZNO and KAMIL E. KALOUSH
Using Small Scale Specimens for AMPT Dynamic Modulus and

Fatigue Tests. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579
XINJUN LI and NELSON GIBSON
A Mechanistic Permanent Deformation Model for Asphalt

Concrete in Compression . . . . . . . . . . . . . . . . . . . . . . . . 617
YEONG-TAE CHOI and Y. RICHARD KIM
SYMPOSIUMSUSTAINABLE ASPHALT TECHNOLOGIES

An Industry Perspective on Sustainable Asphalt
Technologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
HO W A R D M A R K S
Alternative Binders for the Production of Bituminous

Materials. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655
E L HA M F I N I
Recycling with RAP, RAS and Secondary Aggregates . . . . . . . . 661

J O S IA S D AN I EL
Sustainable Energy Use in Asphalt Production . . . . . . . . . . . . 683

AD A M HAN D
Binder Additives For Warm Mix Asphalt Technology . . . . . . . . 685

GAYLON L. BAUMGARDNER and GERALD R. REINKE
Warm Mix Asphalt TechnologyImplementation from a

State DOT Perspective . . . . . . . . . . . . . . . . . . . . . . . . . 711
B IL L S C H I E B E L
INTERNATIONAL FORUMSPECIFICATIONS AND USE OF

BITUMINOUS MATERIALS IN EUROPE
Specifications and Use of Bituminous Materials in Europe . . . . . 713 --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
MIKE SOUTHERN, JEAN-LUC DELORME, JEAN-PA SCAL PLANCHE

and JAN VAN DER ZWAN
List of Award Winners . . . . . . . . . . . . . . . . . . . . . . . . . . 721

List of Officers and Life Members . . . . . . . . . . . . . . . . . . . . 730
Index of Contributors and Discussors . . . . . . . . . . . . . . . . . . 731

CD TABLE OF CONTENTS
In addition to the papers listed above, the following items are also
included in the CD
AAPT List of Members
Presentations of Papers Presented at 2013 Annual Meeting
2013 Annual Meeting Pictures
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Development of a Failure Criterion
for Asphalt Mixtures under Fatigue Loading
Jun Zhang, Mohammadreza Sabouri, Murthy N. Guddati,
and Y. Richard Kim*
Department of Civil, Construction, & Environmental Engineering, North Carolina State

University, Raleigh, NC 27695-7908
ABSTRACT: The failure criterion defines the applicable region associated with the continuum
damage model and is important in characterizing the service life of asphalt mixtures. A proper
failure criterion should consistently predict the failure of a material that reaches macro-fracture.
A previously developed criterion that uses the viscoelastic continuum damage (VECD) model
exhibits high variability and is considered to be inefficient because it requires calibration tests at
different temperatures. In this paper, a new concept that involves released pseudo strain energy is
introduced. This released pseudo strain energy concept focuses on the dissipated energy that is
related to stiffness reduction only and is fully compatible with and predictable using the VECD
model. A characteristic relationship is found between the stable rate of pseudo energy release
during testing and the final fatigue life of the same mixture, independent of strain amplitude and
temperature. Based on these observations, a new failure criterion is proposed. The proposed
failure criterion combines the advantages of the VECD model and this characteristic relationship,
which both originate from fundamental mixture properties, and is able to predict the fatigue life of
asphalt concrete mixtures across different temperatures and strain amplitudes.
KEYWORDS: Failure criterion, fatigue, asphalt, pseudo energy, viscoelastic, damage
1. Introduction
Fatigue cracking due to repeated traffic loading is one of the most significant distresses
found in asphalt concrete pavements. Depending on the dominant mechanism, the
development and propagation of fatigue cracks inside a pavement structure can be
The oral presentation was made by Professor Kim.
This is a reproduction of an article whose final and definitive form has been published in Road
Materials and Pavement Design 2013 Taylor & Francis. The article is available online at:
http://dx.doi.org/ 10.1080/14680629.2013.812843
1
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

ZHANG, SABOURI, GUDDATI, KIM
categorized into two major phases: pre-localization and post-localization. In general, pre-
localization is manifested by the initiation and propagation of micro-cracks, and post-
failure represents the formation and propagation of macro-cracks.
The viscoelastic continuum damage (VECD) model is a continuum damage
mechanics-based model that has been applied successfully in simulating the behavior of
asphalt concrete mixtures during the pre-localization stages under both monotonic and
cyclic loading (Kim and Little, 1990; Lee and Kim, 1998; Daniel and Kim, 2002;
Underwood et al., 2006; Underwood et al., 2010). However, after reaching the fracture
region, the VECD model begins to lose the ability to characterize the material behavior
because of the formation of macro-cracks. The failure criterion defines the regions where
the VECD model is applicable and indicates the occurrence of fracture. In practice, the
failure criterion directly determines the fatigue life of the mixture, which is an essential
parameter in evaluating the mixtures ability to resist fatigue damage. Hence, it is crucial
to develop a failure criterion that can capture the failure of asphalt concrete accurately.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The definition of fatigue failure of asphalt concrete in laboratory tests has always
been controversial, especially in the controlled strain cyclic loading mode when no
catastrophic failure or fracture is observed. Traditional fatigue analysis defines failure as
the point at which the materials modulus value reduces to 50% of its initial value, and
the corresponding number of cycles is denoted as Nf50. This failure criterion is
considered to be convenient for the implementation of continuum damage models but
may not provide a consistent prediction for the damage state due to its arbitrary
assumption (i.e., the 50% reduction in modulus value). Figure 1 (a) presents the point of
Nf50 on the normalized stiffness diagram of an asphalt mixture under fatigue testing, and
it is obvious that the fatigue failure defined by this approach for this mixture is still far
from the real damage capacity of the material.
As an alternative, Reese suggested a new approach by defining failure as the upper
limit of the phase angle (Reese, 1997). During cyclic loading, the measured phase angle
of asphalt concrete usually exhibits a stable increase followed by a sharp decrease. The
number of cycles at failure, Nf, is defined as the cycle when this sharp decrease occurs
(Figure 2). This approach is generally believed to have more theoretical support than the
traditional approach, because the definition of failure is tracked through the materials
viscoelastic behavior, and the huge reverse in phase angle must represent a
transformation in the dominant mechanism inside the material, which is most probably
macro-fracture. Reeses approach also proves to be consistent on the stiffness diagram.
As shown in Figure 1 (b), for all the fatigue tests the failure point as determined by the
phase angle generally is located at the transition point between the stable decrease and
accelerated decrease in the stiffness diagram. Hence, it is concluded that the drop in
phase angle is the better indicator of fracture (compared to the traditional approach), and

Development of a Failure Criterion for Asphalt Mixtures under Fatigue Loading
that the phase angle should be taken as the target failure point that the developed
criterion needs to predict.
1.0 1.0
VTe30LC VTe30LC
0.8 Nf50 Nf
0.8
0.6 0.6
E*/E50
E*/E50
0.4 0.4
0.2 0.2
(a) (b )
0.0
0.0
0.E+00 5.E+04 1.E+05 2.E+05 2.E+05
0.E+00 5.E+04 1.E+05 2.E+05 2.E+05
Number of Cycles Number of Cycles
Figure 1. Determined failure point by different approaches:(a) traditional fatigue
analysis, and (b) phase angle approach.
55
Phase angle(degree)
50
45
40
35
Nf
30
0 200 400 600 800 1000 1200
Number of Cycles
Figure 2. Definition of failure by Reese's approach.

However, the change in time dependency is not included in most continuum damage
models, including the VECD model, because: first, the change in phase angle for asphalt
mixtures usually is observed not to be that significant compared to asphalt binders;
second, it is the degradation of stiffness that is of more interest for engineering purposes;
and finally, the exact mechanism that causes the variation in phase angle, whether it is
nonlinear viscoelasticity or damage due to fatigue cracking or plasticity, remains
unclear. In this way, the VECD model is not able to sense the variation in phase angle or
to recognize its drop. That is, the model itself cannot predict failure automatically, and
so the development of another failure criterion is necessary.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The development of the failure criterion starts with the simple assumption that the
failure of the material occurs at a critical damage state or at an equivalent critical
stiffness value. Hou tried to develop unified failure envelopes through experimental
observations of twelve different mixtures (Hou et al., 2010). The pseudo stiffness at
failure, which is the pseudo stiffness at the point of the phase angle drop, is expressed as
a function of reduced frequency, and the coefficients of the function are related to the
nominal maximum aggregate size (NMAS) of the mixture and whether or not reclaimed
asphalt pavement (RAP) is included. Even though the basic idea behind this approach
seems to be logically reasonable, the high variability in the experimental data makes it
very difficult to develop a convincing envelope among the different mixtures. In reality,
due to the complexity and heterogeneity of asphalt concrete, it is quite possible that a
slight alteration in the composition of the material, such as the asphalt content or
aggregate properties, may induce a substantial change in the distribution of the failure
stiffness. Hence, investigating failure stiffness may not be the best option for developing
a failure criterion.
Another type of research called the dissipated energy approach also is used in the
study of failure criteria for asphalt concrete. The dissipated energy approach tries to
relate the energy associated with damage to the final fatigue life and is said to be able to
reduce the variability commonly observed in fatigue testing (Castelo Branco et al.,
2008). Currently, most of the dissipated energy approaches focus on the quantification of
the dissipated energy via the hysteresis loop in the domain of stress-strain/stress-pseudo
strain. Ghuzlan and Carpenter (2000) developed the concept of ratio in dissipated energy
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
change (RDEC) and state that it is the change in dissipated energy between consecutive
cycles that represents damage. A unique relationship between the plateau value (PV) in
the RDEC and the final fatigue life also is proposed (Shen and Carpenter, 2005). The so-
called dissipated pseudo strain energy (DPSE) is introduced by uncoupling the effect of
viscoelasticity first via the correspondence principle (Schapery, 1984). Kim et al. (2003)
employed the concept of DPSE in the characterization of micro-crack growth. Masad et
al. (2008) applied the rate of change in the DPSE to the J-integral in fracture mechanics
and assessed the fatigue cracking potential of different asphalt mixtures.
Although these approaches have all shown positive results to some extent, one
significant obstacle remains that prevents any of them from being incorporated into the
VECD model as the failure criterion. Up to now, these dissipated energy approaches
have been investigated through experiments only, in which the histories of stress, strain
and phase angle are all known in advance. However, as mentioned earlier, the variation
in phase angle cannot be captured or predicted using the VECD model. Hence, the
dissipated energy inside the hysteresis loop cannot actually be calculated from the
model, and thus, any criterion based on it cannot be implemented.

Given the above considerations, the objective of this paper is to develop a failure
criterion that can be applied using the VECD model and can predict the fatigue failure
that is equivalent to the drop in phase angle in experimental observations with both
consistency and accuracy.
2. Materials and Test Methods
In this study, experiments were performed on four different types of mixtures to

investigate possible failure criteria. The first mixture is a Superpave surface 9.5 mm mix
(S9.5C) that is commonly used in North Carolina, and the other three are 9.5 mm
Superpave mixes from Vermont that are part of the New England RAP project. The
Vermont mixes are designated as VTeXXLC, with XX representing the percentage of
RAP. Table 1 presents a summary of the properties of these four mixtures.
All the specimens were compacted using the Superpave gyratory compactor (SGC)
with a diameter of 150 mm and to a height of 178 mm. To obtain specimens of uniform
air void distribution, these samples were cored and cut to a height of 150 mm and
diameter of 100 mm for testing. Prior to testing, the air void content was measured using
the CoreLok method for each specimen. All the test specimens used in this study have an
air void ratio within the range of 5.5 0.5% for the S9.5C mixture and 6.0 0.5% for
the VTeXXLC mixtures. To minimize the aging effect, specimens were sealed in plastic
bags and stored carefully in a cabinet if they were not tested immediately after
fabrication. No specimens were tested later than two weeks after they were cored and
cut.
Table 1. Summary of study mixture properties.
Target
NM AS
Mix Type Binder AC (%) RAP (%) Air Void
(mm)
(%)
S9.5C PG70-22 5 .2 0 9 .5 0 5 .5
VTe00LC PG64-28 6 .5 0 9 .5 0 6 .0
VTe30LC PG64-28 6 .6 1 9 .5 30 6 .0
VTe40LC PG64-28 6 .5 5 9 .5 40 6 .0
Two main tests were carried out in this study: (1) the dynamic modulus test was
performed to determine the linear viscoelastic characteristics using both AMPT and
MTS 810 closed loop servo hydraulic machines, and (2) the CX cyclic direct tension test
was implemented to describe the viscoelastic damage characteristics using both the
AMPT and MTS 810 machines.
5
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Dynamic modulus testing on the S9.5C mix was performed using the MTS 810
machine in load-controlled mode in axial tension-compression, following the protocol
given in AASHTO T342-11. Tests were completed for all mixtures at -10C, 5C, 20C,
40C, and 54C and at frequencies of 25, 10, 5, 1, 0.5, and 0.1 Hz. The VTeXXLC mix
dynamic modulus testing was performed using the AMPT in load-controlled mode in
axial compression following the protocol given in AASHTO TP 79-11. Tests were
completed for all mixtures at 5C, 20C, 40C, and 54C and at the same frequencies as
for the MTS. For tests using both machines, the testing order is from low to high
temperatures and from high to low frequencies in order to minimize damage to the
specimens. Load levels were determined by a trial and error process so that the resulting
strain amplitudes were between 50 and 70 microstrain to prevent damage to the
specimens.
All direct tension test specimens were glued to metal plates at both ends using epoxy
before they were placed in the machine for testing. Data acquisition programs were
prepared using LabView software for data collection and analysis. Vertical deformations
were measured in the middle of each specimen using four linear variable differential
transducers (LVDTs) at intervals of 90 degrees with a gauge length of 70 mm for the
AMPT specimens and 100 mm for the MTS specimens. DEVCON steel putty was
used to glue the steel end plates and targets for the LVDTs that were used for testing the
specimens.
The viscoelastic damage characteristics were determined by conducting CX cyclic
direct tension tests at three different temperatures and multiple strain levels for each
temperature with the frequency of 10 Hz. Fingerprint dynamic modulus tests were
conducted to check the variability of the test specimens before running the CX cyclic
direct tension tests. Checking the specimen variability is important so that the linear
viscoelastic properties obtained from the dynamic modulus tests can be used properly in
the VECD analysis. The CX cyclic direct tension test allows the specimen to run to
complete failure while still effectively limiting the viscoplasticity. Hence, viscoelastic
damage is the dominant mechanism. Details of the experimental procedure can be found
in Underwood et al. (2010) and Hou et al. (2010).
3. Development of Failure Criterion
The development of the failure criterion in this study starts with mix S9.5C. Because the
investigation focuses on viscoelastic damage, the signals of stress and strain are focused
mainly on magnitude.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

3.1 Preliminaries
For asphalt concrete under controlled CX cyclic loading, the stress and strain
histories can be described using the sinusoidal functions given in Equations 1 and 2.
i sin (t )
= [1 ]
= i sin (t + i ) [2 ]
where i , i , and i are the stress amplitude, strain amplitude, and phase angle measured
at cycle i, respectively.
If the stress and strain values are plotted against each other, the formed hysteresis
loop represents the dissipated energy, which is the total energy consumed during that
cycle. However, this consumed energy does not correspond to damage exclusively; it
also includes the energy that is associated with viscoelasticity. According to the
correspondence principle, the effect of viscoelasticity can be eliminated by replacing the
physical strain with the equivalent pseudo strain. For the physical strain given in
Equation 1, the corresponding pseudo strain R is written as:
= R i E * i si n ( t + ) [3 ]
where Ri is the pseudo strain amplitude at cycle i ; E is the undamaged complex

modulus for the given reduced frequency; and is the phase angle that is related to
viscoelasticity only.
The newly formed hysteresis loop in the stress-pseudo strain space is the DPSE
(WiR . Its value is equal to the area of the enclosed ellipse and can be calculated using
Equation 4.
=Wi R i iR sin (i ) [4 ]
During the controlled CX cyclic tests, the stiffness and phase angle both change with
the cycles, and so the DPSE will also change shape simultaneously. As the cycles
continue, the incline of the loop will gradually lower due to the reduction in stiffness,
and the enclosed area of the loop also will increase in size because of the increase in the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
phase angle. Figure 3 shows the typical trend of the pseudo hysteresis loop during
controlled CX cyclic testing.

CX cyclic test
Increase in
phase angle
Stiffness
reduction
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 3. Pseudo hysteresis loops for controlled CX cyclic tests.
3.2 Rate of Pseudo Strain Energy Release
For Equation 4, the calculation of the DPSE requires information about both the stiffness
and phase angle. In reality, the reduction in stiffness is known to be induced by damage.
However, with respect to the increase in phase angle, details regarding the underlying
mechanism remain unclear.
There are theories that attribute the increase of phase angle to nonlinear
viscoelasticity, damage or a combination of the two (Masad et al., 2008; Si et al., 2002).
Therefore, instead of calculating the DPSE, which involves a parameter that is both
uncertain and unpredictable, the dissipated pseudo strain energy is evaluated only as it
relates to the reduction in stiffness.
During cyclic loading, the maximum stored pseudo strain energy at each cycle
appears at the point of peak stress, which is also the point that corresponds to the
maximum pseudo strain and maximum experienced damage at that cycle. Hence, the
maximum pseudo strain energy reflects the materials current damage state and also its
ability to store energy. Based on the definition of pseudo stiffness for the VECD model
(Figure 4), the maximum stored pseudo strain energy at cycle can be calculated as:
1
(WmRax =
)i ( 0,ta )i ( R 0,ta )i [5 ]
2

where 0, and 0, are the tension amplitudes of the stress and pseudo strain. The
relationship between these two amplitudes is linked through the magnitude-based pseudo
stiffness, which is:
0,= F * R 0,ta [6 ]
ta
Hence, the maximum stored pseudo strain energy at cycle can be rewritten as:
1 [7 ]
(WmRax =
)i ( F )i ( R 0,ta )i 2
2
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 4. Schematic view of pseudo stiffness in the VECD model.
The maximum stored pseudo strain energy at each cycle reflects the ability of the
material to store energy at that particular time. As the damage accumulates, the material
loses the stored energy for the same magnitude of applied pseudo strain due to the
reduction in pseudo stiffness. The difference between the current stored maximu m
pseudo strain energy and the corresponding undamaged state is referred to as the total
released pseudo strain energy, which represents the cumulative loss of pseudo energy
due to the damage process, and is denoted as WCR .

Because the phase angle is excluded from the analysis, the elliptical hysteresis loop
in the stress-pseudo strain space can be simplified into a straight line that represents the
pseudo stiffness only. The concept of total released pseudo strain energy can be
interpreted in a two-step process. During the first step, it is assumed that all the damage
is temporarily frozen and the material is loaded to the applied maximum pseudo strain.
However, because of damage, the material cannot actually hold that amount of energy.
So, during the second step, the frozen damage is released and some of the stored energy
is released. In the stress-pseudo strain plot, the material will return to the point that
corresponds to the current damage state. The released energy during this second step is
the total released pseudo strain energy, which is presented by the shaded triangular area
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
in Figure 5. Its formulation is also given in Equation 8.
1 R
( 0,ta ) (1 Fi )
2
(WCR )i
= [8 ]
2 i
No damage = 1

With damage
Figure 5. Schematic representation of total released pseudo strain energy,WCR .
Equation 8 indicates that the total released pseudo strain energy is affected by two
factors: one is the applied pseudo strain amplitude, R0,ta , and the other is the reduction in
the material pseudo stiffness. In this way, the same amount of total released pseudo
strain energy can be obtained at either a high applied pseudo strain amplitude or with a
relatively large decrease in stiffness. From this point of view, the total released pseudo
strain energy, WCR , is a comprehensive energy measure that quantifies the dissipated
energy using both the external loading and the material itself.
10

Figure 6 shows the history of the total released pseudo strain energy, WCR ,
calculated from experiments using Equation 8. Because WCR measures the dissipated
energy in a cumulative sense, the rate of change in WCR with respect to the number of
cycles also is evaluated in the same plot. A local window of a fixed size, e.g., 50 cycles,
is chosen and moves along with the loading history. The rate of WCR is calculated at each
window by considering the history of WCR inside as an approximate linear function and
the value of the rate equal to the fitted slope of the linear function.
Figure 6. History of and its corresponding rate for controlled CX cyclic tests.
It is interesting to observe that for all the tests, the evolution of is generally
categorized into three regions. During the initial cycles, the rate of diminishes in a
relatively quick fashion, which usually relates to the transition state of on-specimen
stress/strain for CX cyclic test. When the specimen is loaded close to failure, the rate of
also jumps and it suggests a start of unstable damage propagation. If those two
severe regions are neglected, there is a significant stable region for the rate of ,
during which the rate of release of the pseudo strain energy, or released pseudo strain
energy per cycle, is almost constant. To be more precise, the stable region is defined as
the region where the relative variation in rate of per cycle stabilizes within certain
limits, e.g. 10%, between adjacent windows. Given that this stable region occupies the
majority of the loading history, its corresponding stable rate of pseudo strain energy
release characterizes the overall rate of energy loss during this fatigue test and is referred
to as ( . Because the damage propagates very slowly during cyclic loading, each state
of the material is taken as quasi-static, and the rate of energy release should be equal to
the rate of damage accumulation. Given that the total released pseudo strain energy is
calculated based on stiffness reduction, which is related definitively to damage, the
11
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

stable rate of the pseudo strain energy release ( can be regarded as the steady rate of
damage accumulation as well.
3.3 Characteristic Relationship
Because the stable rate of pseudo strain energy release (GR characterizes the overall rate
of damage accumulation during fatigue testing, it is reasonable to hypothesize that a
correlation must exist between GR and the final fatigue life, Nf (number of cycles to
failure), because the faster the damage accumulates, the quicker the material should fail.
Figure 7 presents the relationship between GR and the fatigue life Nf for mix S9.5C. It is
found that except for the case of quick failure with fatigue life less than 1000 cycles (one
at 5C and one at 27C), the GR and the fatigue life Nf are highly correlated, and
surprisingly, this correlation is not sensitive to temperature, which seems to be a
characteristic function of the given mixture.
1.0E+05
S9.5C-5C
S9.5C-19C
1.0E+04
S9.5C-27C
1.0E+03
GR
1.0E+02
1.0E+01
1.0E+00
1.0E-01
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
500 5000 50000 500000
Nf
Figure 7. Relationship between GR and Nf for mix S9.5C.
The concept behind the characteristic relationship is actually quite similar to that of
the damage characteristic proposed in the VECD model. Instead of stating that a unique
relationship exists between damage and stiffness for all viscoelastic damage dominating
cases, the theory states that a unique relationship exists between the rate of damage
accumulation and fatigue life, which means that no matter what the loading condition is,
once the rate of damage accumulation is determined, the corresponding fatigue life for
the given mixture is determined as well.
12

If the characteristic relationship between GR and Nf is proven to be present

consistently, then it can be used as the criterion for the VECD model to predict the
fatigue life for asphalt mixtures. Because the calculation of WCR relies only on the
variation in pseudo stiffness, its value can be evaluated from the VECD model directly.
Moreover, the corresponding stable energy release rate, GR, can be obtained as well. For
consistency, the same methodology is applied for the determination of GR, which
evaluates the rate of WCR in a moving local window first. The value of GR is picked up
when the fluctuations between adjacent windows are lower than a predefined threshold.
Once the value of GR is obtained, the fatigue life of the asphalt concrete can be predicted
immediately through the characteristic relationship. Figure 8 compares the results
between the predicted GR obtained from the VECD model and the measured GR obtained
from experiments. It is found that the two values are generally quite close to each other.
In reality, because GR is a rate-based value, it is sometimes predicted with even less error
compared to the point-to-point prediction of stiffness. The consistency between the
experimental measurements and the model prediction for GR ensures the linearity of the
proposed characteristic relationship in the model application and also confirms the
applicability of this characteristic relationship for the development of a failure criterion.
4000 100000
S9.5C-5C S9.5C-5C
Predicted Stable rate GR
Predicted Stable rate GR
S9.5C-19C S9.5C-19C
3000 S9.5C-27C S9.5C-27C
LOE LOE
2000 100
1000
(a) (b )
0 0.1
0 1000 2000 3000 4000 0.1 100 100000
Measured Stable rate GR Measured Stable rate GR
Figure 8. Comparison of GR between measured and predicted results: (a) arithmetic

scale, and (b) log scale.
4. Application of the Proposed Failure Criterion
4.1 Parameter Determination for the Characteristic Relationship
Because the characteristic relationship between GR and Nf relates to the fundamental

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
properties of the material in terms of damage tolerance, and is independent of
13

temperature, the development of this relationship can be simplified by conducting tests

at only one appropriate temperature. This temperature should be low enough to minimize
the effect of viscoplastic strain during the cyclic testing. The following rule of thumb is
used to determine the test temperature based on the performance grade (PG) of the base
binder:
T ( C ) 0.5 (High temperature binder PG grade - Low temperature binder PG grade) - 2 [9 ]
For a given mixture, the values of GR can be calculated using the VECD model as
follows:
(1) For a given pseudo strain history, predict the evolution of pseudo stiffness F
using the VECD model;
(2) Incorporate the value of F and pseudo strain tension amplitude R0,ta at each
cycle into Equation 8, and calculate the history of total released pseudo strain
energy WCR ;
(3) Identify the stable linear region during WCR and evaluate the stable rate of
pseudo strain energy release GR.
The fatigue life is obtained through experimental observations of the drop in phase
angle (based on Equation 9 where 13C is chosen as the test temperature for the
VTeXXLC mixtures with PG 64-22 binder, and 19C is chosen for the S9.5C mixture
with PG 70-22 binder). Figure 9 through 11 present the characteristic relationship
developed for mixtures VTe00LC, VTe30LC and VTe40LC, respectively. According to
these graphs, a significant relationship is found between GR and Nf for all three mixtures.
From the fits shown in Figure 9 through Figure 11, a power-law relationship is derived
between GR and Nf, which is later used to predict the fatigue life, as presented in Section
4 .2 .
14
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

10000
VTe00LC-13C
1000
100
R
G
10
R2 = 0.95
1
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)
Figure 9. Relationship between GR and Nf for VTe00LC mix at 13C.
10000
VTe30LC-13C
1000
100
GR
10
R2 = 1.00
1
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)
15
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

10000
VTe40LC-13C
1000
R
100
G
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
10
R2 = 0.99
1
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)
By plotting the relationships for all three VTeXXLC mixtures and the S9.5C mixtures
together, as shown in Figure 12, it is interesting to observe that the characteristic
relationships for the three VTe mixtures do not seem to have any significant difference.
This finding leads to the hypothesis that the characteristic relationship is not sensitive to
the percentage of RAP content. However, a difference does exist between the S9.5C
mixture and the other three mixtures, which may indicate that the relationship is affected
by other factors, such as binder characteristics, asphalt content or aggregate properties. It
should be noted that the fact that the characteristic relationship is not affected by the
amount of RAP does not necessarily mean that during a similar loading history the
fatigue life measurements of the mixtures with different percentages of RAP are the
same. For this phenomenon, the material behavior should be simulated in a pavement
system that may result in different stable rates of the pseudo strain energy release (GR)
for different mixtures with different percentages of RAP and, therefore, result in
different fatigue life measurements by using the same characteristic relationship.
16

10000
VTeXXLC-13C
1000 S9.5C-19C
R2 = 0.98
100
R
G
10 R2 = 0.99
0
1.E+03 1.E+04 1.E+05 1.E+06
Nf (Cycle)
R
Figure 12. Relationship between G and Nf for VTeXXLC mixes at 13C and S9.5C at
19C.
4.2 Fatigue Life Prediction
Once the characteristic relationships are obtained, they can be applied to predict the
fatigue life of asphalt mixtures for any other loading conditions. As mentioned earlier,
the characteristic relationship between the stable rate of pseudo energy release GR and
fatigue life Nf is not limited to only one temperature, as long as viscoelastic damage is
the dominant mechanism. Once the VECD parameters and the applied pseudo strain
history are known, the corresponding fatigue life can be readily determined using the
characteristic relationship between GR and Nf, as shown in Figure 9 to Figure 11.
Once the characteristic relationships are obtained, they can be applied to predict the
fatigue life of asphalt mixtures for any other loading conditions. As mentioned earlier,
the characteristic relationship between the stable rate of pseudo energy release GR and
fatigue life Nf is not limited to only one temperature, as long as viscoelastic damage is
the dominant mechanism. Once the VECD parameters and the applied pseudo strain
history are known, the corresponding fatigue life can be readily determined using the
characteristic relationship between GR and Nf, as shown in Figure 9 to Figure 11.
The advantage of using the characteristic relationship is that it can significantly
reduce the number of tests required for the calibration of the failure criterion. In order to
17
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

obtain the parameters of the previously developed criterion that is based on failure
stiffness, fatigue tests had to be conducted at multiple temperatures. With the new
characteristic relationship proposed here, calibration tests are required at only one
temperature, because the relationship among the different temperatures has been proven
to be statistically unique.
In order to verify the applicability and accuracy of the proposed failure criterion,
fatigue test predictions were performed for the three VTeXXLC mixtures by using the
derived characteristic relationship at 13C to predict the fatigue life for the other
conditions.
Table 2 presents the overall prediction results. According to this table, the average
prediction error is only around 20%, which is well within the variability of fatigue life,
whereas the percentage errors in only two cases exceed 50 percent.
400000 1000000
S9.5C S9.5C
VTe00LC VTe00LC
VTe30LC VTe30LC
300000
VTe40LC VTe40LC
Predicted Nf
Predicted Nf
LOE 100000
LOE
200000
10000
100000
(a) (b )
0 1000
0 100000 200000 300000 400000 1000 10000 100000 1000000
Measured Nf Measured Nf
Figure 13. Comparison of measured and predicted fatigue life:(a) arithmetic scale, and
(b) log scale.
Figure 13 presents the final predicted results for all four mixtures. The figure shows
that the proposed failure criterion is able to provide a reasonable prediction across all
temperatures. Hence, the proposed failure criterion is considered to be both efficient and
accurate.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
18

Table 2. Fatigue Life Prediction for VTeXXLC Mixtures at Multiple Conditions.
Specimen
Mix Type Temperature Initial Straina () Nf, Experiment GR Nf, Predicted Error %
Name
1054 343 4097 485 5668 38
7C 1041 260 26343 46 30195 15
1050 224 99482 12 76952 -23
1037 218 174026 3 210218 21
1043 379 14122 253 8981 -36
13C
1051 277 59610 21 52593 -12
VTe00LC 1041 409 5701 291 8147 43
1046 282 127821 4 182418 43
1044 301 107400 6 125981 17
1015 517 13114 167 12085 -8
20C
1045 465 15992 144 13399 -16
1013 678 6065 744 4187 -31
1009 674 6035 842 3833 -36
1026 222 65609 17 70634 8
1068 312 12518 122 16012 28
7C
1052 348 3697 520 5385 46
1070 266 21534 47 32891 53
1102 366 5507 555 5130 -7
1029 436 1898 1835 2087 10
13C
1044 244 33124 48 32478 -2
VTe30LC 1059 202 110263 8 119182 8
1027 244 43159 33 42562 -1
1043 876 1098 4767 1018 -7
1130 1032 897 7872 698 -22
20C 1046 700 3697 1310 2689 -27
1090 428 15713 144 14164 -10
1077 221 164122 4 210872 28
1028 213 230093 2 360499 57
1019 325 10304 179 12491 21
7C
1011 244 83362 10 88401 6
1025 218 130961 6 133333 2
1060 323 14921 180 12443 -17
13C
1033 330 16929 101 18394 9
1003 499 3498 1089 3652 4
VTe40LC
1038 310 83268 8 106295 28
1032 300 105617 7 113921 8
1007 541 12112 225 10687 -12
20C
1031 541 12921 237 10313 -20
1009 716 4298 1248 3331 -22
1010 726 6894 641 5242 -24
a
On-specimen strain at the 50th loading cycle
19
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

5. Summary and Conclusions
In this study, a new energy-based failure criterion is developed for the prediction of
fatigue life using the VECD model. The (power-law) characteristic relationship that
relates the stable rate of pseudo strain energy release GR and fatigue life Nf is found to be
a fundamental property of the mixture. In contrast to the existing failure criteria that
often require characterization tests at multiple temperatures, the proposed failure
criterion requires characterization tests at only a single temperature, thus significantly
reducing the costs associated with testing. Using this failure criterion and the S-VECD
model, the fatigue life of asphalt concrete at different temperatures and strain amplitudes
can be predicted from dynamic modulus tests and CX cyclic direct tension tests at four
strain amplitudes. The effectiveness of the proposed failure criterion is illustrated by
comparing predicted and actual fatigue life data for the four different mixes.
The proposed model is focused on predicting fatigue life under constant repeated
loading conditions at different temperatures. However, it may be possible to build on the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
proposed ideas to develop a complete VECD-based model for predicting fatigue life
under complicated stress histories that are experienced in the field. Development of such
a model is the focus of on-going research.
6. Acknowledgements
This research was funded by Texas A&M University as part of the Asphalt Research
Consortium project and by the University of New Hampshire as part of the New England
RAP project. The authors gratefully acknowledge their support.
7. References
Castelo Branco, V. T. F., E. Masad, A. Bhasin, and D. N. Little. (2008). Fatigue

Analysis of Asphalt Mixtures Independent of Mode of Loading. Proceedings, 87th
Annual Transportation Research Board Meeting, Transportation Research Board of
the National Academics, Washington, D.C.
Daniel, J. S. and Y. R. Kim. (2002). Development of a Simplified Fatigue Test and
Analysis Procedure Using a Viscoelastic Continuum Damage Model. Journal of
the Association of Asphalt Paving Technologists, Vol. 71, pp. 619650.
20

Ghuzlan, K. A. and S. H. Carpenter. (2000). Energy-Derived, Damage-Based Failure

Criterion for Fatigue Testing. Transportation Research Record, TRB, No. 1723,
pp. 141149.
Hou, T., B. S. Underwood, and Y. R. Kim. (2010). Fatigue Performance Prediction of
North Carolina Mixtures Using Simplified Viscoelastic Continuum Damage Model.
Journal of the Association of Asphalt Paving Technologists, Vol. 79, pp. 3580.
Kim, Y. R. and D. N. Little. (1990). One-Dimensional Constitutive Modeling of Asphalt
Concrete. ASCE Journal of Engineering Mechanics, Vol. 116, No. 4, pp. 751772.
Kim, Y. R., D. N. Little, and R. L. Lytton. (2003). Fatigue and Healing Characterization
of Asphalt Mixtures. Journal of Materials in Civil Engineering, Vol. 15, No. 1, pp.
7583.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Lee, H. J. and Y. R. Kim. (1998a). A Uniaxial Viscoelastic Constitutive Model for

Asphalt Concrete under Cyclic Loading. ASCE Journal of Engineering Mechanics,
Vol. 124, No. 1, pp. 3240.
Lee, H. J. and Y. R. Kim. (1998b). A Viscoelastic Continuum Damage Model of
Asphalt Concrete with Healing. ASCE Journal of Engineering Mechanics, Vol. 124,
No. 11, pp. 12241232.
Masad, E., V. T. F. Castelo Branco, D. N. Little, and R. L. Lytton. (2008). A Unified
Method for Analysis of Controlled-Strain and Controlled-Stress Fatigue Testing.
International Journal of Pavement Engineering, Vol. 9, No. 4, pp. 233246.
Reese, R. (1997). Properties of Aged Asphalt Binder Related to Asphalt Concrete
Fatigue Life. Journal of the Association of Asphalt Paving Technologists, Vol. 66,
pp. 604632.
Schapery, R. A. (1984). Correspondence Principles and a Generalized J-integral for
Large Deformation and Fracture Analysis of Viscoelastic Media. International
Journal of Fracture, Vol. 25, pp. 195223.
Shen, S. and S. H. Carpenter. (2005). Application of Dissipated Energy Concept in
Fatigue Endurance Limit Testing. Transportation Research Record, TRB, No.
1929, pp. 165173.
Si, Z., D. N. Little, and R. L. Lytton. (2002). Characterization of Microdamage and
Healing of Asphalt Concrete Mixtures. Journal of Materials in Civil Engineering,
Vol. 14, No. 6, pp. 461470.
21

Underwood, B. S., Y. R. Kim, and M. N. Guddati. (2006). Characterization and

Performance Prediction of ALF Mixtures Using a Viscoelastoplastic Continuum
Damage Model. Journal of the Association of Asphalt Paving Technologists, Vol.
75, pp. 577636.
Underwood, B. S., Y. R. Kim, and M. N. Guddati. (2010). Improved Calculation
Method of Damage Parameter in Viscoelastic Continuum Damage Model.
International Journal of Pavement Engineering. Vol. 11, pp. 459476.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
22

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Effect of Particle Mobility on Aggregate
Structure Formation in Asphalt Mixtures
Nima Roohi Sefidmazgi*, Pouya Teymourpour, and Hussain U. Bahia
Department of Civil & Environmental Engineering, University of Wisconsin-Madison,

Madison, WI, 53706.
ABSTRACT: During compaction of asphalt mixtures, aggregate structure starts building up by

proximity and direct contact of aggregates. In previous studies it has been shown that the
aggregate structure directly affects service performance. However, the mechanisms of aggregate
structure formation are not clearly understood. This study is focused on the mechanisms affecting
aggregate mobility during compaction and the effect of material properties on aggregate structure
formation. At the initial stages of compaction there is a relatively thick layer of mastic (i.e., mix of
binder and filler) between aggregates which allows for a shearing mobility in the mix if the mastic
viscosity is sufficiently low. However, as compaction proceeds, the mastic layer at the proximity
zone of aggregates becomes thinner due to high stress intensity and the higher viscosity of thin
mastic film or the aggregates dry contact effect increases the shearing resistance against
compaction (i.e., mix becomes locked). In this study mixes are compacted at different
temperatures using one base binder and three different modified binders. The quality of the
aggregate structure and packing throughout the compaction is characterized using 2-Dimensional
imaging of mixture sections and the total aggregate on aggregate proximity length is measured as
an indication of aggregate packing level. It is shown that for mixtures to obtain the maximum
packing, the compaction temperature should be picked based on mastic viscosity. The viscosity of
the mastic should be low enough for lubrication, but high enough to provide sufficient film
thickness at proximity zones and prevent locking of the mixture at the early stages of compaction.
KEYWORDS: Compaction, Imaging, Asphalt Mixture, Viscosity, Film Thickness, Mastic,

Proximity Zone.
The oral presentation was made by Mr. Roohi.
http://dx.doi.org/ 10.1080/14680629.2013.812844
23

ROOHI, TEYMOURPOUR, BAHIA
1. Introduction
In construction of asphalt pavements, workability is defined as the ease of placement and

consolidation (i.e., compaction) of the asphalt mixture in the field (Bennert et al., 2010).
Several studies have characterized workability of asphalt mixtures from diverse
perspectives (Bahia et al., 1998; 2001, Delgadilo and Bahia, 2008; Austerman et al.,
2009; Bennert et al., 2010). In the Superpave mix design procedure, mixing and
compaction temperatures are selected based on the rotational bulk viscosity of the
binders (AASHTO T 316, 2007). Relying on the selection of mixing and compaction
temperatures based on binder viscosity alone has been expanded upon in other studies
(Delgadilo and Bahia, 2008). Bennert et al. (2010) suggested that bulk viscosity may not
be sufficient and thus investigated the application of thin film rheology to ranking
asphalt mixture workability and binder selection. Hanz et al. (2011) also investigated the
role of viscosity, and his group reported that viscosity alone cannot explain the effect of
compaction temperatures on density growth. Hanz et al. suggested lubricity as a possibly
better or complimentary binder property to explain density growth and introduced a
lubricity testing fixture for measurement of coefficient of friction between steel balls
with a thin film of binder in between to quantify the effect of warm mix additives on
workability.
Review of tribology literature, however, indicates that the lubrication phenomenon is
not simple and depends on interaction of lubricant film thickness, pressure in the
interface zone, viscosity, and surface texture (Lu et al., 2006). Stribeck developed the
concept of lubrication and the effect of surface roughness and lubricant properties on the
lubrication mechanism. The "Stribeck curve", or "StribeckHersey curve" (named after
Richard Stribeck and Mayo D. Hersey), used to categorize the friction properties
between two surfaces, was developed in the first half of the 20th century. According to
the Stribeck curve, friction regimes for sliding lubricated surfaces have been broadly
categorized into three main regimes: 1) Boundary lubrication, 2) Mixed lubrication, and
3) Hydro-dynamic lubrication (Figure 1). Boundary regime is defined as a lubrication
mechanism in which two surfaces are mainly in direct contact with maximum interlock
of asperities providing a high coefficient of friction. However, in a mixed lubrication
regime, increased lubricant film thickness between two surfaces decreases the
interlocking of surface asperities and, as a result, the coefficient of friction in between.
The hydro-dynamic regime starts with complete separation of two surfaces when the
lubricant film is thick enough compared to surface texture. The coefficient of friction
increases again in the hydro-dynamic regime as depicted in the Stribeck curve (Figure 1)
due to a viscous drag phenomenon (Lu et al., 2006; Johnson et al., 1971).
Johnson et al. (1971) demonstrated that the applied load on a contact of solid-liquid-
solid is carried by a pressure distribution on the contact which is dependent on the
24
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Effect of Particle Mobility on Aggregate Structure Formation in Asphalt Mixtures
distribution of asperities, asperity heights (i.e., texture), liquid film thickness, and
mechanical properties.
Figure 1. Stribeck curve and lubrication regimes: Z, N and p are lubricant viscosity,
shearing speed, and normal pressure.
Therefore, it can be inferred from the literature that explanation of the asphalt
mixture compaction process and workability ranking using binder rheological properties
as a single factor can be misleading, especially for polymer modified binders.
Another challenge to using binder viscosity as the controlling factor is the inclusion
of mineral dust, which is a requirement in mixture design. Adding mineral filler changes
the rheological properties of asphalt binders significantly due to particle interaction and
absorption of asphalt binder (Faheem, 2009). The change in rheological properties, such
as viscosity, is dependent on several factors: filler type and gradation (i.e. surface area),
binder type, filler concentration, etc. (Richardson, 1915). A fundamental challenge in
characterizing mixture workability in terms of mastic viscosity is that a direct correlation
between binder viscosity and mastic viscosity may not exist, although limited research
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
has been conducted on the subject.

To quantify workability and study the effects of temperature, several researchers
have focused on understanding the compaction process using mixture testing (Bahia et
al., 1998; Gudimettla et al., 2003). Bahia et al. (1998) quantified workability of asphalt
mixtures using indices that can be obtained from the densification curve (i.e., Gmm versus
25

number of gyrations) with an energy concept which successfully characterized different

mixtures based on the required applied energy during compaction to achieve a target
density. Gudimettla et al. (2003) used a bucket mixer to measure the torque required for
mixing of material at different temperatures to select the appropriate mixing and
compaction temperatures. Although the general belief is that higher temperature should
provide better workability for asphalt mixtures (since the viscosities of binder and mastic
decrease with an increase in temperature), several studies have shown that the shear
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
resistance and density are not a simple nor consistent function of temperature. In a
number of studies it is reported that the profile of density versus compaction temperature
follows a parabolic trend (Bennert et al., 2010; Hurley and Prowell 2005a, 2005b, 2005c;
Gudimettla et al., 2003; Anderson, 2002). In these studies it is clear that increasing the
compaction temperature causes an increase in density to a specific level as shear
resistance of the mixture decreases, after which higher temperatures could cause shear
resistance to increase, and thus density to decrease. Therefore, the results shown in these
studies indicate that there is, for many mixtures (if not all), a zone of minimum shear
resistance at which density is highest.
The zone of minimum internal shear resistance of mixtures is commonly referred to
as the Tender Zone; difficulties in achieving the desired density in the field are
reported for compaction within this range of temperatures (Marker, 1977; Crawford
1986; Brown et al., 2000; Buchanan and Cooley, 2002). Buchanan and Cooley
demonstrated that for some materials the Tender Zone may occur in the typical range
of compaction temperatures commonly experienced in the field, and recommended to let
the mat cool down and use heavier rollers at lower temperatures. This phenomenon
could not be explained with binder viscosity testing at different temperatures.
The majority of mix designs and compaction literature consider a specific range of
density or air void content (i.e. 3.5-4.5 % for Superpave mix design, AASHTO M 323-
07) as a target for compaction and quantify workability based on that metric alone. The
basic assumption is that performance of the mixture can be indicated using density.
However, several studies have demonstrated that materials with the same density may
perform significantly different under service loading (Coenen et al., 2012; Roohi et al.,
2012; Olard, 2012). Based on the contact mechanism analysis, Zhu and Nodes (2000)
demonstrated that the transmission of load in the asphalt mixture is mainly determined
by the interaction of aggregates and binder at the proximity zones of adjacent aggregates.
According to this concept, Roohi et al. showed that specific microstructural indices, such
as total aggregate proximity length in the aggregate skeleton (i.e., in 2-Dimensional
section images of mixtures), can be a better indicator of mixture load bearing aggregate
structure and performance. The aggregate structure indices show sensitivity to
compaction conditions such as compaction temperature, method, and mix design
properties (Tashman et al., 2001; Bahia and Roohi, 2012).
26

Therefore, there is a need for a more comprehensive understanding of the mechanics

governing the compaction process to better understand the aggregate on aggregate
lubrication, referred to here as aggregate mobility. The mechanisms resisting
aggregate mobility control the formation of the aggregate skeleton, the main load
bearing structure in asphalt mixtures, thus a deeper understanding of this process can be
invaluable for the control and optimization of the construction process and enhancement
of pavement performance.
2. Mechanisms Controlling Particle Mobility during Compaction
When the compaction load is initially applied on loose mix, packing of aggregates
increases rapidly as the applied stress can overcome the flow resistance of the mastic. In
this stage, mastic viscosity could be the dominant factor since aggregate interlock is
minimal. However, with subsequent aggregate skeleton formation, the rate of bulk
material deformation decreases significantly and tends toward zero. Li and Gibson
(2011) used the gyratory compactor to show that the number of gyrations needed for
locking of mixtures is dependent on the material properties and mix design. In addition,
Li and Gibson reported that most of mixes compacted in the study locked at gyrations of
60 or less and the continued compaction to design gyrations of 100 did not change the
density of mixes significantly. However, there is not a clear fundamental explanation for
the difference in locking point of different mixes.
This study is focused on mixture compaction mechanisms based on the geometrical
and mechanical properties of aggregate on aggregate proximity zones. For the purposes
of this study, aggregates are determined to be in proximity if the perimeters of two
adjacent aggregates are within a predefined spacing in a 2-Dimensional section. It is
assumed that when the shearing stress applied on the mixture is higher than the shearing
strength of aggregate proximity zones (i.e., contact stiffness) or the internal friction of
the lubricant in between, the mixture plastically deforms.
Contact friction of two surfaces with a lubricant in between is known to be a function
of normal load applied, film thickness, viscosity of the lubricant, and morphological
properties of the surfaces, such as texture (Lu et al., 2006). In a contact, given other
factors to be constant, when decreasing the viscosity of the lubricant (such as increasing
temperature of asphalt binder), it is hypothesized that two phenomena can happen
simultaneously:
Reduction of friction between the two surfaces as a result of reduction in viscosity of
the lubricant becasuse the energy needed to overcome the internal friction of the
lubricant is lower.
27
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Reduction of film thickness and increase in proximity of the two surfaces due to the
lubricant being squeezed out in the case of low or zero confinement. When the surfaces
of the particles are smooth, the reduction of film thickness to a very thin layer (i.e., in
the nanometer order of magnitude) causes the effective viscosity of the lubricant to
increase, which increases the shearing resistance between two surfaces (Luengo et al.,
1997). In case of rough surfaces such as natural crushed aggregates, high stress
concentrations at the asperities remove the lubricant and the solid surfaces with a high
coefficient of friction will be directly in contact. This process is called dry contacts
in this study. This condition is also termed as boundary lubrication in the tribology
science studies (Lu et al., 2006, Figure 1). The rate of squeezing out of lubricant is
directly related to the viscosity of the material (Macosko, 1994). Therefore, if the
viscosity of the lubricant is not sufficiently high to resist squeezing out, the friction
between two surfaces may increase significantly as a result of this dry contact effect.
Therefore, it can be hypothesized that the effective friction between two surfaces in a
proximity zone is a parabolic function of lubricant viscosity as shown in Figure 2. This
trend is caused by the competing mechanisms of decreasing film thickness and an
increase in viscosity. The initial reduction in friction is caused by increased film
thickness and the final increase in friction is caused by an increase in lubricant viscosity,
as shown in Figure 1.
Figure 2. Proposed schematic of change in friction at proximity zones as a function of

temperature.
Compaction of asphalt mixtures can be envisioned as packing of aggregates that are

lubricated by the asphalt mastic. Therefore, for a better packing of aggregates, the
28
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

viscosity of the mastic should be low enough to allow sufficient workability, but high
enough to maintain adequate film thickness at proximity zones to avoid dry contact or
the thin film viscosity effect.
The focus of this study is on the packing of aggregate particles during compaction.
However, there are some other concerns in compaction temperature selection, such as
binder drain down at very high temperatures, aggregate segregation and low mastic film
thickness (which may cause durability problems). These concerns are outside the scope
of this study.
In this study, it is hypothesized that viscosity of mastic, assuming other factors to be
constant, controls the mobility and structure formation of aggregate skeleton. Therefore,
as shown in Figure 2, changing the compaction temperature will result in different
aggregate packing due to the change in friction at aggregate proximity zones. This
implies the existence of an optimum packing at the lowest level of proximity zone
friction.
3. Study Objectives
The main objectives of this study are to determine the possible mechanisms of aggregate
structure formation during compaction of asphalt mixtures and the effect of mastic
rheology and film thickness on aggregate mobility.
4. Materials and Test Methods
4.1. Mixture Sample Preparation and Testing
To isolate the effect of mastic as the lubricant phase, mixes with significantly different
mastic viscosities at a specified compaction temperature were produced. The gradation
(Figure 3) has been designed for the purposes of this study from aggregate sources
commonly used for HMA production in Wisconsin (i.e., granite).
29
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

100
80
% Passing
60
40
20
0
0 0 .0 7 5 0 .6 1 .1 8 2 .3 6 4 .7 5 9 .5 1 2 .5 1 9 .0
Sieve Size (mm)
Coarse MMaxaxDDenenisstiyty Control Points
Figure 3. Aggregate gradation used in this study for mix design.
To cover a wide range of mastic viscosities, four different binders were used for
compaction: a base neat binder and three modified binders with three different modifiers
(i.e., a styrene butadiene styrene based elastomer modifier (S), a polyethylene based
plastomer (C), and the mix of elastomer and plastomer modifiers (H)). The level of
modification was selected to obtain high true grade of 771C for all modified binders.
The binder designations and performance grades are shown in Table 1.
Table 1. Binder designations and Performance Grades (PGs).
Binder PG
Ne a t 64-22
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
S 76-22
C 76-22
H 76-22
The reference mixing and compaction temperatures were selected as 155C and
145C, respectively, based on the viscosity-temperature profile (i.e., using the Superpave
mix design procedure) for the neat binder and was kept consistent for all mixes.
Additionally, samples were mixed and compacted at several temperatures providing
significantly different orders of magnitude of mastic viscosity. The mixing temperature
30

was always kept 10C higher than the desired compaction temperature. The compaction
temperatures are shown in Table 2. All mixes are compacted with 100 gyrations, which
is the required Wisconsin E-10 mix design level of gyrations (Ndes) with 5.5% asphalt
content by mass.
Table 2. Compaction temperatures.
Mix Compaction Temperatures (C)

Ne a t 65, 115, 145
S 115, 145, 165
C 65, 95, 145, 165
H 65, 115, 145, 165
After compaction, the samples were cut vertically in three sections leading to
attainment of six 2-Dimensional images (Figure 4), with one cutting section at the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
middle of the sample and two at 25.4 mm from the middle. All mixes were scanned
according to the procedure of Roohi et al. (2012), and image analysis was performed for
aggregate structural characterization (i.e., using Image Processing and Analysis Software
(IPAS2), Roohi et al, 2012). The index as a result of the image analysis procedure to
represent the packing condition in mixtures is the aggregate total proximity zone length
in 2-Dimensional images. In the imaging software, proximity is defined when two
aggregate perimeter pixels are within a distance specified by the user (i.e., 0.1 mm in this
study) and all of the pixels of the two aggregate perimeters within this distance are
captured. These pixels form a line referred to as a contact line in Roohi et al. (2012) and
proximity line here (Figure 5). Higher total contact length (i.e., proximity length)
represents higher aggregate interlock and packing in the mixture and better rutting
performance as a result. Based on image analysis results, the total aggregate proximity
length profile for different compaction temperatures (i.e., different mastic viscosities)
was obtained.
31

Figure 4. Cutting sections (Roohi et al. 2012).

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 5. Proximity line.

In addition to compaction at different temperatures, mixes were compacted at
different levels of effort (i.e., 10, 20, 40, 80 and 100 gyrations) at the temperature of
145C and image analysis was performed. The objective was to investigate the
difference in trend of aggregate packing attainment for materials with different
viscosities during compaction. The air void content of mixes at different gyrations is
shown in Table 3.
32

Table 3. Air voids of mixes compacted at different number of gyrations (compaction

temperature = 145C).
Air Voids, %
Mix
10 gy 20 gy 40 gy 80 gy 100 gy
Ne a t 1 2 .1 8 .8 6 .5 6 .1 4 .5
C 1 0 .0 7 .0 5 .8 3 .8 3 .5
S 1 1 .5 7 .8 6 .3 3 .8 4 .0
H 1 1 .3 7 .2 4 .9 3 .8 3 .0
4.2. Binder and Mastic Preparation and Testing
It is hypothesized that the viscosity of an unconfined/semi confined lubricant between

two surfaces controls the internal frictional behavior of asphalt mixtures through the two
aforementioned mechanisms before the boundary lubrication phase (i.e., dry contact) at
which the locking of the mixture occurs. In order to test this hypothesis and study the
effect of lubricant viscosity on aggregate packing, viscosity testing of binders and
mastics was performed.
Four binders were aged using the Rolling Thin-Film Oven (RTFO) for short term
aging during mixing and compaction. The viscosity testing was performed at all
compaction temperatures with shear rates of 0.01, 0.1, 1, 10 and 100 (sec-1) to cover a
wide range of shear rates that were proposed in different studies (Bahia et al., 2001;
Delgadillo and Bahia, 2008) with rheometer limitation considerations. Binders were
tested using a Bob and Cup fixture (Figure 6) with a binder film thickness of 1 mm. The
samples were conditioned at the compaction temperatures for 15 minutes in the fixture.
The viscosity measurements were performed for three minutes for each shear rate
starting with the lowest shear rate (i.e. 0.01 sec-1).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
33

Figure 6. Bob and cup fixture.
Many of the previous studies on asphalt mixture workability and compaction

mechanisms are focused on rheological behavior (e.g., viscosity) of binder as the most
prominent mix constituent that affects workability. Yet, several studies have shown that
adding filler (i.e., passing sieve the No. 200) to binder may significantly change the
mechanical and rheological behavior of the binder (Faheem, 2009; Clopotel, 2012),
since filler particles reinforce the binder and absorb specific molecular components of
the binder (e.g., asphaltenes) depending on the type, gradation, concentration, and
specific surface area of the filler particles. Therefore, for the purposes of this study,
mastics with the same fillers used in mixture production were prepared for viscosity
testing. The binder content of the mastic samples was determined based on the specific
surface area of the aggregates. With a constant binder film thickness assumption for
aggregates of different sizes in the mix, the portion of binder content of mixture that
coated the filler particles was determined (i.e., depending on the ratio of surface area of
filler particles to the surface area of the whole aggregate blend). Based on such
calculations, the binder content of the mastics was 55.3 % by weight. All the mastics
were mixed using un-aged binders and conditioned at the same mixing and compaction
temperatures as the mixtures.
The viscosity testing procedures performed on mastics were the same as binder
viscosity testing using the bob and cup fixture with the same shear rates. Considering
that the size of particles in mastic is finer than 0.075 mm, which is less than 10% of the
sample film thickness in the bob and cup fixture, the same fixture used for binder testing
can be used for mastic testing since there is no interference expected with sample
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
geometry.
34

5. Results and Discussion
A previous study (Roohi et al., 2012) demonstrated that mixtures with the same
gradation design but different binders may have significantly different aggregate
structures. It is hypothesized in this study that the difference in structures can be
explained with mastic viscosity. Therefore, mixtures were produced at the same
temperature (i.e., 145C) and different compaction efforts (i.e., 10, 20, 40, 80, and 100
gyrations) to monitor the aggregate structure formation for different binders. All the
mixes were cut and scanned, and image analysis was performed. The total aggregate on
aggregate proximity zone lengths, as an indication of aggregate internal structure and
packing for images of each mix, were determined and are depicted in Figure 7.
900
Total Proximity Zone Length (mm/100cm2)
800
700
600
500
400
300
200
100
0
0 20 40 60 80 100 120
Gyrations
Neat C H S
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 7. Total proximity zone length for mixes at different number of gyrations.
The results show that the aggregate packing in the neat mixture is mostly obtained
during the first 10 gyrations and the change in aggregate structure with application of
more compaction effort (i.e., up to 100 gyrations) is insignificant. However, for C and
S mixes, the aggregate structure formation took up to 20 gyrations. The number of
gyrations for aggregate structure formation in the mixture designated as H was 40 and
35

the structure was formed in a more gradual manner and locked later compared to the
other mixtures.
To help explain the trends of Figure 7, all four binders and mastics were tested for
viscosity using the bob and cup fixture at the temperature of 145C (i.e., compaction
temperature) and at different shear rates. The results are shown in Figure 8. It should be
noted that the average standard deviation of the viscosity measurement in the rheometer
using the bob and cup geometry is +/- 0.005 Pas.
Binder
10
Ne a t C S H
Viscosity (Pa.s)
0 .1
0.01 0 .1 1 10 100
Shear Rate (sec-1)
( a)
Mastic
1000
Ne a t C S H
100
Viscosity (Pa.s)
10
0 .1
0.01 0 .1 1 10 100
Shear Rate (sec-1)
(b )
Figure 8. Viscosity at 145C: a) Binder, b) Mastic.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
36

The viscosity results show different rankings for binders compared to mastics.
According to the binder results at 145C, Neat binder showed the lowest viscosity and
C, H, and S have higher viscosities, in respective order. However, mastics are shear
rate dependent and the ranking at low shear rates is different than the binder ranking:
Neat, S, C, and H from low to high viscosity.
As noted, the lubricant phase in a mixture is mastic; therefore, it is more reliable to
consider mastic viscosity for workability properties of mixtures. In addition, it was
shown in the NCHRP 9-10 study (Bahia et al., 2001) that the average shear rate
occurring in mixtures after initial gyrations is low according to the axial strain rate. The
mastic viscous behavior is not easily predicted from the binder viscosity alone at lo w
shear rates. Therefore, for a better characterization of aggregate structure formation
during compaction, the low shear rate viscosity values of mastics (i.e., 0.01 to 0.1 (sec-1))
are chosen.
Mastics with higher viscosity can resist the stress condition at proximity zones of
aggregates and resist the squeeze out from proximity zones longer (more gyrations) than
soft mastics, which provide aggregate mobility for a longer period during compaction. In
other words, locking of mixtures due to dry contacts (i.e., direct contact of aggregate
asperities) or thin film high viscosity happens sooner in mixes with very soft mastics.
However, the mastic viscosity should be low enough to allow for lubrication and
shearing deformation between aggregates with the compaction effort applied. This
theory is in agreement with the results. The H mastic showed the highest viscosity
among others at the shear rate of 0.01 and 0.1 (sec-1) and the neat mastic was the lowest,
which explains the gradual formation of structure in the H mixture and rapid locking in
the neat mixture.
It is further believed that the film thickness of mastic (or mortar) between coarse
aggregates decreases rapidly in the neat mix relative to the other mixtures as it is a
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
function of mastic or mortar viscosity (Macosko, 1994). To prove this hypothesis,

mortar film thickness in four mixtures is estimated at the same number of gyrations
depicted in Figure 7. (Mortar is defined here as binder mixed with aggregates passing the
No. 16 sieve, since aggregates finer than 1.18 mm are not clearly visible in the images
used for analysis in this study.)
Aggregates are naturally rough particles. Therefore, the film thickness of material
(e.g., binder, mastic, or mortar) between surfaces of two aggregates is not constant
(Figure 9). In addition, the friction between two aggregates increases significantly when
asperities on the two surfaces begin contacting each other as the coefficient of friction
between bare aggregates is considerably higher than the internal friction of material in
between (simply meaning that more shearing load is needed to drag an aggregate on
another aggregate when there is no lubricant between them). Therefore, the minimum
37

mortar film thickness between each pair of aggregates (i.e., at proximity zones) is
calculated as an index to represent the trend of aggregate contact formation during
compaction.
Figure 9. Black and white image of a mixture representing minimum film thickness
between aggregates 1 and 2.
Histograms of mortar minimum film thickness distribution between aggregates for

the H mix during compaction are shown in Figure 10. The histogram for the 10
gyrations level of compaction is flatly distributed, however by increasing the number of
gyrations, the film thickness of mortar between aggregates was decreased and the
histogram is distributed in a narrower range of film thickness (i.e., less than 0.1 mm
which is considered as proximity zone in image analysis).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
38

20
10 gy
18
20 gy
16
40 gy
14
80 gy
% Occurance
12
10 100 gy
8
6
4
2
0
0 0 .1 0 .2 0 .3 0 .4 0 .5 0 .6
Film Thickness (mm)
Figure 10. Evolution of mortar film thickness index at proximity zones of aggregates for
different number of gyrations.
The same analysis shown in Figure 10 was performed on the other three mixtures.
For a more appropriate comparison of different mixtures in terms of film thickness at
proximity zones evolution during the compaction process, a weighted average of film
thickness for each histogram (i.e., at each number of gyrations) is calculated. Results
show that (Figure 11) the film thickness of mortar between aggregates gradually
decreases for the H mix during compaction up to 40 gyrations; however, the neat mix
lost its mortar film between coarse aggregates before 10 gyrations, which is in
agreement with the total proximity zone length and mastic viscosity data.
39
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

2 .5
H C S Ne a t
Film Thickness (mm) 2 .0
1 .5
1 .0
0 .5
0 .0
10 20 40 80 100
Gyrations
Figure 11. Mortar film thickness index.

Mastic viscosities at 145C and shear rates of 0.01 (sec-1) and 0.1 (sec-1) are plotted
in Figure 12 against film thickness at each gyration level. There is a correlation between
film thickness and low shear viscosity of mastics at compaction efforts of 40 gyrations
or less, that is considered the aggregate structure formation stage at which the
controlling lubrication mechanisms are hydro-dynamic and mixed lubrication. This
confirms the hypothesis that mastics (or mortars) with higher viscosities keep thicker
mastic layers between aggregates and it takes a higher number of gyrations (i.e., longer
time) to squeeze mastic out of proximity zones under the stress conditions.
To further validate the hypothesis, mastic viscosity values at different shear rates are
plotted versus total proximity length of the four mixes compacted to 100 gyrations in
Figure 13. The results show that there is a good correlation between total proximity
length as a representation of aggregate packing with mastic viscosity at low shear rates
(i.e., 0.01 and 0.1 sec-1). However, there is no significant correlation at high shear rates
(i.e., 10 and 100 (sec-1)).
40
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

0.01 (sec-1) shear rate - 145 C

1000
R = 0.78 R = 0.90 R = 0.95

100
Viscosity (Pa.s)
10 10 Gyrations
20 Gyrations
1 40 Gyrations
80 Gyrations
100 Gyrations
0 .1
0 .0
1 .0 1 .5 0 .5
2 .0 2 .5
Film Thickness (mm)
( a)
Figure 12a. Low shear viscosity versus average proximity zones film thickness at
different compaction levels: a) 0.01 sec-1 shear rate, b) 0.1 sec-1 shear rate.
0.1 (sec-1) shear rate - 145 C

100
R = 0.80 R = 0.88 R = 0.95

Viscosity (Pa.s)
10
10 Gyrations
20 Gyrations
40 Gyrations
80 Gyrations
100 Gyrations
1
0 .0 0 .5 1 .0 1 .5 2 .0 2 .5
Film Thickness (mm)
(b )
Figure 12b. Low shear viscosity versus average proximity zones film thickness at
different compaction levels: a) 0.01 sec-1 shear rate, b) 0.1 sec-1 shear rate.
41
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

1 (1/sec) 10 (1/sec)
121 0.1 (1/sec) 0.01 (1/sec)
101
Viscosity (Pa.s)
R = 1.00
81
61
41
R = 0.99
21
R = 0.75 R = 0.13
1
500 600 700 800 900 1000 1100
Total Proximity Zone Length (mm/ 100 cm2)
Figure 13. Viscosity versus total aggregate proximity zone length at 145C.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
It can be concluded that among the four binders used for this study, H provided the
best workability and packing of aggregates since the viscosity of mastic H at the
compaction temperature was high enough to provide sufficient film of mastic between
aggregates and consequently aggregate mobility to a higher number of gyrations.
However, in the neat mixture with the lowest mastic viscosity, the film of mastic at the
aggregate proximities is squeezed out at less than 10 gyrations, which possibly resulted
in dry contacts or a very thin layer of mastic with high viscosity that increased friction
between aggregates significantly. The C and S mixes showed a transition from the
H mix (i.e., high mastic viscosity, gradual locking of mix) to the neat mix (i.e., low
mastic viscosity, rapid locking of mix) from the mastic viscosity and aggregate packing
point of view.
In contrast, using mastics with extremely high viscosities is expected to result in
lower workability and lower mobility of particles. Thus, there should be a high and low
limit for viscosity of mastics within which sufficient aggregate mobility during
compaction is achieved. Compaction outside this range of viscosities should be avoided.
To investigate this concept, mixes were produced and compacted at different compaction
temperatures as shown in Table 2. Compaction of mixes at temperatures from 65C to
165C provided a wide range of mastic and binder viscosities.
The image analysis results of mixes compacted in different temperatures is depicted
in Figure 14. The analysis for the S mix compacted at 65C was not performed since it
had significant aggregate coating problems.
42

As shown in Figure 14, there is a maximum proximity zone length (i.e. Optimum
packing) at a unique temperature for each mixture. The maximum values of total
proximity length for different mixes are approximately the same. This is possibly due to
the fact that the same gradation is used and thus according to the mastic viscosities at
different temperatures, ultimately a maximum packing of particles is achieved. It is
possible that different mastics provided similar viscosity ranges at different temperatures
and at that specific range for the gradation used in the four mixes optimum packing has
been achieved. It has been shown in a previous study that there is a good relation
between total proximity length and rutting performance (Roohi et al., 2011). Therefore,
this finding shows that the structure needed to resist the service loading can be obtained
by compaction of any of the four mixes only by selecting the correct compaction
temperature.
4,500
4,000
3,500
3,000
2,500
2,000
1,500
1,000
60 80 100 120 140 160
Compaction Temp. (C)
Control C S H
Figure 14. Total proximity zone length - compaction temperature profile.
To determine the effect of viscosity on total proximity zone length at different

compaction temperatures, binders and mastics were tested using the bob and cup
geometry in the DSR. The viscosity results for the binders at different temperatures and
shear rates are shown in Figure 15. According to the results, all binders show Newtonian
behavior (shear rate independent viscosity), specifically at high temperatures, although
for H and C at 65C the materials showed slight shear rate dependency of viscosity.
For all practical purposes, the viscosity ranking of binders at temperatures above 115oC
can be considered independent of shear rate.
43
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Neat - binder
10000
1000
Viscosity (Pa.s)
100
10
0 .1
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0.01 0 .1 1 10 100
Shear Rate (sec-1)
Ne a t 6 5 C Ne a t 1 1 5 C Ne a t 1 4 5 C
( a)
C - binder
10,000
1,000
Viscosity (Pa.s)
100
10
0
0.01 0 .1 1 10 100
Shear Rate (sec-1)
C 65 C C 95 C C 145 C C 165 C
(b )
44

S - binder
10,000
1,000
Viscosity (Pa.s)
100
10
0
0.01 0 .1 1 10 100
Shear Rate (sec-1)
S 65 C S 115 C S 145 C S 165 C
( c)
H - binder
10,000
1,000
Viscosity (Pa.s)
100
10
0
0.01 0 .1 1 10 100
Shear Rate (sec-1)
H 65 C H 115 C H 145 C H 165 C
(d )
Figure 15. Bob and cup binder viscosity results at different temperatures and shear
rates: (a) Neat binder (b) C binder (c) S binder (d) H binder.
45
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Mastics were tested at the same shear rates and temperatures as binders using the bob
and cup fixture (Figure 16). The test results for modified mastics show higher shear rate
dependency (Specifically for C and H mastics) compared to binders at different
temperatures. The other important difference of mastic behavior compared to binders is
that the viscosity ranking of mastics does not follow that of binders. Based on the tests
performed on the four mastics, H is the most viscous material and the neat binder
showed the lowest viscosity at different temperatures. Except for the neat mastic,
viscosity measurements at 65C were out of the torque limit of the rheometer.
Neat - mastic
100000
10000 65 115 145
Viscosity (Pa.s)
1000
100
10
1
0 .1
0.01 0 .1 1 10 100
Shear Rate (sec-1)
( a)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
46

C - mastic
1000
95 145 165
100
Viscosity (Pa.s)
10
0
0.01 0 .1 1 10 100
Shear Rate (sec-1)
(b )
S - mastic
100000
115 145 165
10000
Viscosity (Pa.s)
1000
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
100
10
0 .1
0.01 0 .1 1 10 100
Shear Rate (sec-1)
( c)
47

H - mastic
100000
115 145 165
10000
Viscosity (Pa.s)
1000
100
10
0 .1
0.01 0 .1 1 10 100
Shear Rate (sec-1)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(d )
Figure 16. Bob and cup mastic viscosity results at different temperatures and shear
rates: (a) Neat mastic (b) C mastic (c) S mastic (d) H mastic.
As explained in previous sections, low shear viscosity of mastics is proposed as a
factor affecting mixture workability at different temperatures because it correlates better
with aggregate packing. In addition, there is always a blend of binder and filler (i.e.,
mastic) among coarser aggregates acting as a lubricant phase. However, it is shown in
this study that the total aggregate proximity zone length values do not demonstrate a
clear trend with increasing viscosity as caused by reducing compaction temperature. The
total proximity zone length results versus mastic low shear viscosity for mixes
compacted at different temperatures are plotted in Figure 17. In addition to mastic
viscosity, film thickness at the proximity zones (which is also a function of viscosity) is
another important factor affecting friction between aggregates. By increasing the
compaction temperature, viscosity is decreased providing sufficient lubrication between
aggregates and higher total proximity zone length (as long as the viscosity is higher than
a specific range or, in other words, mastic film thickness is still sufficient to avoid dry
contacts). This is demonstrated in Figure 17. However, for viscosities lower than a
specific range, film thickness of mastic becomes very thin within the proximity zones of
aggregates, as shown in Figure 18. Therefore, due to the dry contacts and/or the thin film
viscosity effect, workability of the mixture and aggregate mobility decreases, leading to
48

a decrease in total proximity zone length (i.e. packing of aggregates) after a maximum
value for mixes compacted at higher temperatures.
4,500
4,000
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
3,500
3,000
2,500
2,000
1,500
1,000
0 .1 1 10 100 1000 10000 100000
Mastic Viscosity (Pa.s)
Control C S H
Figure 17. Total proximity zone length as a function of mastic viscosity at different
compaction temperatures at shear rate of 0.01 sec-.1
The data points from the film thickness plots at which the maximum total proximity
zone length is attained are circled in Figure 18.
49

Mortar Film Thickness

5 .0
H C S Ne a t
Film Thickness (mm)
4 .0
3 .0
2 .0
1 .0
0 .0
60 80 100 120 140 160
Compaction Temp (C)
Figure 18. Mortar average film thickness at different compaction temperatures.
In this study, limited numbers of compaction temperatures are selected for proof of
the concept. However, to determine the viscosity range providing sufficient film
thickness and lubrication for optimum workability, compaction of mixes at additional
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
temperatures is needed to accurately determine the compaction temperature (or mastic

viscosity as a material property) leading to maximum packing. In addition, a change in
aggregate gradation design changes the stress distribution in the mix and at the aggregate
proximity zones. Therefore, with the results of this study, it can be concluded that,
different aggregate gradations may lead to different ranges of viscosity to resist the
squeezing out due to the contact normal stress condition. This needs more verification,
but is in agreement with recent findings in the literature that the viscosity as a single
factor is not enough to predict workability of different mixtures (Hanz et al., 2011).
6. Conclusions
This study was focused on understanding the factors that can critically control aggregate
mobility and packing during compaction. The study was limited to one gradation and
four different binders. This was done to challenge the concept that binder viscosity as a
function of temperature is a sufficient parameter to recommend minimum limits for
compaction temperatures. In addition to measuring the density of mixes compacted at
various temperatures, viscosities of binders and mastics were measured, and estimates of
film thickness at aggregate proximity zones using advanced imaging tools were
50

performed. Although many previous studies had considered a desired density as a

compaction process target, this study was focused on the evolution of aggregate structure
during compaction and effect of lubricant rheology (i.e., viscosity) on the aggregate
packing as a more fundamental indication of final product performance. The aggregate
total proximity zone length as result of 2-Dimensional image analysis was used to
characterize aggregate structure of asphalt mixtures. The main findings of this study are:
Most design specifications use binder viscosity as an indication of mixture
workability at the compaction temperature. This study demonstrated that binder
viscosity is not sufficient and that workability of the mixture is more dependent on
viscosity and film thickness of the actual lubricant, which is the binder and filler
mastic. The film thickness that is important is not the overall average but the
thickness within the proximity zones of aggregates. This finding is supported by
many studies in the field of tribology and lubrication mechanics.
Two competing mechanisms are proposed as controlling factors of the compaction
process and aggregate structure formation. Lubricants with lower viscosity provide
easier sliding between the lubricated surfaces. On the other hand, lower viscosity
results in a higher squeezing out rate of lubricant for a constant normal force within
the proximity zones between two surfaces, which increases the chance of dry
contact interaction, or thin film high viscosity effect. The result is an increase in
friction between two surfaces. Therefore, by increasing the compaction temperature,
the aggregate mobility and structural packing usually begin to increase to a peak
value and then decrease as dry contact or thin film viscosity effects become
significant at higher temperatures.
Mixes were compacted to different number of gyrations with the same gradation but
different binders to demonstrate the effect of mastic viscosity on aggregate structure
formation. The rates of increase in total aggregate proximity zone lengths were
significantly different for different mixes during compaction. Additionally, binders
and mastics were tested at different shear rates. The viscosity behavior of mastic at
low shear rates could not be predicted from binder viscosity, and since mastic is the
lubricant phase in asphalt mixtures, mastic viscosity was used to explain the
behavior of the mixtures. With the aid of image analysis, the film thickness of
mortar at proximity zones of aggregates at different compaction stages (i.e., number
of gyrations) was determined. The mastic viscosity at low shear rates in addition to
mortar film thickness between coarse aggregate results show that the mastic with
higher viscosity at compaction temperatures squeezed out of aggregate proximity
zones in a more gradual manner compared to mastics with lower viscosity.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Therefore, compaction of mixes with higher viscosity mastics resulted in better

packing of aggregates, and the mix with the lowest mastic viscosity was locked at
51

the initial stages of compaction and the total proximity zone length was unchanged
for higher number of gyrations.
In order to study the effect of mastic viscosity on aggregate structure, mixes were
compacted at temperatures from 65C to 165C using four binders providing
significantly different viscosity levels. The results showed that with an increase in
compaction temperature or decrease in mastic viscosity, total proximity zone length
in the aggregate skeleton increased up to a certain temperature and then began to
decrease. The temperature at which the optimum total proximity zone length was
achieved varies for different materials. However, the value of optimum total
proximity zone length was approximately the same for all mixtures. This finding
shows that the particle arrangement and packing level for the same compaction
pressure and aggregate gradation is related to mastic viscosity as a fundamental
material property which can be the same for different materials at different
temperatures. However, the number of mixes and compaction temperatures in this
study were not sufficient to define a meaningful range of mastic viscosity for
optimum packing.
7. Acknowledgment
As members of the Asphalt Research Consortium, funding and support from the FHWA
and WRI is acknowledged. The help and comments from Dr. Raul Velasquez, Dr.
Andrew Hanz, and Mr. Dan Swiertz are greatly appreciated. The testing and sample
preparation of Trevor Schultz is also acknowledged. Partial funding from Honeywell
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Inc., USA is acknowledged.
8. References
AASHTO. (2007). Designation T 316, Standard Method of Test for Viscosity

Determination of Asphalt Binder Using Rotational Viscometer. Standard
Specifications for Transportation Materials and Methods of Sampling and Testing,
27th Ed. Washington, D.C., American Association of State Highway and
Transportation Officials.
AASHTO M 323-07, Superpave Volumetric Mix Design, 2007.
Anderson, R. M., Relationship Between Superpave Gyratory Compaction Properties on
the Rutting Potential of Asphalt Mixes, Journal of the Association of Asphalt Paving
Technologists, Vol. 71, 2002, pp. 725738.
52

Austerman, A., Mogawer, W., and Bonaquist, R., Evaluating the Effects of Warm Mix
Asphalt Technology Additive Dosages on the Workability and Durability of Asphalt
Mixtures Containing Recycled Asphalt Pavement, Transportation Research Board
88th Annual Meeting Compendium of Papers CD-ROM. Washington D.C.:
Transportation Research Board of National Academies, 2009.
Bahia, H., Friemel, T., Peterson, P. and Russell, J., Optimization of Constructability
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
and Resistance to Traffic: A New Design Approach for HMA Using the Superpave
Compactor, Journal of the Association of Asphalt Paving Technologists, Vol. 67,
p.189232, 1998.
Bahia, H. U., Hanson, D. I., Zeng, H., Zhai, M., Khatri, M. A., and Anderson, R. M.,
NCHRP Report 459: Characterization of Modified Asphalt Binders in Superpave Mix
Design, TRB, National Research Council, Washington, D.C., 2001.
Bahia, H.U., and Roohi, N., Effect of Compaction Method, Temperature and Pressure
on Aggregate Structure of Hot Mix Asphalt, International Journal of Pavement, in
press, 2012.
Bennert, T., Reinke, G., Mogawer, W., and Mooney, K., Assessment of
Workability/Compactability of Warm Mix Asphalt,- Transportation Research Board
89th Annual Meeting Compendium of Papers. CD-ROM. Washington, D.C.:
Transportation Research Board of National Academies, 2010.
Brown, E. R., Lord, B., Decker, D., and D. Newcomb, Hot Mix Asphalt Tender Zone,
Report No. 2000-02, National Center for Asphalt Technology Auburn, AL, 2000.
Buchanan, M.S., and Cooley, L.A., Case Studies of the Tender Zone in Coarse-Graded
Superpave Mixtures, Report No.: NCAT Report 02-01.: National Center for Asphalt
Technology, Auburn, AL 2002.
Clopotel, C., Filler Reinforcement Mechanisms in Asphalt Mastic, PhD thesis,
University of Wisconsin-Madison, 2012.
Coenen, A.R., Kutay, M.E., Roohi Sefidmazgi, N., Bahia, H.U., Aggregate Structure
Characterization of Asphalt Mixtures Using 2-Dimensional Image Analysis, Road
Materials and Pavement Design, accepted for publication, 2012.
Crawford, C. Tender Mixes - Probable Causes, Possible Remedies, National Asphalt
Pavement Association Quality Improvement Series 108/86.
Delgadillo, R., and H. U. Bahia, Effects of Temperature and Pressure on Hot Mixed
Asphalt Compaction: Field and Laboratory Study, Journal of Materials in Civil
Engineering, v20, no. 6, p. 440448, 2008.
53

Faheem, A.F., Modeling of Asphalt Mastic in Terms of Filler-Bitumen Interaction, PhD

thesis, University of Wisconsin-Madison, 2009.
Gudimettla, J., Cooley, L. A., and Brown, E. R., Workability of Hot Mix Asphalt, NCAT
Report 03-03. National Center for Asphalt Technology, Auburn University, Auburn,
2003.
Hanz, A.J., Mahmoud, E., and, Bahia, H.U., Asphalt Lubricity Test Evaluation and
Relationship to Mixture Workability, Transportation Research Board 90th Annual
Meeting DVD Compendium of Papers, 2011.
Hurley, G., Evaluation of New Technologies for Use in WMA, Masters Thesis,
Auburn, AL: Auburn University, 2006.
Hurley, G.C., and Prowell, B., Evaluation of Aspha-min Zeolite for use in Warm-Mix
Asphalt, NCAT Report 05-04, National Center for Asphalt Technology, Auburn, AL,
2005.
Hurley, G., and Prowell, B., Evaluation of Evotherm for Use in Warm-Mix Asphalt,
NCAT Report 06-02, National Center for Asphalt Technology, Auburn, AL, 2005.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Hurley, G., and Prowell, B., Evaluation of Sasobit Zeolite for Use in Warm-Mix Asphalt,
NCAT Report 05-06, National Center for Asphalt Technology, Auburn, AL, 2005.
Johnson, K.L., Greenwood, J.A., Poon, S.Y., A Simple Theory of Asperity Contact in
Elasto-hydro-dynamic Lubrication, Wear, Volume 19, Issue 1, January 1972, pp.
91-108.
Li, X., and Gibson, N., Mechanistic Characterization of Aggregate Packing to Assess
Gyration Levels during HMA Mix Design, Journal of the Association of Asphalt
Paving Technologists, Vol. 80, pp. 33-64, 2011.
Lu, Xiaobin, M. M. Khonsari, and E. R. M. Gelinck , The Stribeck Curve:
Experimental Results and Theoretical Prediction, J. Tribol. 128, 789,
DOI:10.1115/1.2345406, 2003.
Luengo, G., Schmitt, F.J., Hill, R., Israelachvili, J., Thin Film Rheology and Tribology
of Confined Polymer Melts: Contrasts with Bulk Properties, Macromolecules, Vol.
30, No. 8, 1997.
Macosko, C.W., Rheology - Principles, Measurements and Applications, John Wiley &
Sons, 1994.
Marker, V. Tender Mixes: The Causes and Prevention, Asphalt Institute Information
Series No. 168 (IS-168), June 1977.
54

Olard, F., GB5 Mix Design: High-Performance and Cost-Effective Asphalt Concretes
by Use of Gap-Graded Curves and SBS Modified Bitumens, Road Materials and
Pavement Design, Vol. 13, Suplement 1, pp. 234259, 2012.
Richardson, C., The Theory of the Perfect Sheet Asphalt Surface, Journal of
Industrial. Engineering and Chemistry, pp 463-165, 1915.
Roohi Sefidmazgi N., Tashman L., Bahia H.U., Internal Structure Characterization of
Asphalt Mixtures for Rutting Performance Using Imaging Analysis, Journal of the
Association of Asphalt Paving Technologists, pp.109138, Vol. 81, 2012.
Tashman L., Masad, E., Peterson, B., Saleh, H., Internal Structure Analysis of Asphalt
Mixes To Improve The Simulation of Superpave Gyratory Compaction to Field
Conditions, Journal of the Association of Asphalt Paving Technologists, Vol. 70, pp.
605-645, 2001.
Zhu , H., Nodes, J.E., Contact Based Analysis of Asphalt Pavement with the Effect of
Aggregate Angularity, Mechanics of Materials Vol. 32, Issue 3, pp. 193202, 2000.
9. Discussion
PROFESSOR DAVID ANDERSON: A lot of interesting concepts here, and I like the
approach and things you're doing, but I do have a couple of comments. I think we need a
more realistic understanding of what the mastic is in the mix. Its not minus 200 filler
plus asphalt binder. It probably is and if you do a little bit of looking in the literature,
you will find instances where people talk about minus 3 micron, minus 20 micron. Its
the mineral filler that floats in the asphalt binder. The minus 200 mesh doesnt do that.
The minus 200, minus 375, whatever, is part of the aggregate framework. I think you
really need to revisit your concept of the mastic and do a little bit of homework there,
and I think maybe youll make more progress towards implementing what youve put
together here.
MR. NIMA ROOHI SEFIDMAZGI: Thank you for the comment. What we tried to do
was basically the gradation design in Superpave. Thats absolutely right, and I think we
are going to probably work more on that. As I said, this is just to introduce the concept,
but basically making it more fundamental, as you said, we have to.
PROF. ANDERSON: Also, please tell us how you measure the specific surface. Thats
pretty important, and its not in the paper.
MR. SEFIDMAZGI: The specific surface of the fillers? I dont remember the name of
the method
55
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

PROF. ANDERSON: Dont do it now because youve got more questions here.
MR. SEFIDMAZGI: Okay. But its based on a gas being exposed.
DR. JOHN DANGELO: Again, I would just like to say interesting concepts here, but
two items. One is you're doing all your analysis on a gyratory compactor, and that
compaction process is completely different than the roadway. In the roadway, you're
allowing dilation of the aggregate particles. Youve got very thin films, so you have
rather large reorientation of the aggregate particles during the compaction process that
you dont have in the gyratory.
You talk about the structure that we design mixes for and put up things like 7% air
voids. We dont design mixes for 7% air voids, but we do make specimens at this void
level to do performance testing. The 7% air voids is trying to simulate what is typically
achieved in the field. Theres been a lot of work that has been shown that density does
relate to performance, but they have to be done over a wider range.
MR. SEFIDMAZGI: Thank you for the comments, but I have something to say about the
gyratory compactor. Thats exactly right. I mean the mechanism is different, and were
trying actually to verify that with some failed mix samples. The point is, that we are
designing with gyratory compactors, so if it is a problem, we have to understand it and
solve it in a way finally. People are going to design using that, so although the
mechanism is different, understanding of that probably can help to some extent.
DR. DANGELO: Your own data all converges to identical performance when you get
to little bit higher gyrations and thats what we typically design to.
PROFESSOR REYNALDO ROQUE: My background formerly is in soil mechanics,

geotechnical engineering. When I look at a lot of the work that you're presenting, the
basic principles and associated mechanisms very much relate to the principles that come
out of that very large body of work that is associated with compaction of soil mixtures.
If I take a clayey gravel, for example, where the clay is representing the binder and the
gravel part is well graded and can represent the mixture, we have lots of information on
density curves and optimum moisture contents, which you can translate perhaps to
optimum binder contents. I think the use of things like dry density and porosity
parameters that are well defined and have been used extensively could be very helpful in
making this much more straightforward and clearer and I think more meaningful than
using a lot of new terminology. I was just wondering how much you have looked at that
literature compared to the tribology literature, which I am looking at myself for other
purposes, which primarily addresses flat surfaces on flat surfaces with little grainy
particles. I think thats very different than soil mechanics, which is a field that we have
been studying for decades and decades.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
56

MR. SEFIDMAZGI: Right. Thats a very good point. I honestly did some literature
review in soil mechanic compaction. I started with soil mechanic compaction and,
honestly, that was the starting point to think about it. But what happened later was I want
to understand what is happening in lower scales which basically led me toward this
tribology. But the point is that that is exactly right. I mean what we tried to do here was
proof of concept and understanding what is going on, but to make it useful, thats
absolutely right. There are a lot of concepts in soil mechanics and granular mechanics
that can be applied. The indices or characteristics that can be used are probably more
user friendly for designers or producers.
PROFESSOR CARL MONISMITH: May I suggest, if you have not done so, to look at
the work by Nijboer from 1948. Also the work of Schmidt and Santucci published by the
Highway Research Board (now the TRB) in 1962. These investigations used rolling
wheel compaction either in the laboratory or in the field. So it would be a good idea for
you to look at the results of these investigations.
N.B. A list of these publications was sent to Prof. H. Bahia following the meeting. These,
which I consider to be classics in asphalt technology, are as follows:
1. Nijboer, L.W.J., "Plasticity as a Factor in the Design of Dense Bituminous Road

Carpets", Elsevier Publishing Company, 1948. A paper in the AAPT Symposium on
HMA Compaction in 1970 by Kor Wester makes reference to Nijboer's work with an
application to pneumatic tire rolling in the Netherlands. Nijboers work was based on
steel wheel rollers.
2. Two Papers by R.J. Schmidt and L. E. Santucci, California Research Corporation
(now Chevron Research)
a. Schmidt and Santucci, "Influence of Asphalt Type on Pavement Setting Rate", Bulletin
333, Highway Research Board, 1962, p. 10.
b. Santucci and Schmidt, "Setting Rate of Asphalt Concrete", Bulletin 333, Highway
Research Board, 1962, p.1.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(The work by Schmidt and Santucci was a very comprehensive study using controlled
aggregate gradings, a small hot mix plant, a replica of the HMA paver screed, and full
scale rollers at the site at what is now the Chevron Research Corporation in Richmond,
California.)
MR. SEFIDMAZGI: Thank you very much. Of course I will. Thank you.
57

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Hot Mix Asphalt Pavement Frictional
Resistance as a Function of Aggregate
Physical Properties
John E. Haddocka* and Joan P. OBrienb
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
a
School of Civil Engineering, Purdue University, West Lafayette, IN
b
Parsons Brinckerhoff, Perth, Australia
ABSTRACT: Aggregates were tested for acid insolubility, differential wear, and frictional resistance
and along with physical data were correlated to aggregate frictional resistance. Physical data
included bulk specific gravity, percent absorption, percent loss from impact, and percent loss from
brine freeze/thaw. It was observed that the aggregates used had a negligible amount of insoluble
material and therefore had an insignificant effect on the frictional resistance. Terminal frictional
resistance increased with decreasing bulk specific gravity and increased with increasing percent
absorption and percent loss from impact. Correlation between terminal frictional resistance and
brine loss was moderate. Differential wear measurements were taken on stone-slag aggregate blends
to determine the effectiveness of blends used in pavement mixtures. Differential wear results
indicated that the high friction component of each blend could be improved by blending with softer
carbonate aggregates, which also may be more cost effective.
KEYWORDS: Friction, skid resistance, polish resistance, aggregate, hot mix asphalt.
The oral presentation was made by Professor Haddock.
HTTP://DX.DOI.ORG/ 10.1080/14680629.2013.812842
59

HADDOCK, OBRIEN
1. Introduction
Pavement frictional resistance is important for the safe performance of pavements and
ranks second among pavement performance considerations in the United States and many
other countries according to a study by the National Cooperative Highway Research
Program (NCHRP) (Henry, 2000). Frictional, or skid, resistance is defined as the ability of
a traveled surface to prevent the loss of tire traction (ASTM, 2002b) and is expressed as a
friction or skid number (FN or SN), defined as the ratio between the frictional resistance
acting along the plane of sliding and the load perpendicular to this plane (Jayawickrama et
al., 1996).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The frictional performance of a pavement is dependent on both the micro- and macro-
texture of a pavement, which are partly controlled by the aggregates used (Jayawickrama et
al., 1996; Skerritt, 1993). On a macro scale, aggregates make up approximately 95 percent
of hot mix asphalt (HMA) by weight and therefore greatly contribute to the texture of an
HMA pavement (Henry, 2000). On a micro scale, the physical properties of the aggregate
also contribute to the texture of a pavement.
Like many highway agencies, the Indiana Department of Transportation (INDOT)
assesses frictional resistance of coarse aggregates before use in HMA pavements through a
series of laboratory tests that measure micro-texture. In 1997 INDOT adopted the
Superpave mixture design method and attendant specifications for HMA mixtures. These
specifications control coarse aggregate characteristics key to HMA performance;
angularity, flat and elongated particles, clay content, toughness, soundness, and deleterious
materials (Asphalt Institute, 1995), without specific regard to macro- and micro-texture,
and therefore may have ramifications on the frictional characteristics of an HMA
pavement. Gradation requirements for Superpave-designed HMA mixtures may also affect
pavement friction characteristics.
After several years of using the Superpave mixture design method for HMA, INDOT
decided to re-examine the methods used to screen coarse aggregates for frictional
resistance to ensure adequate pavement friction. Therefore, the objective of this study is to
determine if current tests used by INDOT adequately screen coarse aggregates for
frictional properties.
In order to achieve the objective, pavements constructed using Superpave designed
HMA wearing courses containing coarse aggregates that have a history of consistent
physical properties were identified, and coarse aggregate samples were collected from the
source quarries and tested for total insoluble residue and laboratory frictional
characteristics. Physical property data for these aggregates were also collected from the
INDOT quality assurance (QA) inventory. The QA and total insoluble residue data were
60

Hot Mix Asphalt Pavement Frictional Resistance as a Function of Aggregate Physical Properties
correlated with laboratory frictional resistance data of each aggregate. In-service field
friction data were also collected and compared to the laboratory results.
2. Background
Skid resistance as a factor in pavement performance began to garner attention in the early
20th century when researchers became interested and began to develop equipment to study
the pavement frictional resistance. Since a pavement can be structurally sound and still
have poor skid resistance (Dillard and Alwood, 1957), one of the early research objectives
was to increase skid resistance without affecting the structural soundness of a pavement.
Implementation of the Federal-Aid Highway Act in 1952 stressed the awareness and
importance of skid resistance partly because of the increased speeds and number of
vehicles on the road.
Skid resistance of a pavement is greatly affected by its texture. Pavement texture has
two components important to skid resistance, micro- and macro-texture. A pavement is
considered polished when it loses its micro-texture, and worn when it loses its macro-
texture. In a general sense, Webb (1970) describes micro-texture as the finer intra-particle
asperities and macro-texture as the coarser, inter-particle texture. Generally, it is believed
that micro-texture has a greater effect than macro-texture on the frictional properties of a
pavement.
According to Webbs definition of micro-texture, aggregates are the most influential
portion of pavement micro-texture. Micro-asperities on aggregate surfaces form the micro-
texture of a pavement and determine the resistance of a pavement surface to skidding
(Rogers and Gargett, 1991). The micro-asperities are a function of aggregate mineralogy
and how it reacts to climate and traffic loading (Rogers and Gargett, 1991). Polishing of an
aggregate occurs when the micro-texture is removed by vehicular traffic. The ability of an
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
aggregate to maintain micro-texture is dependent on its petrographic properties such as
mineralogy, texture and fabric (Webb, 1970), and can be related to skid resistance
performance under various traffic volumes (Gramling and Hopkins, 1974). The largest
aggregate size in the HMA mixture is thought to govern skid resistance (Skerritt, 1993).
2.1 Mineralogy
Coarse aggregates provide both micro-texture by the properties of the aggregates

themselves, such as differential hardness of the minerals and irregularity of the surface, and
macro-texture through spacing and gradation. However, mineralogy of an aggregate is
possibly the most important factor in the ability of a pavement to maintain micro-texture,
61

HADDOCK, OBRIEN
and thus skid resistance. The minerals themselves are not what provide frictional
resistance, but it is the differential hardness between minerals in a particular rock type that
offer sufficient skid resistance (Mullen et al., 1974). Although harder minerals are
preferable, even the hardest rock types will polish if a differential hardness between
minerals does not exist (Dahir and Mullen, 1971). It is the difference in polishing rates
between minerals with different hardness that gives aggregates a rough, skid resistant
texture (Furbush and Styers 1972). However, a harder mineral will provide a longer rate of
polishing (Webb, 1970).
Limestone and dolomite tend to have a fairly uniform texture with low hardness, which
makes these rock types susceptible to rapid polishing. In a study by Burnett, et al. (1968),
limestone and dolomite had a coefficient of friction below 0.32 after 5 million vehicular
passes, but some retained coefficients high enough to remain skid resistant after ten million
vehicular passes. It was observed that the limestone and dolomite aggregates that had
quartz grains within the matrix had a sand paper-like texture and higher skid resistance.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Therefore, it is possible to have a skid resistant limestone or dolomite aggregate provided
some percentage of the constituents consists of harder minerals.
West and Cho (2000) indicate that skid resistance of a pavement with this aggregate
type is partly dependent on the impurity of the limestone or dolomite aggregate, which can
be determined by elemental magnesium content, specific gravity, and total acid insoluble
residue. Dolomite is considered more impure with lower elemental magnesium content and
higher total insoluble residue, and limestone is more impure with both higher elemental
magnesium content and total insoluble residue (West and Cho, 2000). West et al. (2001)
emphasize that this relationship can be seen in the specific gravity of the aggregate. A
dolomite mineral has a specific gravity of 2.800 to 2.900 (Chesterman, 1979). It is
speculated that an aggregate of the dolomite rock type will have a higher frictional
resistance with specific gravities further away from the 2.800 to 2.900 value (West et al.,
2001).
Consequently, for limestone and dolomite aggregates, testing for insoluble residue is
important. Limestone and dolomite are made up of mostly calcium, magnesium carbonate
or both, which react with dilute hydrochloric acid. In the insoluble residue test, the
aggregate is dissolved in dilute hydrochloric acid, and the remaining insoluble portion is
usually made up of quartz (silica), feldspar (clays), or other insoluble minerals. As
discussed previously, the presence of quartz or other insoluble minerals in a limestone or
dolomite is significant in giving aggregates of this rock type sufficient skid resistance.
Gray and Renninger (1966) were the first to report that insoluble residue correlated
with an aggregates skid resistance. Burnett, et al. (1968), Furbush and Styers (1972),
Skerritt (1993) and West and Cho (2000) later confirmed this concept, which is now
widely accepted by most researchers. However, there are conflicting reports on the amount
62

or the size of insoluble residue that is needed in order to retain sufficient skid resistance.
Gray and Renninger (1966) recommend at least ten percent insoluble residue; Furbush and
Styers (1972) and Kearney et al. (1972) recommend at least twenty percent. Kearney et al.
(1972) also propose that at least ten percent sand-sized impurities are needed instead of a
total percentage. Skerritt (1993) contends that dolomites with greater than fifteen percent
total acid insoluble residue consistently produce friction numbers above forty even at high
accumulated traffic. It is unclear which theory is more correct, but it appears to be widely
accepted that the grain size of the remaining insoluble residue should at least be retained on
the 0.075 mm (No. 200) sieve (Furbush and Styers, 1972; Gray and Renninger, 1966;
Kearney et al., 1972).
In a study by West and Cho (2000), it was observed that the percent of insoluble
residue smaller than the 0.075 mm (No. 200) sieve increased frictional resistance. The
authors speculated that the tiny clay particles that make up the portion smaller than the
0.075 mm (No. 200) sieve break out of the carbonate matrix creating an irregular surface
and providing the needed micro-texture for good skid resistance (West and Cho, 2000).
West and Cho also report that although the total percentage of acid insolubility correlates
well with terminal polish value (BPN10) and wear index (WI), this correlation is stronger
with limestone aggregates than dolomite aggregates. Wear index is the difference between
the initial friction value (BPN0) and the final polish value (WI = BPN0 BPN10), and is
indicative of polish resistance (West et al., 2001).
Additionally, Abdul-Malak et al. (1993) identified that the presence of dolomite in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
carbonates provides a higher level of skid resistance. This may be because dolomite is a
harder mineral than calcite. Calcite is the most common mineral found in carbonate rocks
(Chesterman, 1979). Dolomite and calcite have a similar chemical composition, but
dolomite includes magnesium [CaCO3 and Ca(Mg)CO3, respectively]. INDOT specifies
dolomite with a minimum of 10.3% elemental magnesium for all pavements with medium
to high equivalent single axle load (ESAL) counts (INDOT, 2012). The dolomite aggregate
is considered more pure as the percent elemental magnesium increases. As indicated earlier
by West et al. (2001), impure limestone and dolomite aggregates have better skid
resistance. This is most likely partly due to the differential in hardness between the calcite
and dolomite in such carbonate aggregates, in addition to the insoluble portion.
2.2 Aggregate Blending
Blending low skid resistant aggregate with high skid resistant aggregate is common
practice in states where aggregates with inherently low skid resistance are locally available.
Transportation of aggregates is costly and therefore local aggregates are most commonly
used. Steel slag and air-cooled blast furnace slag are commonly blended with dolomites in
63

HADDOCK, OBRIEN
Indiana to improve HMA pavement skid resistance. Although the terminal polish value of
slag varies, the skid resistance of a pavement constructed with slag remains relatively
constant (Stock et al., 1996). Optimum blending consists of finding an effective blend of
low and high skid resistant aggregates that provides for safer roads and is also cost
effective (Fu, 2004).
Some studies suggest that the frictional resistance of a blend is directly proportional to
blend percentages (Gramling and Hopkins, 1974; Liang, 2003). Therefore, an increased
terminal polishing value can be obtained by increasing the percentage of aggregates with
higher frictional resistance and lower polish susceptibility. However, it is recommended
that a blend be subjected to polish testing to determine whether or not the terminal polish
value is at an acceptable level according to traffic or ESAL count (Liang, 2003).
According to Fu (2004), differential wear (DW) is an important factor in determining
whether a blend will perform well in the field and whether it will be cost effective.
Differential wear is the difference in particle height measurement between the low and
high friction component before and after polishing (Fu, 2004). These values are obtained
by taking measurements with a height-measuring device before and after polishing tests.
Higher DW values maximize both micro and macro-texture and are indicative of blends
that will perform well in the field (Fu, 2004). In other words, if a blend consists of two
aggregates with similar terminal polish values, the differential wear value will be low and
subsequently the terminal polish value of the blend will be lower. Therefore, it is expected
that it will not effectively enhance the skid resistance of the pavement nor will it be a cost
effective blend.
3. Study Methodology
As shown in Table 1, INDOT divides HMA aggregate frictional requirements by ESAL

categories. The experimental methodology consisted of six steps:
1. Identify consistent aggregate sources;
2. Identify pavements with Superpave-designed HMA surface courses containing
the aggregates identified in Step 1;
3. Collect the friction data from the pavements identified in Step 2;
4. Collect samples and quality assurance (QA) data from the aggregate sources
identified in Step 1;
5. Complete laboratory testing of the aggregate samples; and
6. Analyze the data.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
64

Table 1. INDOT wearing course aggregate specification (INDOT 2012).

Equivalent Single Axle Load
Coarse Aggregate Type < 3,000,000 < 10,000,000 10,000,000
Air-Cooled Blast
Yes Yes Yes
Furnace Slag
Steel Furnace Slag Yes Yes Yes
Sandstone Yes Yes Yes
Crushed Dolomite Yes Yes Note 1
Polish Resistant
Yes Yes Note 1
Aggregates
Crushed Stone Yes No No
Gravel Yes No No
NOTE: Polish resistant aggregates or crushed dolomite may be used when blended with
ACBF or sandstone but cannot exceed 50% of the coarse aggregate by mass (weight), or
cannot exceed 40% of the coarse aggregate by mass (weight) when blended with steel
furnace slag.
3.1 Identification of Aggregate Sources
The stone aggregate sources were selected through reviewing laboratory quality assurance
data conducted by INDOT and also by historical skid performance of Indiana pavements.
The physical property data were assessed for aggregate consistency and competency.
Reviewed test data included Los Angeles abrasion, brine freeze/thaw soundness,
absorption, elemental magnesium, and specific gravity. The purpose of identifying
aggregates having consistent physical properties and performance over many years was to
attempt to control aggregates as a variable. The list of selected aggregates and their
corresponding physical property data are presented in Table 2. All aggregates are dolomite
per Indiana Test Method (ITM) 205, Acceptance Procedures for Dolomite Aggregates
and have been approved by INDOT for use in medium and high ESAL count pavements
per ITM 214, Acceptance Procedures for Polish Resistant Aggregates. The D-prefix for
each source represents dolomite.
As seen in Table 1, pavements with dolomite or crushed stone aggregates that are in the
equal to or greater than ten million ESAL category require blending with slag or sandstone.
Pavements used in this study that contain aggregate blends were blended with slag. The
slag sources are presented in Table 3 and are represented by an S-prefix. The steel slag
specific gravity was not available, but typical values are approximately 3.000. Blends are
denoted by concatenation of the stone and slag aggregate codes that are listed in Tables 2
and 3.
65
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HADDOCK, OBRIEN
Table 2. Stone aggregate source list.

LA Freeze/
Bulk Elemental Abrasion Brine Thaw
Source Specific Magnesium Absorption Loss Loss Loss
ID Gravity (%) (%) (%) (%) (%)
D1 2.749 12.1 1.03 23.34 0.01 -
D2 2.442 11.1 4.53 30.74 11.11 -
D3 2.727 12.5 1.19 19.25 0.36 -
D4 2.554 10.8 3.07 28.54 N/A 1.76
D5 2.489 11.5 3.3 26.98 0.33 -
D6 2.658 10.7 1.89 23.14 26.48 1.08
NOTE: Test procedures are further discussed in subsequent sections.
Table 3. Slag aggregate source list.

Source ID Slag Type Specific Gravity
S1 Steel N/A
S2 Blast Furnace 2.463
NOTE: N/A indicates not available.
3.2 Identification of Pavement Sections
Once consistent aggregate sources were selected, a search was conducted that identified
projects or contracts that placed pavements using the consistent aggregate sources with a
Superpave design mixture formula (DMF). Each district within INDOT was given the list
of consistent aggregate sources. The INDOT district personnel conducted a search of their
records where DMF for contracts that used the selected aggregate sources were obtained.
Since the Superpave mixture design method was not implemented until 1997, all
pavements used for this research were placed after that date.
3.3 Collection of Friction Data
To collect and review the skid data from the INDOT friction inventory database, the road
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
name and mile marker for each contract were obtained from the DMF. If either piece of
information was missing from the DMF, the data were obtained from the INDOT contract
(project) database. Using the road name and mile marker for each contract, a search was
conducted in the INDOT friction inventory database to obtain skid numbers for the
corresponding mile markers or reference posts.
66

Upon review of skid data, it was noticed that there were gaps in skid data resulting in
the narrowing of the list of selected pavements. Pavements selected for this study are listed
by contract number and are presented in Table 4. It was not possible to use all pavements
on this list for all analyses.
Skid testing of pavements is independent of the age of the pavement. In other words, if
a pavement on a state road was placed in October of 1997, the first skid resistance testing
could be as early as April of 1998 or as late as September 2000. Since secondary roads are
skid tested once every three years, the amount of data available was limited for these types
of pavements. A final review resulted in 32 pavement sections that were identified for use
in the research. Road numbers are prefixed with an I-, US-, or SR-, indicating an interstate,
federal, or state road, respectively.
3.4 Collection of Aggregate Samples and Quality Assurance Data

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Once pavements and aggregate sources were selected, the aggregates used in the
pavements were determined and samples for testing were obtained from each quarry. To
initiate collection of samples, each stone aggregate source was contacted and a gradation
was obtained. The aggregate gradation was used to determine the approximate quantity
needed from each source. Subsequently, each of the six selected stone quarries was visited,
and samples were collected. Samples were obtained in accordance with ITM 207,
Sampling Stockpiled Aggregates.
Aggregate producers in Indiana have the option of producing aggregates in accordance
with the INDOT Certified Aggregate Producers Program (CAPP). By doing so an
aggregate producer becomes certified and thus may market aggregate for INDOT
projects without the need for INDOT to individually test and certify each stockpile. As part
of the CAPP program, the aggregate producer generates quality control data while quality
assurance data are generated by INDOT. These data include AASHTO T 96, Resistance
to Degradation of Small-Size Coarse Aggregate by Abrasion and Impact in the Los
Angeles Machine, AASHTO T 85, Specific Gravity and Absorption, and ITM 209,
Soundness of Aggregates by Freezing and Thawing in a Brine Solution. Aggregates that
do not meet the INDOT specifications for ITM 209 are also subjected to AASHTO T 103,
Test for Soundness of Aggregates by Freezing and Thawing, Procedure A. INDOT
considers AASHTO T 103 to be the most accurate method, but the test has a long duration.
Therefore, ITM 209 is used as a quick check of the aggregate. If the aggregate passes the
ITM 209 method it is deemed acceptable. If it fails the method, but passes AASHTO T
103, it is acceptable for use. For this project, the INDOT QA data was gathered for each of
the aggregate sources identified.
67

HADDOCK, OBRIEN
Table 4. Pavement list.

Aggregate
Contract Number Road Source ID ESAL ADT
R-23365 US-31 D5S1 12,000,000 35,000
R-23800 I-265 D6S2 11,000,000 40155
R-23924-A SR-46 D1S1 3,700,000 16240
R-24551 I-65 D6S2 25,000,000 30000
R-24555 I-65 D6S2 44,000,000 114000
R-24561 I-65 D6S2 88,000,000 44390
R-24847 I-65 D6S2 53,000,000 45000
R-24947B US-35 D3S3 5,200,000 10000
RS-23321 US-31 D5S1 3,500,000 8000
RS-23321 US-31 D5S1 3,500,000 17000
RS-23492 SR-44 D5 1,200,000 8000
RS-23600 US-24 D2S4 3,200,000 4870
RS-23611 SR-135 D5 1,400,000 10000
RS-23611 SR-135 D5 1,400,000 19000
RS-23787 US-31 D6S2 11,000,000 5000
RS-23790 US-150 D6S2 3,500,000 12000
RS-23792 SR-60 D6 2,000,000 10000
RS-23793 SR-46 D5 3,500,000 8000
RS-23794 SR-46 D5S1 2,500,000 13500
RS-23794 SR-46 D5S1 2,500,000 5000
RS-24107 SR-56 D6 2,500,000 9000
RS-24107 SR-56 D6 2,500,000 3000
RS-24108 SR-11 D5 3,000,000 3500
RS-24109 SR-45 D6 2,000,000 10000
RS-24109 SR-45 D6 2,000,000 4000
RS-24209 SR-129 D4 2,000,000 7000
RS-24256 SR-46 D4 2,000,000 4000
RS-24389-A SR-46 D5S1 7,500,000 12000
RS-24534-A SR-37 D6S1 7,500,000 22000
RS-24536-A SR-43 D1 2,000,000 2000
RS-24540-A SR-62 D6S1 7,500,000 9000
RS-24769-A SR-135 D6 2,000,000 7000
RS-24771-A US-150 D6 2,000,000 5000
RS-24823 SR-135 D5 500,000 800
RS-24832-A SR-45 D1 2,000,000 1000
RS-25200-A US-35 D3S3 4,700,000 17000
RS-25251 SR-3 D4 2,000,000 7000
68
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

High volume or high ESAL count pavements require an aggregate blend of both
crushed stone and slag aggregate. Therefore, steel and blast furnace slag aggregate samples
were obtained and tested along with the stone aggregate samples.
3.5 Laboratory Testing of Aggregate
Laboratory testing was performed on each slag, stone aggregate, and aggregate blend. Each
aggregate and aggregate blend was subjected to polishing using the British polishing wheel
for ten hours. Initial and terminal polish values were taken using the British Pendulum
Tester in accordance with ITM 214 and Texas test method TEX-438-A, Accelerated
Polish Test for Coarse Aggregates for comparison purposes. Stone aggregates were also
tested for total acid insoluble residue.
4. Experimental Procedures
4.1 Insoluble Residue in Carbonate Aggregates
Testing for total acid insoluble residue in carbonate aggregates was performed in
accordance with ASTM D3042, Standard Test Method for Insoluble Residue in
Carbonate Aggregates. The purpose of the test is to determine the percentage of non-
carbonate material in carbonate rock types and the gradation of insoluble material.
Hydrochloric acid is used to dissolve the carbonate constituents, and the resulting data are
indicators of the differential hardness in carbonate aggregates.
Hydrochloric acid was added to a sample of approximately 500 g (1.1 lb) of aggregate
and the sample agitated until effervescence ceases. The solution was then decanted and 300
mL (1.3 oz) of hydrochloric acid added and again examined for additional reaction. When
a reaction was no longer observed, the sample was heated to approximately 110C (230F)
for one hour to completely react the carbonate portion. If a reaction occurred upon heating,
the procedure was repeated, otherwise the solution was decanted and water added to dilute
the solution until a pH of more than 5.5 was reached.
The sample was next thoroughly washed over a series of nested sieves, ranging in
opening size from 0.075 mm (No. 200) to 4.75 mm (No. 4), and a receiving tank. The
sieves and residue were dried at 110C (230F) and the mass difference between the dry
sieves and the dry sieves plus the residue determined. The decanted solution in the
receiving tank was then poured through a 0.075 mm (No. 200) sieve and filtering paper.
When filtration was complete, the remaining residue was rinsed with clean water, and the
filter paper and residue transferred to an evaporating dish and dried along with the 0.075
mm (No. 200) sieve and residue at 110C (230F). The mass of the remaining residue was
69
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HADDOCK, OBRIEN
determined, the mass of the residue from each sieve and the filtering paper summed, and
the percentage of insoluble residue calculated.
4.2 British Pendulum Tester
The test method for measuring surface frictional properties determines the effects of
aggregate polishing by tire action. According to Kummer and Meyer (1962), the British
Pendulum Tester (BPT) (Figure 1) is a simple machine that provides highly consistent
readings. It works by releasing a pendulum with uniform force by gravity from a given
height. A rubber slider is attached to the end of the pendulum, which comes in contact with
the pavement or test specimen surface upon release. When the pendulum is released and
swings down making contact with the pavement or specimen surface, it pushes a pointer up
along a calibrated measuring device and leaves it at the highest point reached by the
pendulum. The less friction that is encountered by the rubber slider, the higher the
pendulum will reach on the calibrated dial resulting in a lower value. One main advantage
is its insensitivity to operator effects (Dillard and Mahone, 1963; Giles et al., 1962).
Figure 1. British Pendulum Tester (ASTM, 2002a).
The BPT can be used to take initial, intermediate, or terminal polishing readings.
Readings are reported as British polishing numbers (BPN) or as polished stone value (PSV
or PV). Initial readings are reported before a specimen undergoes polishing in a polishing
machine and are designated with a zero subscript (BPN0). Interval readings are taken at
given times during the polishing process and are designated with a subscript of the number
70
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of hours of polishing that the specimen has undergone. The physical meaning of a BPN or
PSV value is the coefficient of friction multiplied by 100. Values in this study are reported
as BPN0 for initial readings before polishing, and BPN10 for terminal readings after ten
hours of polishing.
Testing was performed in accordance with ASTM E303, Standard Test Method for
Measuring Surface Frictional Properties Using the British Pendulum Tester, ITM 214,
and TEX-438-A. The difference between ITM 214 and TEX-438-A is the stiffness of the
rubber slider. ITM 214 uses the ASTM E303 specified rubber slider, which is less stiff
than the TEX-438-A rubber slider. Since testing was conducted on samples prepared for
and subjected to accelerated polishing, the test specimens were curved to fit the British
polishing wheel. Five test specimens were prepared and tested for each crushed stone, slag,
and stone-slag aggregate blend.
Each test specimen was prepared in a mold and had a test surface of 44 by 89 mm (1.76
by 3.56 in.) and an arc to fit a 406 mm (16.2 in.) circle. Coarse aggregate passing the 9.5
mm (3/8 in.) sieve and retained on the 4.75 mm (No. 4) sieve was tested. This grain size
was in accordance with the INDOT Standard Specification. The aggregate was thoroughly
washed and dried to a constant mass at a temperature of approximately 105C (221F).
The mold was coated with a release agent, followed by a bonding agent. A single layer of
densely packed aggregate was placed by hand into the mold and the interstices between the
aggregates filled with sand from one-fourth to one-half the aggregate depth. After the
bonding agent cured, the specimen was removed from the mold. A test coupon and mold
are shown in Figure 2.
Figure 2. British Polishing Wheel test coupon and mold (Liang, 2003).
Particle arrangement has been identified as affecting polish values and splitting the
sample is recommended over handpicking each aggregate (Won and Fu, 1996). For this
study, the split method was used for coupons containing aggregate blends.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
71

HADDOCK, OBRIEN
Before taking the initial friction value (BPN0), the pendulum was leveled, the drag
pointer zeroed, and the slide length adjusted. The test specimen was cleaned to be free of
loose particles and was held firmly in place in order to resist movement by the force of the
pendulum. Water was applied to thoroughly coat the test surface and one test swing
performed without recording the reading. Additional test swings were continued and
recorded until four consecutive identical readings are observed, as recommended by Won
and Fu (1996). The test surface was rewetted with each swing, and the results recorded.
Since frictional resistance of most dry surfaces is sufficient for safe driving conditions,
the BPT is performed with wetted specimens. The BPT frictional resistance therefore more
closely resembles the frictional resistance measured in the field by the full-scale, locked-
wheel method using a ribbed tire rather than a smooth tire (Giles et al., 1962). The width of
the rubber slider acts like the grooving of the ribbed tire allowing water to escape around
the slider when it comes in contact with the specimen. Thus, the BPT is a good measure of
micro-texture rather than macro-texture (Henry et al., 2000).
4.3 British Polishing Wheel
The British Polishing Wheel (Figure 3) was mounted onto the British Polishing
Machine (Figure 4) and used to simulate polishing of coarse aggregate in HMA pavements
by vehicular traffic according to ASTM D3319, Standard Practice for the Accelerated
Polishing of Aggregates Using the British Polishing Wheel. Fourteen specimens, prepared
as described in the previous section, were clamped along the outside edge of a wheel to
create a continuous surface of aggregate. The wheel was then mounted onto the machine
and rotated at a speed of approximately 320 revolutions per minute. A smooth surfaced,
pneumatic, rubber-tired wheel was brought into contact with the surfaces of the specimens
in order to apply a load of 390 N (87.75 lbf). Number 150 silicon carbide grit was fed onto
the specimen surfaces at a rate of 6 grams (0.21 oz) per minute for the entire testing time.
Water was also fed onto the surface at a rate of about 50 to 75 milliliters (0.21 to 0.32 fluid
oz) per minute.
Figure 3. British Polishing Wheel (Liang, 2003).
72
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 4. British Polishing Machine (Liang, 2003).

In the tests reported herein, after ten hours of polishing, the wheel was removed and
cleaned to remove the silicon carbide grit. Two BPN10 polish values were then determined
in accordance with test methods ITM 214 and TEX-438-A.
5. Results and Analyses
5.1 Insolubility of Carbonate Aggregates

Acid insolubility test results are presented in Table 5. The results reveal that the stone
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
aggregates used in this study contain a negligible amount of insoluble material, and
therefore the insoluble portion would have an insignificant effect on the frictional
resistance of the aggregate. As previously discussed, the general consensus of the amount
of total insoluble residue needed to affect the frictional resistance of an aggregate is about
ten to twenty percent. The total acid insoluble residue for all stone aggregates in this study
ranged from none to 2.8%, far below this value range.
Gradations of the insoluble portions of the aggregates were also performed and are
included in Table 5. Less than one percent of the total insoluble residue for all stone
aggregates was smaller than the 0.075 mm (No. 200) sieve, which is silt and clay-sized
particles. Aggregate D3 did not contain any insoluble materials. Aggregate D4 had the
most total insoluble residue at 2.8 percent. Of the 2.8% insolubles in aggregate D4, 97.9%
was sand-sized particles, which are generally considered to provide increased frictional
resistance. The remaining four aggregates, D1, D2, D5 and D6, contained less than one
percent total insoluble residue.
73

HADDOCK, OBRIEN
Table 5. Acid insoluble residue test results for stone aggregates.

Source ID D1 D2 D3 D4 D5 D6
Sieve Size Cumulative Percent Retained
No. 4 80.0 87.0 - 97.9 42.8 0.0
No. 8 80.0 93.5 - 99.3 54.2 33.3
No. 16 80.0 95.7 - 100.0 59.9 83.3
No. 30 80.0 95.7 - 100.0 71.3 83.3
No. 50 80.0 97.9 - 100.0 82.7 83.3
No. 100 100.0 97.9 - 100.0 88.4 100.0
No. 200 100.0 100 - 100.0 99.9 100.0
Total Acid
Insoluble 0.1 0.9 0.0 2.8 0.7 0.1
Residue %
5.2 Laboratory Frictional Resistance
British pendulum data are presented in Table 6. The BPNavg values are representative of the
average of the five coupon specimens prepared for each aggregate and aggregate blend.
The hardness, frictional resistance, and polishing resistance of each aggregate are given in
Table 7. BPN values are indicative of the frictional resistance of an aggregate before and
after polishing. Higher BPN0 and BPN10 values represent higher frictional resistance. The
difference in BPN0 and BPN10 is indicative of the resistance to polishing of the aggregate,
and is known as the wear index (WI). Smaller differences in BPN0 and BPN10 correspond
to a greater resistance to polishing.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
For the stone aggregates, aggregate D2 had the highest initial and terminal BPN values,
and D3 had the lowest initial and terminal BPN values. Both aggregates showed sufficient
field skid resistance. However, data available for aggregate D3 only included HMA
wearing courses that are one to two years in age, which is when frictional resistance tends
to be higher. Aggregate D6 showed the greatest loss in frictional resistance after polishing,
but still retained a relatively good level of frictional resistance, which was also seen in the
field data. Although aggregate D3 had the lowest frictional resistance of all the stone
aggregates, it appeared to retain a similar level in frictional resistance after polishing using
the TEX-438-A rubber slider.
74

Table 6. Polishing test results.

ITM 214 TEX-438-A BPN0-BPN10
Aggregate
BPN0avg BPN10avg BPN0avg BPN10avg WIITM 214 WITEX-438-A
Stone Aggregates
D1 38 30 50 31 8 19
D2 47 36 52 38 11 14
D3 36 25 33 29 11 4
D4 45 33 48 37 12 11
D5 44 34 46 37 10 9
D6 48 33 55 35 15 20
Slag Aggregates
S1 40 34 46 40 6 6
S2 46 33 50 38 13 12
S3 42 35 51 39 7 12
S4 48 33 53 39 15 14
Stone/Slag Blends
D5S1 40 35 44 38 5 6
D1S1 38 34 43 39 4 4
D6S1 40 36 45 41 4 4
D6S2 41 37 45 40 4 5
D3S3 39 30 42 32 9 10
D2S4 46 40 51 45 6 6
Table 7. Summary of aggregate hardness, frictional and polishing resistance.
Frictional
Aggregate Hardness Polishing Resistance
Resistance
Stone Aggregates
D1 Moderate Low to Moderate Low to Moderate
D2 Hard High Moderate
D3 Moderate Low Low to Moderate
D4 Hard High Low to Moderate
D5 Hard High Low to Moderate
D6 Moderate to Hard Moderate to High Moderate --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Slag Aggregates
S1 Hard High High
S2 N/A High Moderate
75

HADDOCK, OBRIEN
A two-way analysis of variance (ANOVA) for the stone aggregate BPN10 data
indicates there are statistically significant differences (=0.05) between the mean BPN10
values for the six stone aggregates and between the two test methods. The interaction
between aggregate type and test method is not statistically significant.
Given the indicated differences between the mean BPN10 values for the six stone
aggregates, the Tukey HSD test for multiple mean comparisons was used to determine
which mean pairs are considered statistically equivalent and which are not (=0.05). The
method indicates the mean pairs of D2-D4, D2-D5, D4-D5, D4-D6, and D5-D6 are
considered statistically equivalent. All other mean comparisons are considered statistically
different.
All slag aggregates showed a high degree of frictional resistance before and after
polishing, but the blast furnace slag aggregates (S2, S3 and S4) appear to have higher
initial frictional resistance. Although aggregate S1, a steel slag, had the lowest initial
frictional resistance of all the slag aggregates, it showed the smallest loss in frictional
resistance after polishing with a difference in initial and terminal BPN values of only six
units. Therefore, S1 is considered highly polish resistant. The difference in initial and
terminal BPN values for the blast furnace slag aggregates indicates that these aggregates
are moderately polish resistant. However, terminal (BPN10) values for blast furnace slag
indicate a relatively good terminal frictional resistance.
A two-way ANOVA for the slag aggregate BPN10 data indicates there are no
statistically significant differences (=0.05) between the mean BPN10 values for the four
slag aggregates. However, as with the stone aggregates, there is again a significant
difference between the two test methods. The interaction between aggregate type and test
method is not statistically significant.
The BPN10 data for aggregate blends (Table 6) was also subjected to a two-way
ANOVA. The result indicates there are statistically significant differences (=0.05)
between the mean BPN10 values for the six blends and between the two test methods. The
interaction between aggregate blend and test method is not statistically significant. Again,
the Tukey HSD test for multiple mean comparisons was used to determine which mean
pairs are considered significantly equivalent and which are not (=0.05). The method
indicates that the means of blends D1S1-D5S1, D1S1-D6S1, D1S1-D6S2, D5S1-D6S1,
D5S1-D6S2, and D6S1-D6S2 are statistically equivalent. All other mean comparisons are
considered significantly different.
While initial, terminal, and difference in BPN values for aggregate blends were
presented in Table 6, Table 8 is a summary of the differences in terminal BPN10 values of
the stone and slag components of each blend. This table illustrates the changes in frictional
resistance of the stone and slag aggregates from blending. Terminal BPN10 values for the
stone and slag aggregate are subtracted from the terminal BPN10 value of the blend.
76
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Positive differences indicate an increase in frictional resistance, whereas negative

differences indicate a loss. Table 9 summarizes the rating of effectiveness of the blend
along with additional comments.
Table 8. Summary of differential wear measurements and differences in BPN10 values

between blends and each blend component.
BPN10 (ITM 214) BPN10 (TEX 438-A) DW
Stone Blen Slag B- Stone Blen Slag B- x0.001
Blend (A) d (B) (C) B-A C (A) d (B) (C) A B-C (in.)
D1S1 30 34 34 +4 0 31 39 40 +8 -1 0.3
D2S4 36 40 33 +4 +7 38 45 39 +7 +6 -0.1
D3S3 25 30 35 +5 -5 29 32 39 +3 -7 -0.2
D5S1 34 35 34 +1 +1 37 38 40 +1 -2 -0.2
D6S1 33 36 34 +3 +2 35 41 40 +6 +1 0.5
D6S2 33 37 33 +4 +4 35 40 37 +5 +3 0
All terminal BPN values for stone aggregates were improved by the blend. Although an
improvement of the stone aggregate terminal BPN value is an objective of blending, highly
effective blends improve the frictional resistance of both components. Therefore, there
could be a more effective blend that is also more cost effective, especially for blend D3S3.
Differential wear (DW) measurements are also presented in Table 8. DW is the
difference in particle height measurements of the low and high friction components. Larger
DW measurements indicate a compatible and effective blend. Small DW measurements do
not necessarily mean the blend does not provide sufficient skid resistance, but it does
indicate that the blend could be improved with some alternative that also could be cost
effective. It is suspected that DW measurements are low because the stone aggregates used
in this study are considered skid resistant per ITM 205, and therefore have terminal polish
values that are too similar with the high friction component. All blends in this study could
be improved by blending the high friction component with a softer, carbonate aggregate.
77
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HADDOCK, OBRIEN
Table 9. Summary of improvement in frictional resistance.

Improvement
in Frictional
Blend Resistance Comments
D1S1 effective Although the DW is negligible, the frictional resistance
of the blend is slightly higher than that of the high
friction component (i.e. the slag). Blend could be
improved by use of a softer carbonate.
D2S4 highly Frictional resistance of the blend significantly exceeded
effective that of the high friction component even though the
DW indicates equal wear rates of the two components.
Frictional resistance could be further improved by
blending either component with a softer carbonate
source.
D3S3 effective Although the frictional resistance was improved, it still
is not sufficient for use in high traffic roads. DW of -
0.0002 indicates the aggregates are incompatible and
have equal wear resistance.
D5S1 marginal Marginal improvement due to similar frictional
characteristics between the slag and stone aggregate.
Frictional resistance could be improved by blending
either component with an aggregate with lesser wear
resistance.
D6S1 effective Although DW is negligible, frictional resistance was
increased from the dolomite low frictional resistance of
the dolomite to the slag moderate frictional resistance.
D6S2 highly The friction level of the blend significantly exceeded

effective that of the high friction component. The frictional
resistance could be further increase with the use of a
softer carbonate. The DW value of 0 indicates the equal
wear rates of the two aggregates.
5.3 Comparison of Terminal BPN and Aggregate Physical Properties

BPN10 values from both test methods (ITM 214 and TEX-438-A) were plotted as a
function of bulk specific gravity, elemental magnesium, absorption, LA abrasion loss,
brine loss, and total acid insolubility results for each stone aggregate.
Figure 5 shows the relationship between bulk specific gravity and terminal BPN10
values with the ITM 214 and TEX-438-A test methods, respectively. The relationship is
moderate to strong with R2 values of 0.68 and 0.83 for the ITM 214 and TEX-438-A test
78
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

methods, respectively. The Standard Error of the Estimate (SEE), a measure of the
accuracy of the estimates from the linear regression models, are 2.46 and 1.67 for the ITM
214 and TEX-438-A test methods, respectively. The plot demonstrates that the terminal
BPN10 value increases with decreasing bulk specific gravity. Review of the literature
revealed that the impurity of a dolomite aggregate is related to the frictional resistance,
which can be determined by the specific gravity, elemental magnesium and total acid
insoluble residue. As stated previously, the specific gravity of the dolomite mineral is
approximately 2.800 to 2.900. Impurity of the dolomite aggregate increases as the
difference between the specific gravity of the dolomite aggregate and of the dolomite
mineral increases. Figure 5 supports this claim.
50
45
y = -26.37x + 103.13
R = 0.83
Average BPN10
40
35
ITM 214
30 TEX-438-A
y = -25x + 96.91
25 R = 0.68
20
2.2 2.4 2.6 2.8
Bulk Specific Gravity
Figure 5. Relationship between BPN10 and bulk specific gravity of stone aggregates.
The literature review also revealed that the purity of a dolomite is reflected in higher
percent elemental magnesium. Figure 6 shows the relationship between percent elemental
magnesium and terminal BPN10 values with the ITM 214 and TEX-438-A test methods,
respectively. The relationship is excellent with R2 values of 0.97 and 0.91 for the ITM 214
and TEX-438-A test methods, respectively; SEE for the ITM 214 is 0.76% while the SEE
for the TEX-438-A is 1.31% . The maximum BPN10 value occurs at 11.3% and 11.2% for
the ITM and TEX-438-A methods, respectively. The current INDOT minimum allowable
value of 10.3% elemental magnesium for dolomite aggregates results in a slightly lower
79
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HADDOCK, OBRIEN
BPN10 value. Also, increasing the percent elemental magnesium beyond approximately
11.2% does not necessarily increase the frictional resistance of the aggregate.
40
y = -5.03x2 + 112.21x - 589.02
R = 0.91
35
Average BPN10
30
ITM 214
25
y = -6.48x2 + 145.81x - 785.18
R = 0.97
20
10 11 12 13 14
Elemental Magnesium, %
Figure 6. Relationship between BPN10 and percent elemental magnesium of stone

aggregates.
50
45
y = 2.42x + 28.45
40 R = 0.81
Average BPN10
35 ITM 214
TEX-438-A
30
y = 2.33x + 25.99
R = 0.68
25
20
0 2 4 6
Absorption, %
Figure 7. Relationship between BPN10 and percent absorption of stone aggregates.
80
--`,`,,,``,,```,,`,,,,,`

The relationships between terminal BPN using both test methods (ITM 214 and TEX-
438-A) and percent absorption of the stone aggregate are illustrated in Figure 7. The plot
shows that terminal frictional resistance increases with increasing percent absorption. The
R2 values are 0.68 and 0.81 for ITM 214 and TEX-438-A test methods, respectively; SEE
values for the regression models are 2.45% and 1.80% for ITM 214 and TEX-438-A test
methods, respectively. An aggregate with a higher percent absorption may have a higher
porosity (Lee et al., 1990), which may provide an irregular surface. As the aggregate
surface is worn down, the irregular surface is renewed providing greater frictional
resistance. This finding is consistent with the percent absorption and specific gravity
relationship. Percent absorption increases with decreasing specific gravity, which indicates
that there is a loss of density with increasing porosity.
Figure 8 is a plot of BPN10 as a function of percent LA abrasion loss for both test
methods. The relationship between terminal frictional resistance and results from the LA
Abrasion test is strong with R2 values of 0.80 (SEE = 1.95%) and 0.83 (SEE = 1.67%) for
ITM 214 and TEX-438-A test methods, respectively. The plot shows that terminal
frictional resistance of the stone aggregate increases with increasing LA abrasion loss.
Blast furnace slag provides good frictional resistance by renewal of the surface with
abrasion; however, blast furnace slags tends to fail requirements for the LA abrasion loss
test and therefore are not tested. For the same reason, frictional resistance for stone
aggregates increases with a higher percent LA abrasion loss.
50
45
y = 0.80x + 14.24
R = 0.83
40
Average BPN10
35 ITM 214
TEX-438-A
30
y = 0.82x + 11.0
25 R = 0.80
20
15 20 25 30 35 40
Los Angeles Abrasion Loss, %
Figure 8. Relationship between BPN10 and Los Angeles Abrasion loss of stone aggregates.
81
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HADDOCK, OBRIEN
The relationship between terminal BPN10 and brine loss and is shown in Figure 9.
Values of R2 are nearly identical for the ITM 214 and TEX-438-A methods at 0.44 and
0.43 respectively; SEE for ITM 214 and TEX-438-A test methods are 3.72% and 3.51%,
respectively The brine loss test simulates freezing and thawing of an aggregate, whereas
LA abrasion loss is an impact test. It appears that impact, like load, has a greater effect on
friction number than mechanical weathering. It is suspected that increased loads (higher
ESAL values) assist in retaining frictional resistance.
50
45 y = -0.03x2 + 0.84x + 32.13

R = 0.43
40
Average BPN10
35
ITM 214
30
y = -0.03x2 + 0.91x + 29.47
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
R = 0.44
25
20
0 10 20 30
Brine Loss, %
Figure 9. Relationship between BPN10 and percent brine loss of stone aggregates.
Terminal polish values were plotted as a function of total acid insoluble residue for
both test methods and are shown in Figure 10. The relationship between the two is good
with R2 values of 0.86 (SEE = 1.81%) and 0.75 (SEE = 2.09) for ITM 214 and TEX-438-A
test methods, respectively. However, as pointed out previously, the amount of total
insoluble residue for the aggregates used in this study was negligible and therefore would
have an insignificant effect on the frictional resistance; the results in Figure 10 should be
viewed with this in mind.
82

50
45
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
y = 36.24x0.021
40 R = 0.75
Average BPN10
35
ITM 214
y = 33.85x0.026
30 R = 0.86 TEX-438-A
25
20
0 1 2 3
Total Acid Insoluble Residue, %
Figure 10. Relationship between BPN10 and total acid insoluble residue of stone
aggregates.
5.4 Correlation of Laboratory Tests and Field Friction Tests

Frictional resistance values of in-service pavements were obtained from the INDOT
database. INDOT measures pavement friction using a standard smooth tire in accordance
with ASTM E274, Standard Test Method for Skid Resistance of Paved Surfaces Using a
Full-Scale Tire, ASTM E524, Standard Specification for Standard Smooth Tire for
Pavement Skid-Resistance Tests, and guidelines set forth (Li et al., 2003). Testing is
conducted at a speed of 64 km/h (40 mph) and reported as a friction (skid) number, FN.
Although not a part of the work reported in this paper, the in-service pavement frictional
number data indicate that pavements older than three years tend to reach equilibrium FN
values and these equilibrium values are approximately the same as before Superpave
mixture designs were implemented.
An attempt was made to correlate the in-service equilibrium FN values to BPN10
measured in the lab. The BPN10 for both ITM 214 and TEX-438-A test methods were
plotted as a function of the equilibrium FN values. As seen in Figure 11 the R2 values for
BPN10 obtained from both tests methods strongly indicate that there is not a relationship.
83

HADDOCK, OBRIEN
However, scatter in FN values from skid tests could be a source of difficulty in observing
any strong relationships. Also, the trailer test speed of 64 km/h (40 mph) may not represent
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
the actual in-service speed of the facility. Additionally, as was stated earlier, the BPN
frictional resistance as measured in the BPT more closely resembles the in-service
frictional resistance measured by the full-scale, locked-wheel method using a ribbed tire
rather than a smooth tire (Giles et al., 1962). Thus, BPN is a good measure of micro-
texture rather than macro-texture (Henry et al., 2000). Since INDOT uses the locked-
wheel, smooth tire, more a measure of macro-texture, it stands to reason that the BPN
would not correlate well with the in-service FN numbers obtained by the INDOT testing.
50
y = -0.10x + 40.09
R = 0.08
40
Average BPN10
30
y = -0.085x + 36.43 ITM 214

20 R = 0.11 TEX-438-A
10
10 20 30 40 50
Average Friction Number
Figure 11. Comparison of field and laboratory friction data.
6. Conclusions and Recommendations
Laboratory testing was conducted to determine total insoluble residue and frictional
characteristics of aggregates. Physical property data collected from INDOT quality
assurance records and total insoluble residue data were correlated with aggregate frictional
resistance results. Findings and conclusions are summarized and recommendations made in
the following subsections.
84

6.1 Relationship Between Physical Properties and Frictional Resistance
Physical property characteristics of each aggregate were plotted against corresponding

laboratory BPN10 results from both the ITM 214 and TEX-438-A test methods. The TEX-
438-A test method had higher coefficients of determination with the Los Angeles abrasion
results, and coarse aggregate bulk specific gravity and water absorption; the ITM 214 test
method had higher coefficients of determination with the amount of elemental magnesium
in coarse aggregates, brine loss and acid insoluble residues.
Comparison of BPN10 values to specific gravity showed that BPN10 increases with
decreasing bulk specific gravity, indicating that impure dolomites have greater frictional
resistance. Additionally, the terminal polish value for dolomite aggregates is higher when
percent elemental magnesium is approximately 11.2%. This also confirms that impure
dolomites have greater frictional resistance. BPN10 of coarse aggregates also increases with
higher percent water absorption and percent LA abrasion loss. The relationship between
BPN10 and brine loss is moderate and suggests that a coarse aggregate brine loss
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
percentage of 14-15% may be desirable. While the relationship between BPN10 and total
acid insoluble residue appears to be moderate, the total acid insoluble residue for all
dolomite aggregates used in this study was less than three percent and is not expected to
have an effect on frictional resistance. Currently, INDOT does not measure acid aggregate
insoluble residue for frictional specification purposes.
Each of the relationships between BPN10 and the aggregate physical property tests used
in the research show a bias between the ITM 214 and TEX-438-A test methods. This bias
is most certainly a result of the different rubber sliders used for each method. The ITM 214
test method uses the ASTM E303 specified rubber slider, which is less stiff than the TEX-
438-A rubber slider. At first glance it may seem that since the bias is consistent, it will not
matter which test an agency chooses to use. However, since each test method is better
correlated to certain aggregate physical property tests, an agency should choose the
laboratory polishing test method that best correlates to the aggregate properties being used
by the agency. For example, for INDOT, the percentage of elemental magnesium is an
important property in identifying dolomite aggregates for use in HMA pavement surfaces.
Thus the use of ITM 214 is appropriate since it correlates to the BPN10 values better than
does the TEX-438-A test method.
British polishing results indicate that harder stone aggregates tended to have higher
frictional resistance but not necessarily high polishing resistance. The frictional resistance
of all stone aggregates was improved by blending. However, a few blend combinations
used in this study show a reduction in frictional resistance of the slag component. Blends
could be improved by blending the high friction component (slag) with a softer carbonate
aggregate. A blend is considered effective if both components increase in frictional
resistance. Blends are considered compatible if differential wear measurements are large.
85

HADDOCK, OBRIEN
However, small differential wear measurements do not necessarily indicate insufficient

skid resistance, but it does indicate that the blend could be improved.
Finally, when the BPN10 data were correlated to in-service equilibrium FN values, no
relationship was observed. This may be due to the fact that INDOT performs locked-wheel
friction testing with the smooth tire, measuring macro-texture of the surface rather than the
micro-texture. If some locked-wheel friction testing were performed with the ribbed tire,
the results might better correlate to the BPN values measured on the coarse aggregates
used in the HMA surface mixture.
Overall, the current testing INDOT uses does appear to adequately screen coarse
aggregates for frictional properties. However, this does not mean that improvements
cannot be made.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
6.2 Recommendations
From the analyses conducted in this study, the following is recommended:

1. Raising the minimum requirement for percent elemental magnesium in dolomite
could help to insure the use of dolomites that offer higher friction resistance. The
relationship between laboratory measured aggregate friction properties and
percent elemental magnesium in aggregate established in this research suggests
that were the required minimum elemental magnesium percentage raised to 11%,
the average BPN10 for the aggregates in this study would be raised from 30.2 to
34.7. This recommendation should be weighed against the number of dolomite
sources that would be disallowed were the specification raised. Two of the six
dolomites used in this study have elemental magnesium values below 11%.
2. Use tests that identify the impurity of carbonate aggregates to identify skid
resistant aggregates. This could include the total acid insoluble residue test along
with a grain size distribution that includes material smaller than the 0.075 mm
(No. 200) sieve.
3. Continue using the ASTM E303 specified rubber slider for BPT as specified in
the ITM 214 test method.
4. Conduct a study to investigate the effect of blending high friction aggregates with
softer carbonate aggregates.
5. Measure at least some in-service pavement friction using the ribbed tire in order
to better correlate the results with laboratory measured aggregate properties.
86

7.0 Acknowledgment
The authors wish to acknowledge the support of Indiana Department of Transportation, the
Joint Transportation Research Program at Purdue University, and the Federal Highway
Administration. The contents of this report reflect the views of authors who are responsible
for the facts and accuracy of the data presented herein. The contents do not necessarily
reflect the official views or policies of the Indiana Department of Transportation or the
Federal Highway Administration at the time of publication.
8.0 References
Abdul-Malak, M.U., Fowler, D.W. and Meyer, A.H., "Mineralogy of Aggregates in

Relation to the Frictional Performance of Seal Coat Pavement Overlays: a
Petrographic Study." Transportation Research Record, No. 1418, 1993. p. 3542.
Asphalt Institute, Superpave Level 1 Mix Design. Superpave Series No. 2 (SP-2),
Lexington, KY, 1995.
ASTM, "Standard Practice for Accelerated Polishing of Aggregates or Pavement Surfaces
Using a Small-Wheel, Circular Track Polishing Machine." Book of ASTM Standards.
ASTM International, West Conshohocken, PA, 2002a.
ASTM, "Terminology Relating to Vehicle Pavement Systems." Book of ASTM Standards.
ASTM International, West Conshohocken, PA, 2002b.
Burnett, W.C., Gibson, J.L. and Kearney, E.J., "Skid Resistance of Bituminous Surfaces."
Highway Research Record, No. 236, 1968, p. 4960.
Chesterman, C.W., National Audubon Society Field Guide to North American Rocks and
Minerals. Alfred A. Knoph, Inc., New York, 1979.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Dahir, S.H.M. and Mullen, W.G., "Factors Influencing Skid-resistance Properties."
Dillard, J.H. and Alwood, R.L., "Providing Skid-resistant Roads in Virginia." Proceedings
of the Association of Asphalt Paving Technologists, Vol. 26, 1957, p. 122.
Dillard, J.H. and Mahone, D.C., "Measuring Road Surface Slipperiness." 66th Annual
Meeting Papers, ASTM 366, Atlantic City, NJ, 1963.
Fu, C.N., "Aggregate Properties and Blending versus Frictional Performance of
Bituminous Pavements." Raba-Kistner Consultants, Inc., 2004.
87

HADDOCK, OBRIEN
Furbush, M.A. and Styers, K.E., "Relationship of Skid Resistance to Petrography of

Aggregates." Pennsylvania Department of Transportation, Harrisburg, PA, 1972.
Giles, C.G., Sabey, B.E. and Cardew, K.H.F., "Development and Performance of the
Portable Skid-Resistance Tester." ASTM Special Technical Publication No. 326:
Symposium on Skid Resistance - 65th Annual Meeting Papers. American Society for
Testing and Materials, New York, NY, 1962, p. 5074.
Gramling, W.L. and Hopkins, J.G., III, "Skid Resistance Studies: Aggregate-Skid
Resistance Relationship as Applied to Pennsylvania Aggregates." Report No. 65-4,
Pennsylvania Department of Transportation, Harrisburg, PA, 1974
Gray, J.E. and Renninger, F.A., "The Skid-Resistant Properties of Carbonate Aggregates."
Henry, J.J., NCHRP Synthesis 291: Evaluation of Pavement Friction Characteristics.
Transportation Research Board - National Research Council, Washington, D.C., 2000.
INDOT, Standard Specifications, 2012.
Jayawickrama, P.W., Prasanna, R. and Senadheera, S.P., "Survey of State Practices to
Control Skid Resistance on Hot-Mix Asphalt Concrete Pavements." Transportation
Research Record (1536), 1996, p. 5258.
Kearney, E.J., McAlpin, G.W. and Burnett, W.C., "Development of Specifications for
Skid-Resistant Asphalt Concrete." Highway Research Record, No. 396. 1972, p. 12
20.
Kummer, H.W. and Meyer, A.H., "Measurement of Skid Resistance." ASTM Special
Technical Publication No. 326: Symposium on Skid Resistance - 65th Annual
Meeting Papers. ASTM Special Publication No. 326. American Society for Testing
and Materials, New York, NY, 1962.
Lee, D., Guinn, J.A., Khandhal, P.S. and Dunning, R.L., "Absorption of Asphalt into
Porous Aggregates." SHRP-A/UIR-90-009, Strategic Highway Research Program,
Washington, D.C., 1990.
Liang, R.Y., "Blending Proportions of High Skid and Low Skid Aggregate". FHWA/OH-
2003/014, Ohio Department of Transportation and US Department of Transportation,
Federal Highway Administration, Akron, OH, 2003.
88
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Mullen, W.G., Dahir, S.H.M. and El Madani, N.F., "Laboratory Evaluation of Aggregates,
Aggregate Blends, and Bituminous Mixes for Polish Resistance." Transportation
Research Record, No. 523, 1974, p. 5664.
Rogers, M.P. and Gargett, T., "Skidding Resistance Standard for the National Road
Network." Highways and Transportation, Vol. 38, No. 4, 1991, p. 1016.
Skerritt, W.H., "Aggregate Type and Traffic Volume as Controlling Factors in Bituminous
Pavement Friction." Transportation Research Record, No. 1418, 1993, p. 2229.
Stock, A.F., Ibberson, C.M. and Taylor, I.F., "Skidding Characteristics of Pavement
Surfaces Incorporating Steel Slag Aggregates." Transportation Research Record, No.
1545, 1996, p. 3540.
Webb, J.W., "The Wearing Characteristics of Mineral Aggregates in Highway Pavements."
VHRC-70-R7, Virginia Highway Research Council, Washington, D.C., 1970.
West, T.R. and Cho, K.H., Development of a Procedure to Identify Aggregates for
Bituminous Surfaces in Indiana. FHWA/IN/JHRP-2000/28, Joint Transportation
Research Program, West Lafayette, IN, 2000.
West, T.R., Choi, J.C., Bruner, D.W., Park, H.J. and Cho, K.H., "Evaluation of Dolomite
and Related Aggregates Used in Bituminous Overlays for Indiana Pavements."
Transportation Research Record, No. 1757, 2001, p.137147.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Won, M. and Fu, C.N., "Evaluation of Laboratory Procedures for Aggregate Polish Test."
Transportation Research Record, No.1547, 1996, p. 2228.
9.0 Discussion
DR. MICHAEL HEITZMAN: Good presentation. I have a couple of comments. One,

when you were trying to put your British pendulum coupons together with blended
aggregates, how did you do the blending? How did you do the coupons for blended
aggregates when typically you only have one aggregate in that coupon?
PROFESSOR JOHN HADDOCK: We actually put them in blended to the desired
percentage.
DR. HEITZMAN: You are individually putting in particles one next to another?
PROF. HADDOCK: Yes.
89

HADDOCK, OBRIEN
DR. HEITZMAN: Okay. I think thats interesting when you're doing blended aggregates.
You observed that two different test methods for friction are going to give you different
values. We are seeing that, as well. We have to be very careful about comparing studies
with respect to absolute friction values because not everyone is running the same tests. My
second comment, I am concerned about the focus on the coarse aggregate properties for
pavement friction when, in fact, the mixture properties on the surface are whats going to
tell us more about the friction. In fact, your one graph (Figure 11 in the reprint) shows that
there was really no correlation between the coarse aggregate properties and the friction
measurements in the field. We have to step beyond coarse aggregate properties and get into
the mixtures. I know Becky McDaniel with the North Central Superpave Center and here
at NCAT, we have that ability now to start looking at those mixture properties on the
surface.
PROF. HADDOCK: Thank you. I agree with the points you have made.
DR. REBECCA MCDANIEL: Well, with that opening, Im Becky McDaniel with the
North Central Superpave Center at Purdue, and I would add to what Mike said. We too
recognize that it is the mixture properties that are key. The Indiana DOT recognizes that as
well, and theyve funded some additional research looking at frictional properties of
mixtures. Weve looked at the contributions of the coarse and the fine aggregate, and the
blending of softer carbonates with high-friction aggregates, which, as John mentioned, is
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
something additional to look at. INDOT has funded research along those lines, and perhaps
I should write a paper based on that. Thank you.
PROF. HADDOCK: Thank you, Becky. I agree with Becky, as well. It is the pavement
that must really be looked at, and I know that Becky has some work going on at the
Superpave Center looking at the pavement. I know that NCAT is doing some things on it,
as well. In Indiana, I might just mention, we have a procedure that we go through. If you
have an aggregate that you would like to use as a friction aggregate, we actually go through
an entire process of lab testing, and then it gets put in the field, and that friction gets
monitored in the field to see what happens. So, I look at these aggregate physical properties
as more of quick checks to make sure that the aggregate might be suitable as friction
aggregates. But in the end, one must really get a measurement of the pavement to tell what
kind of skid resistance the pavement has, Thank you.
90

IlliTCLow Temperature Cracking Model
for Asphalt Pavements
Eshan V. Davea*, William G. Buttlarb, Sofie E. Leonb,
Behzad Behniab and Glaucio H. Paulinob
a
Department of Civil Engineering, University of Minnesota Duluth, Duluth, MN 55812
b
Department of Civil and Environmental Engineering, University of Illinois, Urbana, IL
61801
ABSTRACT: Low temperature cracking (LTC) is a major distress and cause of failure for asphalt
pavements located in regions with cold climate; however, most pavement design methods do not
directly address LTC. The Thermal Cracking Model (TCModel) utilized by the AASHTO
Mechanistic-Empirical Pavement Design Guide (MEPDG) relies heavily on the phenomenological
Paris law for crack propagation. The TCModel predictions are primarily based on tensile strength
of the asphalt mixture and do not account for quasi-brittle behavior of asphalt concrete.
Furthermore, TCModel uses a simplified one-dimensional viscoelastic solution for determination
of thermally induced stresses. This paper describes a newly developed comprehensive software
system for low temperature cracking prediction in asphalt pavements. The software system called
IlliTC utilizes a user-friendly graphical interface with a stand-alone finite-element based
simulation program. The system includes a preanalyzer and data input generator module that
develops a two-dimensional finite element pavement model for the user and which identifies
The oral presentation was made by Professor Dave.
http://dx.doi.org/ 10.1080/14680629.2013.812838
91
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DAVE, BUTTLAR, LEON, BEHNIA, PAULINO
critical events for thermal cracking using an efficient viscoelastic pavement stress simulation
algorithm. Cooling events that are identified as critical are rigorously simulated using a
viscoelastic finite element analysis engine coupled with a fracture-energy based cohesive zone
fracture model. This paper presents a comprehensive summary of the components of the IlliTC
system. Model verifications, field calibration and preliminary validation results are also
presented.
KEYWORDS: Asphalt, thermal cracking, fracture, performance, simulation, cohesive zone,

transverse cracking, viscoelasticity, model, pavement, IDT, DC(T), IlliTC.
1. Motivation and Introduction
One of the main advantages of asphalt concrete over Portland cement concrete (PCC) is
the smoothness and cost savings afforded by continuous paving, i.e., without the need
for transverse joints. Unlike PCC and other infrastructure materials, asphalt concrete is
generally able to undergo thermal cycling without the need for expansion or contraction
joints due to its viscoelastic nature. Under imposed strain, which is constantly occurring
in pavements due to temperature change, viscoelastic materials are able to relax stress
over time. In addition, asphalt is generally a fracture resistant material, owing to its
flexible mastic matrix and particulate composite morphology. Significant energy is
required to initiate and propagate a crack through asphalt concrete, as the asphalt mastic
is tough, strain tolerant, and viscoelastic (stress relaxing), and the aggregates add
strength, crack bridging, and crack surface tortuosity. However, improper selection of
asphalt grade, excessive aging of the asphalt binder, and/or a weak asphalt mixture
(weak aggregates, low cohesion, low adhesion) can all contribute to poor mixture
fracture resistance. Poor mixture fracture resistance can lead to the development of
thermal cracks, which are typically transversely oriented with traffic and periodic in
nature.
Thermal cracking is very serious pavement distress, as it can significantly increase
pavement roughness and because it creates a permanent discontinuity in the pavement
structure. In a recent study by Islam and Buttlar (2012), pavements allowed to reach a
rough condition were shown to increase user costs (vehicle repair, tire, and fuel cost) by
over 5 million dollars per lane mile over a 35 year life cycle. In contrast, by investing
just 1/50th of this cost in additional maintenance, the pavement could be kept in smooth
condition over its lifespan, avoiding these additional user costs. However, in order to
achieve this result, the pavement would need to be properly designed to avoid the
development of medium or high severity thermal cracking, since an improper mixture
92
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

IlliTCLow Temperature Cracking Model for Asphalt Pavements
design could lead to the development of thermal cracks prior to the application of the
first maintenance or rehabilitation treatment.
Asphalt technologists have long recognized the need to control thermal cracking in
asphalt pavements, and the tests and models available to assist in this endeavor have
continuously evolved. Early efforts to control asphalt behavior at low temperatures were
focused on the asphalt binder. A comprehensive review of early binder tests in the US
has been documented by Brown et al. (2009). The penetration test, especially if run at
two temperatures (generally 25C and 4C), provided some control over binder
consistency and temperature susceptibility. The ductility test provided a simple
measure of binder stretch or strain tolerance. However, neither of these tests were
applicable to temperatures below 0 C, where thermal cracking is likely to occur. In
Europe, the Fraass breaking point test is used (as specified by EN 12593:2007), which is
a torture-type binder test designed to determine the temperature at which a thin film of
binder bonded to a small rectangular brass plate becomes intolerant to a bending strain
arising from flexing the brass plate to a specified curvature. Although this test allowed
direct mechanical testing of the binder at temperatures below 0C, its direct relation to
thermal cracking is questionable due to the very high strain level imposed. Superpave
(AASHTO M320) addressed thermal cracking with modern binder tests run at low
temperatures and in fundamental testing configurations. A bending beam rheometer
characterized the stiffness and m-value (a measure of the ability of the binder to relax
stress). An optional direct tension test was also specified, which was designed as a
second referee test to address certain polymer modified binders that possessed higher
stiffness but also high strain tolerance and fracture resistance. However, the system was
never intended to directly control thermal cracking; rather, it was developed as a binder
purchase specification.
Although binder tests are convenient, practical, and important from the perspective
of binder selection, purchase, and quality control, asphalt technologists have
acknowledged the need to address thermal cracking more directly through testing of the
asphalt mixture and modeling of the pavement structure. Canadian researchers
developed limiting mixture stiffness recommendations based upon a comprehensive field
investigation at Ste. Anne (Deme and Young, 1987). During the Strategic Highway
Research Program (SHRP), a mixture-based testing and analysis scheme was developed
to validate the binder test and specification system being developed. A low temperature
mixture creep and strength test was developed, later called the Superpave Indirect
Tension Test (the acronym IDT was developed in early Superpave publications, and is
still commonly used), as specified in AASHTO T-322. Master creep compliance curves,
shift factors, tensile strength, and optionally, mixture coefficient of thermal expansion
and contraction at low temperatures can be obtained with the Superpave IDT. SHRP
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
researchers, working under project A-005 also developed a computer-based thermal
93

cracking model, called TCMODEL, as part of the effort to validate the Superpave binder
specification. The first version of TCMODEL was completed in 1992, near the end of
the SHRP program.
TCMODEL made great strides in modelling some of the key physics underlying the
thermal cracking mechanism. In particular, TCMODEL had a highly sophisticated
viscoelastic pavement response model, which predicted pavement tensile stress versus
depth on an hourly basis throughout the life of the pavement based upon principles of
linear viscoelasticity. Due to limitations in computational power, a phenomenological
pavement cracking model was used for distress prediction. Change in crack length was
predicted using a power-law type model reported by Paris et al. (1961), which uses
change in stress intensity (which was calculated in an approximate manner by
interpolating presolved 2D elastic finite element runs) to predict change in crack length.
In TCMODEL, Paris law parameters were empirically linked to IDT strength and to the
slope of the log mixture compliance vs. log time relationship at long loading times
(mixture m-value). TCMODEL was selected for inclusion in the AASHTO Mechanistic
Empirical Pavement Design Guide, and was improved and streamlined as part of the
NCHRP 1-37A project. For instance, an automated mixture master curve generation
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
program was bundled with TCMODEL. In addition, additional field data, including that
obtained from MnROAD test sections, was used to recalibrate TCMODEL.
With changes in asphalt binder and mixture designs and materials over the past 20
years, such as the increased use of polymers and other additives and the increased use of
recycled asphalt pavement (RAP), it became apparent that the heavy reliance on mixture
tensile strength in cracking predictions was limiting the prediction accuracy of
TCMODEL. New mixture fracture tests were developed to address these new materials,
including the disk-shaped compact tension test, or DC(T), which provided a convenient
means to obtain mixture fracture energy using a fracture mechanics based approach.
Recognizing the shortcomings of TCMODEL in light of modern mixture fracture tests
and computational power, a new thermal cracking model was developed at the
University of Illinois Urbana-Champaign, called IlliTC, as part of a national Pooled
Fund Study on Low Temperature Cracking. IlliTC improves the manner in which
fracture is handled in the simulation scheme, namely; the 1D Paris-Law
phenomenological modeling approach was replaced with a 2D, cohesive zone fracture
modeling approach implemented within a viscoelastic finite element modeling
framework. The cohesive zone approach considers both material strength and fracture
energy in computing crack initiation and propagation using fundamental fracture
mechanics principles. In summary, the new approach used in IlliTC has the following
improvements over TCMODEL:
A 2D model is used instead of 1D.
94

The physics of cracking in a quasi-brittle, heterogeneous particulate composite

are more correctly captured by using a cohesive zone approach, where softening
and fracture have a distinct length scale that is captured.
Asphalt mixtures may have unique combinations of strength and ductility (as
characterized by mixture fracture energy). For instance, some polymer-
modified mixes portray moderate tensile strength and high fracture energy;
some have high strength and lower fracture energy, and some have both high
strength and high fracture energy. Mixtures with higher recycled material
content may have high strength, but low fracture energy. IlliTC can capture all
of these combinations in a direct manner, while TCMODEL could only capture
these effects in an indirect manner.
A user-friendly graphical interface (GUI) has been provided for IlliTC. The
GUI program module within IlliTC is referred to herein as Visual-LTC.
This paper presents the IlliTC model components, model verification, model
calibration, and preliminary validation results. Ongoing research, aimed at adding
additional software capabilities and modeling features to the IlliTC program, are also
described.
2. IlliTC Framework
The software program (IlliTC) provides an intuitive and user-friendly graphical user
interface (GUI), as a convenient gateway to the rigorous viscoelastic finite element /
cohesive zone modeling engine. The overall flow of the IlliTC program along with
various inputs and outputs are graphically illustrated in Figure 1. The code consists of a
GUI, which is represented as a box on the left in Figure 1, and four analysis modules,
which are encased in dotted boxes in Figure 1. The analysis modules include a
preprocessor, an input file generator, a preanalyzer, and a finite element based thermal
cracking prediction engine. An overview of the GUI and analysis modules is presented
in the remainder of this section and further implementation details are presented in
sections 3-5.
To initiate an IlliTC thermal cracking simulation, the user is queried to enter
information pertaining to the project location, design life, pavement structure, and
material properties into the GUI. The GUI handles cumbersome tasks like data
organization and unit conversion, then automatically assembles additional data needed
for analysis from internal databases.
The GUI passes the raw creep compliance data to the preprocessor, which returns the
thermo-viscoelastic material properties in the form of Prony series parameters
(generalized Maxwell model) and time-temperature shift factors. The preprocessor
95
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

consists of two modules, the first one is based on the code Master developed by
Buttlar et al. (1998) and second one is based on the TCModel (Lytton et al., 1993, Roque
et al., 1995a, 1995b). The first module generates the creep compliance master curve
from the raw creep compliance data and fits a generalized Voigt-Kelvin model to the
master curve. The Voigt-Kelvin model for creep-compliance is then converted to a
generalized Maxwell model in the form of a relaxation modulus.
The GUI reads the Maxwell model parameters from the preprocessor output,
organizes it along with other data, and passes it to the input file generator. The input file
generator conducts two main tasks: 1) to develop a finite element mesh for the pavement
geometry specified by the user, called the Geometric Data File, (*.mesh) shown hashed
box in Figure 1, and; 2) to create a material data file, based on information provided by
the user, called the Material Data File (*.mtr), also shown in the hashed box in Figure 1.
Figure 1. Flowchart of new thermal cracking model IlliTC. --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The GUI collects the climatic data necessary to perform the analysis in the chosen
location for the specified duration from internal databases. This data is stored in the
Climatic Information File (*.poly), shown in the hashed box in Figure 1. Then the GUI
passes all input files to the preanalyzer.
96

The preanalysis module (or preanalyzer) serves to minimize the analysis time of
the more rigorous finite element engine. Typically, low temperature cracking analysis is
focused around critical cooling events. Hence, critical cooling events are identified by
the preanalyzer so that the full model can be focused on critical cooling events, typically
reducing computational time by more than 90% (i.e., in a 365 day year, less than 36 day
would typically need to be simulated). The GUI reads the results of the preanalysis
module and extracts the data pertinent to the critical cooling events. These data are then
passed to the finite element model.
The user is not required to have direct interaction with the viscoelastic and cohesive
zone finite element simulation; instead, results are sent back to the GUI, which are
interpreted for the user.
3. Graphical User Interface (GUI)
The GUI, called Visual LTC, collects and compiles the input conditions provided by
the user, executes various analysis modules to conduct finite element analysis, and
interprets and displays the results. Visual LTC was written with the object-oriented
programming language C# (pronounced see-sharp) under Microsofts .NET
framework, and which is intended for the development of deployable software. Visual
LTC was designed to be intuitive to use by practitioners or researchers. A series of
windows are used to query the user for required model inputs. For example, the window
that collects the asphalt layer material properties is shown in Figure 2. User inputs and
options shown in Figure 2 are discussed in more detail in subsequent sections. For
further details on conducting analysis using Visual LTC the reader is referred to Dave et
al., 2011.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
97

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 2. Visual LTC window asphalt layer material properties.
3.1. Communication with Analysis Modules
Data is passed between Visual LTC and the analysis modules via input/output files.
Visual LTC reads the user input then performs the series of converting data, writing
files, executing programs, and reading output shown in Table 1.
98

Table 1. Visual LTC steps.
1. Read and store user input

2. Extract and store climatic information for user specified analysis period
3. Write input files for preprocessor
4. Run preprocessor
5. Read and convert preprocessor output
6. Write input files for input file generator
7. Run input file generator
8. Run preanalyzer
9. Process preanalyzer output to identify critical events and generate finite element
temperature input
10. Run finite element analysis engine
11. Read finite element analysis output
12. Convert crack depth to amount of cracking
13. Display results
3.2. User Types
Visual LTC is intended for use by practitioners and researchers alike. Therefore, two
user types are supported: Standard User, and Advanced User. Both users have
access to all functionality previously described. However, Advanced Users have the
additional capability of adding new asphalt mixes and modifying properties of existing
asphalt mixes. The distinction between these user types is present so that existing
properties are protected from inadvertent user error. The user can easily change from one
user type to the other.
3.3. Visual LTC User Inputs
The main user inputs that are required for low temperature cracking analysis are the
analysis location, analysis duration and the pavement material properties. A series of
Integrated Climatic Model (ICM) simulations were conducted to create a library of
pavement temperature profiles available to the user in Visual LTC. Sets of temperature
profiles were generated for one cold, one intermediate and one warm location in each
participating state of the pooled fund low temperature cracking study, as shown in Table
2. Eventually, IlliTC will be coupled with the ICM model, so that climatic data from
thousands of geographic areas are available within the model. The data required to
generate these libraries are the same as those used in the AASHTO MEPDG system.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
99

Temperature profiles at each location were generated for the following asphalt concrete
thicknesses: 75, 100, 125, 150, 175, 200, 225, 250, 300, 350, and 400 mm.
Table 2. Climatic locations available to user in Visual LTC.
State Climate City Air PG

Temperature Grade
Cold Norfolk -29.5C -28
Connecticut Intermediate Hartford -26C -22
W ar m New Haven -20.5C -22
Cold Elizabeth -37C -34
Illinois Intermediate Urbana -31.5C -28
W ar m Anna -27C -22
Cold De c o r a h -40.5C -34
Iowa Intermediate Des Moines -32C -28
W ar m Fort Madison -30.5C -28
Cold Massena -39C -34
Ne w Yo r k Intermediate Albany -33.5C -28
W ar m New York City -19.5C -16
Cold Westhrope -44C -40
North Dakota Intermediate Bismarck -41.5C -40
W ar m Wahpeton -38C -34
Cold International Falls -43.5C -40
Minnesota Intermediate St. Cloud -41.5C -34
W ar m Worthington -34.5C -34
Cold Minong -46C -40
Wisconsin Intermediate Stevens Point -36.5C -34
W ar m Milwaukee -32C -28
In Visual LTC, the user selects a location that is the most climatically similar to the
analysis location. The user also provides the pavement cross-section. The Visual LTC
extracts the appropriate data from the temperature profiles associated with the location
and pavement cross-section. These data are passed to the preanalyzer and finite element
engine where nodal temperatures are computed. The pavement material property related
user inputs are summarized in Table 3 and discussed below.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
100

Table 3. Summary of pavement material user inputs for IlliTC.

Property Units Test
Tensile strength MP a AASHTO T 322
or Extracted from
DC(T) Test
Fracture energy J/m2 ASTM D7313 1
Option 1 Unit weight g/cm3 AASHTO M323
Mixture VMA 2 % AASHTO M323
Aggregate coefficient mm/mm/C No standardized
of thermal expansion test
and contraction
(CTEC) 2
Option 2 Mixture coefficient mm/mm/C No standardized
of thermal expansion test
and contraction
(CTEC) 3
Creep compliance test data (100 or 1000 1/GPa AASHTO T-322
seconds for 3 temperatures)
Creep compliance test temperatures C AASHTO T-322
1
Fracture energy may be obtained with different test geometry; however the model is
calibrated for the ASTM D7313 (disk-shaped compact tension, DC(T) test procedure
2
Mixture VMA and aggregate CTEC do not need to be entered if Mixture CTEC is provided
3
Mixture CTEC will be calculated if mixture VMA and aggregate CTEC are provided
The tensile strength of asphalt concrete can be determined using the Superpave
indirect tension test (IDT), as specified in AASHTO T 322. Recently, a procedure was
developed to extract mixture tensile strength from DCT test data, which will be
incorporated into IlliTC in the near future. The fracture energy can be determined using
a variety of test geometries, such as disk-shaped compact tension (DC(T)), semi-circular
bend (SCB) and single-edge notched beam (SENB) test, although the model has been
calibrated and validated based on the DCT test geometry. Furthermore, the fracture test
is expected to be performed at a crack mouth opening displacement (CMOD) rate of
0.0167 mm/s and at temperature of 10C above the 98% reliability Superpave PG low
temperature grade, as dictated by the project location.
The user can either directly input the coefficient of thermal expansion and
contraction (CTEC) or provide asphalt mixture volumetric properties. If volumetric
properties are provided, the CTEC is estimated using the approximation equation used
by the AASHTO MEPDG software. The researchers at the University of Wisconsin have
proposed experimental procedures to measure the CTEC of asphalt mixtures (Bahia and
101
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Nam, 2004, 2009, Marasteanu et al., 2007); the use of their procedure is recommended
for added accuracy in prediction. The work by the same research group has shown a
bilinear trend in the volumetric changes that occur in asphalt binder and mixes, with
significantly different CTEC values above and below the glass transition temperature.
This feature is not currently implemented in the IlliTC system; however it is recognized
as one of the tasks for implementation in future versions.
The user can directly enter laboratory measured 100 or 1000 second creep test data
from three temperatures following the AASHTO T-322 test procedure. These data are
passed to the preprocessor, which converts the data into thermo-viscoelastic material
properties in the form of Prony series parameters (Generalized Maxwell model) and time
temperature shift factors.
3.4. Data Storage
A simple and intuitive class structure is employed to store and maintain data required for
low-temperature cracking analysis, i.e. material properties, climatic data, pavement
structure, and project information. The data should be easily accessible by the user and
should not require installation of additional software. A working directory containing
input files stores all of the data necessary for Visual LTC to conduct the analysis.
Furthermore, the user is not required to directly access the files, as Visual LTC creates
and modifies files automatically. The project input file stores general information (i.e.
project name, description, date, etc.), climatic information, and the pavement structure.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Asphalt concrete input files store all material properties associated with the mix. A
working directory can contain many project files, thus giving the user the option of
creating a new project by modifying an existing one. Similarly, the working directory
can contain as many asphalt concrete input files as necessary, which creates a library of
mix designs for the analyst or designer to investigate.
4. Input Generator and Preanalyzer
The IlliTC prediction system is designed to be practitioner friendly and hence all the
necessary inputs for the finite element simulation are generated by the software. The
input generator module handles the creation of finite element mesh and corresponding
data file with all geometric information and also generates a material data file in the
necessary format for the finite element analysis code. The first task of this module is to
develop a finite element mesh for the pavement geometry selected by the user. The
details on the mesh generation were previously discussed by Dave et al. (2011). Briefly,
the code generates a finite element mesh using four node quadrilateral elements (Q4) and
102

it automatically generates a transition mesh with element sizes increasing as the distance
from potential thermal crack increases. A single line of cohesive zone elements are
inserted into the mesh. These elements allow for simulation of a single thermally
induced crack that can span across the pavement thickness. The crack is simulated in the
transverse direction. The input generator takes the material data provided by the user as
well as the viscoelastic parameters determined by the preprocessor and generates the
material data file (.mtr).
The preanalysis module (or preanalyzer) was developed to optimize analysis times of
the finite element engine. A simplified one-dimensional (1D) problem is solved by this
module to identify critical events that are then analyzed by the finite element analysis
engine; hence only critical cooling events are analyzed with the full model. The
preanalyzer module solves the stresses on a restrained 1D viscoelastic body that is
imposed with temperature boundary conditions representative of the temperatures at the
pavement surface. The body is assumed to have same properties as the thermo-
viscoelastic properties of asphalt concrete. The one dimensional viscoelastic solution for
thermal stress can be found in Apeagyei et al. (2008); this solution is implemented in
IlliTC using a recursive-incremental numerical integration method.
The results from the preanalyzer were verified with the analytical solution. The
stresses obtained with the preanlayzer (VE1D) and the analytical solution for a 1D body
imposed with thermal loading shown in Figure 3(b) are compared in Figure 3 (a). The
results show excellent agreement between the preanalyzer results and analytical solution.
( a) (b )
Figure 3. Verification of preanalyzer. (a) Comparison of stresses obtained with the
preanalyzer (VE1D) and the analytical solution. (b) Thermal loading.
103
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The critical cracking events are identified when the thermal stress in the 1D model
exceeds 80% of the tensile strength of the asphalt mix. The thermal stress from the 1D
analysis is assumed to represent the stresses on the pavement surface. The threshold of
80% was selected based on previous experience of researchers in determining the stress
threshold corresponding to the onset of damage. The full analysis with the finite element
engine is performed for the 24 hours surrounding the critical event.
Figure 4 shows an example of the results of the preanalyzer. The winter time surface
temperatures during the five-year analysis period are shown in Figure 4(a) and the
resulting thermal stresses are shown in Figure 4(b). Only the time duration between
October 1st and March 31st are simulated, due to significantly greater potential for
thermal cracking during these months. Four critical events were identified by this
analysis; the full finite element simulation is performed on these events accordingly.
( a)
80% of Tensile Strength 2 4
1
3
(b )
Figure 4. Results from preanalyzer. (a) Winter time surface temperatures for five-year
analysis period. (b) Resulting thermal stresses where four critical events were identified.
104
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

5. Finite Element Analysis (FEA) Engine
The finite element analysis (FEA) method is used extensively in everyday design and
analysis of civil infrastructure. The biggest strengths of FEA are its ability to simulate
challenging geometries, such as pavements, and the relative ease in which it is possible
to incorporate complicated material behavior. For situations involving these
characteristics, analytical solutions would be challenging and often unfeasible. The
IlliTC system uses a finite element code that was developed in-house for simulation of
thermo-viscoelastic problems. It has the capability of simulating cracking in asphalt
concrete through use of a powerful yet computationally efficient cohesive zone fracture
model. Traditional modeling approaches have not provided a direct means for the study
of crack initiation and propagation in asphalt materials. The cohesive zone fracture
approach provides a rational means for modeling cracking in quasi-brittle materials such
as asphalt concrete, as the length scale associated with the fracturing process is
accounted for. The following subsections briefly describe the cohesive zone model and
the thermo-viscoelastic implementation in IlliTCs FEA engine. This is followed by
selected verification examples and brief description of the post-processing methodology
to extract results from the FEA.
5.1. Cohesive Zone Fracture Model
In order to correctly replicate the complex mechanisms underlying cracking in asphalt

concrete, a standard strength of materials type analysis is not sufficient, due to: 1) the
highly non-linear behavior in the vicinity of the crack tip, and 2) the importance of the
crack in the overall structural response (i.e., the need to model the crack as a moving
boundary value problem). For simulation of the crack initiation and propagation, a
cohesive zone model was selected because of its accuracy and efficiency in accounting
for material response ahead of the crack tip in the fracture process zone (region of
micro-cracking, crack pining, crack branching, material softening, etc.). Several
researchers have used this type of approach for simulation of cracking in asphalt
materials, for example Soares et al. (2004), Song et al. (2006), Dave and Buttlar (2010),
Baek et al. (2010), and Kim et al. (2010).
The cohesive zone model provides the relationship between the displacement jump
or the opening along the crack path and the total capacity of material to transfer traction
(or load) across that crack path. In the case of brittle materials, the capacity to carry load
across the crack path is either 100% (uncracked) or 0% (cracked). The threshold of stress
at which this capacity goes from 100% to 0% is the material strength and, in the case of
mode I (opening) type of cracking, it is tensile strength. For quasi-brittle materials the
capacity to transfer load across the crack path also reduces once the material has incurred
105
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

damage. However, unlike brittle materials, for quasi-brittle materials this capacity
reduces gradually. The stress threshold at which the damage begins to occur is still
tensile strength. Finally, when the crack separation (displacement jump) is greater than a
critical value, the material no longer has bearing capacity and the traction is zero. The
relationship between displacement jump and traction capacity across crack can be
defined by different geometric shapes. Various shapes have been proposed for use with
asphalt concrete, such as, bilinear (Song et al., 2006), power law (Song et al., 2008), and
exponential (Dave and Buttlar, 2010).
In this study, the bi-linear CZM described by Song et al. (2006) is being employed;
the graphical representation of the model is shown in Figure 5. Note the unloading and
reloading behavior of the model is also shown on the figure. This behavior assumes that
the damage incurred by the material is permanent and is present when it is reloaded. The
area under this traction and displacement jump curve is the fracture energy of the
material. Thus, the CZM implementation allows for fracture representation of asphalt
concrete through use of two material properties, namely, tensile strength and fracture
energy. An intrinsic cohesive zone modeling approach is used in this work; hence a
penalty stiffness (i.e. initial ascending slope) is introduced in the computational
implementation. The initial penalty stiffness is determined on the basis of the numerical
stability associated with the finite element implementation (Roesler et al., 2007). The
implementation of CZM in IlliTC is limited to mode I cracking. Since the thermally
induced stresses in conventional asphalt pavements are along the longitudinal direction
the predominant failure occurs through tensile failure in mode I. Very small quantities of
shear stresses are generated due to the thermal only loading case and thus mode II or
mixed mode cracking is not expected.
Traction (Capacity) across crack (MPa)
Tensile strength
Area = Fracture Energy
Unloading
Critical displacement
Reloading
Displacement jump across crack (mm)
Figure 5. Bi-linear cohesive zone model (CZM) used in IlliTC.
106
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

5.2. FE Implementation and Verification
The thermo-viscoelastic finite element code was implemented in the C programming

language. The code is based on the incremental-recursive formulation proposed by Yi
and Hilton (1993) and Zocher et al. (1997). The details on the implementation and
formulation have been presented by Dave et al. (2012). Due to the non-linear nature of
the CZM, the FE analysis was implemented using a modified Newton-Raphson solution
scheme. A simple adaptive time increment scheme was utilized that automatically
increases the time increment as long as the convergence error is below the threshold.
The FE implementation was verified using analytical solutions and commercial
programs. Two verification examples are presented herein. The first example involves
the simulation of a boundary value problem which resembles the thermal stress retrained
specimen test (TSRST). The simulation was conducted to include both cooling and
warming temperature boundary conditions to ensure the rigor of IlliTC (some cohesive
zone models have convergence problems upon crack closure). Figure 6 presents the
thermal stresses predicted by IlliTC against those predicted by the commercial FE
program ABAQUS. The figure also shows the temperature boundary condition that was
used in this simulation. The results show very good agreement between IlliTC and
ABAQUS results, providing verification for the thermo-viscoelastic stress prediction
capabilities of IlliTC.
Thermo-viscoelastic predictions coupled with CZ fracture simulation were also
verified, as shown in Figure 7. The verification was conducted on a solid rectangular
body as shown on the plot, where a temperature drop of 20C over duration of 150
minutes was imposed. Once again, very good agreement was observed between the
IlliTC and ABAQUS predictions. Note that there were minor deviations in the predicted
stresses, especially during the damage and cracking process (post-peak). For this
verification example the stress variations between two programs ranged between 1 and
2%. This level of discrepancy is not unexpected, due to differences in the
implementation of the CZM and use of different numerical solvers in the two FE
engines.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
107

Figure 6. Comparisons for thermo-viscoelastic stress predictions using IlliTC and

commercial finite element program ABAQUS.
Figure 7. Comparisons of stress for a thermo-viscoelastic body with CZ fracture for

predictions using IlliTC and commercial program ABAQUS.
108
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

5.3. Post Processing
The output generated by the FEA contains displacements at nodal points and stress and
strain predictions as element integration points. For the average user, the results require
post-processing to generate a simple thermal cracking versus time output. The IlliTC
screens the outputs generated by the FEA and produces a series of text files that contain
information regarding crack opening widths, which is the deformation experienced by
CZM elements in the longitudinal direction. By comparing the deformation of cohesive
zone elements against the critical displacement in the CZM, the location of the crack tip
and location of the point beyond which the asphalt has started to soften is computed. For
example, in a pavement with an asphalt concrete thickness of 150 mm, if the cohesive
zone elements in the top 25 mm of the pavement have experienced deformation in excess
of the critical deformation, the thermal crack has formed to a depth of 25 mm.
Furthermore, if the cohesive zone elements along the upper 75 mm of the asphalt layer
have undergone deformations exceeding the deformation corresponding to the tensile
strength in the local CZM (c.f. Figure 5), the upper 75 mm of asphalt concrete has
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
undergone softening damage. In the case of this example, the upper 25 mm of the
pavement is fully cracked; while the pavement between a depth of 25 and 75 mm has
undergone softening damage. In the later section on field validation an example is
provided containing more details on this topic. In addition, the field calibration section
describes how the post-processed information is used to predict the extent of thermal
cracking using a probabilistic crack distribution model.
6. Probabilistic Crack Distribution Model
To avoid the modeling complexities and computation expense needed to simulate

multiple thermal cracks, the scheme used in the original TCMODEL to translate a single
thermal crack depth prediction to thermal crack density (spacing) was adopted in the
current version of IlliTC (IlliTC v1.0). The modelling of multiple thermal cracks, while
more exact, was not deemed as being worth the added computational expense, since the
point at which crack interaction occurs is well within the range of severe cracking. Since
it is unlikely that a designer would use a high cracking level as a design target, it was
decided that multiple cracks would not be considered in this version of IlliTC. Rather,
the model completes its execution once a high level of cracking is reached (200 m of
transverse cracking per 500 m of pavement, which corresponds to a crack spacing of 10
m).
This probabilistic crack distribution model converts the computed crack depth of a
single modeled crack (viewed as a representative thermal crack, having a crack depth
109

representing an average crack depth) to an amount of thermal cracking (crack frequency)

with the following expression:
C=
f 1 PR ( log C > log hac ) [1 ]
log C / hac
C=
f 1 N [2 ]

Where:
C f = Predicted amount of thermal cracking (m/500m) at a given simulation time
1 = Multiplier representing maximum thermal cracking level
N ( x ) = Standard normal distribution evaluated at x
= Standard deviation of the log of the depth of cracks in the pavement
C = Depth of crack predicted by IlliTC at a given simulation time
hac = Thickness of asphalt layer being simulated (generally taken as the thickness
of all asphalt layers)
Crack amount (m/500m) can be converted to thermal crack spacing by dividing
predicted crack amount, C f , by lane width (typically assumed to be 4 m), and taking
the inverse of this quotient and multiplying by the unit section length (500 m). Citing
the example provided earlier in this section, a crack amount of 200 m corresponds to:
(1/(200/4))*500 or 10 meters. This corresponds to 1000/10 or 100 full-lane-width cracks
per km, which corresponds to approximately 161 thermal cracks per mile. This
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
corresponds to the maximum thermal cracking level predicted by IlliTC in the current
version. Note, similar to the approach taken in the development of TCMODEL, the
parameters 1 and were taken as model calibration parameters. Since thermal
cracks are difficult to detect until they propagate completely through the pavement, it
would be extremely difficult to directly measure and assess the parameter. Thus, its
selection as a model calibration parameter is a practical means to circumvent the need to
directly measure .
110

7. Field Calibration
The MnROAD full scale pavement test sections were used to calibrate IlliTC, namely;
sections 03, 19, 33 and 34. Details about these sections can be found elsewhere
(Marasteanu et al., 2007). A decision needed to be made with regards to the climatic
files used in model calibration, since two approaches were possible: (1) use the actual
time ranges corresponding to the field thermal cracking data for each section simulated,
or; (2) use the climatic files available in IlliTC. The argument for using the actual time
ranges that correspond with the field data is that predicted critical cooling events would
match actual critical events in the crack history data files, leading to more accurate
thermal cracking predictions for model calibration. The argument for using the climatic
files included in IlliTC is that future pavement simulations conducted using IlliTC would
be expected to utilize these climatic files (unless the user takes the effort to modify
IlliTC to utilize alternate climatic files, which is a cumbersome process in the current
version of the software). Considering that most users will likely use the climatic files
provided in IlliTC, and also considering that the model should be re-calibrated to local
conditions rather than rely on the calibration provided herein, it was decided to conduct
model calibration using the climatic files provided in the current version of IlliTC.
The subsequent subsections present the results from the preanalyzer and the finite
element analysis engine for each of the calibration sections. This is followed by a brief
description of actual calibration of parameters in the probabilistic crack distribution
model.
7.1. Preanalyzer Runs
Selected outputs from IlliTCs preanalyzer are provided in Figures 8 (a) through (e), and
summarized in Table 4. By comparing Figure 8 (a) with Figure 8 (b) through (e), it is
clear that the days with the coldest temperatures correspond to the events with the
highest surface tensile stress. Table 4 shows that one critical cooling event was
computed for MnROAD section 03 during the simulated 5-year analysis period, while
four, one, and zero critical cooling events were predicted for sections 19, 33, and 34
respectively. Comparing the number of computed critical cooling events with field
cracking behavior indicates the correlation between mixture viscoelastic behavior (as
captured by the creep compliance master curves) and cracking behavior. The correlation
between mix creep compliance and fracture behavior was also demonstrated via a
statistical analysis in the previous phase of this study (Marasteanu et al., 2007).
111
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(a) Pavement surface temperature using default climatic files in IlliTC for MnROAD site
(in the Category of Moderate Climate within the State of Minnesota).
(b) Thermal stress on pavement surface for MnROAD03 from preanalyzer (Bold
horizontal line indicates 80% of tensile strength).
(c) Thermal stress on pavement surface for MnROAD19 from preanalyzer (Bold
(d) Thermal stress on pavement surface for MnROAD33 from preanalyzer (Bold
112

(e) Thermal stress on pavement surface for MnROAD34 from preanalyzer (Bold
Figure 8. Results from preanalyzer for calibration field sections.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Table 4. Preanalyzer results (number of critical events) compared to field cracking.
Number of Critical
MnROAD Field Cracking
Events (as predicted Binder Grade
Cell (m/500 m)
by Preanalyzer)
PEN 120/150
03 1 182
(PG58-28)
19 4 AC20 (PG64-22) 547
33 1 PG 58-28 91
34 0 PG 58-34 6
7.2. Finite Element Runs
Sample finite element modeling results from the calibration phase of the study for
MnROAD section 19 is provided in Figure 9 (a) through (d). The various aspects of the
cohesive zone based finite element modeling approach can be seen in these stress and
(exaggerated) deformed structure plots, which show the elevation view of the asphalt
layers in the vicinity of the modeled crack. The progression of stress build-up, crack
initiation, and crack propagation can be tracked as follows:
Figure 9(a) shows high surface tensile stress (as indicated by the dark color
contours), and a slight disruption in the contours at the crack interface caused
113

by the early stages of damage (post-peak softening when tensile stress exceeded
material strength at the surface of the pavement) at -23.3C surface
temperature. Mild compression is still present in the lower regions of the
pavement, due to time-lag effects of heat flow.
Figure 9(b) shows that a thermal crack has propagated partially downward
through the pavement at a temperature of -24.3C, and that a fracture process
zone of about 15% of the pavement thickness exists ahead of the current crack
tip location, illustrating one of the features of the cohesive zone modelling
approach (length scale of fracture is directly considered). A compression zone
still exists near the bottom of the asphalt layer.
Figure 9(c) shows a later stage of crack propagation, where the fracture process
zone has grown in size, and demonstrates that a compression zone no longer
exists (which may partially explain the expansion of the fracture process zone)
at a temperature of -25.4C.
Figure 9(d) shows a fully formed crack, occurring around -29.6C. In reality,
IlliTC considers the pavement section as fully cracked prior to this analysis
step, as described below.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
114

Thermal Stress in Longitudinal

Thermal Stress in Longitudinal
Direction near the Crack Path (MPa)
Direction near the Crack Path (MPa)
(a) Thermal stress built-up along (b) Partial depth softening damage
longitudinal direction (Surface temperature = -24.3C)
(Surface temperature = -23.3C)
Thermal Stress in Longitudinal Thermal Stress in Longitudinal

Direction near the Crack Path (MPa) Direction near the Crack Path (MPa)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(c) Partial depth crack (d) Fully formed crack

(Surface temperature = -25.4C) (Surface temperature = -29.6C)
Figure 9. Results from finite element analysis for MnROAD section 19 (calibration field
section). (Tensile strength = 4.22 MPa)
7.3. Model Calibration Discussion
Model calibration in pavement studies acknowledges the significant complexities

associated with pavement materials, construction, climatic effects, traffic loading and
115

performance. Pavement performance model calibration is almost always needed as a

result. In the case of thermal cracking, factors such as construction variability, inability
to model aging and aging gradients with accuracy, approximate nature of tests and
material models, approximate nature of climatic records and pavement temperature
predictions, presence of load-associated effects and damage on pavement, etc., exist and
result in the need for model calibration.
A number of factors were available to be used for model calibration, including:
fracture energy multiplier, tensile strength multiplier, thermal coefficient multiplier,
crack tip definition, and beta ( 1 ) and sigma ( ) parameters from the probabilistic
crack distribution model were readily available for use in model calibration. As a
preliminary approach, it was decided to leave the material property factors as
uncalibrated, and focus on the following three factors for model calibration: crack tip
definition, beta ( 1 ) parameter, sigma ( ) parameter. Crack tip definition refers to the
fact that more than one material state can be considered as the point of crack initiation in
the cohesive zone modeling technique. For instance, in Figure 5, any point along the
post-peak softening curve (declining linear function in the case of the bi-linear cohesive
zone model, which represents the gradual accumulation of material damage and loss of
load carrying capacity across the forming crack as the material separates) could be
selected as the arbitrarily chosen location of the crack tip. The point at where the
softening curve reaches zero traction (the right hand limit of the plot shown in Figure 5)
is arguably the point where the material no longer possesses the ability to heal.
However, it can also be argued that an intermediate point along the softening curve may
be a realistic choice for the crack tip. After examining the finite element results from the
MnROAD calibration finite element runs, it was decided that the crack tip would be
defined as the point in the pavement along the line of cohesive zone elements where a
softening threshold of 75% post-post peak decay of material strength (25% traction
remaining) is reached. In addition, an identical beta factor ( 1 = 400 m of cracking per
500 m of pavement section) as used in the original TCMODEL as calibrated in the
MEPDG would be used. Finally, the sigma parameter in the probabilistic crack
distribution model was calibrated to a value of = 1.1.
The results of the calibrated IlliTC model, using MnROAD pavement sections, are
presented in Table 5. As can be seen, reasonable modeling predictions were achieved
for three out of the four sections evaluated. For instance, MnROAD section 19, which
experienced very high pavement cracking due to the use of an AC-20 binder (PG 64-22)
in a PG XX-34 climate, was predicted to have a maximum level of cracking. Recall that
when the maximum predicted crack depth is reached (crack depth = thickness of
pavement), this implies that the average crack is equal to the pavement thickness. Thus,
half of other pavement cracks will be less than the thickness of the pavement, and
therefore, not yet counted as thermal cracks. Stated otherwise, the probabilistic crack
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
116

distribution model has a maximum cracking level of 200 m of cracking per 500 m
section, when 1 = 400. MnROAD section 33 was found to have a cracking level of 94
m of cracking, as compared to a measured level of 91 m (this was the section that drove
the calibration of = 1.1), and MnROAD section 34 was found to have a cracking
amount of 0 m as compared to a measured cracking level of 6 m. The only poor
prediction that resulted was for MnROAD section 03, where zero cracking was predicted
as compared to 182 m of measured cracking. It should be noted, however, that IlliTC
did indicate that softening damage had begun to occur in this section (although not
enough to reach the 75% softened threshold). It is also acknowledged that the time
period for the IlliTC simulation was shorter than the period of field performance
reported for Section 03. In addition, the master curve data used in the calibration was
less-than-optimal, with data from only two test temperatures being available (three is
preferred). Rather than add additional calibration factors to IlliTC, it was decided that
the aforementioned calibration parameters were sufficient for the calibration of IlliTC.
However, it is recommended that IlliTC be recalibrated to local conditions to arrive at
better model accuracy. Model validation using an independent data set is provided in the
next section of the paper.
Table 5. IlliTC model calibration results.

MnROAD Measured Field Predicted Field
Binder Grade
Cell Cracking (m/500 m) Crackinga (m/500 m)
PEN 120/150
03 182 0
(PG 58-28)
AC-20 > 200 (max. allowable
19 > 200 (547)
(PG 64-22) cracking)
33 PG 58-28 91 94
34 PG 58-34 6 0
a
Predictions are made using non-synchronized climate files
8. Field Validation
Five pavement sections were constructed in Olmsted County, Minnesota during the 2006
construction season, which were used for the analysis portion of the validation process.
The mixes were sampled during the construction process and were characterized
extensively in the lab. Detailed information about the pavement sections and mixture
properties obtained experimentally are documented elsewhere (Marasteanu et al., 2012).
117
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The five Olmstead Co. pavement sections were simulated using IlliTC, and the results
from the preanalyzer and finite element analysis are briefly presented in the following
section.
Table 6 presents the results from the IlliTC preanalyzer routine. As can be seen,
only Section 4 experienced a critical tensile stress level in the five years analyzed. The
preanalyzer thermal stresses for this section are presented in Figure 10. Since this was
the worst section in terms of field cracking, this indicates that the IlliTC program has
correctly ranked the five field sections.
Table 6. IlliTC preanalyzer results and field cracking for the validation sections.
Number of
Field Performance
Critical Events
Validation Section (Transverse
(Predicted by
cracking m/500m)
Preanalyzer)
1 0 23 (low)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
2 0 2 (very low)
3 0 29 (low)
4 1 53 (moderate)
5 0 25 (low)
Figure 10. Preanalyzer results for validation section 4 (Bold horizontal line indicates
80% of tensile strength).
Based on the preanalyzer results, the validation section 4 was selected for detailed
presentation herein. The thermal stresses and damage predicted by IlliTC for Section 4 is
presented in Figure 11. These results show that while zero cracking was predicted,
softening was activated along the cohesive zone fracture elements. As discussed
previously, the IlliTC system uses the probabilistic crack distribution model to predict
field cracking from the finite element analysis. Based on the preanalyzer and finite
118

element analysis the IlliTC field cracking predictions for validation sections are
presented in Table 7.
Softening near the

top of pavement
Thermal Stress in Longitudinal Direction near the Crack Path

(MPa), Tensile strength = 4.37 MPa
Figure 11. Thermal stresses at the end of the critical event for validation section 4.
Table 7. IlliTC predictions and field cracking for the validation sections usaing
calibrations discussed in previous section.
Field Performance
Validation Predicted Crack Predicted Field
(Transverse
Section Depth (mm) Cracking (m/500 m)
cracking m/500m)
1 0 0 23
2 0 0 2
3 0 0 29
4 0 0a 53
5 0 0 25
a
softening was predicted, indicating that thermal cracking would likely result if a longer
analysis period was used.
As evident from Table 7 under the current calibration parameters established in the
previous section, zero cracking was predicted for all sections. Given the fact that most
of the sections have experienced low cracking to date, it can be concluded that IlliTC
under its current calibration is slightly under-predicting the cracking behavior for these
sections. It should also be noted that a limited amount of creep compliance data were
available for these sections (testing at two temperatures instead of the preferred three), so
119
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

errors caused by incomplete compliance data could also have contributed to the under
prediction observed. Given the fact that one of the five sections in the calibration data
set was also under-predicted, the validation trials here may suggest that IlliTC should be
recalibrated to produce higher levels of cracking. However, given the limited validation
data available and since local calibration is recommended before implementing IlliTC in
a given region, further calibration of IlliTC using the current field data was not pursued
herein.
9. Summary, Conclusions and Future Research Tasks
A new thermal cracking model called IlliTC was developed as part of a recently
completed pooled fund study on low temperature cracking. Various components of the
IlliTC model and their verification along with model calibration and preliminary
validation were presented, including: a user-friendly graphical interface called Visual
LTC; a finite element modeling engine involving viscoelastic bulk material and cohesive
zone fracture elements, and; a probabilistic crack distribution model, identical to the one
used in the original TCMODEL program. Model calibration strategies, including a
discussion regarding crack tip definition in the cohesive zone modeling scheme was also
presented.
The model calibration was conducted using four MnROAD sections, three of which
(19, 33, and 34) were found to have very good model predictions after calibration. The
only unreasonable prediction that resulted was associated with MnROAD section 03,
where zero cracking was predicted as compared to 182 m of measured cracking. It
should be noted, however, that IlliTC did in fact indicate that softening damage had
begun to occur in this section (although not enough to reach the 75% softened
threshold). The time period for the IlliTC simulation in this case was shorter than the
period of field performance reported for Section 03. In addition, the master curve data
used in the calibration was less-than-optimal, with data from only two test temperatures
being available (three is preferred). Very limited model validation data was available,
with only limited cracking observed to date in the four validation sections studied
(Olmstead Co., Minnesota). IlliTC predictions were in general agreement with the
observed field cracking (zero cracking predicted vs. very low to low cracking observed
in most sections). The calibration and validation sections indicated that IlliTC should
probably be recalibrated to produce slightly larger cracking predictions; however, it is
recommended that such calibration be performed by highway agencies, designers, or
researchers using local material properties and local field performance data.
120
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Over the course of this study several future research tasks were identified that would
further enhance the capabilities and applications of the IlliTC system. Some of the future
implementation tasks for IlliTC are as follows:
The current model assumes a constant coefficient of thermal expansion and
contraction; the future version can include the bilinear thermo-volumetric
trend observed by researchers from University of Wisconsin at Madison.
The effects of oxidative aging are not included in the current simulations;
the aging will have very pronounced effect on the thermal cracking
performance due to stiffening and embrittlement of asphalt mixtures near
the pavement surface. A material aging model, such as one used in
AASHTO MEPDG, can be utilized in the IlliTC system to include the effect
of aging. Furthermore, at low temperatures asphalt binders may also exhibit
significant non-oxidative stiffening (steric hardening), the effect of steric
hardening on cracking performance should be evaluated.
The IlliTC system does not account for multiple asphalt mix types, the
extension to multiple material types requires minor modifications to the
software without need for new development in finite element analysis code.
It is anticipated that the future version will have this capability.
The current IlliTC utilizes single fracture energy input at one temperature;
the system should be modified to allow users to input fracture properties at
multiple temperatures (or in functional form).
The current program is currently calibrated using global fracture energy for
finite element simulations; however, researchers have shown that use of
local fracture energy improves the prediction accuracy. Local fracture
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
energy can be determined using the raw data from fracture tests and the
viscoelastic properties of the mix, along with inverse analysis by modeling
the test specimen. This capability should be added to a subsequent version
of IlliTC.
An algorithm recently developed to extract mixture tensile strength fro m
DC(T) test data should be added to IlliTC, so that users can avoid the need
to directly input mixture tensile strength data.
The effects of variability in material property inputs on the predicted
thermal cracking performance should be evaluated.
A much more comprehensive calibration and validation of IlliTC is needed;
the current calibration was limited to four pavement sections at one location.
121

With more extensive calibration, the prediction capability of IlliTC should

be improved, which can be evaluated using additional validation sections.
10. Acknowledgement
The authors are grateful for the support provided by the sponsors and partners of Federal
Highway Administration (FHWA) Pooled Fund Study TPF-5(132). Any opinions
expressed herein are those of the writers and do not necessarily reflect the views of the
sponsors.
11. References
Apeagyei, A.K., Dave, E.V., Buttlar, W.G., Effect of Cooling Rate on Thermal
Cracking of Asphalt Concrete Pavements, Journal of Association of the Asphalt
Paving Technologists, Vol. 77, p. 709738, 2008.
Baek, J., Ozer, H., Wang, H., Al-Qadi, I.L., Effects of Interface Conditions on
Reflective Cracking Development in Hot-Mix Asphalt Overlays, Road Materials
and Pavement Design, Vol. 11, No. 2, p. 307335, 2010.
Brown, E.R, Kandhal, P.S., Roberts, F.L., Kim, Y.R., Lee, D.-Y., Kennedy, T.W., Hot
Mix Asphalt Materials, Mixture Design and Construction, Lanham, MD, National
Asphalt Pavement Association, 2009.
Buttlar, W.G., Roque, R., Reid, B., Automated Procedure for Generation of Creep
Compliance Master Curve for Asphalt Mixtures,, Transportation Research Record,
No. 1630, p. 2836, 1995.
Buttlar, W.G., Sahu, R., Behnia B., Dave, E.V., Determination of Asphalt Concrete
Tensile Strength from the Disk-shaped Compact Tension (DCT) Fracture Test, To
be submitted for journal publication, 2012.
Dave, E.V., Buttlar, W.G., Low Temperature Cracking Prediction with Consideration
of Temperature Dependent Bulk and Fracture Properties, Road Materials and
Pavement Design, Vol. 11-SI, p. 3359, 2010.
Dave, E.V., Leon, S.E., Park, K., Thermal Cracking Prediction Model and Software for
Asphalt Pavements, Proceedings of the First T&DI Congress 2011, ASCE
Conference Proceedings, Vol. 398(41167), p. 6472, 2011.
Dave, E.V., Paulino, G.H., Buttlar, W.G., Viscoelastic Functionally Graded Finite
Element Method with Recursive Time Integration and Applications to Flexible
Pavements, Int. J. Numer. Anal. Meth. Geomech., Vol 36, No. 9, p. 11941219,
2012.
122
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Deme, I.J., Young, F., Ste. Anne Test Road Revisited 20 Years Later, Proc. of
Canadian Technical Asphalt Association, p. 254283, 1987.
Islam, M.S., Buttlar, W.G., "Effect of Pavement Roughness on User Costs,"
Transportation Research Record, In Press, 2012.
Kim, Y.-R., Aragao, F.T.S., Allen, D.H., Little, D.N., Damage Modeling of Bituminous
Mixtures Considering Mixture Microstructure, Viscoelasticity, and Cohesive Zone
Fracture,, Canadian Journal of Civil Engineering, Vol. 37, No. 8, p. 11251136,
2010.
Lytton, R.L., Roque, R., Uzan, J., Hiltunen, D.R., Fernando, E., Stoffels, S.M.,
Performance Models and Validation of Test Results, Final Report to Strategic
Highway Research Program; Asphalt Project A-005, SHRP Report A-357, 1993.
Marasteanu, M., Zofka, A., Turos, M., Li, X., Velasquez, R., Li, X., Williams, C.,
Bausano, J., Buttlar, W., Paulino, G., Braham, A., Dave, E., Ojo, J., Bahia, H.,
Gallistel, A., and McGraw, J., Investigation of Low Temperature Cracking in Asphalt
Pavements, Report No. 776, Minnesota Department of Transportation, Research
Services MS 330, St. Paul, MN 55155, 2007.
Marasteanu, M, Moon, K.H., Teshale, E.Z., Falchetto, A.C., Turos, M., Buttlar, W.,
Dave, E., Paulino, G., Ahmed, S., Leon, S., Bahia, H., Arshadi, A., Tabatabaee, H.,
Ojo, J., Velasquez, R., Mangiafico, S., Williams, C., Buss, A., Investigation of Low
Temperature Cracking in Asphalt Pavements National Pooled Fund Study -Phase II,
Report No. MN/RC 2012-23, Minnesota Department of Transportation, St. Paul MN,
2012.
Nam, K., and Bahia, H.U., "Effect Of Binder and Mixture Variables on Glass Transition
Behavior of Asphalt Mixtures," Journal of Association of Asphalt Paving
Technologists, Vol. 73, p. 89119, 2004.
Nam, K., and Bahia, H. U., "Effect of Modification on Fracture Failure and Thermal-
Volumetric Properties of Asphalt Binders," J.Mater.Civ.Eng., Vol. 21, No. 5, p. 198-
209, 2009.
Paris, P.C., Gomez, M.P., Anderson, W.E., A Rational Analytic Theory of Fatigue,
The Trend in Engineering, Vol. 13, p. 914, 1961.
Roseler, J.R., Paulino, G.H., Park, K., Gaedicke,C., Concrete Fracture Prediction Using
Bilinear Softening, Cement and Concrete Composites, Vol. 29, No. 4, p. 300-312,
2007.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Roque, R., Hiltunen, D.R., Buttlar, W.G., Thermal Cracking Performance and Design
of Mixtures Using Superpave(TM), Journal of the Association of Asphalt Paving
Technologists, Vol. 64, p. 718735, 1995a.
Roque, R., Hiltunen, D.R., Buttlar, W.G., Farwana, T., Engineering Properties of
Asphalt Mixtures and the Relationship to their Performance, STP1265, Engineering
123

Properties of Asphalt Mixtures and the Relationship to their Performance, ASTM

International, 1995b.
Soares, J.B., Colares de Freitas, F.A., Allen, D.H., Crack Modeling of Asphaltic
Mixtures Considering Heterogeneity of the Material, Transportation Research
Record, No. 1832, p. 113120, 2004.
Song, S.H., Paulino, G.H., Buttlar, W.G., "A bilinear cohesive zone model tailored for
fracture of asphalt concrete considering viscoelastic bulk material," Eng.Fract.Mech.,
Vol. 73, No. 18, p. 28292848, 2006.
Song, S.H., Wagoner, M.P., Paulino, G.H., Buttlar, W.G., "25 Crack Opening
Displacement Parameter in Cohesive Zone Models: Experiments and Simulations in
Asphalt Concrete," Fatigue and Fracture of Engineering Material and Structures,
Vol. 31, No. 10, p. 850856, 2008.
Yi, S., Hilton, H.H., Dynamic Finite Element Analysis of Viscoelastic Composite
Plates in the Time Domain, International Journal of Numerical Methods in
Engineering, Vol. 37, No. 12, p. 40814096, 1994.
Zocher, M.A., Groves, S.E., Allen, D.H., A Three-Dimensional Finite Element
Formulation for Thermoviscoelastic Orthotropic Media,, International Journal of
Numerical Methods in Engineering, Vol. 40, No. 12, p. 22672288, 1997.
12. Discussion
MR. MOHAMMAD MOLAYEM Prepared Discussion: Thanks Dr. Dave for your
good research and presentation. My question is about the reason for choosing cohesive
zone concepts in your model. Cohesive zones are of course simple meshless
computational frameworks, but the drawback is that the crack path should be known in
advance which is not the case for pavement surfaces. So besides the cohesive zone
methodology, why didn't you incorporate other alternatives such as GFEM/XFEM,
viscoelastic fracture mechanics or even linear elastic fracture mechanics, because as you
are modeling the low temperature cracking, the behavior of asphalt concrete tends to be
brittle irrespective of the nature of the material itself and what happens around the crack
tip at micro-scale level.
PROFESSOR ESHAN DAVE: The first and foremost reason was practicality, to be able
to produce software that runs at relatively faster speeds. When we started working on
this, the extended finite elements had just come into the market and were gaining
popularity. You are correct in saying that has a lot of potential and definitely maybe
someday in the future we would go to that. Some of the other methods that you
mentioned (such as, Viscoelastic fracture mechanics or linear elastic fracture mechanics)
are quite highly dependent on the mesh sizes because you are looking at the strain
124
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

gradients. In that case, then the challenge is that you have to run a mesh dependency
analysis before you simulate your problem. So that was the reason to go with the
cohesive zone model here with traditional finite elements.
PROF. MOLAYEM: Thank you very much.
PROFESSOR REYNALDO ROQUE: Eshan, I think you definitely attack the right
deficiencies, one of them being, of course, the failure limits that were considered in the
original model. My question is whether there any evidence specifically and maybe it
relates to the previous question that this cohesive zone model is applicable to asphalt
mixtures? Is there any real evidence that this is actual behavior or is it just something
that we have taken and used as another way to interpret data without really knowing
whether it is applicable? Frankly, at low temperatures, I dont see that it is applicable.
PROF. DAVE: Thats a very good question. What we have done, and this is was done
probably about seven or eight years ago, we have shown at least at the lab scale where
with the cohesive zone model you can capture or you can at least match a lab behavior
really well. More recently what we have shown is that if you simulate a test using the
cohesive zone model and get materials properties from that, we are able to simulate a
different test using those properties at different loading rates and different temperatures.
So that is one aspect. The other aspect is through a National Science Foundation GOALI
Study, we were able to do something where we took lab measured responses and were
able to predict reflective cracking in asphalt overlays. So that was a slightly different
application, not necessarily a thermal cracking but closely associated one. In that case
also, especially for accelerated pavement test sections, we were able to match the field
behavior pretty well. So we feel pretty confident in using this model.
PROF. ROQUE: The basic mechanism of cracking is what I am questioning. One can
match data and you can do that with most any model but that does not mean that the
mechanism being modeled is representative. In other words, is the reduction in stiffness
or damage really associated because a part of that crack is hanging on or because you are
getting distributed microdamage in a zone around the crack? You are interpreting it one
way, but it may not be that way.
PROF. DAVE: Thats a good question. Currently, there is a study going on at the
University of Illinois where they are trying to use digital image correlation technique
(DIC), again to try to learn more about cracking. I think we need to go to advanced
mechanisms and Prof. Gabriele Tebaldi is here in audience. He is leading a task group
for the RILEM Technical Committee on Mechanisms of Cracking and Debonding in
Asphalt and Composite Pavements (TC-MCD), which is planning an integrated study to
better understand the microscopic damage in asphalt material during cracking. Maybe
computer tomography will tell us something more or DIC will tell us something more to
learn more about what is really happening in the physical sense. Right now those are all
based on what you describe, essentially trying to see whether we can match responses.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
125

As far as knowing what is really happening to the material in physical sense, I think
work is still underway.
PROF. ROQUE: Okay. Thank you.
DR. M. EMIN KUTAY: Very nice presentation, Eshan. I have a question about your
preprocessing stage. Right before the final analysis, you do one-dimensional analysis
and you're choosing the critical values. Does that mean that you do not accumulate
damage during those small events (i.e., variations of temperature/thermal stresses)? So
you're only looking at the critical event?
PROF. DAVE: We are only looking at the critical events, and that is why I mentioned
the capability to have thermal fatigue in future extensions, if such a thing exists, where
we are continuously accumulating damage. The objective for this work was: Can we run
the model in 10 minutes and be done with it to make it more practitioner friendly? So
right now we are skipping over time and only simulating critical events. Although, lets
say you have five critical events and you got a certain amount of damage or cracking on
event one, then we go through an equilibrium phase, so it would remember how much
damage you had and then go to the next event. But theres no healing built into this
model.
DR. KUTAY: In your opinion, would the strength that you measure at the undamaged
state be same as the strength of the mixture after several cyclic thermal stresses are
applied?
PROF. DAVE: I think strength would be greater. Thats why I said in future, we need to
add that type of cyclic or fatigue related effects.
DR. KUTAY: Okay, thanks.
126
--`,`,,,``,,```,,`,,,,

Effect of Short-Term Conditioning and Long-
Term Aging on Permanent Deformation
Characteristics of Asphalt Mixtures
Haleh Azaria* and Alaeddin Mohsenib
a
AASHTO Advanced Pavement Research Laboratory at National Institute of Standards
and Technology, Gaithersburg, MD 20899-8615
b
Pavement Systems Inc., Bethesda, MD
ABSTRACT: Quantification of the effect of short-term conditioning and long-term aging on the
permanent deformation properties of asphalt mixtures is important for a reasonable estimation of
rutting in the field. As the asphalt industry continues to develop more sophisticated mixtures,
measuring the effect of short-term conditioning and long-term aging on performance becomes
even more critical. Significantly more pavements are being built with Reclaimed Asphalt
Pavement (RAP), Shingles (RAS), and warm mix asphalts, which may age very differently from
mixtures with virgin asphalt. The use of warm mix asphalt in particular is rapidly increasing,
which tends to soften the mixture in the early life of the pavement and impose a complex challenge
for determining appropriate laboratory and field conditioning. For these reasons, it is critical that
more precise determination of the effect of aging on permanent deformation properties of asphalt
mixtures be made. This study proposes a new methodology that could be used in conjunction with
the design of conventional asphalt mixtures and any new asphalt technology. Using the
Incremental Repeated Load Permanent Deformation (iRLPD) test, which is the basis of the new
The oral presentation was made by Dr. Azari.
http://dx.doi.org/ 10.1080/14680629.2013.812833
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
127

AZARI, MOHSENI
methodology, the effect of short-term conditioning and long-term aging on the permanent
deformation properties of asphalt specimens can be evaluated. The minimum strain rate (MSR)
property from the iRLPD test, measured before and after the aging process, will show the rate of
aging and rate of change in the mixtures rutting performance. In this study, six mixtures were
tested according to the iRLPD procedure and the aging effect on permanent deformation
properties of the mixtures was quantified. It is shown that different asphalt mixtures age in very
different ways and this will significantly affect their short-term and long-term rutting
performances. In addition, it was found that there exists a significant interdependence between
short-term conditioning and long-term aging of asphalt mixtures. Thus, adequate short-term
conditioning is important for acceptable long-term rutting performance. The MSR master curve
from the iRLPD test provided the necessary information for estimating rutting, considering the
effect of temperature, tire pressure, and aging. In this respect, rut depths of the six mixtures were
estimated using the hourly pavement temperatures for a 20-year period.
Keywords: asphalt mixtures, rutting, permanent deformation, repeated load, strain rate, short-
term conditioning, long-term aging
1. Introduction
The permanent deformation of an asphalt mixture is a major distress that could result in
unsafe driving conditions such as hydroplaning, accelerating moisture damage, and in
some cases steering problems. The selection of an appropriate mixture with respect to
the climatic conditions and traffic levels of the intended location could significantly
reduce the damage due to rutting. In this regard, a laboratory permanent deformation test
should be able to accurately evaluate a mixture design and provide properties for reliable
prediction of rut depth in the field, by incorporating the main factors affecting permanent
deformation. The stiffening of asphalt due to short-term conditioning and long-term
aging is one of the important elements that should be considered in laboratory testing for
reliable selection of an appropriate mixture and accurate prediction of field rutting.
The most commonly used laboratory tests for measuring resistance of asphalt
mixtures to permanent deformation are the Flow Number (FN) test, the dynamic
modulus test, and the wheel-track test. While the effect of short-term conditioning is
incorporated in results of these tests, the effect of long-term aging is not reflected in the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
tests properties. In this study, a new test procedure and analysis method referred to as
Incremental Repeated Load Permanent Deformation Test (iRLPD) (Azari and Mohseni
2012, Azari and Mohseni 2013) is used for determining the resistance of un-aged and
aged mixtures to permanent deformation. Unlike the existing high-temperature
laboratory tests on asphalt mixtures, which are either stiffness-based or failure-based, the
128

Effect of Short-Term Conditioning and Long-Term Aging
iRLPD test is damaged-based, meaning that the level and duration of the load is selected
to cause incremental micro-damage to the material without causing it to fail. This would
resemble the damage experienced from load applications in the field without masking
the sensitivity of the test to damage caused by other factors being studied such as aging.
During the iRLPD test, asphalt mixture specimens are subjected to multiple stresses
in several increments. The property of the test is the strain per cycle at the end of each
test increment, also known as Minimum Strain Rate (MSR), which represents a unit
damage in the field from the passing of a single axle. A sweep of MSR values from
different stress levels is used to develop the MSR master curve. The master curve
provides high temperature performance of an asphalt mixture at any temperature and
stress condition and can be used to predict rut depth in the field. To evaluate the effect
of aging on rut depth prediction, the change in the mixture property with short-term
conditioning and long-term aging are quantified by measuring MSR after various short-
term conditioning and long-term aging durations. The iRLPD test is especially suitable
for this evaluation since the test increments are applied on the same test specimens
before and after each aging duration.
2. Effects of Short-Term Conditioning and Long-Term Aging on Rutting
The effect of short-term conditioning and long-term aging on performance of asphalt

mixtures is acknowledged in AASHTO R 30, Standard Practice for Mixture
Conditioning of Asphalt Mixtures (AASHTO, 2012). For performance testing, the
standard prescribes short-term conditioning of the uncompacted mixtures at 135C for 4
hours and long-term aging of compacted mixtures at 85C for 5 days. Although the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
importance of the conditioning /aging on stiffening of asphalt mixtures is a well known
fact, as addressed in AASHTO R 30, the effect of aging is seldom measured and
incorporated into the mixture design and performance prediction. Several researchers
have been aimed to quantify the amount of laboratory long-term aging; however, the
results have not been very successful. This is partly due to the nature of the performance
tests used, which could not capture the effect of aging. During NCHRP Project 9-23, the
dynamic modulus test was used to measure the stiffening of mixtures due to long-term
aging. However, due to high sensitivity of the dynamic modulus to the slump that
typically occurs during long-term aging, a clear increasing trend in stiffness of the
mixtures in the study was not found (Houston et al., 2005).
In another study, Azari (2011) examined the effect of short-term conditioning and
long-term aging durations on FN test results. Although the trend of increase in FN from
the increase in aging duration was observed, the trend was not significant enough to
conclude an appropriate aging duration. The reason was that due to the destructive nature
129

AZARI, MOHSENI
of the FN test, separate sets of specimens were used to determine the properties of the
un-aged and aged specimens and, as a result, the effect of aging was masked by the
sample-to-sample variability.
3. Objectives
The main objective of this study is to incorporate the effect of long-term aging on
pavement rut depth prediction using iRLPD test results. The second objective is to
examine the effect of short-term conditioning duration on long-term performance of
asphalt mixtures. The third objective is to rank asphalt mixtures with respect to rutting
resistance using the aged MSR master curves. These objectives were achieved by short-
term conditioning of six mixtures for various durations and measuring the MSR of the
compacted specimens, before and after various long-term aging durations.
4. iRLPD Testing and Analysis Methodology
4.1 iRLPD Testing Protocol
The iRLPD procedure, developed by the AASHTO Advanced Pavement Research

Laboratory (AAPRL), was used for measuring permanent deformation resistance of the
mixtures (Azari and Mohseni 2012, 2013). The test follows the FN test protocol,
AASHTO TP 79, (AASHTO, 2012) in terms of test apparatus, specimen fabrication,
load pulse requirements, and data quality criteria; however, the iRLPD test offers
variations in the number of load applications, test properties, and method of analysis.
Figure 1 shows the Asphalt Mixture Performance Tester (AMPT) that is used for
performing both the FN and iRLPD tests. For a typical FN test, the test continues for
10,000 cycles, until flow is reached (at minimum strain rate), or when permanent strain
exceeds 50,000 microstrain. Figure 2 provides a typical output of an FN test, where the
test is performed at a single combination of temperature and stress (either confined or
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
unconfined). The FN test parameter is the number of cycles at the test termination. On
the other hand, the iRLPD test includes several test increments since the test is stopped
after 500 cycles and restarted at either the same stress level or at higher stress levels,
depending on the purpose of the test. Figure 3 shows the output of an iRLPD test, where
three replicate specimens are each tested at a confining stress of 69 kPa and four
different deviatoric stress levels of 200 kPa, 400 kPa, 600 kPa, and 800 kPa in four 500-
cycle test increments. For the iRLPD test, temperature is fixed at 60C. The property of
the test is the minimum strain rate (MSR) at the end of each test increment. Therefore,
130

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
from testing each specimen, several MSR values could be calculated, as pointed out in
Figure 3.
Figure 1. iRLPD test setup using AMPT loading machine.
Figure 2. A typical output of an FN test; graphs show permanent strain and permanent
strain rate versus number of cycles; the three stages of a repeated load test (primary,
secondary, and tertiary) are also shown in the figure.
131

AZARI, MOHSENI
MSR, 4th increment
MSR, 3rd increment

MSR, 2nd increment
Figure 3. A snapshot of the the iRLPD test output; figure shows the permanent strain
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
and permanent strain rate of the four 500-cycle test increments; Figure also shows
minimum permanant strain rate (MSR) at the end of each test increment.
4.2 iRLPD Analysis Method

The analysis of the iRLPD test results involves developing the MSR master curve. The
MSR values from various stresses are plotted versus a parameter TP, which is the
product of temperature (C) and deviatoric stress (MPa). The MSR versus TP is referred
to as the MSR master curve and explains the rutting damage characteristics of a mixture
at any stress and temperature combination. A power model in the form of MSR=a*(TP)b
is fit to the MSR master curve. The coefficient a of the power curve is fixed at 0.001.
Through a series of statistical analyses of the iRLPD data, it was determined that fixing
the a coefficient at 0.001 would result in a better fit to the upper portion of the master
curve, where MSR values are the most important, without adversely affecting the fit to
the lower portion of the curve. Using this approach, each master curve is reduced to a
single number (power coefficient b), which makes it convenient to use the master
curve for mixture ranking and field implementation. Figure 4 shows a plot of MSR
132

versus TP (MSR master curve) for a mixture from Wisconsin (WI-E10). The MSR
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
values were measured at a single temperature of 60C and three stress levels of 400 kPa,
600 kPa, and 800 kPa. The TP values of 24, 36 and 48C-MPa are calculated by
multiplying the test temperatures in C and stress levels in MPa for each test increment
(Azari and Mohseni 2012).
4.3. Logic behind MSR Master Curve
In the initial work by the authors (Azari and Mohseni 2012), the iRLPD test was
conducted at multiple temperatures as well as multiple stresses. From the results of the
tests it was shown that the effect of temperature and stress are interchangeable; meaning
that by increasing the stress, the same effect on MSR is captured as by increasing the
temperature. In this respect, the TP variable was defined to explain both effects of
temperature and stress on MSR. The use of TP made possible performing the test at only
one temperature but multiple stresses to obtain the same master curve as that obtained
from conducting the test at multiple temperatures and stresses and then shifting the MSR
versus stress curves from different temperatures into a single MSR master curve.
MSR Master Curve- WI (E10)

5 0 .0
4 5 .0
MSR (mic r ostr ain pe r c yc le )
4 0 .0
3 5 .0
3 0 .0 y = 0.001x2.5839 Inc r e me ntal
2 5 .0
Powe r Mode l
2 0 .0
1 5 .0
1 0 .0
5 .0
0 .0
0 .0 1 0 .0 2 0 .0 3 0 .0 4 0 .0 5 0 .0 6 0 .0 7 0 .0
TP (C * MPa)
Figure 4. MSR master curve of the WI-E10 mixture (4 hour short-term conditioned,
without long-term aging).
133

AZARI, MOHSENI
4.4. Advantages of iRLPD Testing and Analysis Methodology
The iRLPD is a unique approach compared to previous methods (Kenis, 1978) in the
sense that multiple stresses are performed on the same specimens, which would remove
sample-to-sample variability. This is essential in capturing the effect of factors such as
aging, which could be masked otherwise. In addition, the definition of MSR at the 500th
cycle, as opposed to the slope of the secondary stage, which was used by previous
researchers (Goh et al., 2011; Von Quintus et al., 2012) is a new definition. The slope of
the secondary stage is not a constant value and is changing with the increase in cycle
number while using the slope at the 500th cycle (MSR) sets a fixed criterion within the
secondary stage of deformation and therefore, provides a more robust and less variable
test parameter. Furthermore, creating the MSR master curve and fitting a power curve
model is an advancement in the analysis of the repeated load test data from various
temperatures and stresses. The repeated load data that is typically conducted at different
laboratories at different stresses and temperatures could never have been compared
before; however by creating the MSR master curves, the results of tests can easily be
compared across different laboratories.
In this study, the application of the iRLPD testing and analysis methodology for
measuring the effect of long-term aging on asphalt mixture rutting performance is
provided. Using the MSR of the long-term aged mixtures, the rutting performance of the
mixtures in this study would be predicted.
5. Materials and Mix Design

Six materials from pavement projects with different design traffic levels in various
climatic regions across the United States were used in this study for the high temperature
characterization of asphalt mixtures subjected to various short-term conditioning and
long-term aging durations. Table 1 provides the sources of these materials, the mixture
ID, asphalt grade, aggregate nominal maximum aggregate size (NMAS), location of the
paving projects, design traffic level, and the pavement temperature selected based on the
LTPPBind 50% reliability performance grade at a depth of 20 mm (LTPPBind V3.1,
2005).
134
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 1. List of mixtures evaluated in this study.
Design LTPPBind High

Asphalt Mixture Temperature, 50%
Traffic Mixture
Location Binder NMAS, Reliability, C
Level, ID
Gr a d e mm
MESAL
Monroe County,
PG58-28 1 2 .5 WI-E3 4 9 .1
Wisconsin
3 Wilmington
County, North PG64-22 9 .5 NC 5 8 .6
Carolina
Marathon County,
PG64-28 1 2 .5 WI-E10 4 6 .7
Wisconsin
10
Crawfordsville,
PG64-22 9 .5 IN 5 3 .3
Indiana
Hernando County,
PG67-22 9 .5 FL 6 3 .0
Florida
30
Kern County,
PG70-10 1 9 .5 CA 6 2 .0
California
6. Experimental Testing
The six mixtures were prepared at different short-term conditioning periods. A total of
126 specimens were prepared using seven short-term conditioning periods of 0 hours to
6 hours, in 1-hour intervals. The mixtures were then compacted and tested using iRLPD.
Each specimen was subjected to three test increments at three stress levels before the
long-term aging and one increment after the long-term aging to develop the MSR master
curves for un-aged and aged samples. The long-term aging included heating the
specimens at 85C for 2, 5 or/and 9 days. The one increment of iRLPD after long-term
aging was conducted at the highest stress level that was applied to the specimen before
long-term aging.
135
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

AZARI, MOHSENI
7. Experimental Results
7.1. Effect of Long-Term Aging on Power Coefficient b
MSR master curves for various short-term conditioning and long-term aging scenarios
were developed for all six mixtures. Figure 5 shows the master curves for the IN
mixture test specimens, short-term conditioned for total of three hours at 135C prior to
compaction and long-term aged at 85C for 0, 2 and 5 days after compaction. To develop
the un-aged master curve, the power function MSR=0.001* (TP) b was fit into the un-
aged MSR versus TP data. The un-aged b coefficient was determined from the fit. For
the aged master curve, the measured MSR of the aged specimens and their
corresponding TP were substituted into the power function to calculate the aged b
coefficient. Figure 5 shows that exponent b for this mixture is 2.68 prior to the long-
term aging and 2.48 and 2.37 after 2 and 5 days of long-term aging at 85C, respectively.
The reduction in the b coefficient indicates the increase in resistance to permanent
deformation of the IN mixture due to long-term aging.
IN Mix: MSR Master Curves for 3 Hours Short-

term Conditioning and Various Long-term Aging
20
18 No LT Aging
16 2 Days @ 85 C
14 5 Days @ 85 C
MSR, Microstrains
12
10
MSR = 0.001 TP2.68
8 R = 0.9981
6 MSR = 0.001 TP2.48

R = 0.9958
4
MSR = 0.001 TP2.37
2 R = 0.9991
0 --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0 5 10 15 20 25 30 35 40
TP, Deg-Mpa
Figure 5. Sample MSR master curve for IN specimens, un-aged and long-term aged for
2 and 5 days (LT in the graph stands for long-term).
136

7.2. Effect of Short-Term Conditioning on Power Coefficient b

Figure 6 shows the power exponent b for various short-term conditioned and long-
term aged specimens of the six mixtures. The slope of the fitted lines to b coefficients
reveals important information regarding the susceptibility of the mixtures to short-term
conditioning duration. The slope of the lines for no long-term aging indicate the
susceptibility to short-term conditioning right after construction and the slope of the b
lines corresponding to long-term aging indicates the effect of short-term conditioning
after several years that the road has been in service. Considering the effect of short-term
conditioning right after construction, the E3 mixture is most susceptible to short-term
conditioning as indicated by a slope of -0.079 for the un-aged b line, the NC mixture is
the second most susceptible (slope of -0.0571), and E10 is the third most susceptible to
short term conditioning (slope of -0.0489). This indicates that for these mixtures, the
duration of short-term conditioning is the most critical for right-after-construction rutting
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
performance. The slopes of un-aged b lines for FL and CA are the lowest (-0.010 and
-0.030), indicating the lowest effect of short-term conditioning on the right-after-
construction performance.
The slope of the b values versus short-term conditioning time for the aged mixtures
indicates the susceptibility of the mixtures to the effect of short-term conditioning after
the pavement has been in service for some period of time. As shown in Figure 6, the
slopes of the fitted b lines for the 2-days and 5-days long-term aged WI-E3 mixtures
are the largest (-0.082 and -0.077). Thus, this mixture is considered the most susceptible
to short-term conditioning even after long-term aging. The next most susceptible mixture
to short-term conditioning after long-term aging is WI-E10 with slopes of -0.058 and
-0.041 from 2-days and 5-days of long-term aging, respectively. This indicates that for
WI-E3 and WI-E10, the effect of short-term conditioning remains significant even after
the roads have been in service for several years. FL is the least susceptible to the effect
of short-term conditioning after long-term aging, as indicated by the smallest slopes of
its aged b lines (-0.010 for the specimens aged both 2 days and 5 days). Therefore, the
duration of short-term conditioning has the least effect on the long-term performance of
Florida mixture.
137

AZARI, MOHSENI
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 6. Change in exponent b of master curve with change in short-term con-
ditioning and long-term aging.
7.3. Comparison of Aging Rates using Power Coefficient Ratios

The power coefficient b is used to compare the rate of change in resistance to
permanent deformation of the long-term aged specimens of the six mixtures. In this
respect, the power coefficient b of every mixture for various aging times was divided
by the power coefficient of the un-aged specimen to obtain power coefficient ratios.
Figure 7 shows the power coefficient ratios of the long-term aged specimens of the six
mixtures for the 4-hour short-term conditioning. It is evident from the figure that for
these mixtures, the rate of decrease in power coefficient ratios (aging rate) as a result of
138

long-term aging is very different from each other. As shown in Figure 7, the Wisconsin
mixtures E10 and E3 had the highest aging rate. These two mixtures are used in regions
of Wisconsin where the asphalt is not exposed to very high temperatures that cause
excessive aging (the 50% reliability high pavement temperature from LTPPBind is
49.1C and 46.7C for the WI-E3 and WI-E10 mixtures, respectively). Therefore, in
terms of aging, the two Wisconsin mixtures are appropriate for their environmental
conditions, where rapid aging of the mixtures is not facilitated. Figure 7 also shows that
the CA mixture had the lowest aging rate. This mixture is used in a region where
prolonged heat is known to cause excessive aging. Therefore, in terms of long-term
aging effect, the CA mixture is appropriate for its environmental condition.
Normalized Power Coefficent Ratios for 4 Hour

Short-Term Conditioning
1 .0 5
1
Power coefficient Ratio
0 .9 5
CA
0 .9 E1 0
0 .8 5 E3
FL
0 .8
IN
0 .7 5
NC
0 .7
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0 2 4 6 8 10
Days of Long-term Aging
Figure 7. Normalized power curve coefficients b versus long-term aging for all six
mixtures.
7.4. Reduction in MSR with Short-Term Conditioning and Long-Term Aging

Another means for evaluating the effect of long-term aging on permanent deformation
properties of the mixtures is determining the percent reduction in MSR due to long-term
aging. MSR is a measure of unit damage per load cycle and reduction in MSR represents
139

AZARI, MOHSENI
the reduction in permanent deformation damage. The power equations of the MSR
master curves (MSR= a *TPb) were used to calculate the MSR at a reference temperature
of 60C and a tire pressure of 600 kPa for various short-term conditioning and long-term
aging periods. The percent reduction in MSR due to long-term aging was then calculated
using the un-aged and aged MSR values. Figure 8 shows the reduction in MSR due to
long-term aging for various short-term conditioning durations for all mixtures.
Irrespective of the duration of short-term conditioning, the reduction in MSR after 5
days of long-term aging for the FL and CA mixtures was about 65% and 40%,
respectively. The reduction in MSR after 5 days of long-term aging for the two
Wisconsin mixtures (E3 and E10) was around 75%.
The reduction in MSR of the other three mixtures in Figure 8 shows a somewhat
different trend. While for the CA, FL, and the two Wisconsin mixtures, the reduction in
MSR was the same regardless of the duration of short-term conditioning, the reduction
in MSR of the long-term aged specimens of IN and NC are significantly affected by
short-term conditioning duration. The NC mixture had the highest reduction (80%) in
MSR if it was short-term conditioned for 2 hours, but the lowest reduction (45%) if it
was short-term conditioned for 4 hours. The IN mixture also had a lower reduction in
MSR at 4 hour conditioning compared to 2 hour conditioning. The results of IN and NC
mixtures show significant interdependency between the effect of short-term and long-
term conditioning on these mixtures. This observation indicates the importance of
enforcing a specification for the duration of short-term conditioning in asphalt plants
and/or quantification of the effect of short-term conditioning for accurate long-term rut
depth prediction.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
140

Reduction in MSR with Long-term Aging

(2 Hour short-term conditioning)
1 0 0% CA
90%
80% E 10
70%
Re duct ion in MSR
E3
60%
50% FL
40%
IN
30%
20% NC
10%
0%
0 2 4 6 8 10
Days of L ong-t e r m Aging

1 0 0% CA
90%
80% E 10
70%
Re duct ion in MSR
E3
60%
50% FL
40%
IN
30%
20% NC
10%
0%
0 2 4 6 8 10

1 0 0% CA
90%
80% E 10
70%
Re duct ion in MSR
E3
60%
50% FL
40%
IN
30%
20% NC
10%
0%
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0 2 4 6 8 10
Figure 8. Percent reduction in MSR after long-term aging of asphalt mixtures.
141

AZARI, MOHSENI
8. Performance Prediction Using MSR Master Curves

A computer program was developed to estimate 20-year rut depths using the MSR
master curves of the six mixtures. The program includes estimated hourly pavement
temperature data at a depth of 20 mm. The asphalt layer depth was assumed to be 50
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
mm, which is the depth at which rutting is believed to be developed; thus the
temperature at a depth of 20 mm would correspond approximately to the mid-depth of
the surface layer. The hourly pavement temperature was calculated from hourly air
temperature, wind speed and sky cover using estimated pavement parameters (MEPDG,
2004). Figure 9 shows a screen shot of the software. In the material section, master
curves of the IN mixture are chosen for analysis. The weather data from a weather
station located at Indianapolis was used for estimating the temperature. Traffic was
selected at 20 million ESALs and tire pressure was assumed to be 0.6 MPa. The MSR
values for the short-term conditioning of 4 hours were used for the prediction. The top
bar chart shows the number of hours the pavement was at certain temperatures for the
month of July of the first year. It can be seen that the pavement temperature reached
62C in that month. The bottom bar chart shows the calculated rutting for every degree
centigrade. As temperature increases, the rutting damage also increases. It is shown that
although there were only a few hours with 62C, the damage at this temperature was still
significant. Rutting for one month is the sum of the rutting for all temperatures. The
MSR master curve for every month is different due to the effect of long-term aging; as
the pavement ages the permanent deformation rate decreases. The total 20-year rutting
is the sum of all rutting for all months.
Figure 9. Computer screen of the software for prediction of rutting.
142

Table 2 shows the estimated 20-year rutting for all mixtures using hourly pavement
temperatures at a depth of 20 mm. The MSR master curve for long-term aging was used
to discount the rutting damage for the absorbed heat. Rutting was calculated for three
different short-term conditioning prior to compaction: 2 hours, 4 hours, and 6 hours. As
shown in the table, there is a significant difference between rutting estimated for the
mixtures with 2 hours and 6 hours short-tem conditioning. The highest difference is for
the E3 mixture (110%) and the lowest is for the CA mixture (47%). Therefore,
uncertainty about the length of short-term conditioning before the mixture is placed and
compacted may result in significant under or overestimation of rutting.
Table 2. Estimated rutting for mixtures at their design location and traffic levels;
STC stands for short-term conditioning.
Rutting after 20 years, mm Percent

Increase
Design between 2
Traffic, 2 hour 4 hour 6 hour and 6
Mix Location MESALs STC STC STC hours STC
WI-E3 Madison 3 4 .4 3 .1 2 .1 1 0 9 .5 %
NC Raleigh 3 3 .5 3 1 .9 8 4 .2 %
WI-E10 Madison 10 8 .8 7 .7 5 .1 72.6%
IN Indianapolis 10 4 .6 3 .6 3 .1 48.4%
FL Tampa 30 8 6 .1 5 .4 48.2%
CA Los Angeles 30 2 5 .2 2 2 .2 1 7 .1 47.4%
9. Summary of Findings
The following provides a summary and the conclusions of this study:
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
1) A method for quantifying the effect of short-term conditioning and long-term
aging on the permanent deformation properties of asphalt specimens was proposed. The
MSR master curve from the iRLPD test provides the key information for estimating rut
depth; considering effects of temperature, tire pressure, and aging.
2) Knowledge of the change in material properties during long-term aging is crucial
in estimating asphalt mixture rut depth. The iRLPD test is particularly useful for this
evaluation since it allows testing the same specimens before and after long-term aging
without sample-to-sample variability. The change in the MSR master curve of a single
143

AZARI, MOHSENI
specimen with increase in laboratory long-term aging duration provides the necessary
information for the long-term performance prediction.
3) Different asphalt mixtures may age in very different ways and this will
significantly affect long-term properties of asphalt mixtures. For this reason, it is critical
that the effect of long-term aging be quantified for each asphalt mixture and the
information be used in performance prediction.
4) This study shows that there is a significant interdependency between short-term
conditioning and long-term aging of some asphalt mixtures. The short-term conditioning
of asphalt mixtures affects the long-term aging, and the effect is more significant for
some mixtures than others. This indicates the importance of having a set specification
for short-term conditioning of the mixtures before construction.
5) Since controlling the short-term conditioning that takes place in a silo could be
difficult, it is critical to quantify the effect of short-term conditioning on the permanent
deformation properties of asphalt mixtures. This can be accomplished by conducting the
iRLPD test right after plant mixing and right before mixture lay down. Without the
knowledge of the various aspects of short-term conditioning in a plant and its effect on
material properties, it will be difficult to provide any reasonable estimate of rutting in the
field.
10. Recommendations
The following recommendations are given from the conduct of this study:
1) The effects of long-term aging on other asphalt properties, such as intermediate-
and low-temperature cracking, also need to be determined. Similar to the permanent
deformation application, the iRLPD test method can be used to quantify the effect of
long-term aging on intermediate and low temperature properties by measuring the
change in minimum strain rate (MSR).
2) Duration of short-term conditioning of asphalt mixtures should be controlled in
the field and its effect should be considered in mixture performance. This study has laid
out a method for determining the effect of short-term conditioning on permanent
deformation. In the field, the reduction in MSR of a plant-produced mixture right after
mixing and right before compaction should be monitored and considered in prediction of
long-term performance.
3) Asphalt technology is changing by the day. Significantly more roads are built
using Reclaimed Asphalt Pavements (RAP) and shingles (RAS), which may age very
differently from virgin asphalt pavements. The use of warm mix asphalt is also rapidly
increasing the trend tend to soften the mix from an early age of asphalt pavement. For
144
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

these reasons, it is even more critical that more precise determination of permanent
deformation properties of asphalt mixtures is made. This study proposes a new method
that can be used in conjunction with new asphalt technology.
4) The long-term aging method of asphalt mixtures in AASHTO R30 should be re-
examined and improved to include the environmental temperature and the amount of
heat absorbed by the mixture in the field. Currently, all long-term aging is performed at
a fixed temperature and time and does not distinguish hard and soft binders or the
location where the mixture will be used. For this reason, the laboratory generated data is
only loosely related to the field conditions.
11. Acknowledgement
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The authors would like to thank Mr. Steve Lenker, the Director of the Construction
Materials Reference Laboratory, for his support throughout this study. The authors wish
to acknowledge Mr. Frank Fee for coordinating the acquisition of the materials for this
study. The efforts of State DOT and private agencies, which provided the materials and
the corresponding mixture designs, are most appreciated. The agencies are Florida
Department of Transportation, Gainesville, Florida; North Carolina Department of
Transportation; Heritage Research Group, Indiana; MTE Construction, Wisconsin; and
California Department of Transportation.
12. References
Azari, H., "Analysis of the Effect of Laboratory Short-Term Conditioning on
Mechanical Properties of Asphalt Mixture," TRB 90th Annual Meeting
Compendium of Papers DVD, Paper No. 11-1427, Transportation Research Board,
Washington, DC. 2011.
Azari, H. and Mohseni, A., "Incremental Repeated Load Deformation Testing of Asphalt
Mixtures," TRB 91st Annual Meeting Compendium of Papers DVD, Paper No. 12-
4381, Transportation Research Board, Washington, DC. 2012.
Azari, H. and Mohseni, A., "Permanent Deformation Characterization of Asphalt

Mixtures Using Incremental Repeated Load Testing," TRB 92nd Annual Meeting
Compendium of Papers DVD, Paper No. 13-5160, Transportation Research Board,
Washington, DC. 2013.
AASHTO Standard Practice. AASHTO R 30: Standard Practice for Mixture

Conditioning of Hot Mix Asphalt (HMA), Thirty-Second Edition, American
Association of State Highway and Transportation Officials, Washington, DC. 2012.
145

AZARI, MOHSENI
AASHTO Provisional Standards. AASHTO TP 79: Standard Method of Test for

Determining the Dynamic Modulus and Flow Number for Hot Mix Asphalt (HMA)
Using the Asphalt Mixture Performance Tester (AMPT), Thirty-Second Edition,
American Association of State Highway and Transportation Officials, Washington,
DC. 2012.
Goh S., You Z., Wang H., Mills-Beale J., Ji J., Determination of Flow Number in
Asphalt Mixtures from Deformation Rate During Secondary State, Transportation
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Research Record, No. 2210, Transportation Research Board, Washington, DC., 2011.
Houston W.N., Mirza M. W., Zapata C., and Raghavendra E. S. Environmental Effects
in Pavement Mix and Structural Design Systems, NCHRP Web Document 113,
2005
Kenis W., Predictive Design Procedures, VESYS Users Manual: An Interim Design
Method for Flexible Pavements using the VESYS Structural Subsystem, Federal
Highway Administration, Report No FHWA-RD-77-154., 1978.
LTPPBind V3.1, 2005
http://www.fhwa.dot.gov/research/tfhrc/programs/infrastructure/pavements/ltpp/inst
all.cfm
MEPDG, Guide for Mechanistic-Empirical Design of New and Rehabilitated Pavement
Structures, NCHRP 1-37A Final Report, Appendix DD-4, National Academies,
Washington DC 2004.
Von Quintus H. L., Mallela J., Bonaquist R., Schwartz C. W., and Carvalho R. L.
Calibration of Rutting Models for HMA structural and Mixture Design, NCHRP
Report 719, National Academies, Washington, DC. 2012.
http://onlinepubs.trb.org/onlinepubs/nchrp/nchrp_rpt_719.pdf,
13. Discussion
PROFESSOR MIHAI MARASTEANU : Nice presentation. I have more of a

philosophical question. Aging is good for rutting, so why do we spend so much time on
aging and rutting? Aging negatively affects fatigue and low temperature cracking. The
work you presented is good, but what I would like to understand is what is the ultimate
purpose to come up with a mixture that ages more and is better for rutting?
DR. HALEH AZARI : You want to select the right mixture for a pavement project and
also to be able to predict the performance of a mixture correctly. Rutting and fatigue and
low temperature cracking this method applies for all of them. We are presenting it for
rutting, but, you want to be able to predict performance for any property. Effect of aging
146

needs to be considered in combination with all properties. When designing, the effect of
aging should be considered on both rutting and fatigue and on low temperature cracking
if it is a concern for you.
PROF. MARASTEANU: I understand that. It is ok to accurately predict rutting.
However, it is well known more aging improves rutting and significantly decreases
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
cracking resistance. I would have thought it made more sense to study aging effect on
cracking.
DR. AZARI: The material selection should be a balanced process. In an environment
where the mixture is exposed to prolonged heat, you dont want to place mixtures that
age rapidly. Even though it would be good for rutting, it wont be beneficial for fatigue
cracking. You need to know at what rate your material ages in order to select the right
material for the environment.
PROF. MARASTEANU: Sure. All right. Thank you.
DR. ALAEDDIN MOHSENI : If I may answer. The usefulness of the procedure is that it
tells you if you are going to have a problem with cracking when you place a mix thats
too stiff. If a mix is aging rapidly, then maybe we do not want to place it down when its
too stiff, and its going to help with the cracking later on. So the two work together.
Right now, we dont have a way of distinguishing between a high age material or not. As
shown in some of the graphs here, aging can have tremendous effect on rutting and as a
result on cracking performance in the future.
MR. FRANK FEE: Mixes taken from plant mix where did you obtain these mixes that
you're aging?
DR. AZARI: The mixtures are the same mixtures we used as part of the ETG study, so
they were aggregates and asphalts.
MR. FEE: They were lab mixes.
DR. AZARI: Yes, they were laboratory mixes.
MR: FEE: Okay. So do we have any field performance on these mixes or do you kno w
anything about what they would look like coming out of an asphalt plant in terms of
characterization?
DR. AZARI: All we know about these mixtures is the traffic level they have been
designed for, and we know that they have been performing well for their climatic
condition and traffic level. Some of these mixtures, even without considering the effect
of long-term aging, were predicted to perform well in their projects; but by incorporating
aging, we are making better prediction and less rutting has been predicted for them,
which is closer to the actual performance of those pavements.
147

AZARI, MOHSENI
MR. FEE: But in terms of the actual performance in the field, they were all equal, they
were all good, so we dont really know if we could rank them against field performance
at this time. In other words, they were all good. Youre saying that these rank differently,
but in the field, can we show that? I guess thats where Im going with this.
DR. AZARI: Yes. I mean this ranking is in laboratory without consideration of the
traffic level and the climatic region. In their own climatic region, they all performed
well. But when you compared them in laboratory, some performed better than the others.
For example, the Wisconsin mixture is not performing as well as California, but in their
own climatic conditions and projects they both performed well.
MR. FEE: I think you have good work here, its potential is there. As you know, one
of the concerns we have with this difference between the plant and lab aging is early
pavement failure. A material that gets out of the plant and is too soft. It doesnt age
quickly enough and traffic will rut it quickly. As time goes on, we dont worry about it
as much because it continues to stiffen. We need to get field performance on these
relationships. Youve selected certain temperatures and times for the lab aging. Now see
if you can get those to match something that actually happens in the field.
DR. AZARI: That is why we need to know at what rate every asphalt ages and as it was
shown different asphalt mixtures age at very different rates. Mixture that age slowly can
be placed stiffer than mixtures that age rapidly to avoid early rutting and long-term
cracking.
The long-term aging that is proposed in AASHTO R30 is five days at 85C regardless of
where these mixtures are going to be placed. The laboratory aging should be in
consideration of the amount of heat the mixtures will be exposed in the field. For those
materials that are not exposed to a lot of sun, five days of aging might be too long. For
those that are exposed to excess sun, five days of aging might not be enough. So these
need to be considered, and we are using the Degree-Day parameter from LTPPBind to
determine for each particular climatic condition what should be the amount of aging in
the laboratory.
PROFESSOR GABRIELE TEBALDI: You show a performance about mixture in
Madison, Tampa, or in California. So my question is: Have you adjusted your aging
process related to the environmental condition or how can you do that? Because of
course the aging is strongly related with the climatic condition. I do not understand how
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
you can calibrate that related to
DR. AZARI: Exactly. Thats what I just mentioned. For this study, we used two days of
aging, five days of aging, nine days of aging. It was constant for all of these mixtures.
But the correct way of doing it is to determine how much aging takes place in each
environmental condition. If the temperature is too high and if the sun exposure is too
much then the laboratory aging process should be adjusted. All of these are
determined by the Degree-Day parameter from LTPPBind. Dr. Mohseni, actually, can
148

explain more on the climatic part of the data and on how we incorporate that into the
laboratory aging.
DR. MOHSENI: The actual implementation was that we did estimate the rutting. We
used the climatic data for only one location. That was the Indiana location. Now there
were six mixes from different parts of the United States, but they were actually applied
in only one location. Now if we want to go to different location, then we are actually
working on the Degree-Day concept to adjust the MSR master curve for the location. So
for different locations in the United States, there would be different adjustment factors to
the aging ratios.
PROF. TEBALDI: Related to temperature or with the temperature and other
environment and conditions.
DR. MOHSENI: Degree-Days parameter includes both temperature and time. So the
Degree-Day in LTPPBind shows pretty much the heat quantity. If you go to LTPPBind
for any location, you can get an estimate of the heat quantity per year. We are actually
tying the MSR master curve to that heat quantity. The laboratory aging would still be
five days, but then we might not use the whole property after five days of aging. We are
going to correct that based on the Degree-Day. So for Wisconsin, it might be four days
and for Alabama it might be more like six days of aging that is going to be used.
PROF. TEBALDI: It is only temperature based aging?
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. AZARI: Temperature and duration of the temperature.
DR. MOHSENI: Degree-Days is temperature and duration. So say in Florida, the
temperature may not go very high, but its high for like four months. That means the
Degree-Day would be high.
PROF. TEBALDI: Okay.
149

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluating Photocatalytic Asphalt Pavement
Effectiveness in Real World Environments
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
through Developing Models: A Statistical

and Kinetic Study
Heather Dyllaa, Somayeh Asadib, Marwa Hassana*,

and Louay N. Mohammadc
a
Louisiana State University, Department of Construction Management, Baton Rouge,
LA, 70803;
b
Texas A&M University-Kingsville, Department of Civil and Architectural Engineering,
Kingsville, TX, 78363
c
Louisiana Transportation Research Center, Louisiana State University, Baton Rouge,
LA 70808
ABSTRACT. Photocatalytic asphalt pavements are evaluated in this study as a possible air-pollution
reduction strategy and to comprehend their behavior in the real world environment. Statistical
models from real world experiments or reaction kinetics are two approaches to understand the
photocatalytic reduction in real world environments. The objective of this study was to develop a
statistical model for nitrogen oxide (NO) reduction using data from a field study and to evaluate
the photocatalytic reaction kinetics of NO reduction, which could be used in future theoretical air
pollution model simulations. To achieve this objective, a photocatalytic water-based spray coating
was applied on an existing asphalt pavement site for the field study statistical model and on
laboratory samples for the kinetic study. Based on the field data, the NO reduction was modeled
using statistical regression techniques by creating a model for a non-coated pavement and
photocatalytic pavement. The coefficient of determination was 0.79 and 0.67, respectively. To
improve prediction, other parameters may need to be included into the model and more sampling
time is required. Based on the lab results, the NO reduction was reaction controlled following the
Langmuir-Hinshelwood model. The adsorption equilibrium constant calculated for photocatalytic
asphalt pavements was similar to those of concrete pavements, while the reaction rate constant
The oral presentation was made by Ms. Dylla.
http://dx.doi.org/ 10.1080/14680629.2013.812839
151

DYLLA, ASADI, HASSAN, MOHAMMAD
was significantly lower. While humidity has a negative correlation on both L-H constants,
intensity has a positive correlation. However, interaction between these two parameters exists.
KEYWORDS: photocatalytic asphalt; TiO2; statistical modelling; kinetic modelling
1. Introduction
Negative health effects from vehicle pollution are associated with living, working and
going to school near highways from long-term and short-term exposure. As a result,
many technologies have been implemented to reduce vehicle emissions. However,
increasing urbanization, vehicle use, traffic congestion, and desire for larger cars, often
offset the reductions realized by these new vehicle emissions control strategies (HEI,
2010; Menz, 2002). This has led to the practice of continually reducing vehicle
emissions standards. Another technique of pollution abatement is photocatalytic
pavements. Photocatalytic pavements use a semiconductor photocatalyst that, when
irradiated by sunlight, can reduce air pollutants after they are emitted.
Initial interest in environmental photocatalysis began in the 1970s, initiated by
Fujishima and Hondas research in photoelectrochemical solar energy conversion.
Through biomimicy of plant photosynthesis, Fujishima attempted to replicate the photo-
induced redox reactions, oxidizing water and reducing carbon dioxide, by using a
semiconductor irradiated by UV light (Fujishima and Zhang, 2006). To accomplish this,
the semiconductor is used as an electrode connected to a counter electrode to generate
electrical work to drive the redox chemical reactions (Fujishima and Honda, 1972). By
removing the electrode, environmental photocatalytic oxidation occurs completely
decomposing both organic and inorganic compounds. Since then, interest in
environmental photocatalysis has increased and TiO2 photocatalysts have been applied to
glass, tile, paper, and pavements for self-cleaning materials, water purification, air
purification, sterilization, and oil spill remediation (Fujishima and Zhang, 2006).
For abatement of pollution in roadway microenvironments, pavements are an ideal
substrate due to their close proximity to higher concentrations of pollution and large
surface area (Beeldens, 2006). Titanium dioxide is the preferred photocatalyst used
because of its high stability, super-hydrophilicity, relative cheapness, low toxicity, and
commercial availability (Cassar, 2004; Fujishima and Zhang, 2006; Diamanti et al.,
2008; Toma et al., 2009; Yu, 2009). Titanium dioxide can be applied to both asphalt and
concrete pavements. In laboratory studies, Li et al. demonstrated that an emulsified
asphalt blend with 2.5% TiO2 reduced up to 40% of the nitrogen oxides (NOx) (Li et al.
2009). In addition, Hassan et al. achieved as high as 66% reduction of NOx using a
152
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluating Photocatalytic Asphalt Pavement Effectiveness in Real World Environments
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
spray coat on warm mix asphalt laboratory samples (Hassan et al., 2012). However, in
order to grasp the widespread pollution reduction made possible by photocatalytic
pavements, the next major step is to comprehend its behavior in the real world
environment. Statistical models from field experiments or reaction kinetics are two
approaches to understand the photocatalytic reduction in real world environments.
Therefore, the objective of this study was to develop a statistical model for the NO
reduction using data from a field study and to evaluate the photocatalytic reaction
kinetics of NO reduction, which could be used in future theoretical air pollution model
simulations. Not only are these models important for understanding the significance of
the NO reduction expected, they could be included into the State Implementation Plan
(SIP) air quality calculations and models allowing photocatalytic pavements to be
evaluated as a possible pollution reduction strategy (EPA, 2012).
2. Background
As previously mentioned, titanium dioxide is the preferred photocatalyst used in

photocatalytic pavements. When TiO2 is exposed to energy, from photons, that exceeds
the band gap energy of 3.2 eV, an electron is expelled from the valence band to the
conduction band, leaving a hole behind (Zhao and Yang 2003, Fujishima et al., 2000).
For TiO2, this process is initiated by energy from a UV light wavelength determined by
the electromagnetic radiation, hv, where h is Plancks constant and v is the frequency of
light. The wavelength required for TiO2 irradiation is between 300 to 365 nm (Zhao and
Yang 2003). The production of electron-hole pairs, called excitons, results in redox or
oxidation chemical reactions (Fujishima et al., 2000, Zhao and Yang 2003, Hunger et al.,
2008). In the presence of water, these oxidizing holes, h+, and photogenerated electrons,
e-, create hydroxyl radicals and superoxides respectively, as shown in Equations 1 and 2
(Fujishima et al., 2000).
OH - +h+ OH * [1 ]
O2 e- O2- [2 ]
The resulting hydroxyl radicals and superoxides are key for oxidation or reduction
reactions allowing for degradation of pollutants in the oxidation of NOx to water soluble
nitrates, as shown in equations 3 and 4 below (Beeldens 2008):
Ti O2
NO OH * NO2 H * [3 ]
153

2
2 3 [4 ]
This allows for two techniques to measure photocatalytic efficiency. The first is to
measure the reduction directly by measuring the ambient air pollution concentration, and
the second is to measure the reduction indirectly by measuring the byproducts created
from the degradation process. For nitrogen oxides, the approved method of ambient air
NOx detection is chemiluminescence, which continuously monitors the NOx
concentrations. This is challenging since environmental parameters are constantly
changing. The indirect technique to measure photocatalytic reductions of NOx, is to
measure the NO3 and NO2 deposited on the surface. Nitrates and nitrites are water-
soluble and therefore washed from the surface with water to be quantified (Beeldens,
2008). Water samples are usually collected daily and analyzed for nitrates and nitrites,
thus it does not capture the environmental variability essential for understanding
photocatalytic reduction of NOx in real world environments. Despite this limitation,
nitrate accumulation provides evidence that the NO reduction measured in ambient air is
indeed from a photocatalytic reduction and not adsorption.
2.1. Photocatalytic Models
Few models exist to describe the performance of photocatalytic pavement materials

reduction of NOx. Statistical techniques or theoretical reaction kinetics are two
modeling methods that could be used to model the reduction of NO from photocatalytic
pavements.
Statistical models can be used to describe the pollution in an area based on various
parameters. Statistical models have been important to identify the contribution of
pollution sources to human exposure for air pollution policy and regulation to mitigate
associated risks (Zeng et al., 2010). Statistical approaches include regression,
multiregression, and artificial neural networks. Estimates are calculated by the statistical
relationship to various factors collected at a particular receptor (Sharma et al., 2004).
Source apportionment models, which attempt to identify different sources contribution to
the pollution at a particular receptor, use many of these statistical methods such as
principal component analysis (PCA) and multilinear regression (Vallius et al., 2008).
These models are easy to use, but they are receptor or location specific, require large
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
amounts of historical data, and do not identify the significance of various physical and
chemical processes (Sharma et al., 2004). Nevertheless, they can be used to create
simple models to characterize pollution in roadway microenvironments.
Photocatalytic oxidation reaction kinetics theory is well-established by researchers
for reactor engineering. The most widely used model for heterogeneous surface
154

reactions is the Langmuir-Hinshelwood where at high pressures a zero order reaction

prevails and at low pressures a first order reaction prevails (Wang et al., 2007; Hunger et
al., 2010). Using this model, the reaction rate for NO photodegradation is described as
follows:
k*KCNO
rNO [5 ]
1+KCNO
where: rNO = NO photodegradation rate (mg/m3min),

k = L-H reaction rate constant (mg/m3min),
K = the L-H adsorption equilibrium constant (m3/mg),
CNO = the concentration of NO (mg/m3).
To account for the impact of humidity, competition between the two pollutants for
adsorption exists, thus the equation takes the form presented in Equation 6 (Hunger et
al., 2010):
kKNO CNO
rNO [6 ]
1+KNO CNO +KH2OCH2O
where: CH2O = the concentration of H2O,

KH2O = equilibrium constant for H2O.
Despite knowledge of this relationship, many models found regression techniques to
incorporate the impact of relative humidity more adequate (Hunger et al., 2010;
Bengtsson and Castellote, 2010).
Recently, kinetic studies have been incorporated in photocatalytic pavement studies.
Hunger et al. modeled the kinetics of the photocatalytic degradation of NOx using
photocatalytic concrete paving stones. The impact of the UV-light intensity and
humidity on the kinetic constants were each modeled separately (Hunger et al., 2010).
To incorporate these parameters into one model, Bengtsson and Castellote used non-
linear regression to model the reaction rate constant (Bengtsson and Castellote, 2010).
Nevertheless, to the authors knowledge, the only kinetic studies completed for
photocatalytic pavements have been on concrete pavements.
3. Methodology
To develop the statistical models, a field study was conducted while the reaction kinetics
were determined from laboratory experiments. The same photocatalytic thin film was
applied to both the field study pavement and the laboratory samples. The thin film was
applied as a spray coat in two parts. A primer was first applied as a base coat. Then
155
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

immediately after, an aqueous liquid of suspended anatase TiO2 nanoparticles, 2% by

volume, was applied. The TiO2 nanoparticles are nanorods ranging from average 12 nm
width by 75 nm length diameter.
3.1. Field Study
3.1.1. Field Study Description
A 0.2-mile asphalt pavement located in Baton Rouge, Louisiana, was selected as the
field study location. A distributor sprayed the primer and photocatalytic coat at a rate of
16.1 ml/m2 to 21.5ml/m2 (Figure 1). Electrostatic precipitators were used in the nozzles
to ensure more even distribution of the TiO2 photocatalyst.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 1. TiO2 photocatalytic coating field application.
The NO, NO2, and NOx concentrations, traffic count and climatic conditions were
monitored and recorded for a control area and for a photocatalytic pavement area over a
period of two months during the winter. The equipment used for field data collection
was housed in an air-conditioned trailer. A Thermo 42i NOx analyzer was used for
monitoring NO, NO2 and NOx concentrations as shown in Figure 2. The NOx analyzer
was calibrated in accordance with EPA calibration procedures using the Thermo 146i
gas calibrator with the gas phase titration (GPT) option. The stainless steel tubing was
placed to sample at the middle of the lane at pavement level in order to measure the
concentration that comes in contact with the pavement surface, Figure 3. The
concentrations were stored as minute averages.
To monitor climatic conditions at the site, a Davis Vantage Pro2 weather station was
installed in the field that recorded and stored meteorological data including humidity,
ambient temperature, wind speed, wind direction, rain, dew point, and solar radiation
continuously every 5 minutes as shown in Figure 3-a. Furthermore, a traffic counter was
156

installed to count the number of vehicles per 5 minute intervals per lane as shown in
Figure 3-b.
Figure 2. NOx analyzer and calibrator.
(a ) (b)
( c)
Figure 3. Field Monitoring (a) Weather station, (b) Traffic counter, and (c) NO
collector.
157
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

3.2. Kinetic Study
3.2.1. Laboratory Sample Preparation
Two asphalt pavement samples, 33 x 25.4 x 6.4 cm in size, were prepared using a
conventional hot mix asphalt with a PG 64-16 asphalt binder. The spray coat was
applied to each sample using a crosshatch formation (Figure 4) using a hand spray gun at
a 0.21 mg/cm2 catalyst loading.
Spray 1 Spray 2
Figure 4. Crosshatch spray formation.
3.2.2. Sample Preparation
Heterogeneous photocatalytic oxidation is a reaction that occurs between a gas or liquid

in contact with a photocatalytic solid surface. Thus, in order for the reaction to occur,
the pollutant must transfer to the photocatalytic surface where it is adsorbed. Therefore,
the reaction can be either mass transfer controlled or reaction controlled, determined by
whichever process is slowest. Previous kinetic studies have shown that photocatalytic
oxidation is reaction controlled and are thus modeled by the Langmuir-Hinshelwood. In
order to test whether the photocatalytic degradation of NO was mass transfer limited or
reaction rate limited, the effect of the inlet concentration was varied. The NO
concentration was varied by using inlet concentrations of 150, 250, 350, 550, 1000,
2000, and 3000 ppb, while the flow rate, relative humidity and intensity remained
constant at 1.5 l/min, 20%, and 2.4 mW/cm2 respectively.
In addition, in efforts to create a model for the photocatalytic reduction effective for
various environmental conditions, a kinetic study was completed at three relative
humidity levels (23, 48, 78 5%) and three average light intensities (2.4, 1, 0.5
mW/cm2). A fractional factorial study was completed such that nine different
environmental conditions summarized in Table 1 were tested and the Langmuir-
Hinshelwood (L-H) constants were calculated. In order to construct a robust regression
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`-
158

fit and to calculate the L-H constants, the inlet concentration was varied at four levels:
150, 250, 550, and 3000 ppb (1000 ppb for 0.5 mW/cm2 intensity since there was no
reduction at 3000 ppb) resulting in a total of 36 experimental runs.
Table 1. Experimental laboratory cases to determine L-H constants.
Run
Humidity (%) Intensity (mW/cm2) Concentration (ppb)
Scenario
1 23 2 .4 150, 250, 550, 3000
2 23 1 150, 250, 550, 3000
3 23 0 .5 150, 250, 550, 1000
4 48 2 .4 150, 250, 550, 3000
5 48 1 150, 250, 550, 3000
6 48 0 .5 150, 250, 550, 1000
7 78 2 .4 150, 250, 550, 3000
8 78 1 150, 250, 550, 3000
9 78 0 .5 150, 250, 550, 1000
3.2.3. Experimental Setup
The experimental setup used to quantify NOx removal efficiency was modified from the
Japanese Industrial Standard (JIS TR Z 0018 Photocatalytic materials air purification
test procedure) in order to accommodate larger samples and variations of environmental
parameters (JIS, 2009). The photoreactor is a plug flow model with an air space volume
of 16.75 liters. Further details of the setup can be found elsewhere (Dylla et al., 2010).
The sample was housed inside the photoreactor, to create an enclosed controlled
environment for the duration of the test. The photoreactor was irradiated with 5 UV-
fluorescent black tube lamps (20W, Philips) emitting wavelengths within 300-400 nm
(Zhao and Yang 2003). The distance between the sample surface and the photoreactor
determined the UV intensity measured by UV-A intensity meter (OAI Model 306) at 365
nm. The UV profile was measured for each test condition and the average intensity
159
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

observed was calculated. The highest intensity, 2.4 mW/cm2, had the most variance due
to the sample proximity to the lights. This variance is illustrated in Figure 5 showing the
intensity exposed over the sample area at the highest intensity level.
4 3-4
2-3
E (mW/cm2)
3
2 1-2
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
330
1 0-1
220
0 110
0 59 118 177 0 X (c m )
236 295 354
Y (c m )
Figure 5. Profile of UV-A irradiance over sample surface (2.4 mW/cm2).
While the JIS test procedure requires 5 hours of irradiation, it is noted that steady
state equilibrium is achieved within the first hour of irradiation. Therefore, the test
procedure was shortened until equilibrium was achieved, defined by stability of NOx
concentration for 10 minutes. To prevent deactivation by intermediates adsorbed on the
TiO2 active sites, samples were soaked in deionized water for 24 hours and dried for 48
hours before being retested.
4. Model Development
4.1. Field Study Regression Model Development
Two regression models were developed, one for predicting hourly NO concentration
before TiO2 application and the second for predicting hourly NO concentration right
after application using the data collected from the field study. Nitrates were collected
and measured from the field confirming a photocatalytic reduction of NOx. Results can
be found in Hassan et al. (2012). The regression models were used to compare NO
concentrations per hour under different environmental conditions. Comparing the two
160

models allowed for quantification of the field hourly NO reduction efficiency (Hassan et
al., 2012). The developed regression models relate the measured NO concentration per
hour linearly to the environmental factors as follows:
NO = f (T, H, V, T Ambient, S) [7 ]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
NO Before = 0.96T+0.22H-1.33 T Ambient -10.5V+ 0.02S [8 ]
NO After = 0.31T+0.06H-0.1 T Ambient -0.75V + 0.0003S [9 ]
NO Reduction = NO Before - NO After [1 0 ]
Where:
NO Before = average hourly NO concentration before TiO2 application (ppb),
NO After = average hourly NO concentration after TiO2 application (ppb),
NO Reduction = average hourly NO concentration reduced (ppb),
T = number of vehicles per hour,
H = relative humidity,
V = wind speed (m/s),
T Ambient = ambient temperature (C),
S= solar radiation (W/m2).
Statistical goodness of fit was assessed through the coefficient of determination (R2)
and Root-Mean Square Error (RMSE) as shown in Table 2(a). Table 2(b) illustrates the
range of variation for each parameter investigated in this study. The developed models
are only valid in the input ranges shown in Table 2(b).
Table 2. (a) Descriptive statistic of the developed models and (b) Valid input ranges of
variables in the developed models ranges of variables in developed models.
(a )
Statistical Parameters Untreated Treated
Coefficient of Determination (R2) 0 .7 9 0 .6 7
Root Mean Square Error (RMSE) 1 3 .7 3 .7
Normal Root Mean Square Error (NRMSE) 1 .3 1 3 .0 3
161

(b)
Range of variation
Parameters
Low Level High Level
Number of vehicle per hour 2 100
Humidity (%) 30 90
Wind speed (m/s) 0 3 .0
Outside temperature ( C) 0 35
Solar Radiation (W/m2) 0 1000
The validity of the statistical models was evaluated by comparing statistical results with
observed data not used to build the models. Figure 6 (a and b) compares the results of
the statistical models with independent measurements. Results are presented for two
cases: before TiO2 application and just after TiO2 application. As shown in these
figures, at some points, a good agreement was observed between the hourly predicted
concentrations of NO and observed concentrations while some points had high errors.
This suggests that more data are needed or other parameters may need to be added to
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
improve the prediction from the model.
( a) (b )
Figure 6. Comparison of measured vs. predicted NO concentrations for (a) before and
(b) after photocatalytic coating.
162

4.2. Photocatalytic Reaction Kinetic Model Development
Figure 7 shows the NO reduction per initial concentration. As the concentration

increases, the percentage of NO degradation decreases. Since NO reduction is
dependent on pollutant concentrations, this indicates that the mass transfer is reaction-
kinetics controlled rather than mass-transfer controlled.
100%
80%
y = e-1E-03x
NO Reduction
60%
R = 0.8078
40%
20%
0%
0 500 1000 1500 2000 2500
Initial Concentration NO (ppb)
Figure 7. NO reduction by the influence of initial concentration (1.5 l/min flow, 20%
relative humidity, 2.4 mW/cm2 intensity).
4.2.1. Mass Transfer Limited
Further confirmation can be found by assuming that the reaction is not mass transferred
controlled and diffusion is the limiting step. Assuming instantaneous conversion, the
mass balance can be written. Derivation of the NO mass balance equation for a plug
flow reactor is described elsewhere (Hunger et al., 2010). Integrating the mass balance
equation over the length of the photocatalytic surface the percent reduction can be
calculated by the following equation:

() 2
1 = 2 [1 1 ]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
()
where: CNO,in = inlet concentration (mg/m3),

CNO,out = outlet concentration (mg/m3),
163

Sh = Sherwood number,
= velocity of air (m/s),
h = height of air space in photoreactor (cm),
D = Diffusion Coefficient of NO (m2/s),
L = Length of sample (cm).
Using the variables defined in Table 3, if mass transfer controlled, the percent
reduction would be 92%. This, however, is not the case as shown in Figure 7 where the
reduction was varied by the initial NO concentration and the highest reduction was 73%
when the inlet concentration was at 150 ppb.
Table 3. Mass transfer variables.

Variable Symbol Value
Sherwood Number Sh 4 .4
Length L 3 3 cm
Height h 5 cm
Diffusion Coefficient NO D 1.51x10-5 m2/s
Velocity of Air 0.0015 m/s
4.2.2. Reaction Rate Limited

To solve for the L-H reaction rate constant and the L-H adsorption equilibrium constant,
laboratory experiments were used. For a single pass experimental setup, plug flow is
assumed and the mass balance is written as follows:
dCNO k*KdCNO
rNO -vair * [1 2 ]
dx 1+KdCNO
dCNO
where: = rate of change of concentration per horizontal distance (mg/m2).
dx
Integrating the mass balance, Equation 12, over the length with the boundary conditions
shown in Equation 13, results in the following linear relationship in Equation 14.
CNO CNO,in [1 3 ]
CNO,in V
lnC kKd
NO,out Q 1
- [1 4 ]
CNO,in -CNO,out CNO,in -CNO,out k
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
164

where: Q = flow rate (m3/min),

V = volume of air space (m3).
Thus, from linear regression analysis by graphing lnCNOin/CNOout/(CNOin-CNOout) versus

(V/Q)/(CNOin-CNOout) the k and Kd fitting parameters can be determined from the graph;
where k is the reciprocal of the y-intercept and Kd is 1/(mk), where m is the slope of the
line. Furthermore, the linear relationship is evidence that the reaction is a gas-solid
reaction following the L-H mechanism, rather than a gas-phase reaction (Sleiman et al.,
2009).
The results plotted in Figure 7 were used to calculate and plot the
lnCNOin/CNOout/(CNOin-CNOout) versus (V/Q)/(CNOin-CNOout) as shown in Figure 8. As
shown in this figure, there is a strong linearity thus supporting the Langmuir-
Hinshelwood model and that the mass transfer is reaction-rate limited. In addition,
previous studies have shown that the percent degradation of NO is increased by
increasing the amount of TiO2 catalysis, also suggesting reaction-controlled (Hassan et
al., 2012).
From the regression results, the Langmuir-Hinshelwood constants were calculated as
k=0.11 mg/m3min and Kd=4.70 m3/mg. Compared to previous concrete kinetic studies
conducted in the Netherlands, the reaction rate is much lower and the adsorption
equilibrium constant is higher. Hunger et al. reported for concrete a reaction rate of
k=0.42 mg/m3 and adsorption equilibrium of Kd=2.00 m3/mg for 50% relative humidity
and 10 mW/cm2 (Hunger et al., 2010). The higher adsorption equilibrium rate could be a
result of the lower relative humidity used in this study thus having less competitive
adsorption. Meanwhile, the lower reaction rate could be from a chemical interaction
between the asphalt hydrocarbons and the NOx.
165
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

35
y = 1.9394x + 9.091
30
R = 0.9899
25
20
15
10
0
0 5 10 15
Figure 8. Determination of L-H constants for asphalt pavements.
4.2.3. Effect of Relative Humidity and Intensity on L-H Constants
From Figure 9, it is clear that relative humidity impacts both the L-H adsorption
equilibrium constant and the L-H reaction rate constant. The extent of the impact of
relative humidity is largely dependent upon the substrate material. Titanium dioxide can
be both hydrophobic and hydrophilic. Typically, and during irradiation, TiO2 undergoes
a hydrophilic effect. This promotes a water monolayer that inhibits pollutants to adsorb
on the photocatalyst active sites (Hunger et al., 2010). As seen in Figure 9, in general as
the relative humidity increases, the L-H adsorption equilibrium rate decreases
confirming this theory that the relative humidity competes with photocatalytic active
sites (Figure 9a).
Furthermore, the reduction in photocatalytic active sites also leads to lower L-H
reaction rates, which is illustrated in Figure 9b. The lower reaction rates are explained
by less active sites available on the TiO2 photocatalyst thus reducing the reaction rate. In
addition, due to the nonparallel trends illustrated in both Figures 9a and 9b, there seems
to be a slight interaction between the relative humidity and light intensity. For example
166
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

at the lowest light intensity, the reaction rate seems to have an optimum relative
humidity point.
5 2 .4
Kd (m3/mg) 4 1
3
0 .5
2
1
0
0 50 100
Relative Humidity (%)
( a)
2 .4
0.12 1
k (mg/m3min)
0 .1 0 .5
0.08
0.06
0.04
0.02
0
0 50 100
Relative Humidity (%)
(b )
Figure 9. L-H adsorption equilibrium constant (a) and L-H reaction rate (a) versus
humidity at 2.4, 1, and 0.5 mW/cm2 intensities.
Figure 10 illustrates that the light intensity also had an impact on both the L-H
adsorption equilibrium constant and the L-H reaction rate constant. This is contrary to
previous kinetic studies, which illustrated that the light intensity does not have any
significant impact on the equilibrium constant and only influences the reaction rate
constant (Hunger et al., 2010). Similarly to conclusions from Figure 9, both graphs in
Figure 10 support evidence of an interaction effect between the relative humidity and
intensity factors on both the L-H constants.
167
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

5 20
Kd (m3/mg)
4 48
3 78
2
1
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0
0 1 2 3
Intensity (mW/cm2)
( a)
0.12 20
0 .1 48
k (mg/m3min)
0.08 78
0.06
0.04
0.02
0
0 1 2 3
Intensity (mW/cm2)
(b )
Figure 10. L-H adsorption equilibrium constant (a) and L-H reaction rate (b) versus
light intensity at 20, 48, and 78% relative humidity.
Table 4 is a summary of the regression fits and the resulting L-H constants calculated
for each environmental scenario simulated. Under similar environmental conditions as
the Hunger et al. (2010) kinetic study for concrete, the L-H reaction rate was
significantly smaller at 0.03 mg/m3min compared to 0.42 mg/m3 and the L-H adsorption
equilibrium constant was relatively similar at 1.87 m3/mg compared to 2 m3/mg.
Therefore, a major reason for the lower NO reduction efficiencies in photocatalytic
asphalt pavements compared to concrete pavements is a result of the significantly lower
L-H reaction constants characteristics of photocatalytic asphalt pavements.
168

Table 4. Summary of L-H constants and coefficient of determination.
Run Intercept Kd K Coefficient of

Scenario Slope (m) (b ) (m3/mg) (mg/m3min) Determination (r2)
1 2 .1 7 7 .4 7 4 .6 8 0 .1 1 0 .9 8
2 3 .3 5 1 0 .7 1 3 .2 0 0 .0 9 0 .9 4
3 3 .8 7 1 6 .8 5 4 .3 5 0 .0 6 0 .9 6
4 5 .1 6 2 3 .0 8 4 .4 8 0 .0 4 0 .9 4
5 1 5 .4 8 2 8 .9 0 1 .8 7 0 .0 3 0 .9 8
6 7 .4 3 1 6 .6 5 2 .2 4 0 .0 6 0 .8 6
7 1 1 .5 2 3 3 .5 5 2 .9 1 0 .0 3 0 .9 8
8 2 7 .7 2 4 4 .7 7 1 .6 1 0 .0 2 0 .9 9
9 4 1 .8 0 8 7 .6 6 2 .1 0 0 .0 1 0 .9 6
5. Conclusions
Using data from a field study, the NO reduction was modeled using statistical regression
techniques. The coefficient of determination was 0.79 and 0.67 for a non-coated
pavement and photocatalytic pavement, respectively. To improve prediction, other
parameters may need to be included in the model and more sampling time is required.
As a result, using mathematical models may play an intermediate role in understanding
real-world photocatalytic pavements.
A kinetic study was completed for asphalt photocatalytic pavements being the first
step to creating a chemodynamic model to understand the significance of pollution
reduction from using photocatalytic pavements. The results of the study confirmed that
the NO mass transfer was controlled by the reaction rate. These results are similar to
previous photocatalytic kinetic studies for other substrates. Outcomes of the kinetic
study showed that the adsorption equilibrium of NO for photocatalytic asphalt
pavements was similar compared to concrete substrates; however, the reaction rate
constant was significantly lower. Consequently, the lower reduction efficiencies
observed in past studies for asphalt photocatalytic pavements as compared to concrete
photocatalytic pavements are a result of the lower L-H reaction rate constants.
The results of the parametric study identified that the relative humidity and light
intensity had a significant impact on NO reduction efficiencies. The impact of the
relative humidity was attributed to competition in adsorption sites and reduction of
169
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

active sites, evident by the negative correlation of Kd and k, respectively, with

increasing relative humidity. As for light intensity, there was a positive linear
correlation related to the NO reduction efficiencies. There was interaction between the
relative humidity and intensity variables. These results are important for future work to
develop a NO reaction rate model for photocatalytic pavements to understand the
significance of NO reductions in real world environments.
7. References
Beeldens, A. (2006, June). An environmental friendly solution for air purification and
self-cleaning effect: the application of TiO2 as photocatalyst in concrete.Belgian
Road Research Centre. Proceedings of Transport Research Arena, Europe - TRA,
Gteborg, Sweden.
Beeldens, A. (2008). Air purification by pavement blocks: final results of the research
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
at the BRRC. Transport Research Arena Europe, Ljubljana.

Bengtsson, N. and Castellote, M. (2010). Photocatalytic activity for NO degradation by
construction materials: parametric study and multivariable correlations. Journal of
Advance Oxidation Technologies, 13(3), 341349.
Cassar, L. (2004, May). Photocatalysis of cementitious materials: clean buildings and
clean air. MRS Bulletin, 14.
Diamanti, M.V., Ormellese, M. and Pedeferri, M. (2008). Characterization of
photocatalytic and superhydrophilic properties of mortars containing titanium
dioxide. Cement and Concrete Research, 38, 13491353.
doi:10.1016/j.cemconres.2008.07.003.
Dylla, H., Hassan, M., Schmitt, M., Rupnow, T. and Mohammad, L. (2011). Laboratory
investigation of mixed nitrogen dioxide (NO2) and nitrogen oxide (NO) gasses on
titanium dioxide photocatalytic efficiency in concrete pavements. Journal of
Materials in Civil Engineering, 23(7), 10871093. doi:10.1061/(ASCE)MT.1943-
5533.0000248.
Environmental Protection Agency (EPA). (2012). Nitrogen dioxide implementation
Programs and requirements for reducing oxides of nitrogen.
(http://www.epa.gov/air/nitrogenoxides/implement.html).
Fujishima, A., and K. Honda. (1972). Electrochemical Photolysis of Water at a
Semiconductor Electrode. Nature, 238, 3738. doi:10.1038/238037a0.
170

Fujishima, A., Rao, Tata N. and Tryk, Donald A. (2000). Titanium dioxide
photocatalysis. Journal of Photochemistry and Photobiology C: Photochemistry
Reviews, 1, 1-21. doi:10.1016/S1389-5567(00)00002-2.
Fujishima, Akira, and Zhang, Xintong. (2006). Titanium dioxide photocatalysis:
present situation and future approaches. C.R.Chimie, 9 750760.
doi:10.1016/j.crci.2005.02.055.
Hassan, M. M., Mohammad L., Asadi, S., Dylla, H., and Cooper, S. (2012).
Sustainable Photocatalytic Asphalt Pavements for Mitigation of Nitrogen Oxide
and Sulfur Dioxide Vehicle Emissions. ASCE: Journal of Materials in civil
engineering. Advance online publication. doi: 10.1061/(ASCE)MT.1943-
5533.0000613.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Health Effects Institute (HEI). (2010). Traffic-Related Air Pollution: A Critical Review
of the Literature on Emissions, Exposure, and Health Effects. (Special Report 17).
Hunger, M., Husken, G., and Brouwers, J. (2008). Photocatalysis applied to concrete
products, Part 1: Principles and test procedure. Materials Science, ZKG
International, 61, 7785.
Hunger, M., Husken, G., Brouwers, H.J.H. (2010). Photocatalytic degradation of air
pollutants From modeling to large scale application. Cement and Concrete
Research. 40, 313-320. doi:10.1016/j.cemconres.2009.09.013.
Japanese Industrial Standard (JIS). (2004). Fine ceramics (advanced ceramics,
advanced technical ceramics) Test method for air purification performance of
photocatalytic materials- Part 1: Removal of nitric oxide. JIS R 1701-1, 1-9.
Li, L. and Qian, C. (2009). A lab study of photo-catalytic oxidation and removal of
nitrogen oxides in vehicular emissions and its fieldwork on Nanjin No.3 bridge of
Yangtze River. Journal of Pavement Resource Technology, 2(5), 218-222.
Menz, F.C. (2002). The US experience with controlling motor vehicle pollution:
lessons for China. Int. J. Environment and Pollution, 18(1), 121.
Sharma, N., Chaudhry, K.K., and Chalapati Rao, C.V. (2004). Vehicular pollution
prediction modeling: A review of highway dispersion models. Transport Reviews,
24(4), 409435.
Sleiman, M., Conchon, P., Ferronato, C., and Chovelon, J.M. (2009). Photocatalytic
oxidation of toluene at indoor air levels (ppbv): Towards a better assessment of
conversion, reaction intermediates and mineralization. Applied Catalysis B:
Environmental, 86(3-4),159165. doi:10.1016/j.apcatb.2008.08.003.
171

Toma, F.-L.; Berger, L.M.; Jacquet, D.; Wicky, D.; Villaluenga, I; Miguel, Y.R.;
Lindelov, J.S. (2009). Comparative study on the photocatalytic behavior of
titanium oxide thermal sprayed coating from powders and suspensions. Surface
and Coatings Technology, 203(15), 21502156.
Vallius, M., Ruuskanen, J., and Pekkanen, J. (2008). Comparison of multivariate source
apportionment of urban PM2.5 with chemical mass closure. Boreal Environment
Research, 13, 347358.
Wang, Shaobin; Ang, H.M.; and Tade, Moses O. (2007). Volatile organic compounds
in indoor environment and photocatalytic oxidation: State of the art. Environment
International, 33, 694705. doi:10.1016/j.envint.2007.02.011.
Yu, Q.L. and Brouwers, H.J.H., (2009). Indoor air purification using heterogeneous
photocatalytic oxidation. Part I: Experimental study. Applied Catalysis B:
Environmental, 92, 454461. doi:10.1016/j.apcatb.2009.09.004.
Zeng, F., Shi, G-L., Li, X., Feng, Y-C., Bi, X-H., Wu, J-H., Xue, Y-H. (2010).
Application of a combined model to study the source apportionment of PM10 in
Taiyuan, China. Aerosol and Air Quality Research, 10, 177184.
doi:10.4209/aaqr.2009.09.0058.
Zhao, J. and Yang, X. (2003). Photocatalytic oxidation of indoor air purification: a
literature review. Building and Environment, 38, 645654. doi: 10.1016/S0360-
1323(02)00212-3.
8. Discussion
DR. HONGBIN XIE: I have a question, not exactly from this study but it is relevant.
What is the durability of this coating in the real world, in a field application particularly?
Like how often you will be required to repaint it?
MS. HEATHER DYLLA: We havent done complete long-term durability testing. I
know there are future prospects of projects that are going to do that at LTRC. But we did
test it differently, like the loaded wheel tester, we did the rheological test on the asphalt
pavement. But as far as durability on the environmental side, on how much reduction
can you expect, we have not gotten there yet. It has just been so difficult to test. And Dr.
Hassan wants to add something.
PROFESSOR MARWA HASSAN: From a durability perspective, we did do an initial
study that is currently under review by ASCE, so it is going to be published. For that
field study, we measured the deterioration or the reduction of photocatalytic efficiency
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
over time. Then we did a regression analysis prediction and it is very preliminary.
Results suggest that the service life of TiO2 coating in concrete is between 6 to 11
172

months and between 10 to 16 months in asphalt pavements. But that is very preliminary
because of the varying weather conditions and all the variability that Heather has been
talking about. To get an accurate durability, what we need is to test under accelerated
pavement testing facility conditions, and thats the next step that were going after.
DR. XIE: Okay, thank you.
PROF. HASSAN: No problem.
DR. JEAN-PASCAL PLANCHE: I do not have a question. It is more of a comment for
information. This type of technology has been used in Europe for about 10 years. It
started, I believe, in the concrete industry, to make walls with the painting to reduce
NOx. There is also a French contractor who put together an asphalt and they have made
a field trial, maybe five years ago or something like that. I dont know exactly the
outcome, but just to let you know that this type of thing is already being applied to some
extent. I think it is limited, but I think its used more in the industry for petrochemical
plants or refineries to capture and reduce some exhaust.
MS. DYLLA: Yes. Thanks for the comment. That is true. I didnt go into all the
literature review. Most of our literature review comes from Europe Europe and Japan.
Its far advanced, and they have lots of different products.
DR. PLANCHE: There are some IR patents behind it.
MS. DYLLA: Yes.
PROFESSOR HUSSAIN BAHIA: Very nice study. My question is about the purpose of
spreading it in the wheel path. If there is a durability issue, can we think about putting it
on the shoulders or buildings along the highway? Is the proximity to the source
important? If not, do we need to spray it where the tires could actually erode it?
MS. DYLLA: The reason why we initially put it on the wheel path is we were looking at
different air pollution modeling studies that show the pollution dispersion and it is higher
concentration the closer you are to the source. But, of course, it changes because some
vehicles have their emission spouts up, some have it down. It is very difficult right no w
to say where the best place to put it is. But initially it was the thought of higher
concentrations or at the source or on the pavement level. We have confirmed that by
measuring pavement level concentrations. Weve hit high spikes around 1000 ppb. Once
it gets to like walls, it has been dispersed some, so that is the main reason.
PROF. BAHIA: Thank you.
MS. DYLLA: You are welcome.
PROF. HASSAN: Just to add a little bit about durability. Where to put this
photocatalytic coating, whether it is asphalt or on concrete is not a question that we can
answer in a day. It has to be correlated with urban planning. So you are really looking
173

--`,`,,,``,,``

for interstate or areas that have high traffic volumes so the concentrations of pollution
are high. You are also looking for urban areas that have tall buildings that prevent the
dispersion of the pollutant, which has lots of health impacts. But once you select these
areas, you also want to have it in areas that are away from water bodies so that you are
not producing, for example, nitrates and then worrying about another environmental
issue like eutrophication. Once you decide where these areas are, it can be a combination
of what is on the wheel path and what is on the shoulder of the pavement and maybe the
retaining walls too. What we are saying is we have tested different substrates. Heather is
talking about the field study that chooses the least durable method, which is the spray
coat because, really, you are putting on a spray coat. You can mix it with the asphalt and
the concrete mix, which would actually, if you can create an ultrathin layer or some kind
of an overlay, have a higher durability. These are all expectations that need to be tested
via research and then quantify the long-term durability of it. But it is worth studying
because there is a lot more potential to purifying the outdoors when it is closer to the
pollutant source versus not. Thank you.
DR. ERVIN DUKATZ: Again, compliments on a very fine paper and update of your
research. But one question I have for you is how close is this to practical
implementation? How contractor friendly or DOT friendly is this process and how
available are the materials that if a city, county or state wanted to implement this, ho w
close are you ready to do this?
MS. DYLLA: As far the distributor truck, it was actually really easy. I mean that is no
problem, I think, as far as application for a spray coat. I think the real questions that
remain are how efficient is it in the field? Like if we are looking at sustainability, is it
going to actually have a net benefit? Really understanding that and where do we
implement it? Can we implement it? Yes, we can. Its being done. Roadways? Theyve
done a few roadways, but for sure, if you know the statues in Minnesota where the
bridge fell? They're coated with titanium dioxide. The Super Dome is coated with
titanium dioxide. Not for environmental purposes to clean the air. but it is to keep the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
concrete white.
DR. DUKATZ: Okay. Then the other practical aspect of it is, is the network set up so
that if people try specifying this, are they going to be able to get the titanium dioxide to
actually put in paving quantities? You know, its one thing to do a 200 foot section or
maybe a wall of a building, but when you're paving 10 lane miles or 30 lane miles.
MS. DYLLA: It would definitely create a market. Right now, we have one contractor in
the U.S. that we work with. Most of the manufacturers of titanium dioxide are overseas.
DR. DUKATZ: Thank you.
PROF. HASSAN: To answer your question, what is titanium dioxide? Titanium dioxide
is a very common material that is available. It is available in all kinds of white paint. The
difference is if you use what you put in white paint, it is not photocatalytic because it is
174

made out of rutile crystals. What makes it photocatalytic is if you change the crystal
structure so it is anatase. Or you can also get photocatalytic efficiency if you mix it with
70% anatase, 30% rutile or something of that combination. So it is about creating more
manufacturers that are producing the photocatalytic grade. Also whether you are making
it a nanosize or an ultrafine size. There are major companies worldwide that are
producing this at a photocatalytic grade, although that production line is limited just
because there is no market for it. They can produce it. They have the capacity; just they
are selling it to very few people. The biggest worldwide manufacturer is known as
Nagosa and then the second is called Cristal Global. And all of these companies,
although they are international, found in Europe, they have offices in the U.S. and we
have dealt with them. There is a local company and a local manufacturer that is present
in the U.S. and the company is known as PURETi, and the CEO is Glen Finkel. That is
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
the manufacturer of the spray coat application. So we can do it. In addition, I have talked
to professors from the University of Minnesota who are working in mineralogy, and they
were talking about the amount of titanium dioxide that they are mining that they have no
market for. So once we create that market, it will actually create more job opportunities
and more manufacturing opportunities, and its a win-win situation for the economy.
DR. RANDY WEST: Im going to ask a question. It may be a really stupid question
because any subject matter that deals with chemistry is certainly out of my expertise. But
if it is a stupid question, it wont be the first time I have asked one. It is really about the
conversion of NO to nitrates and conversion of one source of pollution to another source
of pollution. So could you comment about the runoff issue of nitrates into ground water
and those kinds of things?
MS. DYLLA: We have detected that and what has been actually reported is it is lo w
quantities that actually go in, and there is no eutrophication effect extra than what is
normally out there. It is really small. Smaller than what you would get in drinking water
as for the amount of nitrates. But it is something that is a precaution that needs to be
taken and that is very current. Good question.
PROFESSOR LOUAY MOHAMMAD: This is a comment also to Erv to add to what
Heather had stated. A tack coat distributor type of truck can be used to spray a uniform
coverage of this material. The next step is that we are planning to use typical
construction equipment to spray TiO2 on test lanes at our ALF facility. Thus, we believe
that the proposed technology is practical and can be implemented.
MR. CHRISTOPHER DEDENE: This is kind of outside the scope of your project, but
what are your thoughts about capturing the pollution at the source on your car versus
letting it escape the car and then trying to get it on the road?
MS. DYLLA: I actually did an economic study on that for one of my classes,
Environmental Economics, and it actually could be, if you were going to do an economic
study, less efficient. If you put it on the source of your car, everyone has to pay for that
175

service, whereas we are targeting where the pollutions are high. Because if you drive
your car in the middle of nowhere where it is not coming in contact with anybody, it is
not a hazard. It is being naturally dispersed into the environment and it is not as big of an
issue as when you are at higher concentrations, higher congestion, with a higher
population nearby. So were just basing off of that principle. Plus, it would take about 20
years or so to get that implemented so the whole populations fleet of vehicles are using
it and to force that everybody drives that car because if you put it on new vehicles, not
everyones buying new vehicles every year, so that transition is slower. So this gives us
a more immediate fix to the problem for the pollution that is already being emitted right
now. But that is always an option too.
MR. DEDENE: Thank you.
MS. DYLLA: You are welcome.
176
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Laboratory Conditioning Protocols for
Warm-Mix Asphalt
Fan Yina*, Lorena Garcia Cucalona, Amy Epps Martina,
Edith Arambulab, Arif Chowdhuryb, and Eun Sug Parkb
a
Zachry Department of Civil Engineering, Texas A&M University, College Station,
TX 77843
b
Texas A&M Transportation Institute, College Station, TX 77843-3136
ABSTRACT: Economic, environmental, and engineering benefits have led to the rapid
implementation of Warm-Mix Asphalt (WMA) across the world during the past decade. While
WMA technologies have been successfully utilized as a paving material, standard mix design
protocols remain under development. A study of the effect of laboratory conditioning protocols
(prior to compaction) on the performance of HMA (Hot-Mix Asphalt) and WMA mixtures made
with chemical additives, wax additives, or a foaming process was performed. Resilient Modulus
(MR), Dynamic Shear Rheometer (DSR), and image analysis techniques were used to evaluate the
properties of the extracted binders and mixtures conditioned with different protocols. Laboratory
mixed laboratory compacted (LMLC) specimens conditioned for 2 hours at 240F (116C) for
WMA and 275F (135C) for HMA had stiffnesses similar to those of cores collected during the
early life of field pavements. For off-site plant mixed laboratory compacted (PMLC) specimens,
different conditioning protocols are recommended to simulate the stiffness of plant mixed field
compacted (PMFC) cores at construction: reheat to 240F (116C) for WMA with additives and
reheat to 275F (135C) for HMA and foamed WMA.
KEYWORDS: warm-mix asphalt, hot-mix asphalt, laboratory conditioning, stiffness
The oral presentation was made by Dr. Yin.
177
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

YIN, GARCIA CUCALON, EPPS MARTIN, ARAMBULA, CHOW DHURY, PARK
1. Introduction
Hot-Mix Asphalt (HMA) is a well-established paving material with proven

performance on 94% of the more than 2.5 million miles (4.0 million km) of paved roads
in the United States (FHWA, 2008; NAPA,. 2010). It is produced by mixing asphalt
binder and aggregate at an elevated temperature in either batch or drum mix plants and
then is placed by compacting the mix at temperatures ranging from 275F (135C) to
325F (163C) (Kuennen, 2004; Newcomb, 2005). These high temperatures are
necessary to ensure complete drying of the aggregate, subsequent bonding with the
binder and workability for adequate handling and compaction. All of these processes
contribute substantially to good pavement performance. Conversely, these high
temperatures are linked to increased emissions and fumes from HMA plants (Stroup-
Gardiner et al., 2005) and considerable energy consumption in drying the aggregate and
heating all materials prior to mixing.
Economic, environmental, and engineering benefits motivate the reduction of
production and placement temperatures for asphalt concrete paving materials. The latest
technology that has been rapidly adopted for this purpose is Warm-Mix Asphalt (WMA),
which is defined as an asphalt concrete paving material produced and placed at
temperatures approximately 50F (28C) cooler than those used for HMA. It was first
introduced in Europe in the mid 1990s as a way to reduce greenhouse gas emissions, and
then transferred to the United States in the early 2000s largely through the effort of the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
National Asphalt Pavement Association (NAPA).

WMA technologies offer a number of benefits, including decreased energy
consumption, reduced emissions and fumes at the plant, improved working conditions at
the construction site, extended haul distances, longer pavement construction season,
improved workability and compactability, reduced aging, and better resistance to
cracking and ravelling. However, there are several barriers to the widespread
implementation of WMA, such as inadequate coating of aggregates with binder at lower
production temperature, imprecise correlation between laboratory and field performance,
incomplete mix design procedures, and a wide variety of different types of WMA
technologies. Previous studies indicate that laboratory conditioning of loose mixes prior
to compaction to simulate short-term aging during production and construction have a
significant effect on mixture performance and, specifically, conditioning protocols with
longer time at higher temperature substantially increase the mixture stiffness. However,
there is no standard laboratory conditioning procedure for preparing WMA laboratory
mixed and laboratory compacted (LMLC) and off-site plant mixed and laboratory
compacted (PMLC) specimens for performance tests.
The goal of this study was to recommend conditioning protocols consisting of a
combination of time and temperature that produce WMA LMLC and PMLC specimens
178

Laboratory Conditioning Protocols for Warm-Mix Asphalt
calibrated to field cores at construction. Based on previous experience, several

conditioning protocols for LMLC and PMLC specimens were selected, and their effects
on mixture properties were evaluated. Mixture stiffness evaluated in a dry condition
using the Resilient Modulus (MR) test (modified ASTM D7369) was the main parameter
for selecting a conditioning protocol to simulate the stiffness of a pavement in its early
life. Extracted binder stiffness and aggregate orientation of field cores and on-site PMLC
specimens were evaluated using the Dynamic Shear Rheometer (DSR) (AASHTO T
315) and image analysis techniques, respectively.
This paper first provides a brief review of written documentation on previous
research on laboratory conditioning protocols for asphalt mixtures. Next, the laboratory
experimental design is described, followed by test results and data analysis. Finally,
several conclusions are made based on test results, and recommendations for further
research are proposed.
2. Background
The standard practice for laboratory mix design of asphalt concrete paving materials
is to simulate the binder absorption and aging that occurs during construction by short-
term oven aging (STOA) or conditioning the loose mix prior to compaction for a
specified time and temperature. For HMA, the recommended procedure when preparing
samples for volumetric mix design is two hours at compaction temperature (Tc), while
for performance testing it is 4 hours at 275F (135C) (AASHTO R30). In the past few
years, a number of studies have been conducted to evaluate the effect of different
conditioning protocols on the performance of WMA. In general, the majority of these
studies have concluded that an increase in laboratory conditioning temperature and/or
time may reduce the difference in performance between WMA and HMA.
A recent study by Estakhri et al. (2010) evaluated the effect of three conditioning
protocols on WMA mixtures prepared with Evotherm DATTM: 2 hours at 220F (104C),
2 hours at 275F (135C), and 4 hours at 275F (135C). Performance was evaluated
using the Hamburg Wheel Tracking Test (HWTT) (AASHTO T 324) and compared
against HWTT results of HMA cured at 250F (121C). In addition, WMA mixtures
prepared with Advera and Sasobit, conditioned for 2 hours at 220F (104C) and 4
hours at 275F (135C), were also tested and compared against the results of HMA
conditioned for 2 hours at 250F (121C). The results for WMA Evotherm DATTM
showed that the number of passes to generate a 0.5 in. (12.5 mm) rut depth increased
with higher conditioning temperature and longer conditioning time, and that the mixture
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
conditioned for 4 hours at 275F (135C) showed equivalent performance to the control
HMA conditioned at 250F (121C). The HMA showed only a slight decrease in the
179

number of passes to a 0.5 in. (12.5 mm) rut depth when conditioned at 250F (121C)
versus 275F (135C). However, the change for the WMA mixtures prepared with the
three different WMA technologies was significant for the two conditioning temperatures.
The number of passes for all of the WMA mixtures was similar when conditioned at
220F (104C), and all three mixtures sustained much higher numbers of passes to a 0.5
in. (12.5 mm) rut depth when conditioned at 275F (135C). Based on these
observations, a recommendation to condition WMA for four hours at 275F (135C) was
made and was incorporated in the WMA specifications for the Texas Department of
Transportation (TxDOT).
Another study by Estakhri (2012) evaluated the effect of curing time and temperature
on HMA and WMA properties using HWTT (AASHTO T 324) and the Overlay Test
(OT). Two field projects with three different WMA mixtures were included in the study.
HWTT results indicated equivalent rutting characteristics between WMA and HMA
when both types of mixtures were cured for 2 hours at the standard HMA compaction
temperature of 275F (135C). In addition, the WMA mixtures conditioned with
increased curing time and temperature had significantly higher resistance to rutting in the
HWTT tests. Mixture performance measured in the OT was also sensitive to curing time
and temperature. Specifically, a significant reduction in mixture cracking resistance was
shown by both HMA and WMA when the curing time increased from two to four hours.
Based on results from both tests, it was concluded that curing time and temperature have
a significant effect on mixture stiffness.
A study by Al-Qadi et al. (2012) focused on the short term characterization and
performance of WMA with the following technologies: Evotherm DATTM, Sasobit, and
Foaming. The effect of loose mix reheating on mixture properties was evaluated on the
basis of comparison of reheated PMLC specimens versus on-site PMLC specimens.
Laboratory tests used in the study included dynamic modulus (AASHTO TP79-10), flo w
number (AASHTO TP 62-03), HWTT (AASHTO T 324), indirect tensile (IDT) creep
and strength (AASHTO T 322-07), and semi-circular bending fracture tests (ASTM
D7313-07a). Test results indicated that the effect of reheating on the mixture complex
modulus, tensile strength, and rutting resistance was significant. In addition, an elevated
reheating temperature had a more significant effect on test results.
In a separate study, the University of California Pavement Research Center used the
conditioning protocol of 4 hours at Tc for preparing LMLC specimens as part of a
comprehensive accelerated pavement testing (APT) program (Jones, 2011). Results
showed no difference in rut depth between WMA and the control HMA after HWTT
(AASHTO T 324) and full-scale accelerated load tests (using the heavy vehicle
simulator) with this conditioning protocol. However, WMA without conditioning prior
to compaction was more susceptible to rutting. These results confirmed that additional
180
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
laboratory conditioning significantly increases the stiffness of WMA such that
equivalent performance to HMA is eventually achieved.
Research performed at the University of Kentucky (Clements, 2011) explored the
differences in the flow number test (AASHTO TP 62-03) and disc shaped compact
tension test (ASTM D7313-07a) performed on HMA and WMA conditioned at several
intervals (0.5 h, 2 h, 4 h, and 8 h) at 275F (135C) and 240F (114C), respectively.
Since no differences were observed between the performance of HMA and WMA at the
various conditioning times, the author proposed considering WMA and HMA as
equivalent with respect to conditioning time. A complementary study by Clements
(2012) evaluated the performance of WMA as compared to HMA with different
conditioning times prior to compaction. Evotherm 3G was used as the WMA additive,
and four aging times for loose mix were included in the study (0.5 hr, 2 hr, 4 hr, and 8
hr). WMA performances were evaluated and compared to those of HMA on the basis of
dynamic modulus (AASHTO TP 79-10), flow number (AASHTO TP 62-03), HWTT
(AASHTO T 324), and disk-shaped compact tension (ASTM D7313-07a) test results.
Test results indicated that WMA had lower stiffness and higher susceptibility to rutting
than HMA, yet greater fracture energy when tested at 28F (-2C). The difference in
mixture properties was attributed to the lower production temperature of WMA and the
incorporation of chemical additives in the mixture. In addition, it was shown in the study
that, generally, increasing the aging period corresponded with an increase in mixture
stiffness and rutting resistance for both HMA and WMA.
The recently completed National Cooperative Highway Research Project (NCHRP)
on mix design practices for WMA (Bonaquist, 2011) recommended a conditioning
protocol for WMA of 2 hours at Tc for both volumetric mix design and performance
testing as listed in the draft appendix to AASHTO R 35. This conditioning protocol was
selected based on comparisons of maximum specific gravity (AASHTO T 209) and IDT
strength (AASHTO T 283) of LMLC specimens subjected to the mentioned conditioning
protocol versus the results obtained for plant mixed field compacted (PMFC) cores. The
specific gravity comparison showed equivalent maximum theoretical density for LMLC
specimens and PMFC cores, indicating the same binder absorption level. The difference
in IDT strength between LMLC specimens and PMFC cores was also insignificant based
on a paired t-test comparison with a 95% confidence interval. In addition, further
research was recommended to develop a two-step WMA conditioning procedure for the
evaluation of moisture susceptibility and rutting resistance, similar to the conditioning
protocol applied to HMA. The first step would be the conditioning for 2 hours at Tc to
simulate binder absorption and aging during construction, and the second step would
consist of an extended conditioning time at a representative high in-service temperature
but no longer than 16 hours (Bonaquist, 2011).
181

In general, the majority of studies that have been performed to understand the effect
of conditioning prior to compaction on the performance of WMA have concluded that an
increase in laboratory conditioning temperature and/or time may reduce the difference in
performance between WMA and HMA. However, no standard conditioning protocol for
WMA has been established to date.
3. Materials and Testing
This section briefly introduces two field projects in Iowa and Texas from which raw
materials, loose plant mix, and PMFC cores were acquired. The procedures followed to
fabricate LMLC and PMLC specimens are described, and the experimental design of the
study is presented.
3.1. Materials
Materials used in this study are from two field projects located in Iowa and Texas. They
were selected considering a wide spectrum of materials and field conditions: climate,
aggregate type, binder source, inclusion of recycled materials, and WMA technologies.
The climates at the Iowa and Texas field projects are classified as Wet-Freeze and Wet-
No Freeze, respectively, as defined by the Strategic Highway Research Program (SHRP)
Long Term Pavement Performance (LTPP) Environmental Zones (Hadley, 1994).
The Iowa field project is near Adams County on U.S. Route 34. Five fractions of
quartzite and limestone aggregates and river sands from four different producers and
Recycled Asphalt Pavement (RAP) were used. The combined aggregate gradation is
presented in Figure 1. The type of asphalt binder used in this field project is a Superpave
performance-graded (PG) 58-28 binder with a specific gravity of 1.03. The design
optimum binder content was 5.4% (by weight of the total mixture). Evotherm 3G and
Sasobit were selected as the WMA technologies in this study. Evotherm 3G is a
combination of surfactants, waxes, processing aids, polymers, acids, and other materials
that may provide the reduction of frictional forces between the binder and aggregate.
Sasobit is a crystalline, long chain aliphatic polyethylene hydrocarbon, identical to
paraffin waxes that are found in crude oil, except that it has a higher molecular weight.
Due to its ability to lower the viscosity of the binder at high temperatures, Sasobit
improves the binder flow during the mixing process and during laydown operations.
Both WMA additives were blended at 0.4% by weight of binder at the plant. The
transportation time and hauling distance for the project were approximately 15 miles and
25 minutes, respectively.
182
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The Texas field project is on FM 973, near the Austin Bergstrom International
Airport. Three types of aggregates and two sands were used in this field project. The
combined aggregate gradation is presented in Figure 1. A PG 70-22 binder with a
specific gravity of 1.03 was used in this project, and the design optimum binder content
was 5.2% (by weight of the total mixture). Evotherm DATTM and a foaming process
were used as the WMA technologies in this field project. Evotherm DATTM has been
designed to enhance coating, adhesion, and workability at lower production
temperatures. In order to treat the binder with this chemical additive, the binder was
heated to the mixing temperature (Tm) and the additive was blended at 5% by weight of
binder. Foamed binder was produced on-site by injecting 5% water and air into the hot
binder inside a special expansion chamber. In the laboratory, a foaming device that
simulates the air-atomized mixing at the plant was used to produce foamed
binder/mixtures with 5% water. The transportation time and hauling distance for the
project were approximately 25 miles and 40 minutes, respectively.
Compaction temperatures used in the Iowa and Texas field projects are summarized
in Table 1. During construction, on-site PMLC specimens were fabricated in a mobile
laboratory at the production plant. Raw materials, loose plant mix, and field cores were
also acquired and transported to the Texas A&M Transportation Institute McNew
Laboratory for further processing and specimen fabrication.
100%
90%
Cumulative Passing (%)
80%
70%
60%
50%
40%
Iowa
30%
20%
T ex as
10%
0%
#200 #100 #50 #30 #16 #8 #4 3/8 1/2 3/4"
Sieve Size
Figure 1. Combined aggregate gradation for the Iowa and Texas field projects.
183
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 1. Summary of compaction temperatures used in the Iowa

and Texas field projects.
Specimen Type
Location and
Environmental Mixture Type On-Site
Off-Site
Condition PMFC PMLC 1- LMLC
PMLC
2h
HM A 295-300F 295-300F 295F 295F
Iowa WMA with
240-248F 240-248F 240F 240F
(Wet, Freeze) Evotherm 3G
WMA with
235-240F 235-240F 240F 240F
Sasobit
HM A 270-285F 275F 275F 275F
Texas
WMA with
(Wet, No- 230-235F 225F 240F 240F
Evotherm DATTM
Freeze)
WMA with
240-250F 250F 235F 275F
Foaming Process
h: hour(s)
3.2. Specimen Fabrication

To fabricate LMLC specimens, aggregates and binder were heated to the specified Tm
independently and then mixed with a portable mixer. Afterwards, loose mixtures were
conditioned in the oven following a specific conditioning protocol prior to compaction
with the Superpave Gyratory Compactor (SGC). Trial specimens were fabricated to
assure specimens were obtained with air void contents (AV) of 70.5%. In total, 180
LMLC specimens with 70.5% AV were fabricated for the Iowa and Texas field projects
that included six mixtures and five laboratory conditioning protocols described
subsequently. Most LMLC specimens were tested to determine MR approximately 2 to 3
weeks after fabrication.
To fabricate off-site PMLC specimens at the TTI laboratory, loose plant mixes were
taken out of buckets and reheated in an oven to the specified conditioning temperature.
The temperature of the loose mix was monitored using a digital thermometer every 15-
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
minutes after being reheated in the oven for 1 hour. The temperature of the loose mix
was averaged from measurements at five random locations. Generally, the conditioning
184

temperature was usually achieved after 2-3 hours depending on the mixture. After being
reheated to Tc, loose mixes were further conditioned in an oven for a controlled period of
time following the conditioning protocol prior to compaction. A total of 144 off-site
PMLC specimens were fabricated for the two field projects that included six mixtures
and four laboratory conditioning protocols described subsequently. Loose mixes fro m
the Iowa field project were stored for 1 to 2 months and those from the Texas field
project were stored for 3 to 4 months prior to being fabricated. Most off-site PMLC
specimens were tested to determine MR approximately 2 to 3 weeks after fabrication.
For the Iowa field project, PMFC cores were obtained at construction and after six
months in service. To fabricate on-site PMLC specimens, loose mixes were taken from
the trucks before leaving the plant and maintained in the oven for 1-2 hours at the
specified temperature prior to compaction. Therefore, 18 PMFC cores and nine on-site
PMLC specimens were tested in this study. The placement of pavement sections in the
Texas field project was completed in January 2012; therefore, only field cores at
construction were included in this portion of the study. The effect of storage time of the
loose mix prior to compaction was evaluated as a mini experiment, therefore, on-site
PMLC specimens for the Texas field project were maintained in the oven for 0-1 hour
and 1-2 hours at the specified temperature before compaction, and then tested with MR.
Overall, 9 PMFC cores and 18 on-site PMLC specimens from the Texas field project
were tested. PMFC cores and on-site PMLC specimens from both field projects were
tested to determine MR after approximately 1 month and 2 months, respectively, in
storage.
3.3. Experimental Design
Based on previous experience with laboratory tests in evaluating asphalt mixture

stiffness, a non-destructive test was selected to quantify the stiffness of mixtures
conditioned with different protocols. The non-destructive test of choice was MR, which
is cost effective and able to provide an accurate indicator of mixture stiffness in the dry
condition. A minimum of three replicate specimens were utilized, and each replicate was
tested twice (i.e., by rotating the specimen 90 after the first measurement).
As shown in Table 2 and Table 3, five different conditioning protocols were selected
for LMLC specimens prior to compaction and four different ones were applied to off-site
PMLC specimens after reheating to the specified conditioning temperature, together with
one conditioning protocol for on-site PMLC specimens. For LMLC specimens, the
conditioning protocol of 2 hours at Tc for was proposed since it was recommended in the
recently completed NCHRP Project 9-43, and 4 hours at 275F (135C) was proposed
because it is the current standard in the state of Texas. The comprehensive conditioning
185
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

protocol of 2 hours at Tc followed by 16 hours at 140F (60C) and 2 hours at Tc was

proposed during a WMA workshop (Harrigan, 2012) held on May 13, 2011, in Irvine,
California. The other two protocols used were derived by combining common
conditioning temperatures and times. For off-site PMLC specimens, the conditioning
protocol of reheating to Tc was proposed as the least amount of conditioning
time/temperature possible prior to compaction. Additionally, three protocols proposed
for LMLC specimens were also used to prepare off-site PMLC specimens. Volumetrics
of LMLC specimens and on-site PMLC specimens were calculated and compared in
terms of theoretical maximum specific gravity (AASHTO T 209), binder absorption and
film thickness (Saskatchewan Test Procedure STP 204-19). All specimens were tested to
determine MR at a constant temperature of 77F (25C).
Field cores at construction and on-site PMLC specimens were expected to have
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
similar stiffnesses as they experienced approximately the same level of binder aging.
However, their performance in MR tests was significantly different as described
subsequently and thus binder was extracted and recovered from these specimens to
measure the difference in binder stiffness with the DSR test. In addition, images were
acquired from the same specimens through a novel method to evaluate the effect on
mixture stiffness of aggregate orientation produced by different compaction methods.
Finally, the effect of total AV on the stiffness of the specimens was also evaluated.
Table 2. Laboratory conditioning test plan for LMLC specimens.
Laboratory Conditioning Protocols

Location and
2h@Tc +
Environmental Mixture Type 2h@ 4h@ 2h@ 4h@
16h@140F
Condition Tc Tc 275F 275F
+ 2h@Tc
HM A X X X X X
WMA with
Iowa X X X X X
Evotherm 3G
(Wet, Freeze)
WMA with
X X X X X
Sasobit
HM A X X X X X
Texas WMA with
X X X X X
(Wet, No- Evotherm DATTM
Freeze) WMA with
X X X X X
Foaming Process
h: hour(s)
186

Table 3. Laboratory conditioning test plan for PMLC specimens.
Laboratory Conditioning Protocols

On-Site
Off-Site PMLC
Location and PMLC
Mixture
Environmental Reheat + Reheat
Type Reheat
Condition Reheat 16h@140 +
1-2h@Tc +
to Tc F + 4h@27
2h@Tc
2h@Tc 5F
HM A X X X X X
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
WMA with
Iowa Evotherm X X X X X
(Wet, Freeze) 3G
WMA with
X X X X X
Sasobit
HM A X X X X X
WMA with
Texas Evotherm X X X X X
(Wet, No- DATTM
Freeze) WMA with
Foaming X X X X X
Process
h: hour(s)
4. Laboratory Test Results and Data Analysis
4.1. Mixture Volumetrics
Table 4 presents the comparison of volumetrics of LMLC and on-site PMLC specimens
maintained in the oven for 1-2 hours at the specified Tc from the Iowa and Texas field
projects in terms of Rice specific gravity (Gmm), percentage of absorbed binder (Pba), and
effective binder film thickness (FT). The last two parameters were calculated based on
aggregate gradation, binder content and Gmm values.
187

Table 4. Mixture Volumetrics for the Iowa and Texas Field Projects.
Specimen Conditioning P ba
Location Mixture Type G mm FT (m)
Type Protocol (%)
LMLC 2h@275F 2 .4 1 5 0 .8 2 1 3 .2
HM A On-Site
1-2h@Tc 2 .4 4 3 1 .3 2 1 1 .9
PMLC
LMLC 2h@275F 2 .4 0 0 0 .5 3 1 4 .0
WMA with
Iowa On-Site
Evotherm 3G 1-2h@Tc 2 .4 3 4 1 .1 7 1 2 .3
PMLC
LMLC 2h@275F 2 .3 7 4 0 .0 4 1 5 .3
WMA with
Sasobit On-Site
1-2h@Tc 2 .4 3 8 1 .2 4 1 2 .1
PMLC
LMLC 2h@275F 2 .3 9 7 0 .1 0 1 2 .5
HM A On-Site
1-2h@Tc 2 .4 2 0 0 .5 3 1 1 .5
PMLC
WMA with LMLC 2h@275F 2 .3 9 9 0 .1 3 1 2 .4
Texas Evotherm On-Site
DATTM 1-2h@Tc 2 .4 0 8 0 .3 0 1 2 .0
PMLC
WMA with LMLC 2h@275F 2 .4 0 7 0 .2 8 1 2 .1
Foaming On-Site
Process 1-2h@Tc 2 .4 0 0 0 .1 5 1 2 .4
PMLC
The comparison among different specimen types shows that all LMLC specimens
(except for WMA with foaming process from the Texas field project) had lower Gmm,
lower Pba, and higher FT than corresponding on-site PMLC specimens. These results
reveal that on-site PMLC specimens experienced more conditioning prior to compaction.
Additionally, all WMA mixtures from the Iowa field project had lower Gmm and Pba and
larger FT than the corresponding control HMA for a given specimen type, which might
be caused by lower production temperature for WMA. A similar trend was observed for
on-site PMLC specimens from the Texas field project while volumetrics of HMA and
WMA LMLC specimens were opposite. In the case where lower Pba for WMA was
observed, decreased bonding strength between the aggregates and the binders could
potentially affect the interface of these two materials, making WMA more susceptible to
moisture damage. Differences in volumetrics were also evident for WMA with different
technologies.
188
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

4.2. Resilient Modulus (MR)

The MR test is conducted through repetitive applications of compressive loads in a
haversine waveform along a vertical diametral plane of cylindrical asphalt concrete
specimens. The resulting horizontal deformations of the specimen are measured by two
linear variable differential transformers (LVDTs) aligned along the horizontal diametral
plane. MR of the specimen is calculated based on vertical load, horizontal deformation,
and the asphalt mixture Poissons ratio at the test temperature. The MR test equipment
used to perform the measurements is shown in Figure 2. LMLC and PMLC specimens
with different conditioning protocols, and PMFC cores were tested to determine MR in
accordance with the current ASTM D7369 with a modification consisting of replacing
the on-specimen LVDTs with LVDTs aligned along the horizontal diametral plane (i.e.,
gauge length as a fraction of diameter of the specimen = 1.00).
( a) (b ) ( c)
Figure 2. MR test equipment; (a) Loading frame and data acquisition system, (b)
Specimen with mounted LVDTs, (c) Specimen setup in loading frame.
4.2.1. Laboratory Conditioning Protocols for LMLC Specimens

Figure 3 and Figure 4 present the MR results of PMFC cores and LMLC specimens fro m
the Iowa and Texas field projects, respectively. In each graph, PMFC cores are presented
on the left side of the figure and the LMLC specimens with different conditioning
protocols are shown on the right side of the figure. Each bar in these figures represents
the average value of three replicate specimens and the error bars represent one
standard deviation from the average value.
As illustrated in Figure 3 for the Iowa field project, the stiffness of HMA and WMA
with Evotherm 3G PMFC cores increased slightly after six months in service, while
PMFC cores of WMA with Sasobit increased significantly. A general trend was shown
that the longer conditioning protocols for LMLC specimens resulted in specimens with
equivalent or higher stiffness than the MR values measured in the early life of the
pavement. Among the five conditioning protocols applied to LMLC specimens, several
189
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

protocols applied for WMA with Evotherm 3G were able to produce enough aging such
that the stiffness of the LMLC specimens was equivalent to the stiffness of PMFC cores
at construction. In the case of HMA and WMA with Sasobit, equivalent stiffnesses
were obtained between PMFC cores at construction and the LMLC specimens
conditioned with 2 hours at Tc. Additionally, WMA specimens conditioned with 2 hours
at 275F (135C) had significantly higher stiffness than those conditioned with 2 hours
at Tc, while WMA with Sasobit specimens conditioned with 2 hours and 4 hours at Tc
had similar stiffnesses. Thus, it can be inferred that WMA specimens are more
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
susceptible to conditioning temperature rather than conditioning time in terms of

changes in MR.
In the case of the Texas field project, the conditioning protocol of 2 hours at Tc
followed by 16 hours at 140F (60C) plus 2 hours at Tc was not performed on the
LMLC specimens given the high stiffness values obtained for the same protocol in the
Iowa field project and the impractical nature of this protocol. Among the four
conditioning protocols applied to the LMLC specimens, 2 hours at Tc more closely
represented the stiffness of the pavement in its early life. Similar trends to the ones
obtained for the Iowa field project were observed with the stiffness increasing with
higher conditioning temperature and longer conditioning time, and the stiffness of the
mixtures being more sensitive to conditioning temperature than conditioning time
(Figure 4).
( a)
190

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(b )
( c)
Figure 3. MR comparison of Iowa PMFC cores versus LMLC specimens subjected to
different conditioning protocols; (a) HMA, (b) WMA with Evotherm 3G, (c) WMA with
Sasobit.
191

( a)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(b )
192

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
( c)
Figure 4. MR comparison of Texas PMFC cores versus LMLC specimens subjected to
different conditioning protocols; (a) HMA, (b) WMA with Evotherm DATTM, (c) WMA
with Foaming Process.
Based on the results shown, 2 hours at Tc was the recommended conditioning

protocol for LMLC (mix design) specimens to simulate the stiffness of both WMA and
HMA pavements in their early life. A statistical analysis was completed to further justify
this recommendation and account for the variability in the MR results. An analysis of
variance (ANOVA) and Tukey-Kramer Honestly Significant Difference (Tukey's HSD)
test were conducted with a 5% significance level (i.e., = 0.05) to verify the difference
in MR between the conditioned LMLC specimens versus the PMFC cores at
construction. In addition to the main factor of interest, which is conditioning protocol,
the effect of orientation (i.e., rotating the specimen 90 after the first measurement) as
well as the interaction effect between orientation and conditioning protocol was also
tested by using a split plot design analysis. The results confirmed that neither the
interaction effect between orientation and conditioning protocol nor the main effect of
orientation were statistically significant for any of the mixtures considered. The effect of
conditioning protocol was statistically significant for all mixtures except for Texas
HMA.
The results of the Tukey's HSD test on conditioning protocols are shown in Figure 3
and Figure 4 with different capital letters above the MR results. The MR values decrease
as letters change from A to E. Conditioning protocols with different letters have MR
193

values that are statistically different from each other. As shown by the Tukeys HSD
results, for all HMA and WMA mixtures except Iowa WMA with Evotherm 3G and
Texas WMA with the foaming process, LMLC specimens conditioned for 2 hours at Tc
had statistically equivalent stiffnesses as corresponding PMFC cores at construction. For
Texas WMA with foaming process, the least difference in mixture MR stiffness as
compared to that of PMFC cores at construction was shown by LMLC specimens
conditioned with 2 hours at Tc, although significantly higher MR stiffness were indicated
by the Tukeys HSD results. In general, the conditioning protocol of 2 hours at Tc was
able to represent the stiffness of HMA and WMA pavements at their early life.
It is important to note, however, that in most instances Tc is not specified in the mix
design, and it is sometimes arbitrarily selected with different values used for LMLC
specimens, on-site PMLC specimens, and placement temperatures during pavement
construction. Besides, standard conditioning temperatures for HMA and WMA are
desired for practicality. Table 4 shows that Tc, monitored after the paver during
construction, for the Iowa HMA was 295F (146C) and that used for the Texas HMA
was 275F (135C). Since the conditioning protocol of 2 hours at 275F (135C) was
able to provide enough compactability for Iowa HMA and Texas HMA, the standard
laboratory conditioning protocol for preparing HMA LMLC specimens is ultimately
recommended as 2 hours at 275F (135C). In the case of WMA, Tc for most of the Iowa
and Texas mixtures was approximately 240F (116C) with the exception of Texas
WMA with foaming process. Therefore, 2 hours at 240F (116C) was ultimately
recommended as the standard laboratory conditioning protocol for WMA LMLC
specimens.
4.2.2. Laboratory Conditioning Protocols for PMLC Specimens
Figure 5 and Figure 6 present the MR results for PMFC cores, on-site PMLC specimens,
and off-site PMLC specimens for the Iowa and Texas field projects, respectively. In each
graph, PMFC cores are located on the left side of the figure and on-site and off-site
PMLC specimens subjected to different conditioning protocols are shown on the right
side of the figure. Each bar in these figures represents the average value of three
replicate specimens, and the error bars represent one standard deviation from the
average value.
A statistical analysis similar to that used for the LMLC specimens was used to verify
the difference in MR stiffness between PMFC cores versus on-site PMLC specimens and
off-site PMLC specimens subjected to different conditioning protocols. The interaction
effect between conditioning protocol and orientation was statistically insignificant for all
mixtures. The main effect orientation was statistically insignificant for all mixtures
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
194

except for Texas WMA with Evotherm DATTM, but the difference was practically
insignificant. The effect of conditioning protocol, on the other hand, was statistically
significant for all mixtures. The general results of the Tukey's HSD test are shown in
Figure 5 and Figure 6 with capital letters above the bars. Conditioning protocols with
different letters have MR values that are statistically different from each other.
Figure 5 and Figure 6 show that for all the Iowa and Texas mixtures, on-site PMLC
specimens had equivalent mixture stiffness or mixture stiffness with the least difference
(i.e., Iowa HMA and WMA with Sasobit) as compared to those of PMFC cores at
construction. In the case of the Texas field project, besides the 1-2 hours conditioning at
Tc for the on-site PMLC specimens, a separate set was prepared with 0-1 hours of
conditioning time at Tc. The MR stiffness of both sets of specimens was compared using
a student t-test. The results of the comparison indicated equivalent stiffness between the
two sets of on-site PMLC specimens for HMA and WMA with Evotherm DATTM,
demonstrating an insignificant effect on stiffness from the increase in on-site
conditioning time by 1 or 2 hours.
In contrast, the conditioning protocols used on the off-site PMLC specimens yielded
specimens with statistically higher stiffness as compared to the PMFC cores at
construction. This indicates that even reheating the off-site PMLC to Tc is enough to
significantly increase the stiffness of the mixture. In addition, the stiffness of the off-site
PMLC specimens reheated at Tc was in most cases equivalent to the stiffness of the off-
site PMLC conditioned for longer periods after being reheated to Tc.
The smallest difference in mixture stiffness between PMFC cores versus PMLC
specimens corresponded to the on-site PMLC specimens, followed by the off-site PMLC
specimens with the conditioning protocol of reheating to Tc. The latter also required the
least amount of aging prior to compaction. Therefore, the use of on-site PMLC
specimens is recommended as the best alternative, especially when preparing QC/QA
specimens. In cases where on-site PMLC specimens are not available, the next best
option is to employ off-site PMLC specimens reheated to Tc.
As previously mentioned, the Tc of 275F (135C) was able to provide enough
compactability for the loose HMA from both field projects, and most Tc for WMA from
the Iowa and Texas field projects were approximately 240F (116C). Therefore, the Tc
in the recommended conditioning protocol for preparing HMA and WMA PMLC
specimens was standardized at 275F (135C) and 240F (116C), respectively. WMA
with foaming process off-site PMLC specimens required a different conditioning
protocol as compared to WMA with additives because the foaming effect during
production is assumed to be lost after mixing and cooling of the loose mix. Therefore,
the conditioning protocols recommended for preparing PMLC specimens are on-site
with (1) 1 hour at 240F (116C) for WMA and (2) 1 hour at 275F (135C) for HMA.
195
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

When compacting PMLC specimens on-site is not viable, the recommended

conditioning protocol for off-site PMLC specimens is to (1) reheat to 240F (116C) for
WMA with additives and (2) reheat to 275F (135C) for HMA and WMA with foaming
process.
( a)
(b )
196
--`,`,,,``,,```,,`,,,,,`,``,

( c)
Figure 5. MR comparison of Iowa PMFC cores versus on-site and off-site PMLC
specimens subjected to different conditioning protocols; (a) HMA, (b) WMA with
Evotherm 3G, (c) WMA with Sasobit.
( a)
197
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

(b )
( c)
Figure 6. MR comparison of Texas PMFC cores versus on-site and off-site PMLC
specimens subjected to different conditioning protocols; (a) HMA, (b) WMA with
Evotherm DATTM, (c) WMA with Foaming Process.
198
--`,`,,,``,,```,,`,,,,,`,``,````-`-`

4.2.3. Other Factors affecting Mixture Stiffness
PMFC cores at construction and on-site PMLC specimens were expected to have similar
stiffnesses because they experienced approximately the same level of binder aging, with
the PMFC cores possibly aging more during transportation to the pavement site. MR
results from the Texas field project followed this expected behavior, while MR results
from the Iowa field project showed a different trend. For the Iowa field project, the on-
site PMLC specimens showed higher stiffness as compared to the PMFC cores at
construction. These differences were evaluated with respect to binder stiffness, aggregate
orientation, and total specimen AV.
To verify the expectation that binder stiffness has an increasing effect on mixture
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
stiffness, asphalt binder was extracted and recovered (ASTM D2172 and ASTM D5404)
from HMA and WMA with Evotherm 3G/Evotherm DATTM on-site PMLC specimens
and PMFC cores obtained from both field projects. The stiffness of the extracted binders
was then evaluated with the DSR. The difference in mixture stiffness between these two
specimen types was attributed to differences in binder aging that occurred during loose
mix reheating prior to compaction.
DSR tests were performed on the extracted and recovered binders in accordance with
AASHTO T 315 at 77F (25C) to match the MR test temperature. The complex modulus
(G*) was selected as the test parameter to compare the stiffness of the extracted binders.
DSR and MR results of PMFC cores at construction versus on-site PMLC specimens
from both projects are summarized in Figure 7. The bars in the figure represent the
average MR of three replicate specimens, the dots indicate the average G* of three
measurements, and the error bars one standard deviation from the average values.
The results show that all PMFC cores had higher G* values than corresponding on-
site PMLC specimens. Therefore, PMFC cores were expected to be stiffer as compared
to on-site PMLC specimens, assuming an equivalent aggregate orientation and specimen
AV. However, the trend for the MR results was opposite to the G* results, as illustrated
in Figure 7. MR results indicate that the stiffness of the PMFC cores was lower or
equivalent to that of corresponding on-site PMLC specimens. Therefore, factors other
than binder aging, such as different compaction methods and different specimen AV,
affected the stiffness of the mixtures.
199

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
( a)
(b )
Figure 7. MR and DSR results of PMFC cores at construction versus on-site PMLC
specimens; (a) Iowa field project, (b) Texas field project.
A previous study (Boudreau et al, 1992) indicated that different compaction methods
may induce differences in specimen anisotropy and aggregate interlock and that both
200

factors may have significant effects on mixture stiffness. Specifically, field compaction
is expected to give rise to cross-anisotropic materials, indicating that most aggregates
orient along the horizontal direction in the field. These cross-anisotropic materials will
exhibit lower MR values when tested in the horizontal direction than isotropic ones due
to the aggregate orientation.
The difference in aggregate orientation was evaluated via image analysis technique
using a portable scanner to capture a continuous image of the lateral surface of the
specimen as shown in Figure 8. Four on-site PMLC specimens and PMFC cores at
construction from the Iowa and Texas field projects were scanned. The specimens were
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
laid horizontally on an automatic constant speed rotator while the portable scanner was
placed on top of the specimen to scan its lateral surface. Using image analysis software,
several aggregate characteristics including the inclination angle, cutting surface area, and
aspect ratio were measured and used in a modified vector magnitude, , to evaluate the
overall aggregate orientation of the asphalt mixture (Zhang et al., 2011). The parameter
has a range from zero to one with zero indicating full isotropy (i.e., completely
random distribution of particles) and larger values indicating more anisotropy (i.e.,
preferential orientation of the long dimension of the aggregates in the horizontal
direction, which is perpendicular to the direction of compaction).
The results for PMFC cores at construction and on-site PMLC specimens from the
Iowa and Texas field projects are summarized in Figure 9. As expected, the parameter
for the PMFC cores were higher than those for on-site PMLC specimens, indicating
higher anisotropy in the horizontal direction. Therefore, PMFC cores at construction
could have less resistance to the diametral load as applied in the MR test.
Another factor to consider in the comparison of mixture properties conditioned using
the selected protocols is AV content. It is well known that AV content has a significant
effect on mixture stiffness. In this study, all laboratory fabricated specimens (LMLC and
off-site PMLC) had a target AV of 70.5% while the PMFC cores had higher AV, in the
range of 7% to 9%. To evaluate the effect of AV on mixture stiffness, several LMLC
specimens of WMA with Sasobit with AV ranging from 5% to 9% were fabricated and
tested to determine MR. These test results are presented in Figure 10 and show that
mixture stiffness reduced significantly as AV increased from 6% to 9%, while MR was
relatively constant between 5% and 6% AV. Therefore, the higher AV of the PMFC
cores could also explain some of the differences in mixture stiffness as compared to the
on-site PMLC specimens.
In general, for the HMA and WMA evaluated, both the compaction method (i.e.,
anisotropy) and overall AV had a significant effect on mixture stiffness. These factors
help explain the discrepancy in the mixture and binder stiffness observed between
PMFC cores at construction versus on-site PMLC specimens.
201

( a) (b)
Figure 8. Aggregate orientation measurements using image analysis techniques; (a)
Test equipment, (b) Example of the scanned image of the lateral surface of an asphalt
mixture specimen.
Figure 9.Overall aggregate orientation for PMFC cores at construction versus on-site
PMLC specimens from the Iowa and Texas field projects.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
202

Figure 10. Effect of total AV on mixture stiffness for the Iowa WMA with Sasobit
LMLC specimens.
5. Conclusions
The objective of this study was to recommend standard laboratory conditioning
protocols for WMA specimens for performance testing. These protocols are intended to
be used as part of the WMA mix design procedure or the quality control/quality
assurance program for WMA. Different conditioning protocols were selected for
fabricating WMA LMLC and PMLC specimens, and these specimens were tested to
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
determine the effect of the conditioning protocol on mixture stiffness (MR). PMFC cores
at construction were incorporated in the experimental design to represent HMA and
WMA pavements in their early life. Volumetrics, mixture stiffness, binder stiffness, and
aggregate orientation of different HMA and WMA specimens were evaluated. The
following specific conclusions can be made based on this study.
1. Comparison of volumetrics between LMLC and on-site PMLC specimens indicated
that all on-site PMLC specimens (except for WMA with the foaming process from
the Texas field project) have higher Gmm values and binder absorption and lower
effective binder film thicknesses. Thus the loose plant mix experienced more
conditioning prior to compaction than those mixed in the laboratory. The reduction
in mixing and Tc and the incorporation of WMA additives resulted in lower Gmm
values and lower binder absorption as compared to HMA, which may reduce the
203

bonding strength between the aggregates and the binders, making WMA more
susceptible to early age distress such as moisture damage and rutting.
2. MR results showed that stiffness of LMLC specimens increased with higher
conditioning temperatures and longer conditioning time and that WMA was more
sensitive to conditioning temperature than conditioning time. Among the five
selected conditioning protocols for LMLC specimens, 2 hours at Tc was more
representative in terms of stiffness of HMA and WMA pavements in their early life.
Considering the difficulty in accurately defining Tc in the field and the common
range of Tc for HMA and WMA, 2 hours at 275F (135C) and 240F (116C)
instead of 2 hours at Tc are recommended as the standard laboratory conditioning
protocol for HMA and WMA LMLC specimens, respectively.
3. MR results for PMLC specimens subjected to different conditioning protocols versus
PMFC cores at construction showed that on-site PMLC specimen was more
representative in terms of stiffness of HMA and WMA pavements in their early life.
In contrast, the conditioning protocols used on the off-site PMLC specimens yielded
specimens with statistically higher stiffness as compared to the PMFC cores at
construction. Reheating loose mix had a significant effect on the stiffness of off-site
PMLC specimens. Even in the case of HMA and WMA with only reheating to Tc,
the stiffness was higher than the stiffness of PMFC cores at construction.
Considering the difficulty in accurately defining Tc in the field and the common
range of Tc for HMA and WMA, the Tc in the recommended conditioning protocols
for preparing PMLC specimens is standardized at 275F (135C) and 240F
(116C), respectively. Off-site PMLC specimens of WMA with foaming process
may require a different conditioning protocol as compared to WMA with additives
because the foaming effect during production is assumed lost after mixing and
cooling of the loose mix. Therefore, the conditioning protocols recommended for
preparing PMLC specimens are on-site with (1) 1 hour at 275F (135C) for HMA
and (2) 1 hour at 240F (116C) for WMA. When compacting PMLC specimens on-
site is not viable, the recommended conditioning protocol for off-site PMLC
specimens is to (1) reheat to 275F (135C) for HMA and WMA with foaming
process and (2) reheat to 240F (116C) for WMA with additives.
4. The stiffness of the binder extracted from PMFC cores at construction was higher
than the stiffness of the binder extracted from on-site PMLC specimens, as indicated
by DSR testing. The positive effect from binder stiffness, however, is overwhelmed
by the negative effect from aggregate orientation and AV. In other words, the
discrepancy between mixture stiffness in PMFC cores at construction and on-site
PMLC specimens is likely due to mixture anisotropy induced by different
compaction methods and different AV. Based on image analysis techniques, the on-
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
site PMLC specimens showed less anisotropy in the horizontal direction as
204

compared to PMFC cores at construction, resulting in less resistance to the diametral

load in MR test. Higher AV may also significantly reduce the mixture stiffness in
terms of MR. Therefore, mixture anisotropy and overall AV have the most
significant effect on mixture stiffness.
6. Recommendations for Future Research

1. In this study, standard laboratory conditioning protocols to prepare LMLC
specimens and PMLC specimens for performance tests were proposed based on MR
results. Additional mixture properties need to be considered for validation.
2. The effect of the total AV in the asphalt mixture specimen on mixture stiffness was
verified in this study using LMLC specimens of a single WMA technology prepared
with one specific conditioning protocol. Future research into the comprehensive
effects of AV on the stiffness of asphalt mixtures prepared with various WMA
technologies is necessary, with a particular emphasis on exploring the difference in
AV between PMFC cores and LMLC specimens and PMLC specimens.
3. A number of WMA additives are available to reduce the production temperature of
asphalt mixtures. In this study, commonly used WMA additives were used and
evaluated. Future research may include other WMA technologies and verify the
standard conditioning protocols proposed in this study.
7. Acknowledgements
The authors would like to thank Cindy Estakhri, Dr. Jon Epps, and Dr. David Newcomb
at the Texas A&M Transportation Institute and Dr. Robert Lytton at Texas A&M
University for their continuous guidance and helpful comments. Thanks also go to
Stephen Johnson, Stephen Walls, David Zeig, and Rick Canatella at the Texas A&M
Transportation Institute for their work in fabricating specimens for this study. This stud y
was based on NCHRP Project 9-49.
8. References
Al-Qadi, I.L., J. Baek, Z. Leng, H. Wang, M. Doyen, J. Kern, and S.L. Gillen, Short
Term Performance of Modified Stone Matrix Asphalt (SMA) Produced with Warm
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Mix Additives, Report No. FHWA-ICT-12 001, Illinois Center for Transportation,
Urbana-Champaign, IL (January 2012) 45 pp.
205

Bonaquist, R., NCHRP 9-43: Mix Design Practices for Warm Mix Asphalt. NCHRP
Report 691, ISSN 0077-5614 (2011).
Boudreau, R.L., R.G. Hicks, and A.M. Furber, Effects of Test Parameters on Resilient
Modulus of Laboratory-Compacted Asphalt Concrete Specimens Transportation
Research Record, No.1353 (1992) pp. 4652.
Clements, T., Increasing the Loose Mix Aging Time for Warm Mix Asphalt Asphalt
Magazine, Asphalt Institute. As of April 29, 2013:
http://www.asphaltmagazine.com/news/detail.dot?id=717cc83e-e0fc-41dd-
b821-6e84a911737b (March 25, 2011).
Clements, T., P. B. Blankenship, and K. C. Mahboub, The Effect of Loose Mix Aging
on the Performance of Warm Mix Asphalt Journal of the Association of Asphalt
Paving Technologists, No. 81 (2012) pp. 541567.
Estakhri, C., Button, J., and Alvarez, A.E., Field and Laboratory Investigation of Warm
Mix Asphalt in Texas, Report No. FHWA/TX-10/0-5597-2, Texas Transportation
Institute, College Station, TX (July 2010) 144 pp.
Estakhri, C., Laboratory and Field Performance Measurements to support the
Implementation of Warm Mix Asphalt in Texas, Report No. FHWA/TX-12/5-5597-
01-1, Texas Transportation Institute, College Station, Texas (February 2012) 77 pp.
Federal Highway Administration (FHWA), Highway Statistics. As of February 17, 2011:
http://www.fhwa.dot.gov/policyinformation/statistics/2008/ (2008).
Harrigan, E.T., Guidelines for Project Selection and Materials Sampling, Conditioning,
and Testing in WMA Research Studies. Transportation Research Board: Research
Results Digest 370 (February, 2012).
Hadley, W., SHRP-LTPP Overview: Five-Year Report. SHRP-P-416, ISBN 0-309-
05815-5 (June 1994).
Jones, D., R. Wu, B.W. Tsai, Key Results from a Comprehensive Accelerated Loading,
Laboratory, and Field Testing Study on Warm-Mix Asphalt in California.
University of California Pavement Research Center. 2nd International Warm-Mix
Asphalt Conference. St Louis. MO. (October 2011).
Kuennen, T., Warm Mixes are a Hot Topic. Better Roads, James Informational Media,
Inc., Des Plaines, Illinois (June 2004).
206
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

NAPA, Asphalt Industry Update and Overview. As of February 17, 2011:

http://www.hoix.org/index.php?option=com_content&task=view&id=14
&Itemid=33 (2010).
Newcomb, D., "Warm Mix: The Wave of the Future?" Hot Mix Asphalt Technology,
National Asphalt Pavement Association, Lanham, Maryland (July/August 2005b)
pp. 33-36.
Stroup-Gardiner, M., C.R. Langer, and A. Carter, Quantification of Emission Potential
from Asphalt Binders Using Mass Loss and Opacity Measurements International
Journal of Pavement Engineering, Vol. 6, No. 3 (2005) pp. 191-200.
Zhang, Y., R. Luo, R.L. Lytton, Microstructure-Based Inherent Anisotropy of Asphalt
Mixtures. ASCE ISSN 0899-1561 (April, 2011).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
9. Discussion
DR. JOHN DANGELO: Youve collected a lot of data here it is rational and what we
would expect. In your work you selected to duplicate the aging condition of the mix
right at construction. I guess that came from the panel, Im assuming, but this goes
against everything that we have done in the past on early aging. The original SHRP work
was always meant to duplicate six months to a year in the field. A lot of work done after
SHRP up until this project always was to duplicate some field aging, not just plant aging.
Was this driven by the panel for the NCHRP project that aging would be right at the day
of construction?
MR. FAN YIN: Thank you for the question. For the NCHRP project 9-49, there were
three main parts. The first part of the study is about laboratory conditioning that I just
discussed. The objective of this part is to simulate the stiffness of the cores at
construction. Then another part of the project is about the long-term aging of WMA,
which was used to simulate the aging after one summer or after several years in the field.
DR. DANGELO: One year is significantly different than several years. The only issue
that you see here is that if you duplicate aging right at construction, you can look at QC
for air voids. However, if you do any kind of performance testing comparisons to field
performance, it is invalid because any performance characteristics you take from the
roadway, even if it is within a year, have to take into account that aging that took place
in that time period that you relate back to the mixture. So if youre trying to use the
Hamburg, which is being used in a lot of places now, to evaluate early rutting on the
roadway for quality control, is it really valid to use the stiffness of the mix on the day its
placed with no aging? I think youve got significant problems on how to relate that
207

rutting to what you see on the roadway after about a year or two with what happens the
day of construction.
The other thing is, you are recommending specific temperatures. But this is based on
neat asphalts with low absorption aggregates. What about polymer-modified material
with higher absorption aggregates? I think you need to have a lot more work before you
can just make a statement for one set temperature.
MR. YIN: Okay. Thank you for the comments.
PROFESSOR AMY EPPS MARTIN: John, we did actually look at cores at six months,
after a winter for some projects, and cores after a summer, which would be eight months
on one of our projects and 12 months on some of the projects. So we did look at
simulating the stiffness in the early life, which we think is the first year before a summer
of aging when warm mix hasnt caught up to hot mix. So we did consider that. We are
trying to capture the moisture susceptibility of these mixes when they might be in that
vulnerable period, again in that first year of their life. In addition, some of these binders
are polymer-modified. I hope that helps a little bit.
MR. GALE PAGE: Im going to applaud you on the fact that you're looking at
conditioning of specimens. I think thats very important. And I think agencies have a
tendency to forget about the need for conditioning, and the only time that they consider
it is when they get different test results from the contractor. I know back probably about
15-20 years ago, that was an issue, particularly with the determination of air voids, in
Florida. It depended on where you took the samples. Time and location of sampling for
air void calculation was affected not only by absorptive aggregates but also some of the
granite aggregates. So we did a significant study, taking a look at the materials from
stockpile through plant operations, through the paver and then, finally, in the pavement
itself, which is where we need to be concerned about its properties. What is the property
or what are the properties of the mix in place in the pavement? Based on that study, we
implemented a field conditioning of one hour before we did testing for air voids, and the
variability between contractor and agency air voids finally decreased. So conditioning is
important, whether it be for air void calculation, whether it be for resilient modulus, or
whether it be for rutting. I think the important thing to find out is what are the material
characteristics of the pavement in place? Do you have to condition the sample that you
take and where do you take it in order to be representative of the pavement in place?
MR. YIN: Thank you for the comments.
MR. FRANK FEE: I just wanted to address one of John DAngelos points. As you
know, this is a small part of a very much larger study, and one of the phases in that study
is to take a look at the effect of various types of asphalt mix plants on this aging
condition, the short-term conditioning. So the reason for the importance of that initial
stiffness is not to look at the long-term or whats going to happen at a plant, but what is
the baseline that the asphalt plant itself induces in terms of aging. We are using right
208
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

now a standard that was based on pugmill mixing for our laboratory conditioning. That,
we think, is not accurate for many asphalt plants today. So part of this initial stiffness
evaluation has to do with finding out what is the stiffness of those mixes that are coming
out of the plant now. I think we might find that enlightening in terms of where that
baseline is. So thats what that early stiffness is all about, John.
MR. YIN: Thanks for your comments. For another NCHRP project that has been worked
on at TTI, the research focus is to look at the aging characteristic of warm-mix asphalt
and hot-mix asphalt. In that project, we will incorporate different plant types in the test
plan, as you just mentioned, since we anticipate that different plant types may have
different characteristics in term of effects on mixture initial stiffness. Thank you.
MR. ARTHUR JOHNSTON: Very nice presentation. One of the things I gleaned fro m
your presentation is something that weve seen in some of the work that weve done.
That is that the effect in terms of reduction in viscosity, for example, of some of these
additives is relatively temporary, and if you do not get your specimens formed quickly,
they are not very representative of actual conditions. We were looking more at
volumetric properties, and that is really my question or clarification I was going to ask.
You showed that you were getting lab air voids of around 7%. Was that the intent?
Because typically wed be looking, at the design compactive effort, for something closer
to 4%.
MR. YIN: Yes, we tried to make all laboratory specimens with 7% air voids. This study
is based on NCHRP project 9-49, evaluating the moisture susceptibility of warm mix
asphalt. There is evidence showing that warm mix is susceptible to moisture damage at
early life. So for the specimens fabricated in the lab, making them with 7% air void
should be able to speed up the occurrence of the moisture damage in the laboratory.
MR. JOHNSTON: Okay, thank you.
MR. YIN: Thank you.
DR. HONGBIN XIE: I have a question. I just looked at your paper. You didnt show the
detailed data in the presentation. But looking at the paper, it seems like your final
recommendation for the curing or aging is based on the stiffness mainly?
MR. YIN: Im sorry. Can you repeat the question again?
DR. XIE: Your recommendation for the curing, like two hours at 240F. Is that based on
the stiffness?
MR. YIN: Yes, thats right. Its based on stiffness.
DR. XIE: Yes, because I look at your table 4 in terms of volumetrics, which you didnt
mention in your presentation but that would have made a huge difference in selecting the
curing condition. For example, as shown in the Table, the Iowa aggregates apparently
have a much higher absorption than the aggregates from Texas. Just comparing these
209
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

two data from the lab mixed versus the plant mixed the asphalt absorption has a
difference of 1.2%, for example, for the Sasobit WMA. That is the difference between
the plant produced and the lab mix with two hours aging at 275C. So I can see the
difference is going to be even higher if you decrease that temperature to 240F as
recommended by this paper. In that particular case, the warm mix with Sasobit with the
rock from Iowa, could have a difference in absorption of 1.2%. Thats huge, and can
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
directly impact your air voids in your mix design. So my question to you is really that,
do you think it may be more realistic to have a different curing time or different
temperature or combination of these parameters for aggregates with different absorption?
MR. YIN: I think if we want to take into consideration the volumetrics in terms of
simulating the field cores, aggregate absorption should be an important factor. As I
mentioned before, for another NCHRP project that is currently ongoing, we will look at
the effect of binder absorption on mixture stiffness and other performance parameters
DR. XIE: Because in the mix design, regardless of the curing condition for the mix
design, the air voids is the one first thing you look at when you're doing your mix design,
right? So you want to design at 4% air voids, and if your AC content to maintain that 4%
air voids can vary from the plant produced to the lab mix design, like over 1%, that
would make it almost useless for the lab design. In other words, you will have to make a
huge adjustment when you go to production.
MR. YIN: Okay.
DR. XIE: Thank you.
MR. YIN: Thank you.
MR. PHILLIP BLANKENSHIP: The Asphalt Institute also performed research on
WMA oven conditioning. It was presented at 2010 AAPT by Thomas Clements. We
investigated somewhat of a conditioning shift factor, you may call it, looking at half
hour, two hour, four hour, and eight hour loose mix oven conditioning. We found the
stiffening of the mixture, as measured by flow number and Hamburg Wheel Tracker,
revealed a difference of about 1.5 to 2 times more stiffening for HMA as compared to
WMA. Did you take any measurements past the four hour conditioning time?
MR. YIN: No. The longest condition protocol we have is four hours at 275F.
MR. PHILLIP BLANKENSHIP: Okay. That might be worth looking at, just to see how
that conditioning time would look past four hours, especially with WMA. Thank you
very much.
MR. YIN: Thank you for the comments.
PROF. EPPS MARTIN: I just want to say one more thing. We also validated some of
the recommendations that we used at this first stage of the study. In the later stages that
Fan was mentioning, for an additional seven mixtures at two different sites, we did these
210

laboratory conditioning protocols, and we saw that they did produce specimens that gave
us similar susceptibility characteristics in three different standardized lab tests. So we
feel fairly comfortable with these recommendations.
211
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Merits of RAP Dominated Warm Mixed
Flexible Pavement Base Layers
Isaac L. Howarda*, Jesse D. Doyleb, and Ben C. Coxa
a
Department of Civil and Environmental Engineering, Mississippi State University,
Mississippi State, MS 39762-9546
b
Airfields and Pavements Branch, Geotechnical and Structures Laboratory, U.S. Army
Engineer Research and Development Center, Vicksburg, MS 39180-6199
ABSTRACT: Use of Reclaimed Asphalt Pavement (RAP) has increased considerably over the past few
years; approximately 85% of the RAP available is used within either hot mix asphalt (HMA) or
warm mix asphalt (WMA). Within this time frame a number of research efforts have been
performed, but most efforts have focused on RAP contents of 50% or less. This paper describes a
laboratory effort that studied WMA with RAP contents of 50 to 100% in the areas of rutting,
moisture damage, durability, cracking, and mixing uniformity. Lack of a RAP surplus coupled with
performance data presented in this paper indicate that mixtures with more than 50% RAP do not,
for most applications, add value to the highway system in present day. WMA with 50% RAP for
use as an underlying (or base) pavement layer performed adequately in all performance areas
investigated, durability and cracking included.
KEYWORDS: RAP, WMA, wheel tracking, moisture damage, durability, cracking
The oral presentation was made by Dr. Howard.
http://dx.doi.org/ 10.1080/14680629.2013.812834
213
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
1. Introduction and Background
Greater demands on ever dwindling resources are forcing the stewards of our nations
highways to look for more creative pavement material solutions. Creative solutions
using either longer lasting materials for the same cost or less expensive materials
yielding equivalent performance are beneficial. For example, the Mississippi DOT
awarded contracts for 3 million tons of hot mix asphalt (HMA) in 2011, accounting for
approximately 20% of their budget, so even small unit cost changes can have a large
overall effect.
Reclaimed asphalt pavement (RAP) is an area where significant research has
occurred recently; however, relatively little of this research has focused on identifying
the functional limit with regard to the highest viable RAP content producible at a plant
or its corresponding performance. Prior to widespread implementation, any concept
under consideration should be studied to identify its functional limits. These functional
limits could be in the areas of material availability, product performance, or production
capabilities.
State DOT use does not appear to have approached RAP functional limits (material
availability not considered), especially on a large scale. A 2007 survey, where all 50
state DOTs responded, showed the national average RAP usage rate was 12%, and that
the maximum amount of RAP permitted was not being used on a nationwide basis
(Copeland, 2011). Another survey in 2009 from the same author showed about half of
the US states increased RAP use from 2007 to 2009, though fewer than half of US states
used more than 20% RAP in HMA. Hansen and Newcomb (2011) presented material
availability data from a 2010 survey where 47 U.S. states, 196 companies, and 1,027
plants responded. In 2010, 73.5 million tons of RAP were accepted, 62.1 million tons
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
were used in warm mix asphalt (WMA) or HMA (84.5% of the total RAP accepted), 9.8
million tons were used for items such as aggregate base (13.1%), 1.8 million tons were
excess that was likely stockpiled (2.4%), and less than 0.1% was landfilled. The Hansen
and Newcomb (2011) data show considerable RAP use increases on a large scale would
require additional milling, or equivalent means to obtain more RAP as 84.5% of the
material available already goes into HMA or WMA.
The overall goal of this paper is to provide information to facilitate informed
decisions about the use of very high RAP content mixtures. Availability information
was not considered beyond the information presented in the previous paragraph, but the
availability data should be considered in the context of performance when viewing the
results of this paper. The specific objective of this paper is to present mixture laboratory
evaluation results where RAP is the dominant component (i.e., RAP content is 50% or
greater) in WMA base mixtures (i.e., not on the surface). Performance in four key areas
was explored: rutting, moisture damage, durability, and cracking. The ability of asphalt
214

Merits of RAP Dominated Warm Mixed Flexible Pavement Base Layers
plants to produce RAP dominated mixtures was also investigated through literature
review. A paper covering a wide range of performance characteristics for 50 to 100%
plant produced RAP base mixtures was not found during the review of the literature.
2. Literature Review
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Recently, several studies have investigated the use of RAP contents of 25 to 50%. RAP
contents at or greater than 25% are generally referred to as high RAP mixtures (e.g.,
Copeland, 2011). Less work, however, has focused on mixtures intended for plant
production as either HMA or WMA where RAP is the dominant component (i.e., RAP
content is 50% or greater).
2.1. Plant Production of 25 to 50% RAP Mixtures
High RAP (25 to 50%) has been plant produced and used for a variety of applications.
Boggs (2008) performed a demonstration project in South Carolina where 50% RAP
WMA (foamed at 132C) was used for base and surface layers. Measured in place air
voids were approximately 6%, and mat temperatures were as low as 88C. Rut depths
measured in the laboratory on plant mixed material were acceptable at less than 5 mm.
Brown (2011) used 50% RAP WMA for 87 km of shoulder paving in Illinois. The
asphalt was foamed at 121 to 132C, used 2.4% performance grade (PG) 58-28 virgin
binder, and was produced at approximately 245 tons per hour.
Hodo et al. (2009) reported on 0% RAP and 50% RAP WMA (foamed) for city street
surfaces in Tennessee. In place air voids were around 9%, and no distresses were
apparent in either mixture after one year. Recovered binder from the 50% RAP WMA
was continuously graded at 84-18 (PG 64-22 virgin binder); the high temperature
increased by three grades, but the low temperature only decreased by one grade. Plant
mixed and laboratory compacted tensile strength ratio (TSR) values were marginal at
78% for 0% RAP and 82% for 50% RAP. Hamburg test performance was acceptable,
though the average stripping inflection points (SIPs) were 8,900 for 0% RAP and 8,500
for 50% RAP. The Asphalt Pavement Analyzer (APA) rutting performance was
acceptable at less than 4 mm.
West et al. (2009) reported on full scale 0% and 45% RAP HMA surface layer test
sections evaluated at the National Center for Asphalt Technology (NCAT) test track
over two years where 9.4 million equivalent single axle loads (ESAL)s were applied.
The four 45% RAP test sections had: PG 52-28 binder, PG 67-22 binder, PG 76-22
binder, and PG 76-22 binder with 1.5% Sasobit and no drop in mix temperature. The
215

HOW ARD, DOYLE, COX
pavement structure was designed as a perpetual pavement with 56 cm of HMA over

aggregate base. Pavement surface macrotexture results indicated that all RAP sections
had very good raveling performance and that their roughness did not change noticeably
over the two year period. Low severity longitudinal cracking was observed in one case
but it was determined to be reflecting from the underlying pavement. Rutting of the
RAP sections was less than the control section.
Middleton and Forfylow (2009) evaluated four plant produced WMA base mixtures
(foamed at 130 to 135C), one of which contained 50% RAP. No major differences in
APA rutting were observed; all mixtures were acceptable at 4.1 to 5.2 mm. TSR values
were 96% for the 50% RAP mixture and 78% for a 0% RAP control (mixtures with 15%
RAP had TSR values of 83 to 88%). Asphalt binder recovered according to AASHTO T
320 (minus rolling thin film oven [RTFO] aging since materials were plant produced)
showed foamed 50% RAP binder grades increased by about one grade on the high and
low temperature. The 0% RAP mixture graded as a PG 64-22 and the 50% RAP graded
as a PG 70-16. Resilient modulus testing according to AASHTO TP 31-96 showed the
average stiffness ratio values of 50% RAP to 0% RAP to be 1.6, 2.2, and 1.0 at 5, 25,
and 40C, respectively. Overall, 50% RAP increased the resilient modulus; the increase
was temperature dependent.
Copeland et al. (2010) reported on a base mixture with 45% RAP (HMA and WMA
foamed at 132C) in Florida with PG 52-28 binder (continuous grade of 56-30). Binder
grading from recovered 45% RAP HMA resulted in a PG 64-16 (continuous grade of 68-
19), while binder grading from recovered 45% WMA resulted in a PG 52-22 (continuous
grade of 57-27). The 45% RAP WMA was noticeably different than the 45% RAP
HMA and was within 1C of the low end of the continuous grade of the virgin binder.
2.2. Laboratory Performance of 25 to 50% RAP Mixtures

Rutting does not appear to be a formidable issue for high RAP mixtures (e.g.,
Shoenberger and Demoss, 2005; Tabakovi et al., 2010). High RAP contents appear to
offset at least some of the WMA rutting potential, seemingly to an extent that high RAP
WMA should not have a rutting problem in most properly designed cases (e.g., Boggs,
2008; Hodo et al., 2009; Doyle et al., 2011).
A few researchers have investigated moisture susceptibility of high RAP WMA.
Some found WMA with high RAP generally has better performance than WMA with
low or no RAP based on TSR testing (Middleton and Forfylow, 2009; Mejas et al.,
2011) and on Hamburg testing performed at 40C (Mogawer et al., 2011). However, at
least one study found marginal overall moisture damage performance of high RAP
WMA, based on Hamburg and TSR testing of plant-produced mixture (Hodo et al.,
216
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

2009). Austerman et al. (2009) found 25% RAP mixtures performed acceptably as HMA
but had poor performance as WMA, based on Hamburg testing at 50C. It is noteworthy
that none of the original studies to develop the TSR test were performed with either
WMA or RAP (Lottman, 1978; Lottman, 1982; Tunnicliff and Root, 1984; Tunnicliff
and Root, 1995; Epps et al., 2000), which calls into question the TSR to assess moisture
susceptibility of WMA RAP. Overall, these mixed results suggest that additional
research is needed in the area of moisture susceptibility of high RAP content WMA.
Issues associated with moisture susceptibility are also influenced by factors such as
aggregate type and use of anti-strip agents.
Data pertaining to high RAP mixture durability is not as prevalent as rutting,
moisture damage, or cracking. Durability is a long term performance characteristic
defined by the Asphalt Institute (2007) as the ability to resist factors such as asphalt
aging, aggregate disintegration and stripping. Some level of durability prediction can
likely be gained through test data and field performance in the early mixture life. Su
et al. (2009) studied 40% RAP mixtures for airfield surfaces in Japan using the Japanese
raveling test. Laboratory results indicated good performance could be anticipated, and
these results were supported by test sections placed on a taxiway and monitored for three
years, during which time no cracking or raveling was observed.
Doyle and Howard (2011) used the Cantabro test on eight dense graded mixtures
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
with 0 to 15% RAP to evaluate its use for dense gradations since the test has
traditionally been performed on open graded friction course (OGFC) or porous friction
course (PFC). Doyle and Howard (2011) found that, overall, the test was repeatable and
stated the test could be a candidate to assess durability of high RAP mixtures.
Baumgardner et al. (2012) used the Cantabro test on dense graded mixtures to sho w
mass loss (the result of the test) differences between PG 76-22 and ground tire rubber
(GTR) modified binders (mass loss values were 11 to 15%).
Doyle (2011) studied 33 different base mixture designs by obtaining plant mixed and
Superpave Gyratory Compacter (SGC) prepared quality assurance specimens (air voids
were 3 to 5% the majority of the time) taken from Mississippi DOT projects. Sixteen of
these mixtures were 12.5 mm nominal maximum aggregate size (NMAS), 54 specimens
were tested, the average RAP content was 16% (4.3% standard deviation), and the
average Cantabro mass loss was 9.8% (3.0% standard deviation with range of 6 to 16%).
The remaining 17 mixtures were 19 mm NMAS, 42 specimens were tested, the average
RAP content was 18% (4.8% standard deviation), and the average Cantabro mass loss
was 10.6% (2.6% standard deviation with range of 7 to 17%). The base mixture data
from Doyle (2011) is not in any way a designed or controlled experiment, but the data
represent wide ranges of mixtures used statewide in Mississippi, and can be useful for
overall comparisons to high RAP mixtures.
217

HOW ARD, DOYLE, COX
Doyle et al. (2011) tested 24 laboratory prepared airfield surface mixture designs (0
to 50% RAP with HMA and three WMAs) using the Cantabro test and the effect of RAP
content was apparent as mass loss increased with RAP content. Mass loss values ranged
from approximately 5% for the 0 % RAP WMA limestone mixture, to 20% for the 50%
RAP HMA gravel mixture.
A perception of high RAP mixtures is they are more prone to bottom up fatigue
cracking than other alternatives due to the stiff RAP binder. The literature presented in
this paper, however, provides some evidence that these mixtures may not be as fatigue
crack prone as they are perceived to be. Huang et al. (2005) summarized two studies
where HMA fatigue resistance was improved with up to 30% RAP. Santos et al. (2010)
reported better fatigue performance for laboratory and plant produced mixtures
containing 20 to 40% RAP compared to a virgin HMA control. Indirect tensile fatigue
results of Tabakovi et al. (2010) indicated 30% RAP performed significantly better than
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
all other mixtures tested with respect to fatigue resistance and measured strains. Shu
et al. (2008) investigated plant produced fatigue performance for mixtures with up to
30% RAP using Superpave IDT and flexural beam fatigue testing (600 microstrain ()
and 10 Hz). Results were mixed depending on analysis method and test type;
assessments varied from fatigue life may be reduced by RAP inclusion (indirect tensile
(IDT) data) to 30% RAP mixtures might have longer life than other mixtures (beam
fatigue with 50% stiffness reduction criteria).
Cascione et al. (2011) investigated plant mixed high RAP with PG 58-22 virgin
binder using beam fatigue testing. A variety of base and surface mixtures were produced
including 5% post consumer recycled shingles with 20 to 45% RAP, and a 50% RAP
base mixture with no shingles. The base mixtures were designed at 2 to 3% air voids to
be fatigue resistant. Test results indicated all mixtures would likely have adequate
fatigue performance except the 45% RAP and 5% shingles base mixture.
Loria et al. (2011) investigated multiple laboratory and field produced mixtures,
including 50% RAP HMA (PG 58-28 and PG 52-34 virgin binders were used). The
RAP graded at PG 76-10. After field production the PG 58-28 mixture graded at PG 64-
16, and the PG 52-34 mixture graded at PG 64-22 (i.e., low temperature changed two
grades for both virgin binders). Thermal cracking resistance was evaluated with the
thermal stress restrained specimen test (TSRST) by AASHTO TP 10-93, with acceptable
resistance from the 50% RAP mixtures.
Timm et al. (2011) reported on fatigue performance of 50% RAP (HMA and foamed
WMA) and three virgin control mixtures (HMA, foamed WMA, and Evotherm
WMA) placed on the NCAT test track where the total asphalt thickness was 18 cm.
Strains measured over approximately 4 million ESALs at the bottom of the asphalt layer
were strongly temperature dependent, and were corrected for construction differences
218

and shifted to one of three reference pavement temperatures. At 10oC, strains were not
statistically different ( 200 ). At 20oC the 50% RAP WMA had statistically lower
strain ( 250 ) than the virgin controls (310 to 340 ). At 43oC the HMA control had
the highest strain ( 900 ), the WMA controls were lower ( 800 ), and the 50%
RAP HMA and WMA were statistically lower ( 625 and 560 , respectively).
Timm et al. (2011) also performed laboratory beam fatigue testing with plant
produced mixture at 800 and 400 at 20oC (laboratory strain levels were higher than
measured field strains). Results were extrapolated with a fatigue transfer function and
indicated the 50% RAP WMA mixture may have the best performance; however, no
cracking had been observed in the field for any of the mixtures.
2.3 Plant Production of 50 to 100% RAP Mixtures
A presentation was given to the AFH60 committee at the 90th annual meeting of the
Transportation Research Board titled Field Trials of High RAP Mixes (Brad D. Cruea
unpublished data). The information in the presentation was provided to the authors of
this paper along with additional pertinent information for use in this investigation. In the
fall of 2010, mixtures were produced with 50, 60 and 70% RAP in Indiana through a
standard asphalt plant. RAP was introduced through the RAP collar, and virgin
aggregate was typically heated to approximately 370C. The plant produced 60 to 70
tons of each mixture at 250 TPH (600 TPH capacity); stack temperatures of 154C were
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
a potential concern. To produce these mixtures for an extended period of time,
modifications could be necessary to avoid plant damage. All mixtures used PG 64-22
virgin binder, with 44 to 55% binder replacement depending on RAP content. The 60
and 70% RAP mixtures were foamed; the 60% RAP mixture was well coated and
seemed to work well, while the 70% RAP mixture was not well coated and was
questionable to the producer with the information available at the time.
Mallick et al. (2010) reported on 100% RAP with rejuvenators that was plant mixed.
The paper provided technical details and production steps of a plant capable of
producing 75 to 150 tons of 100% RAP per hour. One key element of the plant is a
sophisticated air pollution control technology system. The plant was used to pave city
streets in Queens, New York, in 2002 using a rejuvenator. The mixtures were placed
with conventional equipment and were performing well in 2009 after seven years of
service, leading to the conclusion that good performance of plant mixed 100% RAP is
possible.
219

HOW ARD, DOYLE, COX
2.4. Laboratory Performance of 50 to 100% RAP Mixtures
Su et al. (2009) studied 70% RAP mixtures for airfield surfaces in Japan using the
Japanese raveling test (this study also included 40% RAP). Laboratory results indicated
good raveling performance could be anticipated, and these results were supported by test
sections placed on an airfield taxiway and monitored for three years, during which time
no cracking or raveling was observed.
Mallick et al. (2008) used 75% RAP WMA (Sasobit at 125 C) and two grades of
virgin binder (PG 52-28 and PG 42-42) for base mixtures. The goal was to create
mixtures containing 75% RAP with similar performance as a control mixture with 75%
extracted RAP aggregate and 25% virgin aggregate mixed with neat PG 64-28 binder at
150 C. The PG 64-28 virgin binder mixture had the highest -10C tensile strength of
any of the mixtures, while the 75% RAP mixture with PG 42-42 binder had the lowest.
This indicates a reduction in the potential for low-temperature cracking with the use of a
much softer neat asphalt binder. Laboratory rut depths were less than 4 mm for all
mixtures. Seismic moduli values indicated that the mixture produced with PG 42-42
binder had a significantly lower modulus than mixtures produced with PG 52-28 binder.
Seismic modulus was determined by measuring the travel time of an electric pulse from
a 54-Hz transducer and using this value along with gyratory specimen length, density,
and Poissons ratio to calculate the modulus. Similar performance of 75% RAP and
conventional HMA was possible with the use of very soft grades of asphalt binder and
Sasobit. Similar air voids and comparable mixture stiffness were observed in the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
mixtures as well as an equal or decreased rutting potential.
Mallick et al. (2007) investigated the use of 100% RAP WMA (Sasobit at 125C)
with 2.0% additional PG 64-28 binder as a base mixture. The mixtures were compared
to 100% RAP HMA (150C) without Sasobit. Workability, compactability, resilient
modulus, moisture sensitivity, and indirect tensile strength were considered.
Workability results indicated 100% RAP WMA either increased the workability or was
nearly the same as the 100% RAP HMA. Resilient modulus was measured and no
statistical difference was found between the mixtures. Some differences existed between
tensile strengths of the mixtures. Tao and Mallick (2009) experimented with 100% RAP
and zeolite additives in the laboratory and observed that the stiff RAP asphalt appeared
to hinder the foaming process somewhat compared to what was observed for virgin
binders.
Doyle and Howard (2013a) tested 100% RAP mixtures with varying amounts of PG
67-22 virgin binder. Approximately 500 mixture beams were sawn and tested in the
Bending Beam Rheometer (BBR) from SGC compacted specimens. BBR measured
stiffnesses were generally 7 to 21 GPa, which was not radically higher than low RAP 9.5
mm NMAS control mixtures (9 to 17.5 GPa) tested under the same conditions. BBR
220

measured m-values were on average 0.10 (standard deviation of 0.03) and were
noticeably flatter than the other mixtures tested.
2.5. Summary of Literature Review
Plant produced mixtures with over 50% RAP were rarely identified. Regarding high
RAP WMA performance, rutting does not seem to be an issue, additional moisture
damage information (perhaps additional test methods) is needed, durability assessments
are not prevalent, and cracking does not appear to be as big of a problem as some might
perceive. Little information was found for RAP contents above 50%. The experimental
program was developed using literature review results to evaluate functional limits of
RAP dominated WMA, since it does not appear to be prevalent in literature.
3. Experimental Program
3.1. Materials Tested
Two RAP sources were tested; recovered binder and aggregate properties are provided in
Figure 1. In addition, the extracted coarse aggregate was visually sorted and classified
as one of the two primary aggregate types used in Mississippi (i.e., gravel or limestone).
Using results of the sorting procedure and a correlation developed from 568 Mississippi
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
asphalt mix designs, the relative percentages of limestone and gravel aggregate in the
RAP sources were estimated. The sorting procedure and correlation are further
described in Doyle (2011). R1 was estimated to be 10% limestone and 90% gravel and
sand. R2 was estimated to be 30% limestone and 70% gravel and sand. The sorting
procedure could not differentiate gravel from sand.
The RAP sources were used to create six WMA-RAP dominated mixtures (Table 1)
that were evaluated using the five control mixtures (CM) in Table 2. CM1 is commonly
used in central Mississippi and was used as a control for every parameter investigated.
CM2 and CM3 bracket CM1, in that CM2 is a lower rut resistance, higher durability
design, and CM3 is a higher rut resistance, lower durability design. CM4 and CM5 were
polymer modified, high traffic mixtures that bracket the CM1 NMAS; they were used as
appropriate throughout the paper, CM5 is a surface mixture.
221

HOW ARD, DOYLE, COX
100
Aggregate Property R1 R2 .
NMAS (mm) 9 .5 9 .5
80 Gsb 2.534 2.565
Abs(%) 1 .8 1 .1
Percent Passing
60 T 304 FAA (Method A) 43.7 43.1

D 5821 CAA (2+ faces) 96 96
D 4791 F&E (2:1 ratio) 20.8 42.4
40 R1
R2 Binder Property R1 R2 .
20 Max Density R 29 PG True Grade 118+2 106-4

T 164 AC (%) 5 .5 5 .6
T 316 Visc. @135C (Pas) 52.9 9 .1
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0
0 .0 7 5
9 .5 1 2 .5
0 .1 5
0 .3
1 .1 8 2 .3 6 1 9 .0
0 .6
4 .7 5
Sieve Size mm0.45
Figure 1. Aggregate and asphalt properties of RAP sources tested.
The controls used provide a wide variety of comparison characteristics. The primary
goal of this investigation was not to determine whether RAP dominated mixtures are
superior performers, rather to determine if they could perform in a comparable manner to
conventional mixtures. Other than the aforementioned discussion, there were not any
specific requirements for control mixtures other than they should represent a fairly wide
range of properties. For example, RAP content was 10 to 15%, and RAP asphalt
contents were 5.0 to 6.4% since this was what producers had in their stockpiles at the
time of sampling. The RAP sources were different between RAP dominated and control
mixtures.
The six WMA-RAP dominated mixtures were developed for different purposes. The
two 50% RAP mixtures (50R-1 and 50R-2) were designed as, for practical purposes, the
outer bound RAP content that has been investigated to a considerable extent over the
past few years. The two 75% RAP mixtures (75R-1 and 75R-2) were designed to have
more RAP than has been studied to a considerable extent. The two 100% RAP mixtures
(100R-1 and 100R-2) were designed to investigate applications where plant production
is possible (one case was identified in the literature).
222

Table 1. Properties of RAP dominated WMA mixtures.
RAP Dominated Mixtures

Mixture ID 50R-1 50R-2 75R-1 75R-2 100R-1 100R-2
Prep. Method 1 1 1 1 1 1
Ndes 50 50 50 50 65 65
NMAS (mm) 1 2 .5 1 2 .5 1 2 .5 1 2 .5 9 .5 9 .5
Virgin PG Grade 67-22 67-22 67-22 67-22 67-22 67-22
W MA Sasobit Sasobit Sasobit Sasobit Sasobit Sasobit
TMix (C) 116 116 116 116 116 116
TComp (C) 116 116 116 116 116 116
Cr. Gravel (%) 46 49 24 24 0 0
Limestone (%) 3 0 0 0 0 0
Sand (%) 0 0 0 0 0 0
Hyd. Lime (%) 1 1 1 1 0 0
-12.5 mm (%) 9 3 .0 9 0 .5 9 1 .9 9 4 .3 9 6 .5 9 8 .2
-4.75 mm (%) 5 1 .3 5 3 .2 5 3 .0 6 2 .8 6 0 .1 7 3 .1
-2.36 mm (%) 3 3 .7 3 6 .0 3 6 .3 4 4 .4 4 1 .9 5 2 .8
-0.075 mm (%) 7 .8 7 .7 8 .3 8 .9 8 .4 9 .3
Combined Gsb 2 .4 6 4 2 .4 6 9 2 .4 9 3 2 .5 1 5 2 .5 3 4 2 .5 6 5
H2O Abs (%) 2 .6 2 .3 2 .2 1 .7 1 .8 1 .1
RAP (%) 50 50 75 75 100 100
RAP Source R-1 R-2 R-1 R-2 R-1 R-2
RAP AC (%) 5 .5 5 .6 5 .5 5 .6 5 .5 5 .6
Total AC (%)a 7 .3 6 .1 7 .3 6 .1 7 .4 6 .8
Virgin AC (%) 4 .6 3 .3 3 .3 1 .9 2 .0 1 .2
G mm 2 .2 8 1 2 .3 1 7 2 .2 9 3 2 .3 3 2 2 .3 1 7 2 .3 7 0
--Prep. Method 1 = laboratory-mixed (LM) and laboratory compacted (LC).
a
Total asphalt content (AC) on mixture mass basis including all RAP asphalt.
While the same raw materials were used for all six WMA-RAP mixtures, there was
one difference in the design parameters. The 50 and 75% RAP mixtures were developed
using 50 design gyrations (Ndes = 50) for a research study concerning use of WMA with
RAP for asphalt base mixtures. The 100% RAP mixtures were developed using Ndes =
65 for a separate research study with somewhat different goals. The difference in design
asphalt contents for 100% RAP mixtures using Ndes = 65 compared to using Ndes = 50
were determined from the data provided in Doyle et al. (2012). Therein, volumetrics of
compacted 100% RAP were investigated for a range of temperatures and Ndes levels for
both HMA and WMA. Based on this data, the design asphalt contents with Ndes = 50
would have been roughly 0.1 to 0.2 percentage points higher than for Ndes = 65. It is
believed that performance comparisons made using the 100% RAP mixtures made with
223
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
Ndes = 65 are reasonable for the objectives of this paper and this issue is pointed out
during discussion of results as appropriate.
The same virgin binder grade was used for all RAP dominated mixtures. This
approach leads to different overall binder properties from one mixture to the next. The
rationale for this approach is that PG 67-22 is the primary binder supplied in Mississippi
and is also a prevalent binder in the region. Binders that are stiffer (e.g., PG 76-22) are
readily available, but softer binders (e.g., PG 58-28) are not as readily available. If
readily available binders can be effectively utilized with RAP dominated mixtures, their
likelihood for implementation and their economic advantages improve.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
3.2. Specimen Preparation
As seen in Tables 1 and 2, specimens were prepared in a variety of manners including

laboratory-mixed (LM), where a bucket mixer was used, and plant-mixed (PM). The
temperature during mixing (Tmix) is shown in Tables 1 and 2, alongside the temperature
at the beginning of compaction (TComp). Specimens were either laboratory compacted
(LC) or field compacted (FC) using full-scale construction equipment.
Preparation details of laboratory mixed and laboratory compacted specimens are as
follows. RAP was sampled from producer stockpiles, brought to the laboratory and air
dried under fans to consistent low moisture contents. The RAP was then screened into
11 size fractions (retained on 19.0-mm sieve to passing 0.075-mm sieve). Conglomerate
RAP pieces larger than 19.0 mm were broken apart with a hand tamper to minimize
aggregate breakdown and re-screened. RAP was batched from each size fraction to
reproduce a bulk gradation. The bulk RAP gradations resulted in the extracted aggregate
gradations provided in Figure 1. Virgin aggregates were also screened and separately
batched to the desired gradation. RAP was heated to the mixing temperature for two
hours in a forced draft oven separately from virgin aggregate; virgin aggregates were
heated for a minimum of four hours. Sasobit was pre-blended into the virgin binder at
a rate of 1.0% by binder mass. Additional Sasobit for the RAP asphalt was added just
before mixing. For each specimen, RAP, virgin aggregate and virgin binder in the
desired proportions was blended in a laboratory bucket mixer. The mixture was then
short term conditioned at the compaction temperature for 90 minutes (Mississippi DOT
specified protocol) before compaction.
LC specimens were prepared with either a Superpave Gyratory Compactor (SGC) or
the Linear Asphalt Compactor (LAC) shown in Figure 2 and described in Howard et al.
(2012) Doyle and Howard (2010). Howard et al. (2012) shows that slabs produced in
the LAC are fairly uniform (typical maximum specimen air void range was
approximately 1%), and that slab bulk density can be correlated reliably to air voids
224

measured with AASHTO T 331 or ASTM D6752. An LAC slab is 29 by 62 cm and can
be 4 to 10 cm thick. Slabs produced for this report were 7.0 to 8.2 cm thick and were
compacted to a given air void level using various numbers of passes and hydraulic
system pressures.
Table 2. Properties of control HMA mixtures.
Control Mixtures
Mixture ID CM 1 CM 2 CM 3 CM 4 CM 5
Prep. Method 2&3 2 1 2 2
Ndes 65 50 85 85 85
NMAS (mm) 1 2 .5 1 2 .5 1 2 .5 1 9 .0 9 .5
Virgin PG Grade 67-22 67-22 67-22 76-22 76-22
W MA No No No No No
TMix (C) 163 163 165 160 163
TComp (C) 146 146 146 149 152
Cr. Gravel (%) 39 52 63 73 79
Limestone (%) 35 12 15 4 5
Sand (%) 10 20 6 7 5
Hyd. Lime (%) 1 1 1 1 1
-12.5 mm (%) 9 5 .4 9 4 .3 9 3 .1 8 9 .4 9 9 .7
-4.75 mm (%) 5 3 .7 6 5 .2 6 0 .0 5 1 .0 5 5 .9
-2.36 mm (%) 3 6 .2 5 0 .1 4 0 .5 3 4 .6 3 4 .4
-0.075 mm (%) 5 .9 5 .3 5 .8 5 .2 5 .8
Combined Gsb 2 .5 1 5 2 .5 5 6 2 .5 1 0 2 .4 9 7 2 .5 1 5
H2O Abs (%) 2 .2 1 .3 1 .8 2 .0 1 .8
RAP (%) 15 15 15 15 10
RAP Source stock stock stock stock stock
RAP AC (%) 5 .6 5 .0 6 .4 5 .6 5 .6
Total AC (%)a 5 .4 5 .2 5 .7 4 .9 5 .8
Virgin AC (%) 4 .6 4 .5 4 .7 4 .1 5 .2
G mm 2 .3 7 8 2 .4 1 0 2 .3 5 0 2 .3 7 6 2 .3 5 6
--Prep. Method 1 = laboratory-mixed (LM) and laboratory compacted (LC).
--Prep. Method 2 = plant-mixed (PM) and laboratory compacted (LC).
--Prep. Method 3 = plant-mixed (PM) and field compacted (FC).
a
Total asphalt content (AC) on mixture mass basis including all RAP asphalt.
Specimens were usually compacted to a required air void level that was measured
with AASHTO T 209 and AASHTO T 331 (vacuum sealing method). T331 produces
higher air-void (Va) results than T 166; see Howard and Doyle (2012) for more
225
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
information. Some specimens were compacted to a design gyration level (Ndes).

Compaction specifics for a given evaluation are stated with test results as applicable.
(a) Overall view. (b) Plates lowered onto mix.
(c) PURWheel block (d) Fatigue beam

(29 by 30 by 7.6 cm). (29 by 12.5 by 7.6 cm).
Figure 2. Linear Asphalt Compactor (LAC) and corresponding specimens produced.
3.3. Test Plan
All six RAP dominated mixtures shown in Table 1 were evaluated for rut resistance,
moisture damage, and durability. These six mixtures were compared to all five control
mixtures. This information was used in conjunction with the literature review to elect to
test only 50% RAP mixtures for cracking resistance. Cracking resistance was compared
to select control mixtures.
3.4. Rut Testing with Asphalt Pavement Analyzer
Asphalt Pavement Analyzer (APA) rut testing was performed to 8,000 cycles on SGC
specimens compacted to target air voids of 7% and 10%. The test temperature was
64C, the wheel load was 445 N, and the hose pressure was 690 kPa. Rut depth was
226
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

reported at 8,000 cycles (RDAPA), along with rutting rate (RRAPA) defined as the rut
depth per 1,000 cycles between 2,000 and 8,000 cycles.
3.5. Rutting and Moisture Damage Testing with PURWheel Laboratory Wheel
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Tracker
The original PURWheel was developed in the 1990s at Purdue University. The original
equipment was donated to Mississippi State University in 2007. It was renovated,
modified, and placed into operation alongside the protocols described in Howard et al.
(2010). The renovations and modifications relative to the original equipment and
protocols are also described in Howard et al. (2010).
Figure 3 shows key PURWheel components. Figure 3a shows an overall view of the
components of the system. Two independently controlled wheel carriages mounted with
4-ply pneumatic tires load LAC compacted slabs for 20,000 passes (10,000 cycles).
LAC compacted slabs are sawn in half and each block (Figure 2c) is tested by one wheel
carriage. The tire inflation pressure is 862 kPa, the wheel load is 1,750 N, the wheel
speed is 31 to 35 cm/sec, and the contact pressure at the beginning of the test is 630
kPa with a gross contact area of 2800 mm2. The resulting net contact pressure is 850
kPa. Figure 3b shows one wheel carriage and pneumatic tire raised and the
corresponding tire print used to determine the gross contact area. A scale is located
below the tire print and used to determine the 1,750 N wheel load. During testing, an
LAC specimen is in place of the scale/tire print.
Two protocols were used: (1) PURWheel-dry (PW-dry), specimen tested at 64oC;
and (2) PURWheel-wet (PW-wet), specimen tested submerged in 64oC water. PW-dry
specimen thickness, test temperature (64C is high temperature binder grading
requirement for much of the southeast US), and 6-hour pre-test conditioning mirror test
protocols for the more commonly used APA so that comparisons can be made between
results from the two devices. Howard et al. (2010) is the origin of the PURWheel test
protocol described in this paper. PW-wet is identical to PW-dry with exception of the
specimen being submerged in 64C water as opposed to being surrounded by 64C air.
The rationale for this approach is to observe the difference in behavior caused by water
under load. Figures 3c and 3d show a tested specimen sliced in half after a wet protocol
test. The tire treads can be seen, alongside shear failure within the mixture. Dry protocol
specimens look similar at the conclusion of testing, though their rut depths and shear
failures are less pronounced.
Four slabs (eight test blocks) were compacted per mixture (2 PW-dry and 2 PW-wet).
Data from corresponding LAC slab test blocks were averaged to create one PW-dry or
PW-wet replicate. The PURWheel continuously collects rut measurements during testing
227

HOW ARD, DOYLE, COX
that are adjusted as described in Howard et al. (2010). One passage of the load wheel
over the specimen was defined as a pass, and rut depth (RDPW) was defined as downward
permanent deformation of the specimen surface relative to the original surface.
Specimen surface uplift was not part of the rut depth. RDPW was reported for PW-dry
and PW-wet tests alongside the corresponding number of passes. The number of passes
to achieve 12.5 mm of rutting was denoted P12.5, and was also reported for PW-dry and
PW-wet tests. For PW-dry only, a linear equation was fit to the post-consolidation data
between 4,000 and 16,000 passes (2,000 and 8,000 cycles); the slope of the fit line was
defined as the rutting rate (RRPW) with units of millimeters/1000 cycles for direct
comparison to APA data. For PW-wet only, the stripping inflection point (SIP) was
calculated by fitting linear equations to rut depth data in the creep region and the
stripping region and using the number of passes at the intersection of these equations as
the SIP.
Computer, Control Box
Hood, Tank Assembly

Closed as when a
test is being conducted (a ) (b)
(c) (d)
Figure 3. PURWheel laboratory wheel tracker.
3.6. Indirect Tensile Strength and Tensile Strength Ratio Testing
Specimens were tested for indirect tensile strength (IDT) for three purposes. All IDT
specimens were conditioned in an environmental chamber. Typically, conditioning
228
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

occurred until an additional specimen with an embedded thermometer indicated

equilibrium had been achieved, and testing occurred at a 50 mm/min load rate.
The first IDT purpose was a durability assessment where 150-mm diameter
specimens compacted to Ndes in the SGC were tested at 25C. The second purpose was
tensile strength ratio (TSR) testing of 100-mm diameter SGC compacted specimens
according to ASTM D4867. The third IDT purpose was for use in conjunction with
bending beam rheometer testing where 100-mm diameter by 63.5-mm tall specimens
were SGC compacted to 3 to 5% air voids (in one instance 100-mm diameter field
compacted specimens were cored and trimmed to 63.5 mm tall and tested). Two
replicates per temperature were tested and averaged after conditioning to four
temperatures (-6, -12, -18, and -24oC).
3.7. Cantabro Durability Testing
Cantabro durability testing was performed by placing un-aged 115-mm tall SGC
compacted (to Ndes) temperature conditioned specimens (25C) inside the chamber of an
L.A. Abrasion machine without the charge of steel spheres and subjecting them to 300
revolutions. Specimens were dry when tested. The change in specimen mass was
denoted mass loss or ML, which is expressed as percentage of original specimen mass.
ML can be thought of as an indirect measure of mixture toughness. Three replicates
were averaged.
3.8. Fatigue Testing with Asphalt Pavement Analyzer
LAC compacted slabs were sawn into four beams (Figure 2d) and tested in the APA for
fatigue resistance. A simply supported beam conditioned to 20oC was tracked with a
steel wheel of a given load (1,100 N and 2,000 N were used herein) along the entire
beam length at a speed of 120 passes per minute. Beams were loaded for 50,000 passes
or until failure, which was 1 mm of deflection occurring in one pass. Two replicate
beams were tested at each load, with the average passes to failure (or 50,000 passes) and
the ultimate deflection reported.
3.9. Bending Beam Rheometer Mixture Beams
BBR mixture beams were sawn from SGC compacted specimens (3 to 5% air voids)
using the protocols described in Doyle (2011). In one instance, BBR beams were sawn
from the middle of field compacted slabs. Five successful test replicates were performed
229
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
where a 4.9 N load was applied for 1,000 seconds after 60-minute conditioning in a
methanol bath. Mixture beam stiffness (S) and m-value were averaged for the five
replicates and reported at eight discrete test times.
4. Results and Analysis
This section presents test results by performance category measured in the laboratory.
When viewing these results it is noteworthy that the authors observed problems with
virgin aggregate coating for the 75% RAP mixtures, but not the 50% RAP mixtures.
This is similar to the findings presented earlier in this paper where plant mixed material
with 70% RAP was not well coated (Brad D. Cruea, unpublished data). Inadequate
coarse aggregate coating can lead to a variety of performance problems over time in
service.
4.1. Rutting
Rutting test results are provided in Table 3. Measured values are reported using
terminology from the experimental program alongside adjusted rut depths where the
APA and PURWheel data have been adjusted to 8% air voids on a T 331 basis. Note
that 8% on a T 331 basis roughly corresponds to 7% on the more conventionally used T
166 basis based on data presented in Howard and Doyle (2012).
Rut depth adjustments for air voids were made using the assumption that the change
in rut depth for a given mixture per unit change in air voids was the same between the
APA and PURWheel. Air void targets are easier to achieve in the SGC than in the LAC.
As an example of the adjustment process, CM1 measured APA rut depths were 0.4 mm
different (6.1 minus 5.7 mm), and this was divided by 2.7% (T 331 measured air voids
for CM1 were 9.9 and 7.2%) for an adjustment of 0.15 mm per percent air voids.
PURWheel measured rut depths were reduced by 0.15 mm per percent air voids above
8% and increased by 0.15 mm per percent air voids below 8%. Each mixture had its
own adjustment factor.
Brown et al. (2001) suggested an 8-mm pass/fail rut depth criteria for APA
conditions used in this study. Generally speaking, this criterion would apply to SGC
compacted specimens at 7% air voids measured with T 166. The APA data adjusted to
8% T 331 voids show only CM2 failing the 8-mm standard. APA rut depth decreased
when RAP content was increased as expected. Interestingly, PURWheel data showed
the 75% RAP mixtures rutting on the order of the 50% mixtures. Overall, rutting data
for RAP dominated mixtures agreed with trends found during the literature review.
230
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 3. Rutting test results.
Rut Depth Rutting Rate Adjusted Rut

(mm) (mm per 1,000 cycles) Depth (mm)
Mixture RDAPA RDAPA RDPW RRAPA RRAPA RRPW AP A PW
7% 10% 7% 10%
ID Va Va Alla Va Va Alla 8% Va 8% Va
0 .2 8 ,
CM1 5 .7 6 .1 4.8, 5.5 0 .3 3 0 .2 7 0 .3 3 5 .8 5 .1
1 0 .2 ,
CM2 7 .9 1 0 .2 1 1 .1 b 0 .4 7 0 .5 5 --- 8 .6 ---
0 .3 2 ,
CM3 3 .4 6 .3 4.8, 7.9 0 .3 3 0 .4 3 0 .6 3 4 .5 7 .4
0 .4 3 ,
CM4 4 .0 8 .4 6.0, 8.3 0 .2 5 0 .4 0 0 .6 0 5 .4 9 .2
6 .0 , 0 .3 9 ,
CM5 6 .7 8 .6 1 1 .7 0 .3 3 0 .4 5 1 .0 9 7 .4 9 .5
0 .5 5 ,
50R-1 3 .9 5 .2 7.0, 7.7 0 .2 8 0 .2 8 0 .5 8 4 .8 7 .9
0 .3 3 ,
50R-2 3 .5 3 .6 5.5, 5.9 0 .2 8 0 .2 7 0 .4 2 3 .6 5 .7
8 .3 , 0 .7 3 ,
75R-1 1 .7 2 .0 1 2 .1 0 .1 2 0 .1 3 0 .8 0 1 .9 8 .6
0 .3 3 ,
75R-2 2 .4 3 .7 4.4, 5.4 0 .1 7 0 .3 0 0 .4 2 3 .1 4 .6
0 .2 5 ,
100R-1 1 .1 2 .0 3.9, 4.2 0 .0 5 0 .1 3 0 .2 8 1 .5 3 .5
0 .4 0 ,
100R-2 1 .2 2 .3 4.9, 6.4 0 .0 7 0 .1 8 0 .4 7 1 .5 5 .0
-- APA rut depths are at 8,000 cycles, PW-dry rut depths are at 20,000 passes unless noted.
a
The average value from each slab is shown (each slab is the average of two blocks).
b
Values shown are at 4,000 passes, and specimen failed prior to 10,000 passes.
It is of interest to note that the test results between the APA and PURWheel
correlated to one another. Details of the correlation are provided in Doyle and Howard
(2013b). Ongoing work by the authors is attempting to utilize this correlation to
improve characterization of moisture damage by using APA dry rutting, since it is
usually readily available, in conjunction with PW-wet testing. The moisture damage
assessments in this paper use PW-dry and PW-wet, since both measurements were
available.
231
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
4.2. Moisture Damage
Moisture damage was evaluated in four manners, with three of them using PW-wet
and/or PW-dry data, and the fourth being the TSR. Figure 4 plots cohesion ratio test
results. The cohesion ratio is the ratio of wet and dry rut depths (RDPW) at 5,000 passes
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(PW-wet divided by PW-dry). Figure 4 also shows plots of the rut depth increase that
occurred between the wet and dry tests at 5,000 passes. The higher the cohesion ratio,
the more affected the specimen by moisture on a relative scale. The increase in rut depth
is shown, since a specimen could have a very high cohesion ratio (e.g., 5) from a largely
meaningless change in rut depths (e.g., 0.1 mm dry to 0.5 mm wet). These two values
should be interpreted together.
3 .5 7
6.3 6.2 Cohesion Ratio
Dry to Wet Rut Depth Increase (mm)

3 .0 5.7 6
2.6 Rut Depth Increase
2 .5 5
2.2
Cohesion Ratio
2.0 2.0
2 .0 3.4 4
3.1 1.6 1.6
1.5 1.5 2.8
1 .5 1.4 3
1.3
2.0
1 .0 1.8 2
0.8
1.0 1.1
0 .5 1
-1.2
0 .0 0
Figure 4. Cohesion ratio calculated from PURWheel data.
One criticism of the TSR is that often only the wet to dry strength ratio is presented
without the dry or wet tensile strength. An example would be a dry tensile strength of
350 kPa and a wet tensile strength of 327 kPa; the TSR is 93% but the material probably
will not perform well with such low values of wet and dry tensile strength. TSR data
from the mixtures investigated in this paper are presented in this section, while dry
tensile strength data is presented alongside Cantabro durability data later in the paper.
Calculating the cohesion ratio at 5,000 passes provides an assessment of the cohesive
(or mastic) response of the specimen to moisture. Some field mixtures experience
noticeable moisture damage without stripping or adhesive failure (Little and Jones,
2003). The approach shown in Figure 4 allows one to evaluate the moisture effects due
to cohesion. Cohesive behaviors could be especially important in RAP dominated
mixtures since the interaction of aged and virgin binders would affect the cohesion ratio
232

prior to stripping. RAP particles often have a very stiff and tightly bonded coating that
could protect against stripping, but inadequate partial blending of aged and virgin
binders could lead to cohesive problems in some mixtures.
The cohesion ratio was 1.5 or greater for all control mixtures, and the rut depth at
5,000 passes increased 3.1 to 6.3 mm due to the presence of moisture. The 50% RAP
mixtures were essentially unaffected by moisture at 5,000 passes. The 75% RAP
mixtures had cohesion ratios near the lower bound of the control mixtures, but their rut
depth changes were considerably less than the control mixtures. The 100% RAP
mixtures had cohesion ratios within the band of the control mixtures, though their rut
depth changes were slightly below the control mixtures band.
Increasing the RAP content above 50% increased the cohesion ratio and also
increased the change in rut depth due to moisture presence (both undesirable). R-1
increased the cohesion ratio by 0.24 per every 10% increase, while R-2 increased the
cohesion ratio by 0.07 per every 10% increase (values based on linear regression with R2
of 0.98 or greater). While no definitive explanation is available, the data indicate 50%
RAP outperformed all other mixtures (controls, 75% RAP, and 100% RAP) in terms of
moisture resistance after 5,000 passes in the PURWheel. Additional work should be
performed in this area as many pavements are affected by moisture, but do not strip.
Stripping is a form of adhesion failure, whereas softening of the asphalt mastic is a form
of cohesive failure. Mastic cohesion could be a key issue in high RAP content mixes
since how the aged and virgin binders interact would affect the cohesive behavior of the
mixture.
Figure 5 shows SIP values calculated from PW-wet data. The 50% RAP mixtures
had SIPs within the band of control values, while 75% and 100% RAP mixtures had
higher SIPs that were similar for a given RAP source (higher SIPs are desirable). There
was no meaningful difference between 75% and 100% RAP mixtures. Figure 5 ranks
the RAP mixtures in reverse order of Figure 4, which is not surprising from the
perspective of stripping, since the aggregates are coated with a very stiff binder. These
two figures, when viewed in conjunction with each other, provide evidence that a
pass/fail moisture damage test with only one parameter measured may not provide
performance indicators over the wide range of conditions a typical mixture would
experience when used for different projects.
233
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
20000
16500 17000
15000 13625
12625
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
12125
SIP (Pass)
10375 10375
10000 9375
6050 6125
5000 4125
Figure 5. Stripping Inflection Point (SIP) calculated from PURWheel data.
Figure 6 is a plot of the ratio of PW-wet and PW-dry P12.5 values (wet is divided by
dry, and higher ratios are desirable) and traditionally measured TSR data. A TSR of
80% passes nearly all agency specifications, and only one of the eleven mixtures
evaluated failed the TSR (100R-2). Interestingly, 100R-2 had the highest SIP under
loaded wheel tracking of any of the eleven mixtures studied. The P12.5 ratio led to the
same conclusions as the SIP data presented in Figure 5 and discussed in the previous
paragraph. The P12.5 ratio and TSR data presented in Figure 6 do not correlate to one
another (R2 of 0.13). A key observation is that PURWheel related indicators in Figures
4 to 6 show mixture differences.
120
P12.5 Rat io (%) TSR (%)
100
P12.5 Ratio or TSR (%)
80
60
40
20
Figure 6. Rutting ratio at 12.5 mm calculated from PURWheel data and TSR.
234

4.3. Durability
Figure 7 shows Cantabro and IDT results performed to assess the durability of RAP
dominated base layers. Control mixture (CM) mass losses (ML) ranged from 11.0 to
15.5%, all of which are in the upper half of the distribution of values reported by Doyle
(2011). The mean mass losses of 12.5-mm and 19.0-mm NMAS mixtures reported by
Doyle (2011) were 9.8 and 10.6%, and these mixtures were randomly obtained
throughout Mississippi. With respect to the Doyle (2011) data, one standard deviation
from the mean would encompass ML values of 7 to 13%, and a 95% confidence interval
would be 4 to 16% when rounded to the nearest percent.
3500
30 3000
ML (%) -- ML specimens compacted to Ndes (Va was 4.1 to 5.2% )
-- ML and St specimens tested at 25 oC and were un-aged 2500
St (kPa)
20 2000
ML (%)
St (kPa)
1500
10 1000
500
0 0
Figure 7. Test results for durability performance.
The 50% RAP ML values were 13.4 and 15.8, which are between one and two
standard deviations higher than the mean mass loss from the values reported in Doyle
(2011) and on the upper end of the control mixtures tested in this paper. This indicates
the 50% RAP mixtures are somewhat more prone to mass loss in the Cantabro durability
test than an average MDOT approved material, but are comparable to some of the higher
mass loss materials that are placed in Mississippi. Indirect tensile strength at failure (St)
data provided the same general trend as the Cantabro data. Control mixture St data were
in the 1,100 to 1,900 kPa range, while the 50% RAP was in the 1,900 to 2,100 kPa
range. No database of values was available to compare tensile strength values, but with
the data available it appears the 50% RAP mixtures had a slightly higher tensile strength
than the upper end of the band of values obtained from the control mixtures. Overall,
the 50% RAP mixtures are not unreasonable compared to currently produced and placed
base mixtures in Mississippi in terms of their Cantabro mass loss and indirect tensile
strength.
235
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
The 75% RAP mixtures had mass loss values of 18.1 to 21.4%, which is well outside
the 95% confidence interval of Doyle (2011) and on average approximately 5% higher
than the 50% RAP mixtures, which is a formidable increase considering the standard
deviations from currently placed materials was 2.6 to 3.0%. Indirect tensile ultimate
strengths were 2,500 to 2,800 kPa, which is a considerable increase from the 50% RAP
or the control mixtures. Durability performance of the 75% RAP mixtures is
questionable.
The 100% RAP mixtures had mass loss values of 17.0 to 31.8%. Both of these
values are outside the 95% confidence interval of Doyle (2011), and they differ by an
enormous amount compared to the remaining Cantabro data. Insufficient quantities of
R-1 or R-2 were available to perform additional testing so a third RAP source (R-3) was
used to design a 100% RAP mixture and the average of three replicates was 33.7%,
which is more in line with R-1 than R-2. The 100% RAP mass loss appears to be at least
as high (R-2) and possibly much higher (R-1 and R-3) than the 75% RAP. Indirect
tensile strength was 2,800 to 3,000 kPa, which is moderately higher than the 75% RAP.
Note that the 100% RAP mixes were designed at 65 gyrations as opposed to 50, which
would likely decrease their durability related performance relative to what would occur
with a 50-gyration design.
Higher tensile strengths and lower mass losses (with the exception of 100% RAP)
were observed with R-2 than R-1 for all three RAP contents, even though the R-2 RAP
is less stiff than R-1 RAP (Figure 1). R-1 mixtures always had more virgin binder than
R-2 mixtures, which is an indication that the virgin binder coating RAP particles is
offsetting some of their behavior in terms of stiffness. This observation should not be
taken to indicate that RAP binder properties have no affect on mixture performance, as
they can affect the amount of virgin binder required if they do not re-liven sufficiently at
the production temperature. Doyle et al. (2012) explores this issue with the three RAP
sources investigated in this section.
Overall, RAP dominated mixtures appear to be less durable than a typical MDOT
mixture used in present day. The 50% RAP mixes are not unreasonable in terms of
durability, but 75% and 100% RAP mixes were less durable by noticeable amounts
based on the data in this paper. For non-surface mixes, however, this may not be a first
order concern.
4.4. Cracking Test Results
The cracking investigation considered only 50% RAP mixtures. Insufficient resources
were available to test all mixtures for cracking resistance. The 50% RAP mixtures were
236
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

selected over 75% and 100% RAP mixtures since review of literature indicated they
were more prevalent.
4.4.1. Fatigue Test Results
Fatigue test results are shown in Figure 8, organized into 3 to 5%, 6 to 8% and 10 to
12% air void groups. All RAP dominated mixtures were tested at 3 to 5% voids;
insufficient RAP was available to compact specimens at higher air voids. In addition, it
was desired to test CM1-LC at all air void levels, but the LAC was unable to compact
this mixture to 3 to 5% air voids. CM5 was successfully compacted at all air void levels
and was used to provide continuity to the discussion, especially with regard to CM1.
At 1,100 N and 3 to 5% voids, all RAP dominated and control mixtures lasted 50,000
passes. At 2,000 N and 3 to 5% voids, the RAP dominated mixtures performed in a
comparable manner to CM4 and better than CM5. CM5 outperformed all controls tested
at 6 to 8% and 10 to 12% voids. Since the RAP dominated mixtures performed better
than CM5 at 3 to 5% air voids and CM5 outperformed CM1 at higher air voids,
comparable to favorable performance of CM1 and RAP dominated mixtures seems
logical. The key observation related to RAP dominated mixtures from fatigue testing
was that they were comparable to control mixtures. This observation agrees with the
overall body of literature presented earlier in this paper.
CM1-LC outperformed CM1-FC by factors of 1.7 and 2.8 for 2,000 N and 1,100 N
loads, respectively. Note that CM1-FC was compacted at higher air void levels, since it
was from an emergency paving project. Field compacted material being less load-
resistant than laboratory compacted material is not surprising. This behavior is also
observed in the BBR data presented in the next section.
Values are deflection (mm) at end of test. "---" = no data recorded. Values are deflection (mm) at end of test. "---" = no data recorded.
--- 0.8 0.3 0.2 --- --- 3.1 4.0 --- 6.1 6.1 3.5 0.6 0.8 5.5 4.4 4.2 5.4 --- 3.9
50 50
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
40 40
Passes (103)
30
Passes (103)
30
4.4% Va
20 20
10.9% Va
7.7% Va
11.5% Va
7.6% Va
6.6% Va
10.8% Va
10.9% Va
10.8% Va
11.5% Va
4.4% Va
3.4% Va
4.6% Va
7.7% Va
6.6% Va
3.4% Va
7.6% Va
3.4% Va
4.6% Va
3.4% Va
10 10
0 0
(a) Fatigue tests at 1,100 N (b) Fatigue tests at 2,000 N

Figure 8. APA fatigue test results.
237

HOW ARD, DOYLE, COX
4.4.2. BBR and IDT Test Results
BBR and IDT data are shown in Figures 9 and 10. The 50% RAP and CM1-LC
mixtures were compacted to 3 to 5% air voids; the CM1-FC mixture was compacted to
approximately 10.8% air voids. In general, for mixture stiffness (S) and m-value data,
50R-1 is similar to the CM1-LC control at all test temperatures and 50R-2 is not greatly
different than the CM1-LC mixture. Indirect tensile strength data (St) for these mixtures
at -6, -12 and -18C are similar, but St is lower at -24C.
In contrast, CM1-FC data indicate noticeably lower S and St and higher m-values
relative to the other mixtures at all temperatures. This difference is likely due to a
combination of higher air voids and differences in aggregate structure between
laboratory and field compaction. Not all of the difference is explained by air voids since
the BBR mixture beam test is relatively insensitive to changes in air voids within a range
of 2.5 to 8% (Doyle and Howard, 2013a) and the difference is smaller at -24C where
the aggregate structure contributes less to mixtures stiffness than at warmer
temperatures. Field compacted S is around 70% lower than laboratory compacted at -6C
and around 50% lower at -24C. If the difference were entirely due to air voids, it would
be expected to be constant for all temperatures.
Thermal cracking analysis of the BBR and IDT data was performed to calculate
critical cracking temperature (Tcr) of the mixture as described in Doyle and Howard
(2013a). For 50R-1, 50R-2, CM1-LC and CM1-FC, these temperatures were -30, -28,
-26 and -32C, respectively. Interestingly, the CM1-FC had the best overall performance
and CM1-LC the worst. Fifty percent RAP mixture performance was between that of
the controls. Differences in RAP asphalt properties did not have much effect on the final
composite mixture. As with fatigue data, the primary observation related to RAP
dominated mixtures from BBR and IDT testing was that they were comparable to control
mixtures.
Tcr analysis was performed on ten 9.5-mm NMAS surface mixtures in Doyle and
Howard (2012) consisting of six controls, two 25% RAP and two 50% RAP mixtures.
The 25 and 50% RAP mixtures used the same RAP, virgin aggregate and virgin binder
sources as this paper and selection of control mixtures followed the same general logic
as used herein. Tcr for control mixtures ranged from -27 to -31C, 25% RAP mixtures
averaged -26C and 50% RAP mixtures averaged -22C. In Doyle and Howard (2012),
25% RAP mixtures performed similarly to controls but 50% RAP performed noticeably
worse; while in this paper, 50% RAP mixtures made with the same raw materials
performed about as well as the controls in either data set. By using a slightly different
combination of the same materials, it was possible to improve the Tcr temperature of
50% RAP mixtures by 7C on average to produce values that fell within the range of
current practice 15% RAP mixtures. This evidence suggests that virgin asphalt binder
238
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

grade is not the only variable that can affect cracking potential of high RAP content
mixtures.
100 100
Mixture St (kPa) Mixture St (kPa)
50R-1 4187 50R-1 4233
50R-2 5132 50R-2 4443
CM1-LC 4882 CM1-LC 4721
Mixture Stif f ness (GPa)

CM1-FC 2703 (20.04) CM1-FC 2611
(16.60) (16.94) (12.25)
(13.20) (8.16)
10 10
(5.01) (7.13)
50R-1 50R-1
(3.24) (5.23)
50R-2 50R-2
(1.30)
CM1-LC CM1-LC
(2.76)
CM1-FC CM1-FC
1 1
1 10 100 1000 1 10 100 1000
Time (sec) Time (sec)
(a) -6C test temperature. (b) -12C test temperature.

100 100
(24.68) (17.75)
(20.93) (1 4 . 0 2 ) (22.09)
(18.45)
10 10
(10.06) (11.77) (9.99)
(8.09)
50R-1 50R-1
Mixture St (kPa) Mixture St (kPa) (7.30)
50R-2 50R-2
50R-1 4347 (5 . 1 0 ) 50R-1 2784
CM1-LC 50R-2 4468 CM1-LC 50R-2 3863
CM1-LC 4337 CM1-LC 3792
CM1-FC CM1-FC 2849 CM1-FC CM1-FC 2020
1 1
1 10 100 1000 1 10 100 1000
(c) -18C test temperature. (d) -24C test temperature.
Figure 9. BBR stiffness data and IDT tensile strength by test temperature.
All thermal cracking analysis was performed with mixtures that were not subjected to
any long term conditioning or aging protocols. The mixtures represent performance
potential just after construction. It is not known how these high RAP mixtures would
behave relative to controls after long term field aging. Their cracking potential may
increase at a similar rate to controls, or they may age at a slower rate, since much of the
asphalt binder they are composed of has already experienced considerable aging during
its first period of service; this may be a productive avenue for future research.
239
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
1 .0 0 1 .0 0
(0 . 3 9 8 )
(0.277)
(0.265)
(0.168) (0.187)
m-value
(0.121)
m-value
0 .1 0 (0 . 2 0 8 )
0 .1 0
(0.139) (0.098) (0.137)
(0.088)
50R-1 50R-1 (0.069)
50R-2 50R-2
CM1-LC CM1-LC
CM1-FC CM1-FC
0 .0 1 0 .0 1
1 10 100 1000 1 10 100 1000
(a) -6C test temperature. (b) -12C test temperature.
1 .0 0 1 .0 0
(0.202)
(0 . 1 3 1 )
(0.133)
m-value
m-value
0 .1 0 (0.084) (0 . 0 9 5 )
0 .1 0
(0 . 0 6 8 )
(0.121)
50R-1 (0.078) 50R-1
(0 . 0 6 2 ) (0 . 0 9 3 )
50R-2 (0.047) 50R-2 (0 . 0 4 0 )
CM1-LC CM1-LC
CM1-FC CM1-FC
0 .0 1 0 .0 1
1 10 100 1000 1 10 100 1000
(c) -18C test temperature. (d) -24C test temperature.
Figure 10. BBR m-value data by test temperature.
Availability of RAP in present day does not indicate a significant surplus based on the
survey by Hansen and Newcomb (2011). Lack of a RAP surplus coupled with
performance data presented in this paper indicates mixtures with more than 50% RAP do
not, generally speaking, add value to the highway system in present day (RAP available
can be fully used in other manners). This is especially true when the data presented by
Copeland (2011) is considered, where less than half the US states use RAP contents in
excess of 20%. The data presented in this paper indicates RAP would be better served
by increasing moderate RAP content mixtures (e.g., 20%) to high RAP content mixtures
(e.g., 35%) for several projects rather than using RAP dominated mixtures (e.g., 75%) on
a few projects. An exception could be specific projects (likely fairly small) that occur
240
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

less frequently in areas where RAP is abundant (a large city could be an example). For
these applications, 100% RAP mixtures could be useful.
Specific performance related conclusions follow:
Rutting data for RAP dominated mixtures agreed with trends found during the
literature review; rutting was not problematic.
Adjusted APA and PURWheel rut depths correlated with each other, which could be
useful for future moisture damage assessments.
Fifty percent RAP outperformed all other mixtures (controls, 75% RAP, and 100%
RAP) in terms of moisture resistance after 5,000 PURWheel passes (named
cohesion ratio).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Stripping inflection point (SIP) and P12.5 ratios led to the same conclusions, which
were opposite those from the cohesion ratio (75 to 100% RAP was better than 50%
RAP, but 50% RAP was comparable to control mixtures). This result is not
surprising, since RAP aggregates are coated with a very stiff binder. Their
differences support the position that a pass/fail moisture damage test with only one
parameter measured may not provide performance indicators over a wide range of
conditions, such as a typical mixture used on several projects would experience.
PURWheel related indicators showed differences in moisture related performance.
The 50% RAP mixtures were somewhat more prone to mass loss in the Cantabro
test than an average Mississippi DOT approved material, but were comparable to
some of the higher mass loss materials placed in Mississippi. The 75% and 100%
RAP mixtures had mass loss values outside the 95% confidence interval of control
mixtures.
APA fatigue test results showed 50% RAP mixtures were comparable to control
mixtures at two load levels.
Critical cracking temperature (Tcr) analysis using BBR and IDT data showed 50%
RAP mixture values to be between those of laboratory and field compacted controls.
Plant produced RAP dominated mixtures documented during the literature review
with 70% RAP were not well coated, which supported the visual appearance of 75%
RAP mixtures produced in the laboratory for this project. Inadequate coating is a
concern for long term mixture performance.
6. Acknowledgements
Thanks are owed to the Mississippi DOT Research Division, Ergon Asphalt and
Emulsions, Inc., and Asphalt Technology Consulting. Dr. Thomas D. White is thanked
for assistance in a variety of manners including securing the original PURWheel
equipment and for technical guidance related to wheel tracking and moisture damage.
Permission to publish was granted by the Director, Geotechnical and Structures
Laboratory, U.S. Army Engineer Research and Development Center.
241

HOW ARD, DOYLE, COX
7. References
Asphalt Institute, The Asphalt Handbook: MS-4, 7th Ed., Lexington, KY: Asphalt
Institute, 2007.
Austerman, A. J., Mogawer, W. S., and Bonaquist, R., Evaluating the Effects of Warm
Mix Asphalt Technology Additive Dosages on the Workability and Durability of
Asphalt Mixtures Containing Recycled Asphalt Pavement, Proceedings of
Transportation Research Board 88th Annual Meeting, Paper No. 09-1279,
Washington DC, 1115 January, 2009.
Baumgardner, G. L., Hemsley, J. M., Jordan, W., and Howard, I. L., Laboratory
Evaluation of Asphalt Mixtures Containing Dry Added Ground Tire Rubber and a
Processing Aid, Asphalt Paving Technology: Journal of the Association of Asphalt
Paving Technologists, Vol. 81, 2012, pp. 507539.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Boggs, W., Customers Warm up to Green System, Warm-Mix Asphalt: Contractors
Experiences, Information Series 134, National Asphalt Pavement Association,
Lanham, MD, 2008, pp. 1617.
Brown, E. R., Kandhal, P. S., and Zhang, J., Performance Testing for Hot Mix Asphalt,
NCAT Report 01-05, National Center for Asphalt Technology, Auburn AL, 2001.
Brown, D. C., When Black is Green-Placing Shoulders With Warm Mix and 50 Percent
RAP, Asphalt Pavement Magazine, Vol. 16, No. 6, 2011, pp. 6068.
Cascione, A. A., Williams, R. C., Buttlar, W. G., Ahmed, S., Hill, B., Haugen, D. S., and
Gillen, S., Laboratory Evaluation of Field Produced Hot Mix Asphalt Containing
Post-Consumer Recycled Asphalt Shingles and Fractionated Recycled Asphalt
Pavement, Asphalt Paving Technology: Journal of the Association of Asphalt
Paving Technologists, Vol. 80, 2011, pp. 377418.
Copeland, A., Reclaimed Asphalt Pavement in Asphalt Mixtures: State-of-the-Practice,
Report No. FHWA-HRT-11-021, Federal Highway Administration, 2011.
Copeland, A., DAngelo, J., Dongr, R., Belagutti, S., and Sholar, G., Field Evaluation
of High Reclaimed Asphalt Pavement-Warm Mix Asphalt Project in Florida: Case
Study, Transportation Research Record: Journal of the Transportation Research
Board, No. 2179, 2010, pp. 93101.
Doyle, J. D., Characterization of Reclaimed Asphalt and Performance Based Evaluation
of its use in Recycled Mixtures, Ph.D. Dissertation, Mississippi State University,
Mississippi State, MS, 2011.
Doyle, J. D., and Howard, I. L., Linear Asphalt Compactor Operators Manual, Manual
Number CMRC M 10-1 Version 2, Construction Materials Research Center,
Mississippi State University, 2010.
242

Doyle, J. D., and Howard, I. L., Evaluation of the Cantabro Durability Test for Dense
Graded Asphalt, Proc. of Geo-Frontiers 2011: Advances in Geotechnical
Engineering (GSP 211), Mar 13-16, Dallas, TX, 2011, pp. 45634572.
Doyle, J. D., and Howard I. L., Thermal Cracking Potential of High RAP WMA
Evaluated with Bending Beam Rheometer Mixture Beam Test, Journal of Testing
and Evaluation, Vol. 41, No. 2, 2013a, In Press.
Doyle, J. D., and Howard I. L., Rutting and Moisture Damage Resistance of High RAP
Warm Mixed Asphalt: Loaded Wheel Tracking vs. Conventional Methods, Asphalt
Paving Technology: Journal of the Association of Asphalt Paving Technologists,
Vol. 82, 2013b, Accepted.
Doyle, J. D., Howard, I. L., and Robinson, W. J., Prediction of Absorbed, Inert, and
Effective Bituminous Quantities in Reclaimed Asphalt Pavement, Journal of
Materials in Civil Engineering, Vol. 24, No. 1, 2012, pp. 102112.
Doyle, J. D., Mejias-Santiago, M., Brown, E. R., and Howard, I. L., Performance of
High RAP-WMA Surface Mixtures, Asphalt Paving Technology: Journal of the
Association of Asphalt Paving Technologists, Vol. 80, 2011, 419457.
Epps, J. A., Sebaaly, P. E., Penaranda, J., Maher, M. R., McCann, M. B., and Hand, A.
J., Compatibility of a Test for Moisture-Induced Damage with Superpave Volumetric
Mix Design, NCHRP Report 444, Transportation Research Board, 2000.
Hansen, K. R., and Newcomb, D. E., Asphalt Pavement Mix Production Survey on
Reclaimed Asphalt Pavement, Reclaimed Asphalt Shingles, and Warm-mix Asphalt
Usage: 2009-2010, Information Series 138, National Asphalt Pavement Association,
2011.
Hodo, W. D., Kvasnak, A., and Brown, E. R., Investigation of Foamed Asphalt (Warm
Mix Asphalt) with High Reclaimed Asphalt Pavement (RAP) Content for
Sustainment and Rehabilitation of Asphalt Pavement, Proc., Transportation
Research Board 88th Annual Meeting, Paper No. 09-3789, Washington, DC, 2009.
Howard, I. L., and Doyle, J. D., Comparison of Available Density Measurement
Methods Applicable for Design and Construction, Proc., Transportation Research
Board 91st Annual Meeting, Paper 12-0564, Washington, DC, Jan 2226, 2012.
Howard, I. L., Doyle, J. D., and Barham, J. M., Uniformity, Repeatability, and
Permanent Deformation Resistance of Slabs Produced With the Linear Asphalt
Compactor, Advances in Civil Engineering Materials, Vol. 1, No. 1, 2012, pp. 117.
Howard, I. L., Doyle, J. D., White, T. D., Ivy, J., and Booth, O., PURWheel Laboratory
Wheel Tracker Operators Manual, Manual No. CMRC M 10-2 Version 1,
Construction Materials Research Center, Mississippi State University, 2010.
243
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
Huang, B., Li, G., Vukosavljevic, D., Shu, X., and Egan, B.K., Laboratory
Investigation of Mixing Hot-Mix Asphalt with Reclaimed Asphalt Pavement,
Transportation Research Record: Journal of the Transportation Research Board,
No. 1929, 2005, pp. 3745.
Little, D., and Jones, D. R., Chemical and Mechanical Processes of Moisture Damage
in Hot-Mix-Asphalt Pavements, Moisture Sensitivity of Asphalt Pavements: A
National Seminar, Transportation Research Board, 2003.
Loria, L., Haji, E. Y., Sebaaly, P. E., Marton, M., Kass, S., and Liske, T., Performance
Evaluation of Asphalt Mixtures with High Recycled Asphalt Pavement Content,
2208, 2011, 7281.
Lottman, R. P., Predicting Moisture-Induced Damage to Asphaltic Concrete, NCHRP
Report No. 192, Transportation Research Board, 1978.
Lottman, R. P., Predicting Moisture-Induced Damage to Asphaltic Concrete Field
Evaluation, NCHRP Report No. 246, Transportation Research Board, 1982.
Mallick, R. B., Bradley, J. E., and Bradbury, R. L., An Evaluation of Heated Reclaimed
Asphalt Pavement Material and Wax-Modified Asphalt for Use in Recycled Hot
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Mix Asphalt, Transportation Research Record: Journal of the Transportation

Research Board, No. 1998, 2007, pp. 112122.
Mallick, R. B., Kandhal, P. S., and Bradbury, R. L., Using Warm Mix Asphalt
Technology to Incorporate High Percentage of Reclaimed Asphalt Pavement
Material in Asphalt Mixes, Transportation Research Record: Journal of the
Transportation Research Board, No. 2051, 2008, pp. 7179.
Mallick, R. B., OSullivan, K. A., Mingjiang, T., and Frank, R., Why Not Use
Rejuvenator for 100% RAP Recycling? Transportation Research Board 89th Annual
Meeting, Paper 10-1838, Washington, DC, 2010.
Mejas-Santiago, M., Doyle, J. D., Howard, I. L., and Brown, E. R., Moisture Damage
Potential for Warm Mix Asphalt Containing Reclaimed Asphalt Pavement,
Proceedings of 2nd International Warm Mix Conference, St. Louis, MO, 1113
October 2011.
Middleton, B., and Forfylow, R. W., Evaluation of Warm-Mix Asphalt Produced with
the Double Barrel Green Process, Transportation Research Record: Journal of the
Mogawer, W. S., Austerman, A. J., Bonaquist, R., and Roussel, M., Performance
Characteristics of Thin Lift Overlay Mixtures Containing High RAP Content, RAS,
and Warm Mix Asphalt Technology, Proc., of Transportation Research Board 90th
Annual Meeting, Paper No. 11-1859, Washington DC, 23-27 January 2011.
244

Santos, L.G.P., Baptista, A.M.C., and Capito, S.D., Assessment of the Use of Hot-Mix
Recycled Asphalt Concrete in Plant, Journal of Transportation Engineering, Vol.
136, No. 12, 2010, pp. 11591164.
Shoenberger, J. E., and Demoss, T. A., Hot-Mix Recycling of Asphalt Concrete
Airfield Pavements, International Journal of Pavement Engineering, Vol. 6, No. 1,
2005, pp. 1726.
Shu, X., Huang, B., and Vukosavljevic, D., Laboratory Evaluation of Fatigue
Characteristics of Recycled Asphalt Mixture, Construction and Building Materials,
Vol. 22, 2008, pp. 13231330.
Su., K., Hachiya, Y., and Maekawa, R., Study on Recycled Asphalt Concrete for use in
Surface Course in Airport Pavement, Resources, Conservation and Recycling, Vol.
54, 2009, pp. 3744.
Tabakovi, A., Gibney, A., McNally, C., and Gilchrist, M. D., Influence of Recycled
Asphalt Pavement on Fatigue Performance of Asphalt Concrete Base Courses,
Journal of Materials in Civil Engineering, Vol. 22, No. 6, 2010, pp. 643650.
Tao, M., and Mallick, R. B., Effects of Warm-Mix Asphalt Additives on Workability
and Mechanical Properties of Reclaimed Asphalt Pavement Material,
No. 2126, 2009, pp. 151160.
Timm, D. H., Willis, J. R., and Kvasnak, A., Full-Scale Structural Evaluation of
Fatigue Characteristics in High RAP and Warm Mix Asphalt, Transportation
Research Record: Journal of the Transportation Research Board, No. 2208, 2011,
pp. 5663.
Tunnicliff, D. G., and Root, R. E., Use of Antistripping Additives in Asphaltic Concrete
Mixtures. NCHRP Report No. 274, Transportation Research Board, 1984.
Mixtures: Field Evaluation, NCHRP Report No. 373, Transportation Research
Board, 1995.
West, R., Kvasnak, A., Tran, N., Powell, B., and Turner, P., Testing of Moderate and
High Reclaimed Asphalt Pavement Content Mixes: Laboratory and Accelerated Field
Performance Testing at the National Center for Asphalt Technology Test Track,
No. 2126, 2009, pp. 100108.
245
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
8. Discussion
MR. DALE DECKER: In the words of one red neck to another, all yall done good.
DR. HOWARD: Thank you.
MR.DECKER: Thats a joke between Isaac and me.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. HOWARD: Yes.

MR. DECKER: At any rate, one of the things, Isaac, that I have been struggling with and
trying to get some data on I guess I havent done too much looking but its kind of a
concern. Where is the point at which the percentage of asphalt RAP is outweighed by the
extra expense to heat the material? Because we are having to superheat the aggregate to
get the RAP dry. Have you looked at that at all or has anybody looked really at that? I
havent seen any data like that. But what is the energy cost increase as the RAP cost
increases and where is the point of diminishing return?
MR. FRANK FEE: We have not looked at that.
MR. DECKER: You understand what I am saying, though, right?
DR. HOWARD: I do.
MR. DECKER: My guess is there is probably going to be some point at which that could
come into play.
DR. HOWARD: I would expect there is, but we have not looked into it.
MR. DECKER: Okay.
DR. HOWARD: Our biggest concern from the perspective of the economics of the
situation was the value of the RAP binder and trying to maximize its value. When we
looked at it, all factors considered, you make good point, as the RAP content goes up,
the utilization of that binder does become a little bit more of a challenge in some
applications. If we can comfortably produce RAP mixes in the 30 to 35% range, and use
up the RAP that is available, we felt like that is a very good balance of bitumen use and
plant wear and tear. One of the concerns we had is the wear and tear on plants. If you
significantly increase RAP content, what is that going to do to the plant itself, and that is
definitely not an area where I have great expertise by any stretch of the imagination, but
it is something we were concerned about.
MR. DECKER: Your overall conclusion about the percentage is certainly correct as
well. I mean there are some states, particularly on the East Coast, that are just buried in
RAP at this point, and that is part of the challenge.
246

DR. HOWARD: In some of those areas, what we were trying to recommend was 100%
RAP mixes can be made, and they have been made. One study cited documents some
pretty good data on the 100% RAP mixes, and our investigation suggests that if you
want to increase RAP contents above around 50%, going all the way to 100% seems
logical. Progressively increasing RAP content
MR. DECKER: Okay. Thank you very much, Isaac.
DR. HOWARD: Thank you.
DR. MANSOUR SOLAIMANIAN: I guess what I ask is kind of related to what Dale
asked on superheated aggregates. Maybe you mentioned and I missed it. If I did, Im
sorry. But in terms of preparing your specimen at no RAP, 50% RAP, and 100% RAP,
did you use the same conditioning temperatures or did you change the temperature?
DR. HOWARD: When you say conditioning, are you talking about mixing and
premixing, or are you talking about short-term aging?
DR. SOLAIMANIAN: Short-term aging (duration), as well as the temperature magnitude.
Correct, both of them.
DR. HOWARD: Laboratory short-term aging was for 90 minutes according to
Mississippi DOT standard protocol.
DR. SOLAIMANIAN: But was it all the same duration and temperature for all of them?
DR. HOWARD: Four of the five control mixes were plant produced. We sampled them
out of a paver, and so the conditioning was whatever occurred at the project.
DR. SOLAIMANIAN: Okay. So your compaction temperatures were the same or did
you change it for different RAP contents?
DR. HOWARD: Compaction temperatures varied. For the hot mix controls, compaction
temperatures were 146 to 152oC. For the RAP dominated warm mixes, compaction
temperatures were 116oC. We were trying to simulate warm mix versus hot mix on the
compaction. Preparation of all laboratory materials followed a very similar structure.
DR. SOLAIMANIAN: I guess in comparing your zero RAP versus 50 versus 100, what
were the temperatures?
DR. HOWARD: We did not have any zero RAP.
DR. SOLAIMANIAN: Your 50% RAP versus 100% RAP, was it the same temperature?
DR. HOWARD: Yes, if you are asking if when we made a 50% RAP mix did we use the
same temperatures as we did on the 100% RAP mix.
DR. SOLAIMANIAN: Thats right.
DR. HOWARD: Yes, we did.
247
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
DR. SOLAIMANIAN: Well, what I was trying to say was when you have your control,
50% RAP, and 100% RAP, if you use the same temperature, that might impact in a way
what you observe in your percent mass loss being, for example, in your 50% RAP. So
there might be some optimization of compaction in terms of temperature so that you get
a more coherent material and minimize that loss that you observed in your 50%.
DR. HOWARD: The 50% RAP mixes had less mass loss than the 75 or 100% RAP
mixes, but that is a good comment. Thank you.
DR. SOLAIMANIAN: Sure.
MR. FRANK FEE: Isaac, what was the percent replacement binder in the RAP mixes
that you used? Do you know the percentage of new binder that you put in there?
DR. HOWARD: Virgin binder contents were 1.2 to 2.0% for 100% RAP mixes, 1.9 to
3.3% for 75% RAP mixes, and 3.3 to 4.6% for 50% RAP mixes. The binder
replacement would change quite a bit, depending on the quantity.
MR. FEE: Im sorry, what was the 100% RAP, how much?
DR. HOWARD: Virgin binder contents were 1.2 to 2.0% for 100% RAP mixes. We did
not do percent replacement, rather determined how much binder by mass was needed for
the mix.
MR. FEE: Which theoretically would have been 1% over optimum, right?
DR. HOWARD: If all the material relivens and if the gradation doesnt change, but
neither happens The binder contents that we added agreed very well with some of those
other assessments. But, absolutely, the entire RAP does not reliven, especially in these
stiff binders. When the low PG grade number is a +2, you're not getting all that bitumen
back.
PROFESSOR GABRIELE TEBALDI: In your work you consider the amount of RAP
from 20% to 100%. But RAP is a very heterogeneous material. In what way can you link
the different kind of RAP, or in what way can you be sure that the RAP that you have
when you start your mix is the same RAP when you finish producing your mix? How
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
can you say that RAP is more or less the same in terms of amount of bitumen, aggregate
source, and so forth? Do you have a protocol to evaluate the RAP?
DR. HOWARD: Im not sure that I understand your question.
PROF. TEBALDI: My question is, if you have RAP, RAP can come from different
ways. Also, in the same road, you can have different aging, different kind of particles
sizes due to the milling process. So in what way can you analyze and classify the RAP to
be sure that in your mix design you can achieve the correct grading curve, for example,
or the correct amount of bitumen?
248

DR. HOWARD: We looked at that in some other studies; it was not a part of this paper.
But what we did in this paper is we sampled the RAP sources. We just went and got an
enormous sample and stored it for years. We had the same RAP sample for an extended
period of time. We sieved the RAP into eleven different sizes and re-created the
gradation of interest. For this study, we just made the assumption that we were taking a
homogeneous sample. Now for plant production, obviously that wont work, but that
wasnt really what we were trying to do in this paper.
PROF. TEBALDI: So you characterized your RAP in terms of size of particle?
DR. HOWARD: Yes, we considered RAP particle sizes. We measured asphalt content,
true grade, viscosity, angularity, gradation all sorts of properties. But as far as, if you
took a scoop and dipped into the barrel that is minus 9.5 mm and plus 4.75 mm, we
assumed one scoop was identical to another scoop. And there is some approximation in
that, but thats what we did.
PROF. TEBALDI: Thanks:
DR. HOWARD: Am I answering your question?
PROF. TEBALDI: Yes.
MR. ALEXANDER JACKSON: I saw that it said your nominal max aggregate size was
up to 19 mm, and I was wondering if there was any concern about cutting a 6.25 mm
thick BBR beam from those materials.
DR. HOWARD: We didnt cut any BBR beams from the 19 mm mix.
MR. JACKSON: Okay. So it was just the 9.5 that the BBR beams were made from?
DR. HOWARD: We cut BBR beams from a 12.5 mm control and 50% RAP mixes.
MR. JACKSON: Okay
DR. HOWARD: Raul Velasquezs dissertation in 2009 from the University of
Minnesota is a good source for additional information associated with aggregate size
effects in BBR mixture beams.
MR. JACKSON: Okay. Is there any concern about getting a representative sample size
from that?
DR. HOWARD: There perhaps is some. Velasquez (2009) and Zofka et al. (AAPT
Volume 74) are good resources on this issue. We used the literature suggestions on
several of those items when we started, but we felt it was a reasonable approach, for low
temperatures. As the temperature increases, that becomes an increasingly problematic
issue. For example, if you wanted to do torsion bar testing at high temperatures, the
aggregate size can become more of a concern than it would be at lower temperatures.
MR. JACKSON: Thank you very much.
249
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HOW ARD, DOYLE, COX
PROFESSOR IMAD AL-QADI: In comparing to the control, was the comparison

including VMA and other volumetrics?
DR. HOWARD: We were meeting thresholds. We tried to get reasonably comparable
ranges of overall volumetric conditions. With regard to VMA in high RAP mixes,
calculation is not trivial, for example what is the RAP Gsb and how much of the binder
has not relivened and how is it characterized? All volumetric properties were not
intentionally paired.
PROF. AL-QADI: Right. It is important to compare other volumetrics such as VMA, in
addition to air voids. My second question is you said that the fatigue results are similar.
Is that when binder change is considered as in the RAP case? Did you use different
binder?
DR. HOWARD: We used the same virgin binder grade in all the RAP mixes.
PROF. AL-QADI: So you get the same or similar fatigue life even with 50% RAP using
the same binder?
DR. HOWARD: The fatigue testing we did was not a standard three-point beam fatigue
test. We dont have that equipment, and we do have the APA set up to run fatigue. We
exposed mixes to 50,000 cycles, which is what the APA equipment will do, and we did it
at different loads. Under that test and those conditions, we did not see great differences.
If you conducted some of your other types of fatigue tests, you might observe different
behaviors, but the data we collected generally aligned with what we were seeing in the
literature for those ranges of RAP contents. Some of those authors did use beam fatigue
type equipment, and so we stopped the investigation.
PROF. AL-QADI: So you were using the APA?
DR. HOWARD: That is correct.
PROF. AL-QADI: And at what temperature?
DR. HOWARD: At 20C.
PROF. AL-QADI: Okay. Thank you.
DR. HOWARD: Thank you. Yes, we did less for fatigue than we did anywhere else with
regard to the overall characterization
DR. BAOSHAN HUANG: I just want to make a very quick comment on your APA
fatigue test. As we all know, if you were using a finger to tap four million times, you
dont see the difference for two materials but that does not mean that the two materials
do not have any fatigue life difference. So in a similar way with the APA test probably
you're in the range kind of like rationale because youre really not doing the right way to
test the fatigue. We have done some work with APA testing viscoelastic properties as
well as the fatigue testing by measuring its resilience as well as creep like properties
250
--`,`,,,``,,```,,`,,,,,`

after every load cycle. That way you might be able to capture the true fatigue properties
of asphalt mixtures. Thats my comment. Thank you.
DR. HOWARD: Thank you. Yes, we are not endorsing the APA as the fatigue
characterization tool. One item to note though is once we increased the load from 1100
N to 2000 N, some of the control and RAP dominated mixes failed prior to 50,000
passes so we started to be able to see some differences in the mixes, although it wasnt a
detailed fatigue characterization. More work would be needed to make detailed
statements. Thats why I didnt.
251
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluation of High RAP-WMA Asphalt
Rubber Mixtures
Walaa S. Mogawera*, Alexander J. Austermana, Louay Mohammadb,
and M. Emin Kutayc
a
University of Massachusetts, Highway Sustainability Research Center, Fall River, MA
02723, USA
b
Louisiana State University, Civil and Environmental Engineering, Baton Rouge, LA
70803, USA
c
Michigan State University - Civil and Environmental Engineering, East Lansing, MI
4 8 8 2 , US A
ABSTRACT: This study focused on evaluating the stiffness, fatigue cracking, reflective cracking,
rutting, moisture damage, and workability of asphalt rubber surface mixtures with RAP contents
up to 40% with and without a WMA technology. RAP increased the stiffness of the mixtures;
however, WMA mitigated that increase. The addition of RAP had an adverse effect on the
resistance to fatigue and reflective cracking of the mixtures. This effect was magnified with the use
of WMA regardless if RAP was incorporated in the mixture. Therefore it was recommended to
further investigate the proper drop in temperatures for asphalt rubber mixtures that incorporate
WMA. All mixtures passed the rutting and moisture damage test. The WMA technology improved
the workability of the mixtures.
KEYWORDS: asphalt rubber, gap graded, reclaimed asphalt pavement, warm mix asphalt, fatigue
1. Background
The hot mix asphalt industry is constantly seeking technological improvements to

produce sustainable, cost effective, and environmentally friendly mixes. A logical
approach to achieve such mixtures is to use readily available recycled materials like
Reclaimed Asphalt Pavements (RAP) and Ground Tire Rubber (GTR) (Cooper, 2008).
Also, to advance environment stewardship, the industry has been using Warm Mix
The oral presentation was made by Professor Mogawer.
http://dx.doi.org/ 10.1080/14680629.2013.812846
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,
253

MOGAW ER, AUSTERMAN, MOHAMMAD, KUTAY
Asphalt (WMA) technologies that allow for the production of asphalt mixtures at
temperatures in the range of 17 to 54C (30 to 100F) lower than typical hot mix
asphalt.
RAP is comprised of aggregates and asphalt binder from Hot Mix Asphalt (HMA)
mixtures that have been removed and reclaimed from an existing pavement. The
aggregates in the RAP are coated with aged (oxidized) asphalt binder. RAP has been
used successfully in surface HMA mixtures since the 1970s at percentages generally
around 20%. A concern associated with the use of higher RAP contents is that the
resultant mixture might become too stiff and consequently might be prone to failures in
the field (Bonaquist, 2005; McDaniel et al., 2007; Daniel et al., 2011). The increased
stiffness is due to the aged binder in the RAP. In mixtures with higher amounts of RAP,
blending of the RAP and virgin binders will result in a binder that is stiffer than the
virgin binder used for production. Softer asphalt binders or rejuvenating additives can
be used to counteract the stiffness of the RAP binder, therefore enabling the use of
higher percentages of RAP in HMA. It has been documented that rejuvenating agents
can be carried by GTR, because of its absorptive properties, to revitalize the properties
of the RAP binder (Cooper, 2008).
GTR is recycled tire rubber which has been ground into very small particles and used
as an asphalt modifier. GTR is introduced into HMA mixtures through a wet or a dry
process. A wet process refers to blending the GTR with the liquid asphalt at elevated
temperatures. A dry process refers to mixing GTR rubber into the mixture as a small
part of the aggregate or filler rather than blending the rubber with the liquid asphalt. The
result of modifying asphalt binders with GTR using a wet process is rubberized asphalt
or asphalt rubber. Rubberized asphalt is a term applied to rubber modified asphalt with
less than 15 percent by total weight of the liquid asphalt. Asphalt Rubber (AR) is defined
by American Society for Testing and Materials (ASTM) Specification D6114-97 as:
a blend of paving grade asphalt cements, GTR and other additives, as needed, for
use as binder in pavement construction. The rubber shall be blended and interacted
in the hot asphalt cement sufficiently to cause swelling of the rubber particles prior
to use. It has been found that at least 15% GTR by weight of the total blend is
usually necessary to provide acceptable properties of asphalt rubber (ASTM,
2009).
Asphalt rubber is typically used when designing a gap-graded HMA and requires a
higher amount of asphalt binder than a dense graded mix. Because of the high binder
content, asphalt rubber gap-graded mixtures have several positive qualities such as:
improved fatigue cracking resistance, improved ability to retard reflective cracking,
resistance to aging and oxidization, resistance to surface-initiated cracking, and
resistance to rutting due to higher viscosity and softening point (Walker, 2011; Xiao et
al., 2007).
254
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluation of High RAP-WMA Asphalt Rubber Mixtures
The term warm mix asphalt refers to mixtures that are produced and placed at
significantly lower temperatures than conventional HMA mixtures. WMA mixtures can
be produced by means of various technologies such as foaming, emulsion-based
products, moisture-based agents, wax-based additives, and surfactants (Bonaquist,
2011). The lower production and placement temperatures provide several benefits such
as reduced emissions and odors, decreased energy consumption for production, and
improved environmental working conditions at plants and paving sites (Bonaquist,
2011). Additionally, WMA has been reported to improve the workability of high RAP
content asphalt mixtures.
Designing a gap-graded asphalt rubber surface mixture with high RAP contents
(greater than 20%) and WMA would meet the industry goal of producing a sustainable,
cost effective, and environmentally friendly mixture. However, because the high amount
of RAP will introduce a high amount of aged RAP binder into the mixture, the
performance qualities of such mixture should be evaluated to confirm that RAP binder
did not negatively impact the performance of the mixture. A previous research study
(Mogawer et al., 2011a) suggested that the combination of asphalt rubber, RAP, and
WMA technology can be utilized without negatively impacting the fatigue cracking, low
temperature cracking, and moisture damage of the mixture. That research study limited
the RAP content to 25% and recommended that the use of RAP contents over 25% for
these types of mixtures required further study. The research study presented herein
focused on developing and evaluating the performance of asphalt rubber surface
mixtures with RAP contents up to 40% with and without the use of a WMA technology.
Because the high RAP content utilized may lead to a mixture with increased stiffness
and corresponding reduction in compactibility, the workability and potential workability
improvements resulting from the use of WMA technology were also evaluated.
2. Objectives
The main objective of this study was to design and evaluate an asphalt rubber gap-
graded (ARGG) mixture that incorporated high RAP contents and a WMA technology.
The effect of higher RAP contents and WMA on the performance of the mixture in terms
of stiffness, cracking, rutting, and moisture susceptibility were measured. Additionally
the effect of the RAP and WMA on the mixture workability was also evaluated.
Specifically the objectives were:
1. Design an asphalt rubber gap-graded (ARGG) mixture incorporating 0% (control),
25%, and 40% RAP content with and without WMA.
2. Evaluate the effect of high RAP contents and WMA technology on the stiffness of
ARGG mixtures by measuring the dynamic modulus.
255
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

3. Evaluate the fatigue characteristics of each mixture using beam fatigue, uniaxial
tension-compression test, and the semi-circular bending test.
4. Evaluate reflective cracking performance using the Texas Overlay Tester.
5. Measure the rutting and moisture susceptibility of the mixtures using the Hamburg
Wheel Tracking Device (HWTD).
6. Evaluate the workability of each mixture based on the principal of torque.
7. Determine the effects of using up to 40% RAP and/or WMA technology on the
performance of asphalt rubber surface mixtures.
3. Experimental Plan
In order to achieve the objectives of this study, an experimental plan was developed as
shown in Figure 1. The plan consisted of designing a 12.5 mm Nominal Maximum
Aggregate Size (NMAS) ARGG control mixture in accordance with Arizona
Department of Transportation (ADOT) materials specification Section 413 Asphaltic
Concrete (Asphalt-Rubber) (ADOT, 2011). This specification was followed because it
is one of few specifications for ARGG that is being used by multiple state agencies. An
ARGG mixture was designed in conformance to this specification without RAP and was
referred to as the control mixture. The control mixture was then re-designed with 25%
RAP and 40% RAP. The aggregate gradation requirements and target volumetric for the
control and RAP mixtures were the same. All mixtures were then subsequently verified
with the inclusion of a WMA technology. Overall, six mixtures were evaluated in this
study.
The effect of RAP and WMA on the performance of the mixtures was measured in
terms of stiffness, fatigue cracking, reflective cracking, rutting, and moisture damage.
Emphasis was placed on the fatigue cracking analysis because the mixture durability
might have been negatively impacted due to poor blending of the RAP and virgin
binders thus yielding mixtures with hardened aged asphalt binder. Also, mixtures with
asphalt rubber and high RAP contents might be less workable and difficult to compact
because of the aged binder in the RAP. It has been reported that the workability of such
mixtures can be improved by the incorporation of a WMA technology to the mixture
(Mogawer et. al, 2011a). Accordingly, the effect of WMA on the workability of the
ARGG with RAP mixtures used in this study was evaluated.
256
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Reclaimed
Asphalt 12.5mm Asphalt Asphalt
Rubber (AR) Rubber Gap Pavement
Binder Graded Mixture
Virgin
Aggregates
Mixtures Prepared without Mixtures Prepared with

WMA Technology WMA Technology
Mix: 177C (351F) Mix: 160C (320F)
Age/Compact: 154C (309F) Age/Compact: 141C (286F)
WMA Technology
Mixtures 1.0% SonneWarmix
25% 40%
Control
RAP RAP
Mixture
Mixture Mixture
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Performance Testing Other
Dynamic Modulus
|E*| - Master Curve
Cracking Workability
UMass Dartmouth
Characteristics
1. Beam Fatigue Asphalt
2. Uniaxial Fatigue Workability
3. Semi-Circular Bending Device
4. Overlay Tester
Moisture Susceptibility
& Rutting
Hamburg Wheel
Tracking Device
Figure 1. Experimental plan.
257

4. Materials
4.1 Asphalt Rubber Binder
An Asphalt Rubber (AR) binder obtained from a regional asphalt supplier was used for
all mixture designs. This AR binder was fabricated using a PG58-28 base binder
incorporating 17% rubber through a wet process. The AR binder conformed to the
requirements of ASTM D6114 Type II specifications (ASTM, 2009). Based on the
recommendation of the AR binder supplier, the mixing temperature was 177C (351F)
and the compaction temperature was 154C (309F).
4.2 Warm Mix Asphalt Technology
To determine if asphalt rubber mixtures incorporating RAP and WMA technology can
be produced and compacted at lower temperatures while maintaining the performance
characteristics and workability of the control mixture, a wax-based WMA technology
known as SonneWarmix was used.
SonneWarmix was chosen because it had been previously utilized in numerous field
projects in Massachusetts and other New England states. This technology was added at
the manufacturer recommended dosage rate of 1.0% by weight of total binder (Virgin
binder + RAP binder). Mixtures incorporating the WMA were fabricated at lower
mixing and compaction temperatures (160C (320F) and 141C (286F) respectively)
than the control mixture (177C (351F) and 154C (309F) respectively) without the
technology. These temperatures correspond to the same temperatures that the asphalt
rubber supplier had been using when producing similar mixtures with the same WMA
technology.
4.3 Aggregates and Reclaimed Asphalt Pavement
The virgin aggregates were from a crushed stone source in Wrentham, Massachusetts.
Four different aggregate stockpiles were obtained: 12.5 mm crushed stone, 9.5 mm
crushed stone, stone sand, and stone dust. Each stockpile was tested to determine the
aggregate properties in accordance with AASHTO specifications (AASHTO, 2010a).
The aggregate properties of each stockpile are shown in Table 1.
The RAP was obtained from the same contractor that supplied the virgin aggregates.
The binder content of the RAP was determined by the ignition method in accordance
with AASHTO T 308 (AASHTO, 2010a). The aggregates remaining after ignition were
258
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

then tested in accordance with AASHTO aggregate test specifications (T 11, T 27, T 84
and T 85) in order to determine the gradation and specific gravity of the RAP
aggregates. The properties of the RAP are shown in Table 1.
Table 1. Average virgin aggregate and RAP stockpile properties.

Stone Stone
Sieve Size 12.5 mm 9.5 mm RAP
Sand Dust
19.0 mm 100 100 100 100 100
12.5 mm 8 2 .8 9 9 .4 100 100 9 7 .7
9.5 mm 2 3 .9 9 3 .8 10 100 8 6 .3
4.75 mm 1 .2 2 9 .7 9 9 .8 9 9 .7 6 4 .3
2.36 mm 1 .1 5 .2 8 3 .7 8 3 .7 4 9 .4
1.18 mm 1 .1 2 .8 5 4 .3 5 7 .1 3 8 .3
0.600 mm 1 .1 2 .3 3 3 .8 3 8 .6 2 8 .6
0.300 mm 1 .1 2 .1 1 9 .0 2 4 .9 1 8 .6
0.150 mm 0 .9 1 .8 9 .4 1 5 .9 1 1 .4
0.075 mm 0 .8 1 .5 4 .3 1 0 .9 7 .7
Bulk Specific Gravity, Gsb
2 .6 4 1 2 .6 4 2 2 .6 4 4 2 .6 2 9 2 .6 3 8
(AASHTO T84/T85)
Absorption, % 0 .3 9 0 .4 3 0 .5 3 0 .6 0 0 .7 6
Binder Content of RAP, % (AASHTO T 308) = 4 .8 0
5. Mixture Design
For this study a 12.5 mm ARGG control mixture was developed in accordance with
ADOT materials specification as previously outlined (ADOT, 2011). The same mixture
was redesigned with the incorporation of 25% RAP and 40% RAP. The design mixture
gradation and combined aggregate properties for each design are shown in Table 2.
Based on the aggregate stockpile used and the amount of fine material in the RAP, it was
not possible to design a mixture meeting the gradation requirements for the No. 200
sieve. The mixture gradations were slightly above the 2.5% upper limit for the sieve but
were within the specification limitations once the production tolerance is considered
(1%).
Mixtures specimens were compacted using the Superpave Gyratory Compactor
(SGC) with a compactive effort of 75 gyrations. The gyration level corresponded to
design Equivalent Single Axle Loads (ESALs) of 0.3 to <3 million using the Superpave
design methodology.
Table 2. Mixture gradations and combined aggregate properties.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
259

ADO T
ADO T
Specification
25% 40% Specification
Sieve Size Control Section 413
RAP RAP Production
Asphaltic
Tolerance
Concrete
19.0mm 100 100 1 0 0 .0 100 4
12.5 mm 9 2 .6 9 2 .3 9 2 .6 80-100 4
9.5 mm 6 5 .4 6 6 .2 6 5 .4 65-80 4
4.75 mm 3 3 .9 3 3 .1 3 3 .9 28-42 4
2.36 mm 2 0 .3 1 9 .5 2 0 .3 14-22 3
1.18 mm 1 3 .5 1 4 .2 1 5 .8 - -
0.600 mm 9 .3 1 0 .4 1 1 .5 - -
0.300 mm 6 .2 6 .9 7 .5 - -
0.150 mm 4 .1 4 .3 4 .6 - -
0.075 mm 2 .9 3 .0 3 .1 0-2.5 1
Volumetric specimens were batched, mixed and short-term aged at the compaction
temperature for two hours in accordance with AASHTO R 30 (AASHTO, 2010a). After
aging, specimens were compacted in the SGC. The volumetric properties for each
mixture are shown in Table 3.
The 25% RAP mixture without the WMA technology was the first RAP mixture
designed for the study. This mixture exhibited air voids of 2.6% which was outside of
the specification range of 5.51.0%. For these mixtures, the dried RAP was placed on
the heated virgin aggregate for two hours at the mixing temperature prior to
commencement of mixing. The cause of the low air voids was probably the RAP heating
time, where two hours was not enough for proper heating of the RAP prior to mixing to
allow thorough comingling of binders. Thus the RAP heating time was increased to four
hours for all subsequent RAP mixtures for this study. The increase in the heating time
helped to achieve a percent air voids closer to the target of 5.51.0%. There was
insufficient material to re-fabricate the 25% RAP mixture with the four hours of RAP
heating prior to mixing. However, the performance specimen volumetric properties for
the 25% RAP mixture with the four hour heating were consistently within the
appropriate air void tolerance.
260
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 3. Mixture properties.

Properties Control 2 5 % RAP 4 0 % RAP Spec.
Total Binder Content, % 8 .0 8 .0 7 .7 0 -
Binder from RAP, % 0 1 .2 6 1 .8 6
Virgin Binder Added, % 8 .0 6 .7 4 5 .8 4 -
Air Voids,% 5 .8 2 .6 3 .9 5.51.0%
Voids in Mineral Aggregate % 2 1 .9 1 9 .0 1 8 .9 19% min.
Voids Filled with Asphalt, % 7 3 .7 8 6 .5 7 9 .7 -
Binder Absorbed, % 0 .7 3 0 .8 6 1 .2 0 0 -1.0%
Dust to Binder Ratio 0 .4 0 0 .4 2 0 .5 0 -
Control + 2 5 % RAP 4 0 % RAP
Properties Spec.
WMA + WMA + WMA
Total Binder Content, % 8 .0 8 .0 7 .7 0 -
Binder from RAP, % 0 1 .2 6 1 .8 6 -
Virgin Binder Added, % 8 .0 6 .7 4 5 .8 4 -
Air Voids,% 5 .9 4 .7 4 .8 5.51.0%
Voids in Mineral Aggregate % 2 2 .0 2 0 .9 2 0 .1 19% min.
Voids Filled with Asphalt, % 7 3 .3 7 8 .0 7 6 .3 -
Binder Absorbed, % 0 .7 3 0 .7 7 0 .9 0 0 -1.0%
Dust to Binder Ratio 0 .4 0 0 .4 1 0 .4 5 -
- Not Applicable
WMA = Warm Mix Asphalt Technology (1.0% SonneWarmix by total weight of binder)
6. Stiffness - Dynamic Modulus
Complex dynamic modulus |E*| testing was conducted to determine changes in mixture
stiffness due to the incorporation of RAP and/or the WMA technology. In order to
determine the dynamic modulus, test specimens were placed in the Asphalt Mixture
Performance Test (AMPT) device and subjected to a sinusoidal (haversine) axial
compressive stress at the various temperatures and frequencies. The resultant
recoverable axial strain (peak-to-peak) was measured. From these data the dynamic
modulus was calculated.
Replicate dynamic modulus specimens were fabricated in the SGC for each mixture.
Specimens incorporating WMA technology were produced at the lower mixing and
compaction temperatures noted previously. All specimens were aged for four hours at
the compaction temperature in a loose state prior to compaction. Each specimen was
subsequently prepared for dynamic modulus testing in accordance with AASHTO PP 60
Preparation of Cylindrical Performance Test Specimens Using the Superpave Gyratory
261
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Compactor (AASHTO, 2010b). The final test specimens had an air void content of 7.0
1.0%. Dynamic modulus testing was conducted in accordance with TP 62
Determining Dynamic Modulus of Hot-Mix Asphalt (HMA) (AASHTO, 2010b). Each
specimen was tested at temperatures of 4C, 20C, and 40C (39F, 68F, and 104F) and
loading frequencies of 10 Hz, 1 Hz, 0.1 Hz, and 0.01 Hz (40C only). Figure 2 shows
the results of the dynamic modulus testing for all the mixtures. The error bars shown on
Figure 2 indicate the 95% confidence interval. Thus, error bars that overlap indicate that
the modulus values are not significantly different.
The mixture master curves for each mixture were then developed from the dynamic
modulus data in accordance with AASHTO PP 61 and AASHTO PP 62 Developing
Dynamic Modulus Master Curves for Hot Mix Asphalt (HMA) (AASHTO, 2010b) at a
reference temperature of 20C (68F). Figures 3 and 4 show the mixture master curves
created from the dynamic modulus testing data.
Examination of the dynamic modulus and master curves data showed similar results.
First, the stiffnesses of the mixtures were statistically not significantly different at the
low test temperature of 4C at all frequencies tested. This is represented in the master
curves by the curves overlapping on the right side of the graph. Second, the master
curves indicated that the addition of the 25% or 40% RAP to the control mixture
increased the mixture stiffness at the intermediate and high test temperatures. Figure 2
showed that there was a significant increase in the stiffness at 20C and 40C (68F and
104F) however, not at all frequencies. The same observation held true when comparing
the control mixture incorporating the WMA technology to the same mixture
incorporating both levels of RAP. Comparison of the dynamic modulus and master curve
data for the same mixtures with and without the WMA technology suggests that the
control mixture and control mixture with the WMA technology exhibited similar
stiffness. For the remaining mixtures, generally there was a decrease in the measure of
mixture stiffness for the mixtures incorporating the WMA technology, although this
difference was not always significant. This occurrence may be a result of less aging due
to reduced mixing and compaction (aging) temperatures associated with the mixtures
incorporating the WMA technology.
Overall, the stiffness data indicated that addition of RAP to the control mixture
resulted in an increase of mixture stiffness. Generally, the stiffness increase in the
mixtures containing RAP was mitigated through the use of a WMA technology and
corresponding reduced aging temperatures. The addition of the WMA technology to the
control mixture had little to no effect on the stiffness of the mixture.
262
--`,`,,,``,,```,,`,,,,

100,000
Control
25% RAP
Average Dynamic Modulus, MPa
40% RAP
10,000 Control + 1% WMA
25% RAP + 1% WMA
40% RAP + 1% WMA
1,000
100
10
1
4
20
20
20
40
40
40
40
C
C
C
-0
-1
-1
-0
-1
-1
-0
-1
-1
-0
.1
.0
0H
0H
.1
.0
0H
.0
H
0H
1
z
H
1H
z
z
z
z
z
Figure 2. Dynamic modulus comparison - all mixtures (20C reference temperature).
100,000
Dynamic Modulus E*, MPa
10,000
1,000
Control
25% RAP
100
40% RAP
10
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04 1.0E+06
Reduced Frequency, Hz
Figure 3. Mixture master curve comparison - WMA technology mixtures (20C Reference
Temperature).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
263

100,000
Dynamic Modulus E*, MPa
10,000
1,000
Control + WMA
25% RAP + WMA

100
40% RAP+ WMA
10
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04 1.0E+06
Figure 4. Mixture master curve comparison mixtures without WMA technology (20C
Reference Temperature).
7. Fatigue Cracking Four Point Bending Beam Fatigue Tests
One of the most common and historically used laboratory test procedure to evaluate the
fatigue cracking resistance of asphalt mixtures is the four point flexural beam fatigue
test. This flexural fatigue test is the only standard test method for fatigue testing of
HMA. The typical test protocols for conducting this test are AASHTO T 321
Determining the Fatigue Life of Compacted Hot Mix Asphalt (HMA) Subjected to
Repeated Flexural Bending (AASHTO, 2010a) and ASTM D7460 Determining
Fatigue Failure of Compacted Asphalt Concrete Subjected to Repeated Flexural
Bending (ASTM, 2009). In order to investigate the relative fatigue performance of the
mixtures, the four point bending beam fatigue tests were conducted on the mixtures
following the AASHTO T 321 procedure.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Slabs with dimensions of 150 mm wide, 180 mm tall, and 450 mm long were
fabricated for each mixture using the IPC Global Pressbox slab compactor. From each
slab, beams with dimensions of 63 mm wide, 50 mm tall, and 380 mm long were cut
such that the sides had smooth faces. The air voids of the final cut specimens were
71%. Beam specimens were conditioned at the test temperature of 15C (59F) for at
264

least two hours prior to testing. The 15C (59F) temperature was selected as it
represents the intermediate temperature for the Northeast.
Each beam fatigue test was conducted in strain control mode at a loading frequency
of 10Hz applied using a sinusoidal waveform. Specimens were tested at strain levels of
300 microstrain (), 500 , 700 , and 900 . The number of cycles to failure was
determined by fitting an exponential function to the flexural stiffness versus number of
cycles and then evaluating the number cycles that it took to decrease the initial stiffness
by 50%. The results of the testing are shown in Table 4 and Figure 5.
Generally, the beam fatigue results (in strain control mode) indicated that the
resistance to fatigue cracking of the mixtures decreased with the incorporation of higher
amounts of RAP. The same trend was also apparent with the incorporation of the WMA
technology. Furthermore, at each strain level, the number of cycles to failure for each
mixture dropped when WMA was incorporated relative to the same mixture without the
WMA. For the 300, 500, and 700 , the drop in the number of cycles to failure when
WMA was used was up to fifty percent. For the mixtures incorporating WMA, the
mixing and compaction temperatures were dropped 17C and 13C respectively. This
drop in the temperature might have caused the RAP and AR binders not to comingle
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
sufficiently leading to mixtures with hardened aged asphalt binders. Accordingly, it is

recommended to further investigate and develop a procedure to determine the proper
allowable drop in temperatures for asphalt rubber mixtures that incorporate RAP and/or
WMA. Also the type and dose of the WMA technology may have impacted the cracking
resistance of these ARGG mixtures. The dose utilized may not have been enough to
realize the full benefit of the WMA technology for the mixtures tested. Thus further
investigations into the optimal type and proper dose of the WMA technology are needed.
265

Table 4. Four point bending beam fatigue test results.

Number of Cycles to 50% Initial Stiffness, Nf
Strain Level, Control + Control +
Control
2 5 % RAP 4 0 % RAP
300 6,025,590 3,724,655 2,390,822
500 614,053 677,983 289,898
700 544,687 197,625 46,895
900 25,567 24,984 16,255
Number of Cycles to 50% Initial Stiffness, Nf

Strain Level, Control + Control + 25% Control + 40% RAP
WMA RAP + WMA + WMA
300 2,946,065 1,759,123 1,526,473
500 705,290 775,690 306,746
700 196,372 99,901 51,134
900 21,616 27,026 4 ,6 9 7
8. Fatigue Cracking Uniaxial Compression /Tension Tests
In order to further evaluate the fatigue cracking performance of the mixtures, a uniaxial
cyclic fatigue push-pull test was performed on cylindrical asphalt samples, similar to
samples used for dynamic modulus testing. Perceived advantages of this type of testing
over the cyclic four point bending beams are that samples can be easily prepared using
the Superpave gyratory compactor. Also, the test can be easily performed using the new
Asphalt Mixture Performance Tester (AMPT). Additionally, the Viscoelastic
Continuum Damage (VECD) theory can be used to compute fatigue lives (Nfs) of the
mixtures at different temperatures and frequencies other than the actual tests performed.
266
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

7,000,000
AASHTO T321 Nf to 50% Reduction in Stiffness
6,000,000
5,000,000
4,000,000
3,000,000
2,000,000
1,000,000
0
Control +
25% RAP +
40% RAP +
Control +
25% RAP +
40% RAP +
25% RAP
40% RAP
25% RAP
40% RAP
Control
Control
WMA
WMA
WMA
WMA
WMA
WMA
300 s 500 s
1,000,000
AASHTO T321 Nf to 50% Reduction in Stiffness
900,000
800,000
700,000
600,000
500,000
400,000
300,000
200,000
100,000
0
Control +
25% RAP +
40% RAP +
Control +
25% RAP +
40% RAP +
25% RAP
40% RAP
25% RAP
40% RAP
Control
Control
WMA
WMA
WMA
WMA
WMA
WMA
700 s 900 s
Figure 5. AASHTO beam fatigue results at four strain levels (15C test temperature).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
267

A proposed standard practice for continuum damage fatigue analysis that was
developed by the Asphalt Research Consortium (ARC) was used in this study. The
procedure is described in details by Haggag et al. (2011). The main aim of the proposed
approach is to get plots of the ratio of damaged modulus to the undamaged modulus (C)
as a function of the damage parameter S which is collapsed by choosing the
appropriate value for the continuum damage constant . The damage ration can be
calculated using Equation 1:
E* n
C= [1 ]
E* LVE
where: C = damage ratio

|E*|n = damaged modulus at cycle n
|E*|LVE = undamaged (linear viscoelastic or LVE) dynamic modulus
For each mixture a total of four AMPT specimens were fabricated and prepared for
the uniaxial compression-tension test and subsequent VECD analysis. The final test
specimens had an air void content of 7.01.0%. Each specimen was then tested at a
specific temperature and strain level in the AMPT as outlined in Table 5. The tests were
controlled using on-specimen LVDT strains.
Table 5. Uniaxial compression-tension test matrix.
Micro-Strain ()
Specimen No. Temperature C
(Peak-to-Peak)
1 10 200
2 10 260
3 20 200
4 20 260
A beta version of software (called PP-VECD) developed by Michigan State

University was used to analyze the data obtained during the uniaxial compression-
tension test. The software allows calculation of the damage characteristic curves, i.e., C
versus S curves, of mixtures from uniaxial tension-compression (push-pull) fatigue data
run at a constant frequency and temperature (Kutay et al. 2008; Kutay et al. 2009). Once
the C vs. S curves are computed, the software perform simulations of strain controlled
tests and computes the number of cycles to failure (Nf) based on 50% reduction in |E*|
criterion. The Nf values can be computed at a variety of temperatures, strain levels and
frequencies and calibration constants of traditional fatigue life equation (Nf = a()b(E)c)
can be computed (Mbarki et al. 2012).
268
--`,`,,,``,,```,,`,,,,,

Figure 6 shows the Nf values at different strain levels (same as beam fatigue tests
above) for the frequency of 10Hz and temperature of 15C. As shown in Figure 6, the Nf
decreases with increasing percentage of RAP in both hot mix and warm mix asphalt
specimens which agrees with the results of beam fatigue tests.
Unlike the beam fatigue results, the VECD analysis showed that the number of
cycles to failure slightly increased when WMA was incorporated. This phenomenon
agrees with the dynamic modulus results shown in Figure 2, where the WMA mixtures
were generally slightly less stiff than the HMA mixtures. It is known that soft mixtures
generally perform better in fatigue, assuming a good blending of RAP and virgin binder
is achieved. It is noted that the loose mixture samples for push-pull tests were made at
the University of Massachusetts Dartmouth, then transported to Michigan State
University for testing. As a result, during sample preparation, all samples were reheated
in the oven for 2 to 3 hours for compaction and sample preparation. During this
reheating process, the RAP and virgin binder may have mixed better as compared to the
beam fatigue samples.
Figure 6. VECD analysis results number of cycles to failure Nf that corresponds to 50

% reduction in stiffness (i.e., |E*|)
269
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

9. Fatigue Cracking Semi-Circular Bending (SCB) Test
Because of the disagreement between the beam fatigue and VECD analysis fracture
results, another test was used to evaluate the fatigue cracking of the mixtures. Fracture
resistance potential was assessed using the Semi-Circular Bending (SCB) approach
proposed by Wu et al. (2005). This test characterizes the fracture resistance of HMA
mixtures based on fracture mechanics principals, the critical strain energy release rate,
also called the critical value of the J-integral, or Jc. To determine the critical value of the
J-integral (Jc), semi-circular specimens with at least two different notch depths need to
be tested for each mixture. In this study, three notch depths of 25.4 mm, 31.8 mm and 38
mm were selected based on an a/rd ratio (the notch depth to the radius of the specimen)
between 0.5 and 0.75. Test temperature was selected to be 25C which is a typical test
temperature utilized previously for this type of testing. The semi-circular specimen is
loaded monotonically until fracture failure under a constant cross-head deformation rate
of 0.5 mm/min in a three-point bending load configuration. The load and deformation
are continuously recorded and the critical value of the J-integral (Jc) is determined using
Equation 2 (Wu et al., 2005). Triplicate specimens were used for this test. Specimens
were compacted to an air void level of 7 0.5%.
U U 1
J c = 1 2 [2 ]
b1 b2 a 2 a1
where: a = the notch depth

b = sample thickness
U = the strain energy to failure
High Jc values are desirable for fracture-resistant mixtures. A threshold of a

minimum Jc of 0.55 to 0.65 kJ/m2 is typically used as a failure criterion for this test.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 7 shows the effect of RAP content on the mean critical strain energy (Jc) values
for the HMA and WMA mixtures evaluated in this study. It is noted that, for mixtures
with WMA technology, the Jc values slightly increased as the RAP level increased from
0% to 25%, then decreased at RAP level of 40%. However, there was a noticeable
decrease in Jc values as the RAP contents increased for the WMA mixtures. This shows
good agreement with the beam fatigue data which suggested fatigue life was reduced as
the percentage of RAP was increased and/or WMA technology was utilized. Again, this
might be attributed to the temperature drop associated in this study with the use of WMA
or the type and dose of the WMA technology utilized. Accordingly, it is recommended
to further investigate the proper allowable drop in temperatures for asphalt rubber
mixtures that incorporate RAP and/or WMA as well as the proper type and dose of the
WMA technology utilized.
270

0. 60
0.53 0.54 HMA
0.49
0. 50 WMA
0.42
0. 40
Jc (KJ/m2)
0.28
0. 30
0.25
0. 20
0. 10
0. 00
ARGG Control ARGG Control + 25% RAP ARGG Control + 40% RAP
Mixture Type
Figure 7. SCB test results.
10. Reflective Crack Testing - Overlay Test
In order to evaluate the resistance of the mixtures to reflective cracking, the mixtures
were tested in the Texas Overlay (OT) tester. This test is a displacement controlled test.
In this test, a trimmed gyratory compacted specimen is glued with epoxy onto two
plates as shown in Figure 8. The joint between the plates is located at the midpoint of
the specimen length. The glued test specimen is placed into the OT device. During
testing, one of the plates remains stationary while the other is displaced. The moving
plate is pulled, thereby opening the joint between the plates, to a known displacement.
The plate is then pushed back to the original location, thereby returning the joint
between the plates to its original position. The opening and closing (displacement) of
the joint between the plates occurs in 10 seconds (5 seconds to open the joint and 5
seconds to close the joint). Each opening and closing motion is one cycle. During each
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
cycle the load required to move the plates to the specified displacement is recorded. The
device is set to terminate the test when the load is reduced a certain percentage from the
271

load recorded for the first cycle or the sample reaches a specific number of cycles
without reaching the required load reduction.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 8. Specimen setup in Overlay Test (OT) device.
For this study, the Texas Department of Transportation specification (Tex-248-F) for
testing bituminous mixtures with the OT (TxDOT, 2011) was followed. Specimens were
then fabricated in the SGC and then trimmed. The air void level of the trimmed
specimens was 7.01.0%.
All mixtures for this study were tested with a joint opening (displacement) of 0.06
cm (0.025 inch), test temperature 15C (59F), and a failure criteria of 93% reduction in
the load measured during the first cycle or 1,200 cycles (whichever occurs first). The
average results of the testing are shown in Table 6. Generally, mixtures exhibiting more
cycles to failure exhibit more cracking resistance.
Table 6. Test results from overlay test.

Average OT Cycles
Mixture
to Failure
Control 351
25% RAP 43
40% RAP 54
Control + 1% WMA 275

25% RAP + 1% WMA 64
40% RAP + 1% WMA 21
272

Generally, the results from the OT test indicated that the reflective cracking
resistance of the mixture decreased with the incorporation of higher amounts of RAP
(except for the 40% RAP mixture which performed slightly better than the 25% RAP
mixture). The same trend was apparent regardless if the mixture incorporated WMA
technology or not.
Comparing the results for the mixtures with and without the WMA technology, a
slight improvement in the reflective cracking resistance was noted for the 25% RAP
mixture with the WMA technology. Conversely a slight reduction in the reflective
cracking resistance was noted for the control and 40% RAP mixtures. These data agrees
with the results of the beam fatigue and SCB testing which showed a reduced cracking
resistance for the mixture incorporating WMA. The reason for the push-pull tests
showing different trends than the overlay tester, beam fatigue, and SCB for mixtures
incorporating WMA is unknown. Overall, the results indicated that mixtures
incorporating RAP were more susceptible to reflective cracking as compared to the
control mixtures.
11. Rutting and Moisture Susceptibility Hamburg Wheel Tracking Device
The mixtures in this study were evaluated for their rutting and moisture susceptibility to
determine if high RAP content and/or the WMA technology had any effects on their
performance. It is not known how the use of asphalt rubber, RAP and WMA technology
affects mixture moisture susceptibility. Previous research (Austerman et al., 2009;
Mogawer et al., 2011b) has suggested that the addition of a WMA technology may
increase the moisture susceptibility of conventional mixtures. Therefore, in order to
understand the performance of these mixtures, they were subjected to moisture
susceptibility testing in a Hamburg Wheel Tracking Device (HWTD).
Testing was conducted in accordance with AASHTO T 324 Hamburg Wheel-Track
Testing of Compacted Hot-Mix Asphalt (HMA) (AASHTO, 2010a). The test is used to
determine the failure susceptibility of the mixture due to weakness in the aggregate
structure, inadequate binder stiffness, or moisture damage (AASHTO, 2010a). In this
test, the mixture is submerged in heated water (typically 40-50C) and subjected to
repeated loading from a 705N steel wheel. As the steel wheel loads the specimen, the
corresponding rut depth of the specimen is recorded. The rut depth versus numbers of
passes of the wheel is plotted to determine the Stripping Inflection Point (SIP) as shown
in Figure 9. The SIP gives an indication of when the test specimen begins to exhibit
stripping (moisture damage).
273
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

-2 Stripping
Inflection
-4
Point
(SIP)
-6
Rut Depth (mm)
-8
- 10
- 12
- 14
Number of Number of
- 16 Passes to Passes
Stripping Failure, N f
- 18 Inflection Point
(SIP)
- 20
0 2 ,0 0 0 4 ,0 0 0 6 ,0 0 0 8 ,0 0 0 1 0 ,0 0 0 1 2 ,0 0 0 1 4 ,0 0 0 1 6 ,0 0 0 1 8 ,0 0 0 2 0 ,0 0 0
Number of Passes
Figure 9. Determination of HWTD Stripping Inflection Point (SIP).

Gyratory specimens for this study were fabricated using the SGC to an air void level
of 7.01.0% as required by AASHTO T 324. Testing in the HWTD was conducted at a
test temperature of 50C (122F). The specimens were tested at a rate of 52 passes per
minute after a soak time of 30 minutes at the test temperature. Testing terminated at
20,000 wheel passes or until visible stripping was noted. Table 7 shows the results of
the moisture susceptibility testing.
All mixtures evaluated in this study passed the moisture susceptibility testing in the
HWTD. Additionally, the magnitude of the average total rut depth observed at the end
of each test was less than 1.10 mm (0.043 in.).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
274

Table 7. Moisture susceptibility and rutting HWTD test results.

Average Rut
Average Rut
Stripping Depth at
Mixture Depth at 20,000
Inflection Point 10,000 Cycles
Cycles (mm)
(mm)
Control NONE 0 .8 8 1 .0 9
25% RAP NONE 0 .4 1 0 .5 1
40% RAP NONE 0 .2 3 0 .2 8
Control + 1% WMA NONE 0 .4 5 0 .6 5

25% RAP + 1% WMA NONE 0 .1 4 0 .2 3
40% RAP + 1% WMA NONE 0 .8 5 0 .9 6
NONE = Mixture passed 20,000 cycle test with no SIP.
12. Mixture Workability Asphalt Workability Device (AWD)
Because of the potential decrease in mixture workability due to the incorporation of RAP
in the mixtures, workability evaluations of each of the mixtures were completed. These
evaluations were conducted using a HMA workability device developed by the
University of Massachusetts Dartmouth Highway Sustainability Research Center
(HSRC). This device is known as the Asphalt Workability Device (AWD) and has been
used previously to evaluate high percentage RAP mixtures as well as mixture
incorporating WMA additives (Austerman et al., 2009; Mogawer et al., 2009).
The AWD operates on the torque measurement principles that have been previously
established (Gudimettla et al., 2003). The AWD rotates the loose HMA mixture at a
constant speed (15 rpm for this study) and separately records the resultant torque exerted
on a pug mill style paddle shaft embedded into the mixture. Concurrently the surface
and internal temperatures of the mixture are recorded. As the mixture cools in ambient
conditions, the torque exerted on the shaft increases thereby giving an indication of the
workability of the mixture at different temperatures.
Each of the mixtures in this study was mixed and aged (four hours) at the mixing and
compaction temperatures previously outlined respectively. After completion of aging,
the loose mixture was tested in the AWD.
From the AWD test data for each mixture, a best fit exponential line was fitted to the
raw data. This fit line was then utilized to develop a model curve plotted over the AWD
test temperature range in which torque and temperature data were collected. This
temperature range included the anticipated field placement and compaction temperatures
275
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of the mixture. The model curves are shown in Figure 10. Note that mixtures exhibiting
lower torque values are considered more workable.
70
Control
60 Control + 1% WMA
25% RAP
25% RAP+ 1% WMA
50
40% RAP
T o rq u e (N -m )
40% RAP + 1% WMA

40
30
20
10
80 90 100 110 120 130 140 150 160
Temperature (C)
Figure 10. Workability test results.

The data for the mixtures without the WMA technology showed the expected trend
that as the amount of RAP incorporated into the mixture was increased there was a
corresponding decrease in mixture workability (i.e. increase in torque) over the
temperatures tested. The mixtures with the WMA technology, for the majority, showed
the same trend with the exception of the control with WMA at higher temperatures
showed marginally higher torque values than the 25% and 40% RAP mixtures. Overall,
comparison of the mixtures with and without WMA indicated that the addition of the
WMA greatly improved the workability of the mixtures with RAP to a level similar to
the control mixture without RAP and WMA.
276
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

13. Conclusions
Based on the research conducted to design and evaluate an asphalt rubber gap-graded
mixture that incorporated high RAP contents and a WMA technology, the following
conclusions were made:
For the majority, an ARGG control mixture incorporating RAP (25% and 40%) with
and without a WMA technology was able to be developed and met the desired
volumetric specifications. However, the amount of time the RAP was heated prior
to specimen mixing had a significant effect on the mixture volumetrics. A four hour
heating time was recommended.
The dynamic modulus and mixture master curve data indicate that addition of 25%
and 40% RAP to the control mixture resulted in an increase in mixture stiffness at
20C and 40C (68F and 104F). This stiffness increase was partially mitigated
through the use of the WMA technology and corresponding reduced aging
temperatures. However, the addition of the WMA technology to the control mixture
had little to no effect of the mixtures stiffness.
Generally, the results the beam fatigue tests in strain control indicated that the
resistance to fatigue cracking of the mixtures decreased with the incorporation of
higher amounts of RAP. The same trend was also apparent with the incorporation of
a WMA technology to the RAP mixtures.
The push-pull test showed that the Nf at strain levels similar to the ones used in the
beam fatigue tests at a frequency of 10Hz and temperature of 15C decreased with
increasing percentage of RAP which agrees with the results of the beam fatigue
tests. Unlike the beam fatigue results, the VECD analysis showed that the number of
cycles to failure slightly increased when WMA was incorporated to the mixtures.
The SCB test showed that control mixtures without RAP generally exhibited better
fracture resistance than similar mixture containing RAP. Furthermore, the test
showed that mixtures containing RAP exhibited higher critical strain energy values
than similar mixtures that used WMA technology. These data agree with the beam
fatigue test results thereby indicating the RAP mixtures incorporating WMA had
less fracture resistance. The VECD analysis showed an opposite trend for WMA
mixtures and the reason for this occurrence is unknown.
Reflective cracking test results from the Overlay Tester indicated that mixtures
incorporating 25% and 40% RAP were more susceptible to reflective cracking as
compared to the control mixtures. For the RAP mixtures the addition of the WMA
technology did not yield a better reflective cracking resistance. These data agree
with the beam fatigue and SCB test results,
277
--`,`,,,``,,```,,`,,,,,`,``,

The addition of any percentage of RAP and/or the addition of the WMA technology
to the ARGG control mixture had no significant impact on the moisture
susceptibility or rutting potential of the mixture.
The workability evaluation indicated that as the amount of RAP incorporated into
the ARGG control mixture was increased there was a corresponding decrease in
mixture workability. Comparison of the mixtures with and without the WMA
technology indicated that the addition of the WMA greatly improved the workability
of the ARGG control mixtures with RAP to a level similar to the ARGG control
mixture without RAP or WMA. Therefore, the use of a WMA technology should be
considered when field compactability may be an issue with these types of mixtures.
It is recommended to further investigate and develop a procedure to determine the
proper drop in temperatures for asphalt rubber mixtures that incorporate WMA as
the temperature drop may have been the leading contributing factor to the reduced
cracking performance of the mixture incorporating WMA technology.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Overall, the data indicate that high amounts of RAP and/or the use of WMA
technology may reduce the cracking resistance of the ARGG mixture. The use of
RAP and/or WMA did not negatively impact the moisture susceptibility or rutting of
the mixture. The workability of the mixtures was improved when using the WMA
technology. Further investigations are needed to validate these results with different
types of AR binders and WMA technologies.
14. Acknowledgements
The authors would like to acknowledge Mike Nichols of Aggregate Industries and Mark
Gabriel of All States Asphalt, Inc., for supplying the aggregates and asphalt rubber
binder for this project respectively. The authors would also like to thank Chris Strack,
Sonneborn, Inc., for supplying the WMA technology for this study. Finally, the authors
would like to thank UMass Dartmouth undergraduate research assistant Mr. Michael
Roussel for his assistance with the laboratory testing for this project.
15. References
American Association of State Highway and Transportation Officials (AASHTO).

(2010a). Standard Specifications for Transportation Materials and Methods of
Sampling and Testing. American Association of State Highway and Transportation
Officials, Washington, D.C. 30th Edition.
278

American Association of State Highway and Transportation Officials (AASHTO).

(2010b). 2010 AASHTO Provisional Standards. American Association of State
Highway and Transportation Officials, Washington, D.C. 14th Edition.
American Society for Testing and Materials (ASTM). (2009). Annual Book of ASTM
Standards. ASTM International, Section 4, Volume 04.03.
Arizona Department of Transportation (ADOT). (2011) Section 413 Asphaltic Concrete
(Asphalt-Rubber)- Standard Specifications 2000. Retrieved from
http://www.azdot.gov/Highways/ConstGrp/Contractors/PDF/2000_Standard_Spec
/Sec401_417.pdf Accessed June 15th, 2011.
Austerman A.J., Mogawer W.S., Bonaquist R. (2009). Investigation of the Influence of
Warm Mix Asphalt Additive Dose on the Workability, Cracking Susceptibility, and
Moisture Susceptibility of Asphalt Mixtures Containing Reclaimed Asphalt
Pavement. Canadian Technical Asphalt Association (CTAA) Proceedings.
Moncton - New Brunswick, pg. 5171.
Bonaquist R. (2005). Laboratory Evaluation of Hot Mix Asphalt (HMA) Mixtures
Containing Recycled or Waste Product Materials Using Performance Testing,
Publication FHWA-PA-2005-006+98-32(19), Pennsylvania Department of
Transportation, Office of Planning and Research.
Bonaquist R. (2011). Mix Design Practices for Warm Mix Asphalt. National
Cooperative Highway Research Program (NCHRP), Report 691.
Cooper S. B. (2008). Characterization of HMA Mixtures Containing High Recycled
Asphalt Pavement Content with Crumb Rubber Additives. Master Thesis, Louisiana
State University..
Daniel J., Pochily J. (2009) Can We Make More Use of RAP? A Study of Extracted
Binder Properties in NH, VT and Maine RAP Mixes. Presented at the 2009
Northeast Asphalt User Producer Group (NEAUPG) Annual Meeting, South
Portland, ME October 8th 2009. Retrieved from
http://www.neaupg.uconn.edu/pdf/09_Binder_Study_daniel-pochily.pdf. Accessed
June 6th, 2011.
Gudimettla J., Cooley L., Brown E. (2003). Workability of Hot Mix Asphalt. Report #03-
03, National Center for Asphalt Technology (NCAT), Auburn, AL.
Haggag, M.M., Mogawer, W.S., Bonaquist, R. (2011). Fatigue Evaluation of Warm-
Mix Asphalt Mixtures: Use of Uniaxial, Cyclic, Direct Tension Compression Test.
Vol. 2208, pp. 2632.
279
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Kutay, M.E., Gibson, N.H. and Youtcheff, J. (2008). Conventional and Viscoelastic
Continuum Damage (VECD) based Fatigue Analysis of Polymer Modified Asphalt
Pavements. Journal of the Association of Asphalt Paving Technologists, Vol. 77,
pp. 395434.
Kutay, M.E., Gibson, N.H., Youtcheff, J. and Dongre, R. (2009). Use of Small Samples
to Predict Fatigue Lives of Field Cores: Newly Developed Formulation Based on
Viscoelastic Continuum Damage Theory. Transportation Research Record:
Journal of the Transportation Research Board, Vol. 2127, pp. 9097.
Mbarki, R., Kutay, M.E., Gibson, N. and Abbas, A. (2012). Comparison between
Fatigue Performances of Horizontal Cores from Different Asphalt Pavement Depths
and Laboratory Specimens. Road Materials and Pavement Design,
DOI:10.1080/14680629.2012.685843, pp. 111.
McDaniel R., Shah A., Huber G., Gallivan V. (2007). Investigation of Properties of
Plant-Produced RAP Mixtures. Transportation Research Record: Journal of the
Transportation Research Board, No. 1998, Transportation Research Board of the
National Academies, Washington, D.C., pp. 103111.
Mogawer W.S., Austerman A.J., Roussel M. (2011a, October). Performance
Characteristics of Asphalt Rubber Mixtures Containing RAP and Warm Mix
Asphalt Technology. Proceedings of the 2nd International Warm-Mix Conference,
St. Louis, MO.
Mogawer W.S., Austerman A.J., Kassem E., Masad E. (2011b). Moisture Damage
Characterization of Warm Mix Asphalt Mixtures. Journal of the Association of
Asphalt Paving Technologist Vol. 80, pg 491526.
Mogawer W.S., Austerman A.J., and Bonaquist R. (2009) Laboratory Development and
Field Trials of Thin-Lift Hot Mix Asphalt Overlays Incorporating High Percentages
of Reclaimed Asphalt Pavement with Warm Mix Asphalt Technology. Canadian
Technical Asphalt Association (CTAA) Proceedings. Moncton - New Brunswick,
pg. 7397.
Texas Department of Transportation (TxDOT). (2011). Test Procedure for Overlay Test,
TxDOT Designation Tex-248-F. Retrieved from ftp://ftp.dot.state.tx.us/pub/txdot-
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
info/cst/TMS/200-F_series/pdfs/bit248.pdf Accessed June 15th, 2011.

Walker D. (2011). Understanding How Tires are used in Asphalt, RubberForm
Recycled Products, LLC. Retrieved from http://www.rubberform.com/news.
Accessed June 15th, 2011.
280

Wu Z., Mohammad L.N., Wang L.B., Mull M.A. (2005). Fracture Resistance
Characterization of Superpave Mixtures Using the Semi-Circular Bending Test.
Journal of ASTM International, Vol. 2, No. 3.
Xiao F., Amirkhanian S., Juang and C. H. (2007). Rutting Resistance of Rubberized
Asphalt Concrete Pavements Containing Reclaimed Asphalt Pavement Mixtures.
American Society of Civil Engineers (ASCE) Journal of Materials in Civil
Engineering, 19, 475.
16. Discussion
DR. HAIFANG WEN: Walaa, very nice work.

PROFESSOR WALAA MOGAWER: Thank you.
DR. WEN: Just two points. Number one is on the fatigue by the bending beam. You use
the same strain level for different RAP percentages, right? Then your finding is when
you increase the RAP, the cycle of fatigue life is reduced. But if you think about it, I
have a pavement with the RAP. If the RAP percent is increased, the stiffness is higher,
right? When you have high RAP, the strain level will be lower in the pavement for that
layer. So actually if I use the same material, the same pavement with different RAP, the
strain level is different. When I put real traffic on that, the fatigue life will be different
from what you have here using the same strain level. So from that point of view, I think
when we look at fatigue, we should really look at not just the same strain level
laboratory testing, but a model in which you include the strain level and the fatigue life.
PROF. MOGAWER: I agree. Thats a very, very good point. I just want to mention that
even at the lower strain levels, were seeing the same trend. But I agree with you. That is
a very valid point.
DR. WEN: Right. The number of cycles dropped a lot from a few million for 400
microstrain to a few thousand for 100 microstrain. For instance, I would compare in the
field 300 microstrain for 0% RAP to 100 microstrain for 40% RAP. So thats one point.
The second one is just a general comment. When we are using RAP, if you look at the
specifications for most DOTs, high RAP is put it on the bottom, to be safe. Actually if
you look at the performance based on your findings and others, the high RAP mix is
pretty good at resisting rutting which should be used on the surface of the asphalt layer
and not as good at bending, which should not be used at the bottom of the layer. So from
that point of view, we should have actually put a high RAP mix on the surface and not
on the bottom. There are my comments.
PROF. MOGAWER: Thank you. Very good points.
281
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DR. ALAEDDIN MOHSENI: On your Hamburg test, what temperature did you test it
at?
PROF. MOGAWER: 50C.
DR. MOHSENI: Okay. And related to the previous comment, I dont think 50C would
be enough to show the effect of moisture on the RAP material. You probably need to
increase the temperature because what you are actually testing is the stiffness of the
material more related to rutting. The Hamburg is a combination of both rutting and
moisture susceptibility, and somehow you need to separate the two effects. So what
you're getting there is pretty much what you already know, which is it is stiffer and it
doesnt really give you moisture damage.
PROF. MOGAWER: Again, that is a good point. However, for the no rubber mixtures
we test them at 40C and 45C. I think going above 50C is not going to relate at all to
what we see in the Northeast. So if we tested at much higher, like 55C or 60C, I dont
know what it would add in terms of looking at the mix performance.
DR. MOHSENI: Well, Ive tested them at 65C and even 70C, and if you go to high
temperatures, then they will show the effect.
PROF. MOGAWER: Okay.
DR. JOHN DANGELO: Great presentation. You always do a nice job of laying out
details, and it is easy to follow, and I really appreciate that.
PROF. MOGAWER: Thank you, but here comes the question.
DR. DANGELO: I have a question for you. In the development of your mixes, you
went to 25% RAP and 40% RAP. If you go back and look at how the asphalt-rubber
works, very, very high binder contents, you use 17% rubber in them, now when you try
to replace the binder volume with RAP binder, it is a whole different animal. I mean, I
would like to see in your data what are the actual binder volumes that you saw in the
different mixtures your control against the 25% RAP and the 40% RAP as a
comparison. With rubber mixes you might have a lot less actual binder with these RAP
mixtures compared to the virgin rubber mix. Additionally, you use a 58-28 as your base
binder with the rubber and now you are replacing it with whatever that RAP binder is
which is no longer 58-28 but its probably 70-16 or 76-10 possibly.
PROF. MOGAWER: Right. These are all very good points, John. We put the
volumetrics for this mix in this paper, and you are absolutely right because when we
started doing our mix designs, we realized that the method of incorporating the RAP did
impact the volume of the effective binder in the mixtures.
DR. DANGELO: Thank you.
PROF. MOGAWER: Thanks, John.
282
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DR. BRIAN HILL: We did a considerable amount of fracture testing with the DC(T)
with different RAP levels and four different additives that we had for warm-mix asphalt,
and we found more of an effect of the warm-mix technology than the temperature, I
think. The temperature reduction has a slight effect but not as much as the warm-mix
technology that you're using.
PROF. MOGAWER: Thank you.
DR. HILL: So, when you go back through and you test different technologies, if you add
in a chemical additive, I think you will see a considerable difference, if you look at those
SCV results.
PROF. MOGAWER: Generally, I agree that different WMA technologies can lead to
different performance. We used a wax that happens to be the one used by the producer
of the asphalt rubber used in this study. Thanks.
DR. HONGBON XIE: Actually, I have a couple questions. The first one is gradation
blend. You showed it in your paper and also the presentation. You seem to have some
difficulty to meet the spec on No. 200. And actually, I look at that because we obviously
have similar issues. And I also noticed that you have a 7.7% passing for the RAP itself. I
assume the 40% RAP is aggregate blend?
PROF. MOGAWER: Im sorry?
DR. XIE: When you say 40% of RAP, is that 40% replacement by the RAP aggregate?
PROF. MOGAWER: Absolutely, yes.
DR. XIE: So based on that, I am asking you how can you get 3.1% passing the No. 200
sieve, apparently you had a calculation error in there. If you do the calculation, you will
probably get like a 4 or 5% passing on the No.200, but in your paper it shows only
3.1%. So Im not sure if you washed all other stockpile aggregates completely or why
the combined gradation has 3.1% passing on No.200.
PROF. MOGAWER: Again, I hope Im following right. The 3.1% that you see, thats
the blend of the RAP and whatever stockpile aggregates we did use.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. XIE: Right.

PROF. MOGAWER: So Im not following how you're doing the calculations, but the
gradations that we do are normally washed, but we do not wash all our aggregates to
start blending specimens. For this study, because of the high minus No.200 in the RAP
we did individual sieving and then blended the aggregates based on the target percentage
for each aggregate and RAP stockpile.
DR. XIE: What Im trying to say is if you have 60% or so virgin aggregates, which
maybe have an average of 1 or 2% dust in it, right? Then you have 40% RAP, which has
7.7% dust and your combined dust is going to be over 4% because 40% x 7.7 equals to
283

3.1. Then you have 60% of aggregate that contains maybe 2% dust and that give you an
additional 1% dust so the total combined will have 4% or higher dust.
PROF. MOGAWER: I see your point. I dont have all the numbers in my mind. I have to
go back and check, but I know for the RAP, we had to some sieving for the RAP, if my
memory is correct, and we had to eliminate the top ones. But the 3.1and I have the
numbers with meits the blending of the aggregates and the RAP. So I will check the
number and see if that is right.
DR. XIE: Yes. Check the calculation on that. The second question is the compaction
effort. You did mention that is 75 gyrations of gyratory, but you did not mention the
other compaction parameters. We found it out because we started using gyratory doing
the rubber mix design in California just recently. We found out that it actually required
higher pressure as well as squaring time, basically the holding time at end of
compaction. Do you use a different pressure other than the standard 600?
PROF. MOGAWER: No, just the 600.
DR. XIE: And you do not have to hold the specimen?
PROF. MOGAWER: No. Maybe just for few minutes.
DR. XIE: Because we have similar gradation on the rubber type G material and based on
our limited experience, we have to hold the sample for like 20 or 30 minutes at least
after compaction. Otherwise, its going to rebound. This is significant, actually.
PROF. MOGAWER: The asphalt rubber already came to us pre-reacted. When we are
adding the rubber to the binder in our lab and when we mix it, that is a different study.
That is when we have to hold the specimens for few minutes in the mold.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. XIE: Okay. So you blend the rubber binder in the lab.
PROF. MOGAWER: No, not for this study. The supplier of the binder gave us the
gallons of the asphalt rubber already.
DR. XIE: That is the same binder we have. Basically, the rubber binder
PROF. MOGAWER: Again we didnt add rubber in the lab. Generally, if we do add the
rubber to the asphalt binder in the lab, we might have issues with compaction depending
on the amount or rubber and type of mix.
DR. XIE: Just for the reference, basically, because apparently we are not the only ones to
have this kind of issue with rubber design in California. So actually the current
specification for rubber design allows the pressure to go up to I think its 850 kPa
instead of standard 600 kPa, as well as the holding time is up to 90 minutes because
apparently this is really common. You have to increase pressure, increase the holding
time. Because at the 300 F compaction temperature, the rubber is still so soft basically
when you remove it from the mold, theres going to be rebound.
284

DR. MOGAWER: Thank you.

DR. XIE: Thank you.
DR. MOGAWER: Thank you very much.
DR. GEOFFREY ROWE: Walaa, very nice presentation, as always. Always nice to
listen to you. I wanted to add just a little bit on the fatigue testing and the four-point
bending results that you presented. Now you noted that you did use the AASHTO T 321
analysis, and when you analyzed the data in that AASHTO T 321, there are some
definitions regarding the cutoff of the data, the assumption of the initial flexural stiffness
and those aspects. When you have a modified material incorporating rubber, the
experience in our task group has been that we wouldnt necessarily start to see the point
of crack initiation in that definition in AASHTO T 321. Your results and all other things
being equal with a lower stiffness material in a bending beam, you would expect a
higher fatigue result, and you commented on that. With the study weve done in the
Binder ETG with waxes, as we put in wax in the material, we start to see some marginal
reduction in fatigue performance. So pretty much that would suggest to me that you
might almost be a wash with your results. Then what I do is I look at your results in your
table and to be honest, some of the results switch. If you look at the results of the 40%
RAP, the number of cycles to failure arent always consistent one with the other. When
you look at this data and you look at the results that you show in terms of bar charts you
showed in the slides. You showed two bar charts where you show a linear scale of
fatigue life. But when were measuring fatigue life, really were interested in the log of
the fatigue life. Normally speaking, were looking at variability on a log scale and not a
linear scale. I wonder by presenting it on a linear whether you're amplifying differences
that might be there that when you look at the variance and the statistical difference
between these test results, whether you really have that true statistical difference. So my
question for you is what is the number of test specimens that youve used in determining
your fatigue life and do you really see a statistical difference in the fatigue life when you
compare different strain levels from material to material?
PROF. MOGAWER: Okay, the replicates were three per strain level. So for every strain
level, we tested three beams. I think that was one of your questions. In terms of
statistical difference, I have to go back and look at all the data, but I believe there was
some statistical difference. So Ive got to go back and look at it, Geoff.
DR. ROWE: It would be interesting to know that. Obviously, I have a difference of how
I would analyze that fatigue data in looking at that, and I talked about that on Sunday,
but regardless of that, there is still whether we express the data on log scale how you
determine the averages and how you do something like a T-test to look at statistical
difference between the data sets. So that is sort of what I was interested in.
PROF. MOGAWER: Thats why I was very impressed on Sunday to know that there is
software available to do the analysis.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
285

DR. ROWE: Thanks, Walaa. Nice presentation

PROF. MOGAWER: Thank you very much, Geoff.
DR. RANDY WEST: Walaa, thank you very much for presenting this paper and this
research to the group.
PROF. MOGAWER: Thanks, Randy.
DR. WEST: Just a couple of comments. One is to really kind of follow up on what Dr.
Wen mentioned. The fact that when we are adding some of these materials, for example
RAP or shingles, it will increase the stiffness of the mix, and thats going to change the
response of the pavement to the loading. So comparing just equivalent strain levels I
think can be misleading. The comment that Id like to make is that in some of the work
that weve done at NCAT comparing a variety of different cracking tests, like youve
done in this study, weve also seen conflicting results if you just look at the results at a
face value. What that says to me is that we have to go beyond just laboratory tests,
really, to find out what is the true answer about which mix is better. This means we need
go into the field for more of this kind of work, which is what were doing at the test
track. Im not standing here to try to advertise that, although obviously thats a big part
of the work that were doing. But we do need to do more field-controlled work with
these mixes to understand which of the tests are giving us the best indication of long-
term performance.
PROF. MOGAWER: Thanks, Randy. Very good comments.
DR. NELSON GIBSON: Did you have different failure criteria for the beam fatigue and
uniaxial tests, and, if so, could that be why you had conflicting results?
PROF. MOGAWER: No, we used the same, 50% drop, but I dont know why were
getting the differences. Im not in a position to say why these two tests are giving
differences.
DR. GIBSON: Okay.
PROF. MOGAWER: Maybe Dr. Emin will.
DR. M. EMIN KUTAY: I can only speculate on why there are differences. The uniaxial
fatigue testing, the way that we run the test, its a push-pull test with a zero mean,
meaning that the strain level on the specimen has been kept constant on specimen. We
run the tests using the AMPT. And given this being a push-pull test rather than pull-pull
test, we are applying a lot more compression during the cycles. Again, this is a
speculation, but during the compression, healing might be something that we are
inducing a lot more than the other tests. So the healing is actually happening during the
rest periods. We know that it happens during the rest periods, but by applying
compression, maybe we are indirectly including healing into this overall ballpark
estimation of the fatigue life. So it may be possible that this healing is somehow
286
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

included in the push-pull test, which is not included in the others. You know, when you
look at this warm-mix asphalt increases the fatigue life, which typically most people
expect, but in the other tests, you dont see that. So it is possible that maybe the warm-
mix additive is helping towards healing. I dont know.
PROF. MOGAWER: Thats fair. Thank you very much.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
287

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Rutting and Moisture Damage Resistance of
High RAP Warm Mixed Asphalt: Loaded
Wheel Tracking vs. Conventional Methods
Jesse D. Doylea*, and Isaac L. Howardb
a
Airfields and Pavements Branch, Geotechnical and Structures Laboratory, U.S. Army
Engineer Research and Development Center, Vicksburg, MS 39180-6199, USA
b
Department of Civil and Environmental Engineering, Mississippi State University, P.O.
Box 9546, 235 Walker Hall, Mississippi State, MS 39762-9546, USA
ABSTRACT: An increased potential for moisture damage and rutting have been the two main
problems with warm mix asphalt (WMA) implementation. The use of high reclaimed asphalt
pavement (RAP) contents (25% or more) in WMA may alleviate these problems. At present,
susceptibility to moisture damage and rutting are usually tested for independently; however, these
distress mechanisms can be linked for in-service pavements. An enhanced loaded wheel tracking
test performed on dry and wet specimens, the PURWheel, is investigated in this paper to evaluate
the interaction of traffic and moisture. The PURWheel is also compared to conventional rutting
and moisture damage tests. PURWheel results are used to evaluate performance of 25 and 50%
RAP-WMA. Rutting and moisture susceptibility of the high RAP-WMA was comparable to current
practice low RAP content hot mix asphalt (HMA).
KEYWORDS: loaded wheel tracking, WMA, RAP, rutting, moisture damage, PURWheel
The oral presentation was made Dr. Doyle.
http://dx.doi.org/ 10.1080/14680629.2013.812841
289
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
1. Introduction
Adoption of warm mixed asphalt (WMA) in place of hot mixed asphalt (HMA) by the
highway industry has progressed rapidly in the last few years. Most state Departments of
Transportation (DOTs) have used WMA to some extent or allowed its use as an
alternative to HMA; a number of state DOTs now use WMA for the majority of their
paving projects. In recent years, other changes have occurred (e.g., mixtures with high
amounts of recycled material) that have been considered in conjunction with WMA.
Ever increasing demands on raw materials and energy have resulted in more new options
available to produce asphalt concrete than ever before.
Considerable research has been performed on WMA with generally promising results
(e.g., Prowell et al., 2007; Diefenderfer and Hearon, 2010). Reduced binder aging
associated with lower WMA mixing temperatures has the potential to result in rutting
problems soon after construction. Some evidence has been provided in literature for
increased rutting with WMA (Hurley and Prowell, 2006; Abbas and Ali, 2011);
however, other studies have indicated adequate WMA rutting performance (Hurley and
Prowell, 2005; Prowell et al., 2007). Numerous researchers have reported increased
susceptibility to moisture damage for WMA in laboratory testing (Hurley and Prowell,
2005; Hurley and Prowell, 2006; Prowell et al., 2007; Kvasnak et al., 2009). However,
other studies have observed improved moisture susceptibility test results with plant
produced WMA (Kvasnak et al., 2009; Prowell et al., 2007), and a survey of state DOTs
did not find any evidence of moisture damage problems in field projects based on 30
responses (Mogawer et al., 2011b).
Research has also been performed on WMA with reclaimed asphalt pavement (RAP)
contents in the 20 to 50% range (e.g., Hodo et al., 2009; Copeland et al., 2010; Willis et
al., 2011); in general, these meet the definition of high RAP content as 25% or more of
total mixture (Copeland, 2011). These mixtures have been studied for multiple reasons,
but alleviation of rutting and moisture damage seems to be two of the most prevalent
non-economic motivations for adding considerable RAP quantities to WMA. Evidence
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
suggests that use of high RAP in conjunction with WMA may offset any increased
rutting potential of the WMA (Doyle et al., 2011; Mallick et al., 2008; Boggs, 2008;
Hodo et al., 2009). A few studies showed that high RAP-WMA generally has decreased
moisture susceptibility relative to low or no RAP-WMA based on tensile strength
ratio (TSR) testing (e.g., Middleton and Forfylow, 2009; Mejas-Santiago et al., 2011)
and based on Hamburg testing performed at 40C (e.g., Mogawer et al., 2011a).
However, others have found marginal overall moisture damage performance of high
RAP-WMA (Hodo et al., 2009); while (Austerman et al, 2009) reported that 25% RAP
HMA performed acceptably while 25% RAP WMA did not based on Hamburg testing at
290

Rutting and Moisture Damage Resistance of High RAP Warm Mixed Asphalt
50C. In light of these mixed results, moisture susceptibility of high RAP-WMA should
be evaluated while taking rutting resistance into consideration.
The problems of rutting and moisture susceptibility of asphalt mixtures and the
difficulty of translating laboratory test results to field performance have been recognized
for over 100 years (Richardson, 1912). Although many advances have been made over
the last century, the aforementioned items are still areas of concern as new mixture types
and combinations of materials are developed. The current state of practice is to
independently evaluate rutting and moisture damage, but these behaviors can be coupled
for in-service pavements. One of the most common methods to measure rutting is the
Asphalt Pavement Analyzer (APA) executed according to AASHTO T 340 or with
agency specific procedures. One of the most common methods to measure moisture
susceptibility is the tensile strength ratio (TSR) conducted according to AASHTO T 283,
ASTM D4867, or with agency specific procedures. An additional moisture damage test
method that has gained popularity recently is the Hamburg Loaded Wheel Tester
(HLWT) performed according to AASHTO T 324 or with agency specific procedures.
The APA, TSR, and HLWT each operate under different principles and the ability of
each test method to perform its desired function has been debated. Generally speaking,
moisture susceptibility testing seems to be more of a point of contention than rut testing.
Of the techniques available to characterize moisture damage, loaded wheel tracking
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
seems to be the most promising.
Kiggundu and Roberts (1988, p. 3) defined moisture damage as the progressive
functional deterioration of a pavement mixture by loss of the adhesive bond between the
asphalt cement and the aggregate surface and/or loss of the cohesive resistance within
the asphalt cement principally from the action of water. Moisture damage in asphalt
mixtures falls into two general categories: (1) loss of cohesion within the binder film
(i.e., softening of the binder in the presence of moisture); and (2) loss of adhesion
between the binder film and the aggregate particles (Hicks, 1991). Loss of adhesion in a
mixture is generally visually apparent (e.g., stripping), whereas loss of cohesion in a
mixture is less readily observed, but it can be measured by loss of mixture strength. An
important factor contributing to moisture damage in the field is the presence of traffic in
addition to merely the presence of water (Lottman, 1982; Hicks, 1991; Little and Jones,
2003; Lu, 2005). Moisture damage is often a progressive failure mechanism and, in
some situations, manifests itself in the form of accelerated permanent deformation before
or without loss of adhesion (Little and Jones, 2003). A test method capable of detecting
any increased rutting in the presence of moisture as well as the potential for stripping
would be useful. The test could provide information on two of the most prevalent forms
of moisture susceptibility and also information related to rut resistance. Ideally, such a
test method could account not only for the interaction between water and traffic on an in-
service pavement but also be able to assess the influence of water and traffic separately.
291

DOYLE, HOW ARD
The PURWheel laboratory wheel tracker is a test method with the potential to
perform the needs described above. The original PURWheel was developed in the 1990s
at Purdue University. The original equipment was donated to Mississippi State
University in 2007 where it was renovated and modified. The equipment now operates
under the protocols described in Howard et al. (2010). Key items different from the
original equipment were an increase in test temperature to simulate the environmental
conditions typical of the southeastern U.S., the addition of insulation to maintain the
higher test temperatures, and changes to the specimen conditioning protocol to ensure
thermal equilibrium during the desired test temperature. Items that remained the same
were the load application mechanism and the deflection measurement method. One
motivation for the renovations was to evaluate southeastern U.S. climates and a second
motivation was to update the dry rutting test protocol to more closely match the APA.
This modified device is still referred to as the PURWheel, although as discussed, notable
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
differences exist with respect to the approaches previously published (e.g., Pan and
White, 1999; Stiady et al., 2003). Only two PURWheel devices are in existence.
However, the specific test parameters and testing protocols described in this paper are
believed to be unique to the prototype PURWheel described herein.
The primary objective of this paper was to investigate rutting and moisture damage
using the PURWheel for WMA containing 25 to 50% RAP. A secondary objective was
to compare PURWheel test results to the commonly used rutting and moisture
susceptibility methods (APA, TSR and HLWT). Loading characteristics of the wheel
tracking tests were compared to assess the mixture response. Thereafter, APA, TSR and
PURWheel data were collected in the laboratory and used to evaluate the performance of
WMA with high RAP content relative to current practice mixtures and the relative merits
of the PURWheel with respect to conventional practice tests.
2. Review of Test Method Development
2.1. History of APA and PURWheel Dry Protocol
The PURWheel was originally developed at Purdue University. For surface mixtures,
typical test parameters were 3.8 cm thick specimens and 50 to 60C test temperatures
(Pan and White, 1999; Stiady et al., 2003). Subsequent modifications to the original
equipment after its acquisition by Mississippi State University (Howard et al., 2010)
included additional insulation on the environmental chamber, changes to the heating
system, and improved air circulation within the chamber; a new specimen conditioning
protocol was implemented as well. Afterward, thermocouples were imbedded into
dummy specimens to monitor temperature. Results indicated that specimen mid-depth
temperature was maintained at 641C for the entire duration of a test and that mid-
292

depth temperature was uniform along the direction of loading. Temperature of the
specimen near the top surface was observed to vary slightly more due to cycling of the
air heating equipment and air circulation patterns within the environmental chamber,
although the variability was not more than 2C. In addition, the electro-mechanical
interface hardware was updated and new control software (Hawk) was incorporated.
This enhanced version of the PURWheel was intended to evaluate specimens in
conditions representative of the southeastern U.S.
Around the same time the original PURWheel was being developed, the Georgia
Loaded Wheel Tester (GLWT) was used as a basis for development of the Asphalt
Pavement Analyzer (APA) (Collins et al., 1996). Extensive additional work (e.g.,
Kandhal and Cooley, 2003) led to the final test parameters in use today. No directly
comparative studies have been performed to relate the APA to the original PURWheel or
to the enhanced PURWheel. However, both the APA and the original PURWheel were
shown to have reasonable relationships to field rutting measured at the Westrack
experiment, with R2 = 0.76 for APA and R2 = 0.80 for PURWheel (Federal Highway
Administration [FHWA], 1998). Similarities between APA and enhanced PURWheel
dry protocol testing consist of specimen thickness (7.6 cm), test temperature (64C for
LTTPBind PG 64 regions) and specimen conditioning time (6 hr). Major differences in
the test methods are load application method, specimen compaction method, and
specimen size. Differences in test parameters will be discussed in more detail later.
2.2. History of TSR, Hamburg Loaded Wheel Tester and PURWheel Wet Protocol
The original research performed by Lottman (1978) to develop the TSR test used
California Kneading (i.e., Hveem) compacted specimens (100 by 63.5 mm) compacted
in two lifts; the mixtures tested were based on in service pavements in 15 states
constructed between 1960 and 1972. None of the mixtures contained RAP, and all were
typical HMA mixtures for the time with the only additive being, in some cases, lime or
anti-strip. The use of freeze-thaw cycle(s) as part of the TSR conditioning protocol was
simply a method to induce accelerated damage to the specimens to match field
conditions and was not intended to relate to the occurrence of freeze-thaw zones
(Lottman, 1978). The follow-on field validation study was conducted on in-service
pavements, located in seven states, constructed between 1975 and 1977 of conventional
HMA without any additives other than lime or anti-strip; a minimum TSR value of 0.80
was recommended by Lottman (1982).
Since then, Tunnicliff and Root (1984) proposed a similar procedure using 100 by
62.5 mm Marshall Hammer compacted specimens that led to the ASTM D4867 test
method (Tunnicliff and Root, 1995). The follow-on field validation study included eight
293
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
pavements in eight states and resulted in a minimum recommended TSR value of 0.75
(Tunnicliff and Root, 1995). None of the mixtures contained RAP or any additives other
than lime or anti-strip. The current Mississippi DOT (MDOT) test method for moisture
damage is MT-63 and which resembles ASTM D4867 without optional freeze-thaw
conditioning except for Superpave gyratory compactor (SGC) compaction method and
use of larger specimens; MDOT requires TSR values greater than 0.85.
The original TSR test developed by Lottman (1984) was modified in light of the
research by Tunnicliff and Root (1984) which eventually led to the AASHTO T 283 test
procedure (Epps et al., 2000). Laboratory work was performed by Epps et al., (2000)
with mixtures of known field performance from five states to refine T 283. Results
indicated that compaction method and specimen size were important factors. The 100
mm Marshall compaction and 150 mm SGC compaction resulted in similar TSR values;
however, 150 mm SGC compaction resulted in generally higher TSR values than 100
mm SGC or 100 mm Hveem compaction. The extent of loose mixture conditioning,
specimen saturation level and freeze-thaw conditioning also affected results. AASHTO
T 283 was modified to include 16 hr of loose mix aging and one freeze-thaw
conditioning cycle based on the studys recommendations.
Table 1 provides a comparison of the two nationally accepted TSR test methods and
MT-63. There are several differences between the test methods, many of which will
affect results (e.g., compaction method, specimen size, loose mixture conditioning,
saturation level, and use of a freeze-thaw conditioning cycle). Since the development of
the TSR test, mix design has changed to Superpave (i.e., the SGC), and a variety of new
mix types have been introduced; the use of RAP, WMA or other modifiers is common.
The TSR test is currently used by a majority of state DOTs to evaluate moisture damage
(Mogawer et al., 2011b); yet researchers have observed that: (1) TSR is not a reliable
indicator of field performance for AASHTO T 283 (Zaniewski and Viswanathan, 2006,
p. 67) and (2) there is no relationship between TSR and specific pavement distresses
that are related to moisture damage (surface ravelling and rutting) for ASTM D4867
(Kanitpong and Bahia, 2008, p.9).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
294

Table 1. Comparison of current TSR test protocols for lab-mixed specimens.

Test Method ASTM D4867 AASHTO T 283 MDOT MT-63
Specimen compaction Marshall, Hveem or
Marshall, Hveem or SGC SGC
methods allowed Gyratorya
Specimen size 100 62.5 mm or 100 63.5 mm or
150 95 mm
requirements larger 150 95 mm
Hyd. lime addition dry or wet not discussed wet
Air void targetb 7 1% 7 .0 0 .5 % 7 .0 1 %
Pre-compaction loose 1-2 hr @ Cool for 2 hr, 16 hr @ 1.5 hr @
mixture conditioning comp temp 60C, 2 hr @ comp temp comp temp
Post-compaction cool to room temp room temperature cool to room temp
conditioning for max of 24 hr for 24 3 hr for 12 to 24 hr
Saturation range 55-80% 70-80% 55-80%
Freeze-thaw cycle Opt., 15 hr at -18C Yes, 16 hr at -18C No
Conditioning 24 hr @ 60C 24 hr @ 60C 24 hr @ 60C
TSR single-lab (d2s) None provided 9%c None provided
TSR multi-lab (d2s) 23% 25%c None provided
Minimum TSR typically 75% typically 80% 85%
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
a
May use either Gyratory Testing Machine (GTM) or Gyratory Shear Compactor (GSC).
Does not explicitly allow SGC compaction method.
b
Air voids are determined by traditional saturated surface dry method (D2726 or T 166).
c
Reported by Azari, 2010.
The HLWT was developed in Germany and has been studied by Aschenbrener
(1993, 1995). It was observed that the average stripping inflection point (SIP) for
mixtures as determined by the Hamburg test provided an excellent correlation to field
performance with respect to moisture damage. Pavements with good field performance
had average stripping inflection points (SIP)s generally greater than 10,000 passes and
pavements with poor field performance had average SIPs less than 3,000 passes
(Aschenbrener, 1993). Results indicated that mixture moisture resistance was improved
by increasing the binder grade or increasing the short term aging time; different binder
sources of the same grade were also observed to have different performance
(Aschenbrener, 1995). HLWT testing has been observed to better correspond to field
performance than AASHTO T 283 TSR testing (Azari, 2010). Nielson (2010) found that
moisture damage as evidenced by SIP only occurred above a critical stripping
temperature (CST) that was binder grade dependent. For PG 64 binder and PG 70 binder
grades, CST ranges were reported of 49 to 54C and 54 to 55C, respectively. The
295

DOYLE, HOW ARD
HLWT has been adopted by the Texas DOT and a few other agencies; Texas
specification criterion is defined as minimum number of passes to reach 12.5 mm rut
depth when test is performed at 50C. The criterion is 10,000 and 15,000 passes for PG
64 and PG 70 binders respectively; however it has been suggested that these are too
conservative and alternate criterion be used of 5,000 and 10,000 passes for PG 64 and
PG 70 binders respectively (Rand, 2006).
Development of the original PURWheel was an attempt to enhance the original
HLWT by: (1) incorporating a constant loading speed instead of sinusoidal; (2) using a
pneumatic wheel instead of a solid steel wheel; (3) using larger specimen sizes to
minimize boundary effects; (4) measuring rut depths over the entire specimen length
during testing; (5) incorporating wheel wander; and (6) the flexibility to test in both
hot/dry and hot/wet conditions (Pan and White, 1999). The renovations to the original
PURWheel have already been described. For the wet test protocol, the same
conditioning period before testing and test temperature of the dry protocol was used (6
hr and 64C). Temperature of the water bath was measured with thermocouples to be
uniform from top to bottom and to be 64 1C (Howard et al., 2010). PURWheel testing
has been stated to give a better indication of moisture damage potential than T 283 TSR
testing (Pan and White, 1999).
3. Comparison of Loaded Wheel Testers
Table 2 is a summary of loading details for APA, HLWT and PURWheel tests.
Unlike the APA and HLWT wheel testers, the PURWheel applies loads to specimens
through a pneumatic tire. Load applied by the PURWheel was directly measured with a
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
scale (Figure 1a). Figure 1b shows a tire print where individual treads are clearly visible.
The gross contact area of the tire print inside the ellipse is approximately 2800 mm2.
Additional details are provided in Howard et al. (2010).
Initially, a three dimensional finite element analysis was undertaken to investigate
differences in rutting behaviors measured by the APA, HLWT and PURWheel tests.
However, results were not fully developed for the desired comparison purposes and an
experimental approach was undertaken instead. In general, the greater wheel widths of
the PURWheel and Hamburg may be less sensitive to the effects of individual large
aggregate particles than the APA. In addition, the steel wheel load application of the
Hamburg will not conform to the contours of the deformed specimen surface during
testing, unlike the pneumatic load application of the APA and PURWheel.
296

Table 2. Comparison of APA, HLWT and PURWheel test loading parameters.

Parameter AP A HLWT PURWheel
Load Application Sinusoidal Sinusoidal Linear
Speed (cm/sec)a 59 31 33
Load Application Method pressurized hose steel wheel pneumatic tire
Load (N) 445 705 1750
Load Area Width (mm) 2 7 .5 b 4 7 .0 5 4 .0
Strip Load (N/mm) 1 6 .2 1 5 .0 3 2 .4
Contact Area (mm2) 645c 970d 2800 (gross)
Contact Pressure (MPa) 0 .6 9 c 0 .7 3 d 0.63 (gross)
a
Speed at the point where specimen deformation is measured.
b
Value assumes full embedment of the hose; actual width will be narrower at start of test.
c
Reported by Kandhal and Cooley, 2003.
d
Reported by FHWA, 2012.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Gross Area
2800 mm2
Contact Print
Scale
54.0 mm
(a) Load and contact measurement (b) Completed contact area print
Figure 1. Measurement of load and contact area for PURWheel.
In addition to the finite element effort, laboratory testing was conducted to compare
rutting behaviors of the three tests. Samples of a plant-produced 9.5 mm nominal
maximum aggregate size (NMAS) coarse-graded surface mixture containing 10% RAP
297

DOYLE, HOW ARD
designed with 85 gyrations with a polymer-modified PG 76-22 binder were used for
testing. (Designed properties were Gmm = 2.356, AC% = 5.8%, VMA = 15.0, VFA = 73
and dust to effective binder ratio of 1.13.) Specimens of this mixture were compacted to
7% air voids on an AASHTO T 331 basis for rut testing in the APA at 50 and 64C, in
the Hamburg in a dry condition at 50 and 64C and in the PURWheel in a dry condition
at 64C. Figure 2 shows these rut depth test data plotted as a function of passes. APA
testing produced the lowest rut depths for a given temperature, which is understandable
given its considerably higher loading speed. Hamburg testing produced the greatest rut
depths for a given temperature, which can be explained by the slow loading speed and
high contact pressure. PURWheel testing produced moderately higher rut depths than the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
APA and considerably smaller rut depths than the Hamburg for the same temperature.
20
Ham 64C
PW 64C
16
Rut Depth (mm)
APA 64C
Ham 50C
12
APA 50C
0
0 4 8 12 16 20
Thousand Passes
Figure 2. Experimental APA, dry HLWT and dry PURWheel rutting data.
An interesting observation from the Hamburg test data was that most of the
permanent deformation of this mixture was due simply to rutting since there was 4.8 mm
of total deformation in a dry test at 50C; while the same mixture in a customary
Hamburg wet test at 50C only had 4.9 mm of deformation (data not shown) and no
stripping inflection point or evidence of moisture damage. This dry rutting behavior is
not accounted for in traditional wet Hamburg testing, which provides more of an overall
performance assessment since all rutting deformation is grouped together without
distinction between dry rutting and moisture-related rutting. On the other hand, dry
PURWheel testing of the same mixture at 64C produced 8.8 mm of deformation, while
wet PURWheel testing indicated moisture damage with an average SIP of about 6,000
passes (not shown). These results are in line with the findings of Nielson (2010) that
little or no moisture damage occurs at lower test temperatures.
298

4. Experimental Program
A research approach was developed to support the objectives of this paper. The specific
goal was to determine if the rutting and moisture susceptibility performance of 25 and
50% RAP-WMA was similar to the performance of mixtures currently used in
Mississippi. With this in mind, control mixtures were selected to represent a band of
performance results for current practice mixtures. The performance of 25 and 50% RAP
mixtures was then compared to this band to address the primary objective. Results from
individual performance tests were compared across mixtures to address the secondary
objective. Details of the materials selected, specimen preparation and testing methods
are provided in the following sections.
4.1. Materials Tested
All testing was performed on 9.5 mm NMAS surface mixes where the coarse aggregate
type was predominately crushed gravel. The high-RAP content WMA mixtures
evaluated contained 25 and 50% RAP; two RAP sources were tested. One warm mix
technology (Sasobit) was used at 1.0% of the total binder mass including all RAP
binder. Performance of these four mixtures was compared to three control mixtures with
15% RAP selected to encompass typical behaviors in Mississippi.
The two RAP materials tested (Figure 3) varied in terms of their binder grades and
gradations, and were selected to determine the influence of RAP properties (binder in
particular) on rutting and moisture susceptibility. Figure 3 shows that both gradations
have a hump near the 0.6 mm sieve that is often indicative of a high natural sand content.
Note the especially high viscosity and PG temperature grades for the R1 RAP source. R1
represents a high design traffic mixture; the material was milled from the surface of a
22.5 km stretch of Interstate 55 near Grenada, MS. The material was originally placed in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
1992 and used both polymer modified and non-polymer modified binders. R2 was
selected to represent an intermediate design traffic mixture commonly used on lower
volume roads and state highways. The material was milled from a 12 km stretch of State
Highway 25 in Monroe County, Mississippi.
299

DOYLE, HOW ARD
100
Aggregate Property R1 R2 .
NMAS (mm) 9.5 9.5
Gradation Type Coarse Fine R1
80 Gsb 2.534 2.565
Abs (%) 1.8 1.1
R2
Percent Passing
Max Density
60
40
Binder Property R1 R2 .
20 T 164 AC (%) 5.5 5.6
R 29 PG True Grade 118+2 106-4
T 316 Viscosity @ 135C (Pas) 52.9 9.1
0
0 .1 5
0 .0 7 5
9 .5
0 .3
0 .6 1 .1 8 2 .3 6 4 .7 5 1 2 .5 1 9 .0
Sieve Size mm0.45
Figure 3. Properties of RAP sources tested.
Table 3 summarizes the seven mixtures investigated for use on highway surfaces. In
general, the relative performance boundaries of asphalt mixtures in Mississippi are
represented by 50 design gyration and 85 gyration mixtures. The 50 gyration mixtures
are intended for low traffic volumes and have the highest effective asphalt content for a
particular gradation, resulting in a flexible pavement that is more crack resistant but less
rut resistant. The 85 gyration mixtures are intended for high traffic volumes and have the
lowest effective asphalt content for a particular gradation, resulting in stiff, more rut
resistant but less crack resistant pavements. The goal when selecting control mixtures
CM1 to CM3 for this study was to encompass the range of potential cracking and rutting
resistance of current practice mixtures to the best extent possible. These mixtures were
plant produced and sampled during highway surface paving for MDOT. The 85 gyration
mixtures contained neat binder (CM3) and polymer-modified binder (CM2). The four
high RAP content WMA mixtures were developed in the laboratory with 65 design
gyrations. For the four recycled mixtures, the same virgin aggregate stockpile
percentages were used with each RAP source at the 25% and 50% RAP contents to
control the effects of virgin aggregate gradation. Virgin PG 67-22 binder with 1.0%
Sasobit was used for all high RAP mixtures.
300
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 3. Properties of highway surface mixtures tested.

Mixture ID CM 1 CM 2 CM 3 RM 1 RM 2 RM 3 RM 4
W MA No No No Ye s Ye s Ye s Ye s
TMix (oC) 157 132 160 116 116 116 116
TComp (oC) 116 138 138 116 116 116 116
Ndes 50 85 85 65 65 65 65
P G Gr a d e a 67-22 76-22 67-22 67-22 67-22 67-22 67-22
RAP ID Stockb Stockb Stockb R1 R2 R1 R2
RAP AC (%) 4 .6 5 .5 5 .6 5 .5 5 .6 5 .5 5 .6
RAP (%) 15 15 15 25 25 50 50
Crushed Gravel (%) 43 37 37 56 56 40 40
Limestone (%) 0 31 37 8 8 0 0
Sand (%) 9 10 10 10 10 9 9
Hydrated Lime (%) 1 1 1 1 1 1 1
-9.5 mm (%) 92 96 96 93 95 92 96
-No 4 (%) 67 64 60 59 63 58 65
-No 8 (%) 40 41 37 40 42 40 45
-No 200 (%) 5 .5 6 .1 6 .0 7 .2 7 .5 7 .3 7 .8
Combined Gsb 2 .5 3 3 2 .4 8 0 2 .5 1 8 2 .4 6 2 c 2 .4 6 9 c 2 .4 7 7 c 2 .4 9 2 c
Virgin AC 5 .6 5 .0 5 .0 4 .7 4 .2 3 .7 3 .4
Total AC 6 .3 5 .8 5 .8 6 .1 5 .6 6 .5 6 .2
Gmm 2 .3 6 2 2 .3 3 2 2 .3 6 7 2 .3 0 6 2 .3 3 4 2 .3 1 1 2 .3 3 8
a
PG grade of virgin asphalt prior to mixing.
b
Stock of RAP at production facility, properties were not monitored and the source of the stock was
different for CM1, CM2, and CM3.
c
Gsb of RAP aggregate used to calculate combined value for the gradation was determined by the
procedure recommended in Doyle et al., 2012.
4.2. Specimen Preparation and Test Methods
After mixing and short term conditioning for 1.5 hr, SGC specimens were prepared with
a Pine Instrument model AFGC125X calibrated to an external angle of 1.25 0.02 deg.
The internal angle was verified to be within allowable tolerance of 1.16 0.02 deg.
Slabs were compacted in the Linear Asphalt Compactor (LAC) shown in Figure 4a.
After mixing and short term conditioning for 1.5 hr, the asphalt mixture was placed in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
the mold to prevent segregation and a series of 47 steel plates were placed on top (Figure
4b). Compaction parameters were 18 passes and a 2,413 kPa system pressure. The
rectangular slabs were 29 by 62 cm and a nominal thickness of 7.6 cm (Figure 4c). The
compacted slabs were then cut in half to produce two PURWheel test specimens referred
301

DOYLE, HOW ARD
to as blocks (Figure 4d) or were cored to produce specimens for APA testing. Additional
operational details are provided in Doyle and Howard (2010), while Howard et al.
(2012) provides an evaluation of the compactor.
( a) (b )
( c) (d )
Figure 4. Linear Asphalt Compactor (LAC).
Theoretical maximum specific gravity (Gmm) of each mixture was determined on

duplicate specimens according to AASHTO T 209 and the average value was reported.
Bulk specific gravity (Gmb) of compacted cylindrical specimens was determined
according to AASHTO T 331 and used to determine specimen air voids (Va); research by
the authors has indicated that this test method produces the most reliable and consistent
measurements of Gmb over a wide range of specimen air voids (Howard and Doyle,
2012). For slab compacted specimens, air voids were estimated from measurement of the
bulk density and correlation to equivalent air voids on an AASHTO T 331 basis using
the correlation developed in Doyle and Howard (2010).
To evaluate rutting potential, the Asphalt Pavement Analyzer (APA) was used to test
each mixture in the experimental design according to AASHTO TP 63. SGC compacted
302
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

specimens were tested at nominal air void levels of 7 and 10%; LAC compacted
specimens were tested at their respective air void levels. Wheel load was 445 N and hose
pressure was 690 kPa; the test was run for 16,000 passes (i.e., 8,000 cycles) or 14 mm of
rutting, whichever occurred first. A pass was defined as the load wheel moving from one
end of the specimen to the other, i.e., two passes is one cycle. Rut measurements during
testing were recorded automatically by the APA device and rut depth (RDAPA) was
defined as downward permanent deformation of the specimen surface relative to the
original surface, i.e., uplift of the specimen surface was not considered part of the overall
rut depth. Rutting rate was determined by fitting a linear equation to the post
consolidation rut depth data at 2,000 and 8,000 cycles (i.e., 4,000 and 16,000 passes); the
slope of that line was defined as the rutting rate parameter (RRAPA) with units of
mm/1000 cycles.
To evaluate moisture damage, ASTM D4867 was performed on each high RAP
mixture. Six 100 mm diameter and 62.5 mm thick specimens were tested. A freeze-thaw
cycle was not performed as part of the conditioning process. Air voids of all specimen
sets were between 6 and 8% as by the test method. For control mixtures, TSR values
determined by MDOT MT-63 during mix design were used.
Two protocols were used for PURWheel testing: (1) PURWheel dry protocol (PW-
dry), specimen tested at 64C; and (2) PURWheel wet protocol (PW-wet), specimen
tested submerged in 64C water. The tested LAC-produced specimens were 7.6 0.6 cm
thick and approximately 30 cm square, the air voids of slab specimens were between 6
and 10%. Figure 5 shows key components of the PURWheel. Figure 5a shows an overall
view where the hood/tank assembly, control box, and computer powered by Hawk
software are visible. Two independently controlled wheel carriages mounted with 4-ply
pneumatic tires load the slabs for 20,000 passes. Figure 5b shows a slab plastered into
the mold and ready to be tracked by one of the two wheel carriages. Figure 5c shows an
expanded view of one of the carriages with the pneumatic wheel retracted after a test is
complete. Figure 5d is a detail view of a PW-dry protocol specimen after testing. Figure
5e is a detail view of a PW-wet protocol specimen after testing.
303
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
Computer, Control Box

Hood and Tank Assembly
Shown closed as when a test
(a ) is conducted
(b) (c)
(d) (e)
Figure 5. PURWheel laboratory wheel tracker.
Four slabs (eight test blocks) were compacted per mixture (2 PW-dry and 2 PW-wet).
Data from corresponding LAC slab test blocks were averaged to create one PW-dry or
PW-wet replicate. Dry and wet PURWheel testing was performed according to the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
procedures in Howard et al., (2010). Specimens were conditioned at the test temperature
304

of 64C for 6 hours before testing. Test parameters for PURWheel wet protocol testing
were the same as for the dry protocol test with the addition of sufficient 64C water to
the test chamber to fully submerge the specimens. The tire inflation pressure was 862
kPa, the wheel load was 1,750 N, and the gross contact pressure at the beginning of the
test was approximately 630 kPa. The test was run for 20,000 passes or 20 mm of rutting,
whichever occurred first. A pass was defined as one passage of the load wheel over the
specimen. Raw rut measurements during testing were recorded by the PURWheel device
and adjusted as described in Howard et al., (2010). Rut depth (RDPW) was defined as
downward permanent deformation of the specimen surface relative to the original
surface, i.e., uplift of the specimen surface was not considered part of the overall rut
depth. The number of passes to 12.5 mm of rutting was determined and denoted P12.5.
In PW-dry testing, rutting rate was determined by fitting a linear equation to the post-
consolidation data between 4,000 and 16,000 passes (i.e., 2,000 and 8,000 cycles); the
slope of that line was defined as the rutting rate parameter (RRPW) with units of
millimeters/1000 cycles for comparison to APA data. For PW-wet test data, linear
equations were fitted to rut depth data in the creep region and stripping region; slopes of
these lines were denoted SC and SS respectively with units of millimeters/1000 passes.
The calculated intersection of these two equations was stripping inflection point (SIP).
Air voids of slabs tested in the PURWheel were targeted to be 82% on a Corelok
(AASHTO T 331) basis or 71.7% on a submerged basis (AASHTO T 166). This air
void level is in line with southeastern U.S. DOT specifications where average target, full
pay maximum, and removal void levels are approximately 7, 8, and 10% measured via T
166 (Howard and Doyle, 2012). This criterion allows air voids to be in the range stated
by Terrel and Al-Swailmi (1993) to be favorable to moisture damage (7 to 11%). Testing
slabs at air void levels representing the higher end of permissible values is more
indicative of cases likely to be susceptible to moisture damage in service. If a mixture
performs adequately at the upper end of allowable voids, it should in turn perform well
at the lower end of allowable voids with the same aggregate blend and asphalt content.
5. Test Results and Analysis
Table 4 summarizes results of the 28 PURWheel tests performed. Note that mixture
CM1 performed poorly in both PW-dry and PW-wet tests; the test terminated early when
all specimens exhibited excessive rutting prior to 2,000 passes. Post-test visual
observations of the specimens revealed failure of the mix in shear as evidenced by the
sharp vertical edges of the wheel path. When a mixture did reached 20,000 passes but
did not achieve 12.5 mm of rutting or have a SIP, a value of 20,000 passes was used for
averaging.
305
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
5.1. Effect of Compaction Method on APA Results
The effect of specimen compaction method on APA rut performance was assessed b y
comparing LAC compacted specimens to SGC compacted specimens at the same level
of air voids. The comparison consisted of drawing a line between SGC specimen data
points to predict rutting performance as a function of air voids (Figure 6a). Based on the
equation of this line, estimates of the rutting performance at the level of air voids of
LAC compacted specimens were calculated and the difference between measured
performance of LAC specimens and predicted performance of SGC specimens at the
same level of air voids was then determined (differences were +0.055 and -0.100 for air
voids of 6.6 and 6.8%, respectively). The same procedure was performed with rut depth
data and the process was repeated for all mixtures.
A relative frequency histogram summarizing these data are presented in Figures 6b
for rutting rate; a similar histogram was produced for rut depth data (n = 14, mean = -
0.44, Std. dev. = 1.89). Based on one sample two-tailed Students t-tests at a 5%
significance level (t0.025 = 2.160), the mean of the differences was not statistically
different than zero for rutting rate (t = -0.721) or total rut depth (t = -0.087). Indicating
that for this specific data set, mean APA rutting performance for SGC and LAC
compacted specimens was similar. A similar data analysis approach was taken in
Howard et al., (2012) for a larger data set and it was observed that total rut depths of
LAC compacted specimens were generally similar to or slightly higher than for SGC
compacted specimens at similar air void levels.
5.2. Comparison of APA and PURWheel Dry Protocol Rutting

Based on the observations in the previous section, PW-dry test results for LAC-
compacted specimens were compared directly to APA results for SGC specimens. Both
PW-dry and APA data were adjusted to 8% air voids on T 331 basis, corresponding to
7% air voids on a T 166 basis (Howard and Doyle, 2012). This adjustment was made by
assuming that, for a given mixture, the change in rut depth or rutting rate per unit change
in air voids was the same for APA and PW-dry tests. The shift factor was determined by
dividing the change in rutting by the corresponding change in air voids. For example,
mixture CM2 at 16,000 passes had 2.0 mm APA level rutting at 6.8% air voids and 4.4
mm rutting at 9.5% air voids; the shift factor was then (4.4- 2.0)/(9.5-6.8) = +1.037 mm
of rutting for each 1% change in air voids. For PW-dry replicate 1, air voids were 6.9%
and rutting was 5.2 mm; the rut depth was then increased by (8-6.9)*1.037 = 1.1 mm for
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
a shifted rut depth of 6.3 mm (5.2+1.1 = 6.3). The effect of using air void shift factors
based on APA data instead of shift factors based on PURWheel data are trivial, as shown
in Figure 7. For each mixture, shift factors based on APA rut depths at 16, 000 passes
306

Table 4. Dry and wet PURWheel test results.

Dry Protocol (PW-dry) Wet Protocol (PW-wet)
Va RDPW by Passes (mm) Va RDPW by Passes (mm)
Mixture ID Re p (%)a P12.5 RRPW 4k 5k 10k 15k 16k 20k Re p (%)a P12.5 SC SS SIP 5k 10k 15k 20k
CM1 1L&R 9.2 500 --- --- --- --- --- --- --- 1L&R 9.1 550 --- --- --- --- --- --- ---
2L&R 9.1 350 --- --- --- --- --- --- --- 2L&R 9.2 450 --- --- --- --- --- --- ---
Avg. 9.2 425 --- --- --- --- --- --- --- Avg. 9.2 500 --- --- --- --- --- --- ---
CM2 1L&R 6.9 20000 0.35 3.1 3.5 4.4 5.1 5.2 5.5 1L&R 7.0 8550 1.02 5.86 8000 6.5 14.1 --- ---
2L&R 8.8 20000 0.58 4.3 4.8 6.4 7.7 7.8 8.5 2L&R 7.3 20000 0.22 --- 20000 3.5 4.6 5.6 7.7
Avg. 7.9 20000 0.46 3.7 4.1 5.4 6.4 6.5 7.0 Avg. 7.2 14275 0.62 5.86 14000 5.0 9.4 5.6 7.7
CM3 1L&R 8.0 20000 0.26 3.1 3.3 4.1 4.7 4.7 5.1 1L&R 7.4 20000 0.12 --- 20000 3.0 3.6 4.2 4.7
2L&R 11.5 20000 0.62 4.6 5.1 6.8 8.1 8.3 9.2 2L&R 4.8 13100 0.72 9.79 12750 5.4 5.7 6.8 7.6
Avg.b 8.0 20000 0.26 3.1 3.3 4.1 4.7 4.7 5.1 Avg.b 7.4 20000 0.12 --- 20000 3.0 3.6 4.2 4.7
RM1 1L&R 10.3 18200 0.97 4.4 5.1 7.7 9.7 10.2 11.7 1L&R 9.5 10500 0.95 4.52 11250 6.1 13.0 17.1 ---
2L&R 9.0 20000 0.51 3.4 3.8 5.1 6.3 6.5 7.3 2L&R 9.1 15200 0.64 7.90 15750 4.7 7.7 12.4 ---
Avg. 9.7 19100 0.74 3.9 4.4 6.4 8.0 8.4 9.5 Avg. 9.3 12850 0.79 6.21 13500 5.4 10.4 14.8 ---
RM2 1L&R 7.0 17100 0.81 5.3 5.9 8.1 9.8 10.2 11.7 1L&R 9.1 4500 1.94 9.02 4000 12.0 --- --- ---
2L&R 10.4 20000 0.53 4.4 4.7 6.2 7.3 7.5 8.2 2L&R 8.9 11400 1.19 7.02 4000 10.0 7.8 10.5 13.6
Avg. 8.7 18550 0.67 4.8 5.3 7.1 8.5 8.8 9.9 Avg. 9.0 7950 1.56 8.02 4000 11.0 7.8 10.5 13.6
RM3 1L&R 8.1 20000 0.13 1.8 1.9 2.2 2.5 2.5 2.6 1L&R 8.2 12750 0.54 3.72 11500 3.8 7.5 15.3 ---
2L&R 8.7 20000 0.20 2.1 2.3 2.9 3.3 3.4 3.4 2L&R 8.3 14450 0.37 10.10 8500 3.5 12.7 3.7 4.3
Avg. 8.4 20000 0.16 1.9 2.1 2.6 2.9 2.9 3.0 Avg. 8.8 13600 0.46 6.91 10000 3.6 10.1 9.5 4.3
RM4 1L&R 6.4 20000 0.45 3.2 3.6 4.8 5.7 5.9 6.6 1L&R 6.7 10200 0.65 5.06 9250 4.9 9.9 --- ---
2L&R 8.0 20000 0.57 3.8 4.2 5.7 6.9 7.2 8.0 2L&R 7.6 14000 0.60 10.80 7500 4.3 3.2 7.1 10.0
Avg. 7.2 20000 0.51 3.5 3.9 5.2 6.3 6.5 7.3 Avg. 7.2 12100 0.63 7.93 8375 4.6 6.6 7.1 10.0
Note: Dashes (---) in a column entry indicate that the data in question is not applicable or could not be calculated for that test.
a
Air voids on a AASHTO T 331 basis.
b
Replicate 2L&R was not included in average since air voids were outside of desired tolerances.
307
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
0.6
RRAPA = 0.0248Va + 0.08
RRAPA (mm/1000 cycles)
SGC data
0.5
LAC data
0.4
Linear (SGC data)
0.3 +0.055
0.2
0.1 -0.100
0.0
0 2 4 6 8 10 12 14
Va (% )
(a) Example RRAPA data for CM2
25
Relative Frequency (% )
t0.025 = 2.160 n: 14
20 t = -0.721 Mean: -0.016
Std. dev.: 0.083
15
10
0
-0.2 -0.1 -0.1 0.0 0.1 0.1 0.2
[LAC RRAPA - SGC RRAPA] (mm/1000 cycles)
(b) Histogram for rutting rate
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 6. Effect of compaction method on APA results.
(8,000 cycles) were used to shift PW-dry rut depth data at 16,000 cycles and similarly
for rut depth data at 5,000 passes and rutting rate.
Figures 8 to 10 are equality plots showing APA and PW-dry data for rut depth at
16,000 passes, rut depth at 5,000 passes and rutting rate, respectively. These figures use
data from Table 4 alongside eleven additional base mixtures tested and analyzed with the
same methods of this paper to compare APA and PURWheel rut depths adjusted to 8%
air voids on a T 331 basis. The eleven additional base mixtures consisted of four control
mixtures, two 50% RAP mixtures, two 75% RAP mixtures and two 100% RAP
308

mixtures. The same raw materials (R1 and R2 RAP sources, virgin aggregates, and
virgin binder) were used as in this paper and the raw data from those mixes is presented
in Doyle (2011).
10
Rut Depth at 5,000 Passes
PW rutting at Va = 8% with APA shift factors
Rut Depth at 16,000 Passes

8
Rutting Rate
Plot uses data from mixtures CM2,

2
RM2, RM3 and RM4 where PW-dry
air voids reasonably bracket 8%
0
0 2 4 6 8 10
PW rutting at Va = 8% with PW shift factors
Figure 7. Comparison of shift factors based on APA or PURWheel data.
10
RDPW Adjusted to 8 % T3 3 1 Air Voids (mm)
9 .5 mm (Tab le 4 )
1 2 .5 mm (Do yle, 2 0 11)
8
Eq u ality Lin e
Series4
6 Series5
4
In tercep t Set to Zero :
RDPW = 1 .4 7 (RDAPA) R2 = 0 .2 9
2
In tercep t No t Set to Zero :
RDPW = 0 .9 0 (RDAPA) + 2 .6 R2 = 0 .5 1
0
0 2 4 6 8 10
RDAPA Adjusted to 8 % T3 3 1 Air Vo ids (mm)
Figure 8. 16,000 pass rut depth comparison of APA to PW-dry adjusted to Va = 8%.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`-
309

DOYLE, HOW ARD
6
RDPW Adjusted to 8 % T3 3 1 Air Voids (mm)
9 .5 mm (Tab le 4 )
1 2 .5 mm (Do yle, 2 0 11)
Eq u ality Lin e
4 Series4
Series5
In tercep t Set to Zero :

2 RDPW = 1 .3 8 (RDAPA) R2 = 0 .3 3
In tercep t No t Set to Zero :

RDPW = 0 .6 2 (RDAPA) + 2 .3 R2 = 0 .4 0
0
0 2 4 6
RDAPA Adjusted to 8 % T3 3 1 Air Vo ids (mm)
Figure 9. 5,000 pass rut depth comparison of APA to PW-dry adjusted to Va = 8%.
1
RRPW Adjusted to 8 % T3 3 1 Air Voids (mm/1 0 0 cy cles)
9 .5 mm (Tab le 4 )
1 2 .5 mm (Do yle, 2 0 11)
Eq u ality Lin e
0.75
Series4
Series5
0.5
Intercept Set to Zero:

RRPW = 1.73(RRAPA) R2 = 0.35
0.25
Intercept Not Set to Zero:
RRPW = 1.31 (RRAPA) + 0.12 R2 = 0.39
0
0 0.25 0.5 0.75 1
RRAPA Adjusted to 8 % T3 3 1 Air Vo ids
Figure 10. Rutting rate comparison of APA to PW-dry adjusted to Va = 8%.
310
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

A total of 15 mixes were plotted; CM1 and one of the control mixtures of Doyle
(2011) could not be used due to early rutting failure, one of the 75% RAP mixtures from
Doyle (2011) was a noticeable outlier when the data were plotted (inclusion of this
mixture did not change the overall prediction by a large amount but it was believed the
data were more representative without its inclusion). Figures 8 to 10 include data for 60
individual PW-dry test specimens (15 mixtures with 4 specimens each). The 9.5 mm
surface mixes and the 12.5 mm base mixes were plotted separately, but they did not
differ by a considerable amount so one set of prediction equations was developed.
Figures 8 to 10 do show the 9.5 and 12.5 mm mixes with different markers. In general,
PURWheel data indicates greater rut depths and higher rutting rates than APA data, a
result that aligns with conclusions in the loaded wheel tester comparison discussed
earlier. The data in Figure 9 is used in the moisture damage analysis presented in the
following section.
5.3. Analysis of PURWheel Moisture Susceptibility Data
The first step in analysis of PW-wet data is illustrated in Figure 11a for mixture RM3
and consisted of fitting lines to the data in the first steady-state (creep) region and the
second steady-state (strip) region. The Stripping Inflection Point (SIP) was calculated as
the intersection point of the two lines; the calculations are demonstrated on the figure
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
and the vertical dashed arrow shows were this occurs. Values were rounded to 500
passes since selection of these line fits are subjective. Determination of the number of
passes to 12.5 mm of rutting (P12.5) was straightforward and is shown by the solid
horizontal and vertical arrows; values were rounded to 100 passes.
It was observed that the values for SIP were generally very similar to values for P12.5.
Figure 11b investigates this relationship for 17 mixtures from Table 4 and Doyle (2011);
mixture CM1 from Table 4 failed too quickly to develop any SIP values and was not
used. Figure 11 includes data for 68 individual PW-wet test specimens (17 mixtures with
4 specimens each). It is apparent that for the specific combination of test temperature
and loading conditions in the PW-wet test, P12.5 can safely be used as a surrogate for SIP
in further analysis. Doing so allows for direct comparison of PW-wet and PW-dry data
since P12.5 values are obtained in the PW-dry test but SIPs are not.
Three indicators were defined from PURWheel testing to evaluate moisture
susceptibility: (1) cohesion ratio for loss of cohesion due to moisture; (2) SIP for loss of
adhesion due to moisture; and (3) P12.5 ratio to provide an overall assessment of mixture
performance in the presence of moisture. All of these performance measures are used in
the next section to assess performance of high RAP content WMA.
311

DOYLE, HOW ARD
25
Mix RM3
Strip Region
Rut Depth (mm)
20 y = (10.08E-3)x - 78.6 Replicate 2L
15
10 Creep Region
y = (0.607E-3)x + 1.48
5
SIP = 8500 P12.5 = 8900
0
0 5000 10000 15000 20000
Pass
(a) Example calculation of SIP and P12.5
20000
9.5 mm
12.5 mm (Doyle, 2011)
15000 Equality Line
Linear (All data)
P12.5 (pass)
P12.5 = 1.01SIP
10000 R = 0.94
5000
0
0 5000 10000 15000 20000
SIP (pass)
(b) Relationship of SIP to P12.5
Figure 11. Analysis of PW-wet data and comparison to TSR.
Cohesion ratio (Eq. 1) is the ratio of PW-wet to PW-dry rut depth (RDPW) at 5,000
passes and was used to evaluate the cohesive (or mastic) response of the specimen to
moisture. PW-dry data were shifted to the same air voids of PW-wet data in the manner
already discussed. A higher cohesion ratio is considered indicative of greater rutting in
the presence of moisture without loss of adhesion. Some in-service pavements
312
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

experience noticeable moisture damage without stripping or adhesive failure (Little and
Jones, 2003). Cohesion of the mastic could be especially important in high RAP content
mixtures since the interaction of aged and virgin binders would affect the cohesion ratio
prior to stripping. RAP particles often have a very stiff and tightly bonded coating that
could protect against stripping. Nevertheless, inadequate partial blending of aged and
virgin binders could lead to cohesive problems in some mixtures.
PW-wet RD5 ,000 pass

Cohesion Rati=
o [1]
PW-dry RD5 ,000 pass
The higher the cohesion ratio, the more the specimen is affected by moisture on a
relative scale. However, the overall increase in rut depth due to moisture is also relevant
since a specimen could have a very high cohesion ratio (e.g., 5) from a largely
meaningless change in rut depths (e.g., 0.1 mm dry to 0.5 mm wet). These two values
should be interpreted together. One criticism of the TSR that is often made is that only
the wet to dry strength ratio is presented absent the dry or wet tensile strength. An
example would be a dry tensile strength of 350 kPa and a wet tensile strength of 327
kPa; the TSR is 93% but the material probably will not perform well with such low
values of wet and dry tensile strength.
SIP has already been defined and was considered indicative of loss of adhesion for a
mixture since that is when loose aggregate begins to appear with the binder coating
removed. A lower SIP is indicative of poorer adhesion in the presence of moisture.
Based on HLWT testing at 50C, Aschenbrener (1993) found that SIP values less than
3,000 were indicative of mixtures with poor field performance and SIPs greater than
10,000 were indicative of good field performance. Based on the work of Nielson (2010)
and observations made during comparison of loaded wheel testers, it is expected that
SIPs might be more likely to occur with PW-wet testing at 64C than for HLWT testing
at 50C.
Finally, P12.5 ratio (Eq. 2) is the ratio of PW-wet and PW-dry P12.5 values. It is
considered indicative of overall mixture performance in the presence of moisture. The
key difference in P12.5 ratio and TSR measures of moisture damage potential is that the
P12.5 ratio incorporates the effects of traffic in combination with moisture to evaluate the
potential susceptibility of a mixture to moisture damage. Numerous studies in the
literature have indicated that the effects of traffic are an important factor in moisture
susceptibility of pavements (Lottman, 1982; Hicks, 1991; Little and Jones, 2003; Lu,
2005).
313
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
PW-wet P12.5
P12.5 Ratio (%=) 100 [2]
PW-dry P12.5
5.4. Performance of High RAP Content WMA

Based on the analysis techniques previously discussed, Figure 12 presents rutting
performance results of all high RAP-WMA surface mixtures tested. Values for CM1
were not available for PW-dry since the mixture failed so quickly in PURWheel testing.
In terms of rutting, performance of the high RAP content WMA mixtures was affected
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
by RAP source as well as RAP content. Based on all the data, mixtures made with the
R1 RAP source (i.e., RM1 and RM3) generally had better performance than all the
controls. On the other hand, performance of mixtures made with R2 RAP (i.e., RM2 and
RM4) was better than CM1 but similar to or not quite as good as CM2 or CM3. For R1
RAP mixes, 50% RAP had less rutting than 25% RAP, as expected; for R2 RAP, PW-
dry testing predicted better performance in 50% RAP, but APA testing did not offer the
same prediction. With the one exception, data from APA and PW-dry testing provided
the same rankings and assessments for mixture performance; PW-dry data provided the
more logical ranking of mixtures in the excepted case. Overall, rutting performance of
the high RAP content WMA was similar to or better than current practice mixtures,
results in line with recent studies (e.g., Hurley and Prowell, 2006; Abbas and Ali, 2011).
Based on the observations and analysis techniques previously discussed, Figure 13
presents moisture damage performance results of all the high RAP content WMA
surface mixtures tested. Cohesion ratio and SIP values for CM1 were not available since
the mixture failed so quickly in PURWheel testing. The cohesion ratio (Figure 13a) was
less than 1.5 for the control mixtures and rut depths increased no more than 1.3 mm in
the presence of moisture. Performance of mixtures RM1, RM3 and RM4 was similar to
the controls with cohesion ratios on the order of 1.5 or less and rut depth increases of 1.5
mm or less. On the other hand, cohesion ratio of mixture RM2 was 2.0 and rutting
increased by 5.5 mm due to presence of moisture.
Figure 13b provides average SIP values for each mixture. SIP values for the control
mixture were well above the 10,000 pass criterion of Aschenbrener (1993). Mixtures
RM1, RM3 and RM4 have similar or slightly lower SIP values than control mixtures,
but still above the 10,000 pass criterion. RM2 has a lower result than the other mixtures
of 8,000 passes, indicating potential for increased susceptibility to moisture damage
relative to the other mixtures.
314

1.0
Rut Rate (mm/1000 Cycles)

APA Data Shifted to Va = 8% on T331 Basis
0.8 0.73
PW-Dry
0.6 0.57
0.48 0.48
Not Available
0.4 0.36
0.28 0.26 0.26
0.22 0.23
0.2 0.16 0.16
0.11
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0.0
CM1 CM2 CM3 RM1 RM2 RM3 RM4
Control Mixtures 25% RAP 50% RAP
(a) Rutting rate
20
APA Data Shifted to Va = 8% on T331 Basis
16
PW-Dry
Rut Depth (mm)
11.9
12
9.4
Not Available
8 6.7 7.2
4.7 4.4 4.7 4.7
3.5 3.7 3.7
4 2.2 2.4
0
(b) Rutting depth at 16,000 passes
Figure 12. Rutting performance of high RAP content WMA.
Figure 13c presents P12.5 ratio and TSR values for all mixtures. All of the TSR values
range from 93 to 98%, indicating that the mixtures will likely have similarly good
resistance to moisture damage. On the other hand, P12.5 ratios indicate a wider range of
potential performance. Values for controls CM1 to CM3 were 118, 71, and 100%,
respectively; however, the 118% for CM1 is misleading since rapid failure of the mix in
shear did not provide time for any moisture damage to develop and the overall
performance of that mixture in PURWheel testing was poor (i.e., the ratio of 500 to 425
passes is 118% but the mix failed in about 500 passes relative to the 20,000 passes
desired). For the high RAP mixtures, P12.5 ratios were 67, 68, and 61% for RM1, RM3
315

DOYLE, HOW ARD
6 5.5
Cohesion Ratio
5 Rut Depth Increase
Cohesion Ratio 4
Not Available
3
2.0
2 1.4 1.3 1.3
1.6 1.5
1.2
0.9
1 0.7
- 0.4 - 0.3
0
(a) Cohesion ratio

20
20.0
SIP (Thousand Passes)
15 14.0 13.5 12.9

11.5
Not Available
10 8.0
0
(b) SIP
120 118 TSR P12.5 Ratio

95 100 97 97 98
P12.5 Ratio or TSR (% )
100 94 95 93
80 71 67 68
61
60
43
40
20
0
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

(c) P12.5 ratio and TSR
Figure 13. Moisture damage performance of high RAP content WMA.
316

and RM4, respectively, indicating moisture damage performance similar too or only
slightly less than that of control mixtures. Overall, moisture damage data indicates that
three of the four high RAP content WMA mixtures would have acceptable performance
relative to current practice while the fourth may be slightly more susceptible to moisture.
Based on all the PURWheel data, mixture RM2 may have the potential for increased
moisture susceptibility compared to the control and other high RAP mixtures. This is not
evident from TSR test data, providing evidence that the current PURWheel equipment
and test parameters can provide a more discerning assessment of mixture performance.
Additional research is needed to quantify what PURWheel test values with the enhanced
protocols best indicate good field performance.
The enhanced version of the PURWheel test was used to assess rutting and moisture
damage performance of seven mixtures with up to 50% warm mixed RAP content.
Based on the study results the following conclusions and recommendations are made:
PW-dry and APA were equally good indicators of rutting potential. APA rutting
performance was not affected by compaction method. Using APA dry rut data (since
it is readily available) in conjunction with PW-wet data could be an efficient
combination.
Rutting performance of the high RAP content WMA was similar to or better than
current practice control mixtures. RAP source influenced rutting performance.
PW-wet testing showed a better ability to discern moisture damage performance
potential than the TSR test and also appeared to provide more useful moisture
damage information in terms of cohesion and adhesion.
The current PURWheel equipment and test parameters are capable of providing
moisture damage assessments while including the effects of traffic.
Moisture damage performance of high RAP content WMA appeared to be similar to
or not greatly worse than current practice control mixtures.
Additional research is needed to relate PURWheel parameters to field performance
and for a wide variety of mixture types.
7. Acknowledgements
This research was funded by the Mississippi DOT under State Study 212. Scott
Glusenkamp of APAC Mississippi is thanked for assistance in location of field mixtures.
Dr. Thomas D. White is thanked for assistance in a variety of manners including
317
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
securing the original PURWheel equipment, and for technical guidance related to wheel
tracking and finite element modeling. Permission to publish was granted by the Director,
Geotechnical and Structures Laboratory, U.S. Army Engineer Research and
Development Center.
8. References
Abbas, A. R., and Ali, A., Mechanical Properties of Warm Mix Asphalt Prepared Using
Foamed Asphalt Binders, Report No. FHWA/OH-2011/6, Ohio Department of
Transportation, 2011.
Azari, H., Precision Estimates of AASHTO T 283: Resistance of Compacted Hot Mix
Asphalt (HMA) to Moisture-Induced Damage, NCHRP Web-Only Document 166,
Transportation Research Board, 2010.
Aschenbrener, T., and Currier, G., Influence of Testing Variables on the Results From
the Hamburg Wheel-Tracking Device, Report No. CDOT-DTD-R-93-22, Colorado
Department of Transportation, 1993.
Aschenbrener, T., Evaluation of Hamburg Wheel-Tracking Device to Predict Moisture
Damage in Hot-Mix Asphalt, Transportation Research Record: Journal of the
Transportation Research Board, No. 1492, 1995, pp. 193-201.
Austerman, A. J., Mogawer, W. S., and Bonaquist, R., Evaluating the Effects of Warm
Mix Asphalt Technology Additive Dosages on the Workability and Durability of
Asphalt Mixtures Containing Recycled Asphalt Pavement, Proceedings of
Washington DC, 1115 January 2009.
Boggs, W., Customers Warm up to Green System, Warm-Mix Asphalt: Contractors
Experiences, Information Series 134, National Asphalt Pavement Association, 2008,
pp. 1617.
Collins, R., Shami, H., and Lai, J.S., Use of Georgia Loaded Wheel Tester to Evaluate
Rutting of Asphalt Samples Prepared by Superpave Gyratory Compactor,
No. 1545, 1996, pp. 161168.
Copeland, A., Reclaimed Asphalt Pavement in Asphalt Mixtures: State-of-the-
Practice, Report No. FHWA-HRT-11-021, Federal Highway Administration, 2011.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Copeland, A., DAngelo, J., Dongr, R., Belagutti, S., and Sholar, G., Field Evaluation
of High Reclaimed Asphalt Pavement-Warm Mix Asphalt Project in Florida: Case
Study, Transportation Research Record: Journal of the Transportation Research
Board, No. 2179, 2010, pp. 93101.
318

Diefenderfer, S. D., and Hearon, A. J., Performance of Virginias Warm-Mix Asphalt

Trial Sections, Report No. FHWA/VTRC 10-R17, Virginia Transportation
Research Council, 2010.
Doyle, J. D., Characterization of Reclaimed Asphalt and Performance Based Evaluation
of its use in Recycled Mixtures, Ph.D. Dissertation, Mississippi State University,
2011.
Doyle, J. D., and Howard, I. L., Linear Asphalt Compactor Operators Manual,
Manual No. CMRC M 10-1 Version 2, Construction Materials Research Center,
Mississippi State University, 2010.
Doyle, J. D., Howard, I. L., and Robinson, W. J., Prediction of Absorbed, Inert, and
Effective Bituminous Quantities in Reclaimed Asphalt Pavement, Journal of
Materials in Civil Engineering, Vol. 24, No. 1, 2012, pp. 102112.
Doyle, J. D., Mejas-Santiago, M., Brown, E. R., and Howard, I.L., Performance of
High RAP-WMA Surface Mixtures, Asphalt Paving Technology: Journal of the
Association of Asphalt Paving Technologists, Vol. 80, 2011, pp. 419457.
Epps, J. A., Sebaaly, P. E., Penaranda, J., Maher, M. R., McCann, M. B., and Hand, A.
J., Compatibility of a Test for Moisture-Induced Damage with Superpave
Volumetric Mix Design, NCHRP Report 444, Transportation Research Board,
2000.
Federal Highway Administration (FHWA), Performance of Coarse-Graded Mixes at
WesTrack-Premature Rutting, Report No: FHWA-RD-99-134, Federal Highway
Administration, 1998.
Federal Highway Administration (FHWA), Hamburg Wheel Tracking Device
Equipment Bituminous Mixtures Laboratory (BML) Labs Research Asphalt
Pavements - FHWA, Federal Highway Administration,
http://www.fhwa.dot.gov/pavement/asphalt/labs/mixtures/hamburg.cfm, Accessed
23 July 2012.
Hicks, R. G., Moisture Damage in Asphalt Concrete, NCHRP Synthesis of Highway
Practice No. 175, 1991.
Hodo, W. D., Kvasnak, A., and Brown, E. R., Investigation of Foamed Asphalt (Warm
Mix Asphalt) with High Reclaimed Asphalt Pavement (RAP) Content for
Sustainment and Rehabilitation of Asphalt Pavement, Proceedings of
Washington, DC, 11-15 January 2009.
Howard, I. L., and Doyle, J. D., Comparison of Available Density Measurement
Methods Applicable to Design and Construction, Proceedings of Transportation
319
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
Research Board 91st Annual Meeting, Paper No. 12-0564, Washington, DC, 22-26
January 2012.
Howard, I. L., Doyle, J. D., and Barham, J. M., Uniformity, Repeatability, and
Permanent Deformation Resistance of Slabs Produced with the Linear Asphalt
Compactor, Advances in Civil Engineering Materials, Vol. 1, No. 1, 2012, pp. 1
17.
Howard, I. L., Doyle, J. D., White, T. D., Ivy, J., and Booth, O., PURWheel Laboratory
Wheel Tracker Operators Manual, Manual No. CMRC M 10-2 Version 1,
Construction Materials Research Center, Mississippi State University, 2010.
Hurley, G. C., and Prowell, B. D., Evaluation of Sasobit for use in Warm Mix
Asphalt, NCAT Report No. 05-06, National Center for Asphalt Technology, 2005.
Hurley, G. C., and Prowell, B. D., Evaluation of Evotherm for use in Warm Mix
Asphalt, NCAT Report No. 06-02, National Center for Asphalt Technology, 2006.
Kandhal, P. S., and Cooley, Jr., L.A., Accelerated Laboratory Rutting Tests: Evaluation
of the Asphalt Pavement Analyzer, NCHRP Report 508, Transportation Research
Board, 2003.
Kanitpong, K., and Bahia, H., Evaluation of HMA Moisture Damage in Wisconsin as it
Relates to Pavement Performance, International Journal of Pavement Engineering,
Vol. 9, No. 1, 2008, pp. 917.
Kiggundu, B. N., and Roberts, F. L., Stripping in HMA Mixture: State-of-the-art and
Critical Review of the Test Methods, NCAT Report No. 88-02, National Center for
Asphalt Technology, 1988.
Kvasnak, A., West, R., Moore, J., Nelson, J., Turner, P., and Tran, N., Case Study of
Warm Mix Asphalt Moisture Susceptibility Birmingham, Proceedings of
Washington, DC, 11-15 January 2009.
Little, D., and Jones, D. R., Chemical and Mechanical Processes of Moisture Damage
in Hot-Mix-Asphalt Pavements, Moisture Sensitivity of Asphalt Pavements: A
National Seminar. Transportation Research Board, 2003.
Lottman, R. P., Predicting Moisture-Induced Damage to Asphaltic Concrete, NCHRP
Report No. 192, Transportation Research Board, 1978.
Lottman, R. P., Predicting Moisture-Induced Damage to Asphaltic Concrete Field
Evaluation, NCHRP Report No. 246, Transportation Research Board, 1982.
Lu, Q., Investigation of Conditions for Moisture Damage in Asphalt Concrete and
Appropriate Laboratory Test Methods, Ph.D. Dissertation, University of California
at Berkley, 2005.
320
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Mallick, R. B., Kandhal, P. S., and Bradbury, R. L., Using Warm Mix Asphalt
Technology to Incorporate High Percentage of Reclaimed Asphalt Pavement
Material in Asphalt Mixes, Transportation Research Record: Journal of the
Mejas-Santiago, M., Doyle, J. D., Howard, I. L., and Brown, E. R., Moisture Damage
Potential for Warm Mix Asphalt Containing Reclaimed Asphalt Pavement,
Proceedings of 2nd International Warm Mix Conference, St. Louis, MO, 1113
October 2011.
Middleton, B., and Forfylow, R. W., Evaluation of Warm-Mix Asphalt Produced with
the Double Barrel Green Process, Transportation Research Record: Journal of the
Mogawer, W. S., Austerman, A. J., Bonaquist, R., and Roussel, M., Performance
Characteristics of Thin Lift Overlay Mixtures Containing High RAP Content, RAS,
and Warm Mix Asphalt Technology, Proceedings of Transportation Research
Board 90th Annual Meeting, Paper No. 11-1859, Washington DC, 2327 January
2 0 1 1 a.
Mogawer, W. S., Austerman, A. J., Kassem, E., and Masad, E., Moisture Damage
Characteristics of Warm Mix Asphalt Mixtures, Asphalt Paving Technology:
Journal of the Association of Asphalt Paving Technologists, Vol. 80, 2011b, pp.
491526.
Nielson, J., Development of a Testing Temperature to be used with the Hamburg Wheel
Tracking Device on Asphalt Mixtures that Utilize Performance Grade Binders, MS
Thesis, University of Utah, 2010.
Pan, C., and White, T., Conditions for Stripping Using Accelerated Testing, Report
FHWA/IN/JTRP-97/13, Joint Transportation Research Program, Purdue University,
1999.
Prowell, B. D., Hurley, G. C., and Crews, E., Field Performance of Warm-Mix Asphalt
at National Center for Asphalt Technology Test Track, Transportation Research
Record: Journal of the Transportation Research Board, No. 1998, 2007, pp. 96
102.
Rand, D. A., Hamburg Wheel Test, Technical Advisory Dated August 16, 2006, Texas
Department of Transportation Construction and Bridge Division, 2006.
Richardson, C., The Modern Asphalt Pavement, 2nd Ed., New York, John Wiley &
Sons, 1912.
Stiady, J. L., Hand, A. J. T., Noureldin, A. S., Galal, K., Hua, J., and White, T. D.,
Validation of SHRP Asphalt Mixture Specifications Using Accelerated Testing,
Final Report No. 176, Indiana Department of Transportation, 2003.
321
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DOYLE, HOW ARD
Terrel, R. L., and Al-Swailmi, S. Role of Pessium Voids Concept in Understanding

Moisture Damage to Asphalt Concrete Mixtures, Transportation Research Record:
Journal of the Transportation Research Board, No. 1386, 1993, pp. 3137.
Mixtures, NCHRP Report No. 274, Transportation Research Board, 1984.
Mixtures: Field Evaluation, NCHRP Report No. 373, Transportation Research
Board, 1995.
Willis, J. R., West, R., Nelson, J., Taylor, A., and Leatherman, K., Combining Warm
Mix Asphalt Technologies with Mixtures Containing Reclaimed Asphalt
Pavement, Proceedings of 2nd International Warm Mix Conference, St. Louis,
MO, 1113 October 2011.
Zaniewski, J., and Viswanathan, A. G., Investigation of Moisture Sensitivity of Hot
Mix Asphalt Concrete, Report Submitted to West Virginia Division of Highways,
2006.
9. Discussion
DR. JOHN DANGELO: Very nice presentation. One thing that you might want to look
at when evaluating loaded wheel testers, especially when comparing the APA and PUR
wheel and maybe the Hamburg, is some work weve done a few years ago. The stress
states in the contact areas in those devices are all very, very different.
DR. JESSE DOYLE: Yes, they are.
DR. DANGELO: Because of the different stress states, theres not a linear stress strain
relationship for the results specifically, when you start looking at modified systems.
When you are using neat asphalts, youre basically dealing, at high temperature, with
Newtonian materials that give you a relatively good linear relationship. But when you
get into modified systems, there are actually yield stresses in the polymers. Depending
on which device you use, you can get to those yield stresses very, very easily, and then
youll see very different results that dont necessarily match up. I did some of that in an
AAPT paper quite a few years ago showing some of that. But if you continue on with
this with these evaluations, its very important to look at those different stress states of
the equipment, because theyll significantly affect the rutting.
DR. DOYLE: Thank you very much for your comment.
PROFESSOR CHRISTOPHER WILLIAMS: I can answer Johns question, actually. We
did a lot of other fundamental material tests on mixes and modeled in the PURWheel,
and so we had pretty good agreement with our models there that we did in the 1990s. At
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,
322

any rate, separate from that, I spent a lot of time with the PURWheel, as you might
imagine, and I was wondering after you ran the PURWheel test, did you cut the
specimens in half and try to look at a cross-section?
DR. DOYLE: Thank you very much for your comment. Yes, we did cut the specimens
in half, though there is not any analysis of that data in this paper. We may do some work
looking at image analysis or something in the future. Its not in this paper, but we did cut
them in half and tried to make some assessments there.
PROF. WILLIAMS: Because, you know, when you cut them in half, it actually shows
how you get the same kind of shear planes, densification bulb, and things that we see in
the field mixes in the field, in fact, which goes to answering Johns question. But it also
is kind of telling on the stripping that you see there.
DR. DOYLE: Right. Thank you very much for your comment.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
PROFESSOR IMAD AL-QADI: Thanks for the presentation. I have two quick
questions. The first one, you concluded in your study that the APA test would give a
good indicator or better indicator than IDT. In the last slide of your presentation, I
noticed that IDT test was much more consistent while the variability is very high in the
control. I meant, it went from 74% to 118%, while the IDT results in all tests were kind
of consistent. Any comment on that?
DR. DOYLE: I think that in terms of the TSR and the IDT testing, were not seeing any
effects of traffic and perhaps that the effects of simulated traffic in the APT test that is
starting to differentiate the mixtures a little more so than the tensile test.
PROF. AL-QADI: Okay. My other question is related to the inflection point you used as
a moisture damage indicator. In my opinion, when a material reaches some kind of a
shear flow, regardless if moisture damage exists or not, its going to have an inflection
point. Why would you associate that with moisture damage?
DR. DOYLE: I didnt show any of dry test example data, but for that data we didnt see
any shear flow inflection point where the mixture transitions from secondary flow to a
tertiary flow region in our dry data, while we did generally see those inflection points in
the wet test data. So we took that shear flow inflection point as being somewhat
indicative of a loss of internal cohesion.
PROF. AL-QADI: Yeah, but such moisture damage doesnt occur in a short period of
time. Does that make sense? Just a comment.
DR. DOYLE: Well, the test temperature and applied stress conditions are the same in the
dry and wet tests. The only difference is the presence of water. Thank you very much for
your comment
PROF. AL-QADI: Okay. Thanks.
323

DOYLE, HOW ARD
DR. DOYLE: Thank you.

DR. ISAAC HOWARD: One point to Dr. Al-Qadis comment, keep in mind, we run the
exact same test wet and dry. For example, the shear flow. If it does not occur dry and it
does occur wet, theres only one difference in the combination, and so attributing those
behaviors to moisture, at least to me, makes sense. Now if we ran the dry test at one
combination of conditions like in the APA and then ran the wet test in another
combination of conditions, like in the original PURWheel, its hard to tie those two
together. But what were doing now is were running a test at 64C in air, running the
same test on the same type of specimen at 64C water. If you dont see the flow
conditions dry but you see them wet, then I think its reasonable to assume that its the
moisture that caused those conditions. Now if we had very different conditions in the dry
versus wet, it would be much more difficult to say that.
With regard to the control mixes, those are a variety of different products, and we did
that intentionally. The CM1 that was used did not perform well in wheel tracking. CM1
versus CM2 versus CM3, those were different mixes, and we were trying to see a range
of behaviors. The key point is the TSR said that mix was fine. And that mix was not fine
in the context of wheel tracking. The other item that I think is important to remember
and Dr. Doyle didnt have a chance to cover it in his presentation, but one of the items
that we would like to do moving forward is look at using the APA as your dry rutting
indicator and then scale that to what the PURWheel estimate would be, run the
PURWheel in a wet condition only and use that difference as your moisture indicator to
cut down on some of the testing time. And then to your question, Dr. DAngelo, there is
an assessment in the paper of those different loads and contact conditions. The Hamburg
is noticeably different than the other two, but the other two are at least somewhat
scalable and do not have steel in contact with the mix. So, I guess more comments than
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
anything.
PROFESSOR ESHAN DAVE: Really good work. I enjoyed your presentation. When
you did your TSR, do Mississippi specs require you to freeze your sample, or is it the
modified version without the freezing part?
DR. DOYLE: It is the TSR version without the freeze-thaw conditioning cycle. Thats
what is required by Mississippi specifications, so we followed that, since it was a state
sponsored study. We did not look at any additional TSRs using the freeze-thaw
conditioning. That would be something to consider, though, but we were not able to do
that in this study.
PROF. DAVE: We recently finished a study, obviously, I am in Minnesota, and we are
looking at more of colder climate conditions. What we learned is that freezing and
thawing has really detrimental effects on the cracking performance of mixes. So I think
that it would be good to qualify in the paper that these moisture damage results are more
324

in context of rutting and warmer climates. Depending on where you are in the country,
your mechanisms might be very different and so will be the corresponding effects.
DR. DOYLE: Thank you very much for your comment. Yes, our intention was certainly
to consider pavements in the southeastern United States where freeze-thaw conditions
are not common. So we were sort of regionalizing our interests, yes. Thank you.
PROF. DAVE. Thanks.
325
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Effect of Long-Term Aging on RAP Mixtures:
Laboratory Evaluation of Plant Produced
Mixtures
Jo Sias Daniel*a, Nelson Gibsonb, Sean Tarboxc, Audrey Copelandd,
and Adrian Andriescue
a
Department of Civil Engineering, University of New Hampshire, Durham, NH 03824
b
Federal Highway Administration, FHWA Office of Infrastructure R&D, McLean,
Virginia 22101
c
Woodard & Curran, Portland, ME 04102
d
National Asphalt Pavement Association, Lanham, MD 20706
e
ESC, Inc., TFHRC, McLean, VA 22101
ABSTRACT: As the use of Reclaimed Asphalt Pavement in asphalt concrete mixtures increases, it is
important to understand how the addition of already aged asphalt binder affects the overall
properties and performance of the mixture. In this study, four plant produced mixtures containing
0%, 20%, 30 %, and 40% RAP were long term oven aged in the laboratory to three levels. Mixture
testing included uniaxial complex modulus and fatigue. Recovered binder testing included PG
grading and rheological characterization. The mixture testing showed that the RAP mixtures
stiffen due to laboratory aging at a slower rate than virgin mixtures and that the impact of the
presence of RAP on material properties decreases with aging time. The relative fatigue
performance of the mixtures changes dramatically in stress versus strain control, indicating the
importance of linking mixture and pavement design. The recovered binder properties show
increase in the high continuous PG grade, minimal impact at the intermediate continuous PG
grade and a slight increase in the low continuous PG grade with aging time. The complex shear
modulus shows an increase in stiffness and a decrease in phase angle with aging. The binder and
mixture properties generally show the same qualitative trends with respect to RAP content and
aging level; quantitative comparisons of several parameters show some correlation, indicating
that binder properties may be used to evaluate expected mixture properties measured in the lab.
KEYWORDS: RAP, recycled pavement, aging, fatigue, viscoelasticity, binder, asphalt mixtures
The oral presentation was made by Professor Daniel.
http://dx.doi.org/ 10.1080/14680629.2013.812840
327
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DANIEL, GIBSON, TARBOX, COPELAND, ANDRIESCU
1. Introduction
Asphalt concrete mixtures undergo aging during their in-place service lives. The
aging process stiffens the asphalt, changing its mechanical properties and the resulting
performance under traffic loading. The major factor contributing to the increase in
stiffness of asphalt concrete mixtures over time is the oxidation of the asphalt binder at
the molecular level. Glover et al. (2009) recently performed a thorough literature review
on binder oxidation in pavements and its impact on pavement performance. Hardening
of asphalt binders in pavements due to oxidation has been measured over time by testing
recovered binder (Woo et al., 2007, Woo et al., 2008). Research has indicated that binder
oxidation can have a significant negative impact on fatigue performance of asphalt
mixtures measured using laboratory fatigue tests (Walubita et al., 2005, Walubita et al.,
2006). Researchers have also performed tests, such as the indirect tensile (IDT) (Said,
2005), shear modulus (Al-Khateeb, et al., 2005), and beam fatigue (Raad, et al., 1999)
testing, on specimens from field pavements showing stiffness increases with aging.
Long term oven aging to simulate aging in the field has been used to evaluate stiffness
changes over time (Daniel et al., 1998, Harvey and Tsai, 1997, Morian et al., 2011). The
impact of the measured increases in stiffness on the fatigue performance of the pavement
has been shown to be influenced by the binder type (Harvey and Tsai, 1997), as well as
the pavement structure (Raad, et al., 1999, Molenaar et al., 2010). Aging has also been
shown to reduce the stress relaxation capacity of the binder (Molenaar et al., 2010)
which is part of what causes decreases in damage and cracking resistance.
The use of Reclaimed Asphalt Pavement (RAP) in Hot Mix Asphalt (HMA) has
become regular practice for much of the country. RAP effectively cuts the cost of HMA
production as a source of both quality aggregate and asphalt. As the cost of virgin
materials continues to rise, much of the paving industry has begun to look more
seriously at increasing the use of this material. RAP is generated when the materials
from pavements in service are removed or milled and processed. The asphalt binder in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
the reclaimed asphalt concrete has oxidized to some extent, so the inclusion of RAP in a
new mixture introduces the already aged binder into the mixture.
The objective of this paper is to evaluate how the properties of RAP mixtures change
with laboratory oven aging compared to a virgin mixture. This builds on the work in
Tarbox and Daniel (2012) in which the measured changes in dynamic modulus of four
plant produced mixtures containing 0%, 20%, 30%, and 40% RAP subjected to long
term oven aging in the laboratory were evaluated. The current study evaluates the same
mixtures and includes characterization of recovered binders from the mixtures, refined
analysis with the Global Aging System (GAS) model using measured binder properties,
additional rheological analysis of the mixture master curves and fatigue analysis.
328

Effect of Long-Term Aging on RAP Mixtures
2. Materials and Handling
2.1 Mixtures
Plant produced mixtures were obtained from Pike Industries, Inc. in Portsmouth, NH.
The mixtures were sampled at the time of production (October 2010) at the asphalt plant
and stored in air tight five gallon metal containers until specimens were fabricated in the
laboratory. RAP mixtures containing 20%, 30%, and 40% RAP and a virgin mix were
obtained from a drum plant. All mixtures used a PG 64-28 base binder. The mixtures
were New Hampshire Department of Transportation (NHDOT) Type E 12.5mm
Superpave surface course mixtures. The design was for less than 0.3 million ESALs
using a gyratory compactive effort of Ndes=50. Each mix had the same target gradation to
eliminate variables between mixes with different RAP contents. The gradations and
asphalt contents for the various mixtures are shown in Table 1. The same aggregate
gradation was targeted for all mixtures; the 30% and 40% RAP mixture have gradations
that are slightly coarser in the #30-#100 size range. The RAP source for all mixtures had
an asphalt content of 4.79%. The RAP binder ratio (% RAP binder/ % total binder) for
each mixture is shown at the bottom of Table 1. Because all mixtures were fabricated
with the same RAP source, the % RAP value will be used to identify the different
mixtures in this paper. Table 2 summarizes the production and mix design parameters
for each mixture.
2.2 Specimen Fabrication and Conditioning
The plant produced mixtures were brought back to the laboratory and reheated
following established protocols. The process involves reheating five gallon steel buckets
of loose mix to ten degrees less than the plant discharge temperature for two hours. Once
reheated, the mix was split into appropriate specimen sizes. It was then heated to the
recommended compaction temperature and compacted to approximately 7% air voids
using a Superpave Gyratory Compactor. Once cooled overnight, compacted specimens
were cut and cored to obtain final test specimen dimensions (100 mm diameter by 150
mm tall), and a target air void content of 6%.
Compacted and cut test specimens were oven aged to three levels in the laboratory
following existing Strategic Highway Research Program (SHRP) recommendations for
laboratory aging (Bell et al., 1994, Bell and Sosnovske, 1994). The recommended
durations for aging included two, four, and eight days at 85C. These levels correspond
to approximately four, eight, and sixteen years of service life. Specimens were placed
sideways in a metal cradle and flipped once a day to achieve uniform aging. Once aged,
four mounting points for the LVDTs were mounted using Devcon 10240 five minute
epoxy. The gauge length was set to 70mm along the vertical axis of the specimen using
an alignment jig.
329
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 1. Mixture gradations, asphalt contents.
% Passing
Sieve Size 40%
Virgin 2 0 % RAP 3 0 % RAP Target
RAP
19.0 mm (3/4") 100 100 100 100 100
12.5 mm (1/2") 9 9 .1 9 9 .2 9 7 .7 9 9 .6 9 8 .7
9.5 mm (3.8") 8 9 .6 8 9 .1 8 6 .8 8 7 .7 8 6 .0
4.75 mm (#4) 6 2 .0 6 1 .4 5 8 .9 5 5 .2 5 7 .5
2.36 mm (#8) 4 2 .3 4 3 .4 4 1 .7 3 8 .8 4 2 .9
1.18 mm (#16) 3 1 .7 3 2 .8 3 1 .7 2 9 .9 3 3 .4
0.6 mm (#30) 2 4 .0 2 5 .8 2 2 .8 2 2 .2 2 4 .6
0.3 mm (#50) 1 5 .5 1 5 .3 1 3 .7 1 4 .0 1 4 .5
0.15 mm (#100) 6 .9 7 .3 5 .8 6 .5 6 .8
0.075 mm (#200) 3 .0 3 .7 2 .6 3 .2 3 .6
Total % AC 5 .7 5 .7 5 .7 5 .7 5 .7
% Virgin AC 5 .7 4 .7 4 4 .2 6 3 .7 8 -
RAP binder ratio 0% 1 6 .8 % 2 5 .2 % 3 3 .7 % -
Table 2. Production and mix design parameters.

Parameter Virgin 2 0 % RAP 3 0 % RAP 4 0 % RAP
Discharge Temp
166 158 168 168
(oC)
Compaction Temp
149 154 157 157
(oC)
VMA (%) 1 4 .9 1 4 .5 1 4 .4 1 4 .5
VFA (%) 7 4 .8 7 9 .9 8 1 .3 8 2 .1
2.3 Extracted Binders

Asphalt binder was extracted and recovered from previously tested compacted
specimens. The specimens were heated to 150C, broken apart into loose mixture and
then immediately cooled. The cooled, loose mix was soaked in toluene, centrifuge
extracted and then recovered according to ASTM D5404 in a rotary evaporator using
vacuum and nitrogen at key steps in the recovery distillation process. The final rinse of
the aggregates used a mix of 85% toluene and 15% ethanol. Each specimen had a mass
330
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of about 2,700 grams that yielded approximately 120 grams of binder. Two or three
specimens for each RAP content and age condition were extracted and recovered. The
recovered binder was tested using Fourier Transform InfraRed (FTIR) spectroscopy to
determine whether there were any traces of toluene at the end of the recovery process.
To ensure the removal of air bubbles and possible traces of solvent, the recovered
asphalt was also kept in a vacuum oven at 165C for 45 minutes.
3. Testing and Analysis Methods
3.1 Mixture Testing and Analysis
3.1.1 Complex Modulus
Complex Modulus testing was performed with an Asphalt Mixture Performance

Tester (AMPT) servo hydraulic machine manufactured by IPC Global. Testing
conformed to the specifications put forth by AASHTO TP 62 Determining Dynamic
Modulus of Hot-Mix Asphalt Concrete Mixture. Three replicate specimens were tested
for each condition. These specimens were tested at 4.4C, 21.1C, 37.8C and 54.4C at
25, 10, 5, 1, 0.5 and 0.1 Hz. Specimens were conditioned in a separate environmental
chamber prior to testing and a dummy specimen of the same dimensions was used to
assure testing specimens had reached the appropriate testing temperatures to within
0.1C.
Dynamic modulus testing was performed in unconfined uniaxial compression. The
average dynamic modulus isotherms were shifted to a generalized logistic function [1] to
construct the master curve at a reference temperature of 21.1C. The time-temperature
shift factors were allowed to free-shift, meaning no underlying shape of the shift factor
versus temperature curve was assumed.

log| | 1/ [1 ]
1++(log )
where: |E*| = Dynamic Modulus

r = reduced frequency
, , , , = fitting parameters
Similarly, phase angle master curves were constructed with a reference temperature
of 21.1C and fitted with the logistic function:
+()
90 1+1/ [2 ]
1++()
where: = Phase Angle
331
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The fitting parameters determined from the dynamic modulus master curve are used
to determine the frequency at which the inflection point of the master curve occurs. The
inflection point frequency is defined as follows:

10 [3 ]
The inflection point frequency corresponds to the peak of the phase angle master curve
and indicates where the material behavior transitions from being dominated by the
binder to the aggregate skeleton, essentially transitioning from the more viscous regime
to a more elastic regime. This is similar in concept to the crossover frequency in binders
where the phase angle is equal to 45o and the material behavior transitions from viscous
to elastic.
The shift factor curves determined from the dynamic modulus master curve
construction are fit using the Kaelble modified Williams Landel Ferry (WLF) form
described by Rowe and Sharrock (2011):

log 1 [4 ]
2 +| | 2 +| |
where: aT = shift factor

T = temperature
Tr = reference temperature
Tk = Kaelble defining temperature
C1, C2 = fitting parameters
The Tk value represents an inflection point in the shift factor curve and is similar to
that represented by Td in the work of Anderson et al. (1994) for asphalt binders. The Tk
for asphalt mixtures tends to be higher and when low temperature data is not available,
the inflection temperature is assigned to the lowest temperature isotherm (Rowe and
Sharrock, 2011). In this study, Tk was set to 4.4oC. The C2 coefficient describes the
slope of the log aT versus temperature curve and therefore is an indication of the
temperature susceptibility of the mixture. A higher C2 value occurs when the slope of
the shift factor curve is shallower. A shallower curve occurs when individual isotherms
are not shifted less along the reduced frequency axis, indicating a reduced temperature
susceptibility of the asphalt mixture.
3.1.2 Fatigue
Fatigue testing was performed in uniaxial tension on the AMPT using the UTS032
software. Due to material limitations, only the unaged and 8 day aging conditions were
evaluated in fatigue. Fatigue testing was conducted for two replicate specimens at two
strain amplitudes for each mixture. Analysis was done using the simplified viscoelastic
continuum damage (S-VECD) model developed by Underwood and Kim (2010). The
332
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

data from the UTS032 software was exported and used in conjunction with the
companion ALPHA Fatigue software. The ALPHA Fatigue software conducts S-VECD
analysis to estimate the number of cycles to failure for different temperatures, loading
frequencies, strain amplitudes, or stress amplitudes. The software also estimates strain
levels associated with the endurance limit for each mixture. From this information, the
relative performance of each mixture with respect to fatigue can be evaluated.
3.3.3 Global Aging System
One main component of aging in an asphalt mixture is the change in viscosity of the
asphalt binder as it oxidizes. The Global Aging System (GAS) (Mirz and Witczak, 1995)
is a method of predicting how aging will affect dynamic modulus according to the
increase in viscosity of the binder. The GAS accounts for the effect of depth and can
adjust for air voids of the pavement. Depth and air void adjustment were not considered
in this research because each specimen represents surface mixture and air voids were
kept constant. The GAS is currently used by the Mechanistic-Empirical Pavement
Design Guide (M-E PDG) software to predict the physical properties of asphalt
pavements over time.
The GAS model predicts the change in viscosity of the binder due to age and then
shifts the unaged dynamic modulus curve according to the change in viscosity of the
binder. The model uses the ASTM viscosity temperature relationship as follows:
loglog() log [5 ]
where: = viscosity (cP) maximum value of 2.7X1010 Poise
TR = temperature (Rankine)
A = regression intercept
VT S = regression slope
The A and VTS parameters were determined from binder shear modulus and phase
angle measured from extracted and recovered binders for each mixture and aging
condition using the following relationship:
| | 1 4.8628
[6 ]
10
where: = viscosity (Pa*sec)

|G*| = binder complex shear modulus (Pa)
= binder phase angle
The viscosity of aged binder was calculated with the following equation:
log log(=0 )+
log logaged [7 ]
1+
where: A = 0.004166+1.41213(C)+(C)log(Maat)+(D)loglog(t=0)
333
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

B = 0.197725+0.068384log(C)
C = 10(274.4946-193.831*log(Tr)+33.9366*log(Tr)^2
D = -14.5521+10.47662 log (TR) 1.88161 log(TR)2
aged = aged viscosity (cP)
t=0 = viscosity at mix/lay-down (cP)
Maat = mean annual air temp (F); 42 used in this study
TR = temperature (Rankine)
T = time in months
The unaged dynamic modulus master curve was then shifted using the following
equation to calculate reduced time of loading according to the aged viscosity:
log t r log() c(log() log ) [8 ]
where: t = time of loading at the given temperature of interest
tr = time of loading at the reference temperature
c = determined previously from unaged master curve
= aged viscosity at temperature of interest
tr = unaged viscosity at reference temperature
In this study, the predicted aged master curve using the GAS model is compared to
that measured using long-term oven aging in the laboratory.
3.2 Binder Testing
The recovered binders were characterized for rheological properties and ductile
failure properties. No further aging in the Rolling Thin Film Oven (RTFO) or Pressure
Aging Vessel (PAV) was conducted on the recovered RAP binders. Low, intermediate
and high temperature Performance Grade (PG) was determined following AASHTO T
313 and T 315. Complex shear moduli and phase angles were measured from 5oC to
35oC at 5oC intervals and then at 58oC, 70oC, 82oC and 88oC. At each temperature, the
data were measured at 33 frequencies between 0.1 radians per second and 100 radians
per second evenly spaced in logarithmic scale. Modulus and phase angle master curves
were also utilized to determine the parameter G/(/G) at 44.7oC and 10 radians per
second. This parameter provides a measure of age-induced loss of healing and
embrittlement as proposed by Glover et al. (2005) and Anderson et al. (2011). Ductile
failure properties were characterized at 25oC and 100 mm/minute in a force-ductility
apparatus with Double-Edge-Notched Tension (DENT) specimens based on the LS-299
test method from the Ministry of Transport Ontario. A calculated Critical Tip Opening
Displacement (CTOD) parameter is determined, which is proposed to provide a measure
of ductile strain tolerance in the presence of stress concentrations (Gibson et al., 2012).
334
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Master curves were developed by shifting the complex shear modulus to an

alternative form of the CAM model [9]. The time-temperature superposition shift factors
were allowed to free-shift, meaning no underlying model was used to force the shape of
the log shift factor versus time relationship:

| |

[9 ]

1+

where: R is reduced frequency

me, k and C are fitting parameters
The resultant shifting for the modulus was applied to the measured phase angle and
compared, but not fit, to the alternative form of the CAM model for phase angle:
90
[1 0 ]

1+

Once the free-shifted time-temperature superposition shift factors were determined,

the Kaelbel modified WLF function [4] was fit to the free-shifted time-temperature
superposition shift factors rather than the WLF or Arrhenius model. The high and low
temperature PG was also determined from the fit master curves and compared to the
conventionally determined PG grades using semi-log interpolation.
4. Results and Discussion
4.1 Mixture
4.1.1 Complex Modulus
The dynamic modulus master curves at the different laboratory aging conditions for the
various RAP mixtures are shown in Figure 1. The curves represent a sigmoidal function
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
fit through the data averaged from three replicate specimens. The predictions from the
GAS model are presented in the figures (as dashed lines) and will be discussed in a
following section. Statistical analysis was performed to determine if the master curves
were significantly different. This was done by conducting one-tailed paired t-tests on
the raw data at all measurement points using a 95% confidence interval.
In general, the dynamic modulus increases with aging time for all of the mixtures.
The virgin mixture shows the largest increases with aging time and the 40% RAP
mixture shows the smallest increases. These differences are greatest in the intermediate
and high temperature ranges. The statistically significant differences are with the eight
day aging level in the low and intermediate temperature ranges.
335

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 1. Dynamic Modulus Master Curves and GAS Predictions for (a) 0% RAP, (b) 20% RAP, (c) 30% RAP, (d) 40%
RA P.
336

The phase angle master curves at the different laboratory aging conditions for the
various RAP mixtures are shown in Figure 2. The curves represent a sigmoidal function
fit through the data from three replicate specimens. There is no phase angle data
available for the 30% RAP mix at two days of long term oven aging due to testing
issues. For each RAP content, the phase angle master curve gets flatter and shifts to the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
lower frequencies with aging level. At reduced frequencies above approximately 0.01
Hz, aging causes a decrease in phase angle (more elastic behavior) and at reduced
frequencies below 0.01 Hz, aging causes a more viscous response. The phase angle
master curves at the 8 day aging level are significantly different than all of the other
aging levels.
When the phase angle curves are compared for each level of aging, the peak phase
angle is approximately the same magnitude for the four different RAP contents (the
curves have the same shape and do not get flatter). However, the curves shift to the
lower frequencies with increasing RAP content. At frequencies above the peak phase
angle, the addition of RAP decreases the phase angle and at frequencies below the peak
phase angle, the addition of RAP increases the phase angle.
The blackspace diagrams for the various RAP levels are shown in Figure 3. For
each RAP content, the blackspace diagrams for the various aging levels converge at
higher stiffness values. At lower stiffness values, the higher aging levels have lower
phase angles. The lower phase angles indicate a reduced capacity for relaxation,
meaning that the aged materials will be more susceptible to cracking in these stiffness
ranges. However, the materials will have a higher resistance to rutting. From a visual
observation of Figure 3, the differences in phase angle at lower stiffnesses are larger for
the 0% RAP mix than the 40 % RAP mix. The 20% RAP and 30% RAP mixtures
appear to be similar to the 0 % RAP mixture.
Comparing the various RAP contents at each aging level, similar trends were
observed. The differences in phase angle at lower stiffnesses appear to be greatest for
the unaged condition and lowest for the 8 day aging condition. This indicates that the
properties of the RAP mixtures become more similar with aging time.
337

Figure 2. Phase angle master curves for (a) 0% RAP, (b) 20% RAP, (c) 30% RAP, (d) 40% RAP.
338
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 3. Blackspace diagrams for each RAP content (a) 0 % RAP (b) 20 % RAP (c) 30% RAP (d) 40% RAP.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`-
339

Figure 4a shows the inflection point frequency determined from the dynamic
modulus master curve sigmoidal fit parameters. The virgin mixture shows a constant
decrease in the inflection point frequency with aging while the RAP mixtures only show
the decrease at the later aging times. As aging level increases, the four RAP contents
follow very similar slopes, indicating that they are changing at similar rates. The unaged
mixtures show a clear decrease in inflection point frequency with increasing RAP
content. The mixtures aged to various levels show a very flat curve. This indicates that
the impact of the RAP materials on the hardening of the mixtures decreases with aging
time.
The C2 parameter from the Kaelble modified WLF fit of the shift factor curve is
shown in Figure 4b. The C2 parameters have relatively flat curves (except the the 30%
RAP four day aging point) and the RAP mixtures have higher C2 values than the virgin
mixture. The increase in the C2 parameter indicates a shallower shift factor curve, which
indicates the mixture is less temperature susceptible. There is an increase in the C2
parameter from the virgin to 20% RAP content at all aging levels and then the trend is
unclear.
Figure 4. |E*| Fit parameters (a) Inflection point frequency (b) Shift factor curve
Kaelble modified WLF C2 parameter.
4.1.2 Comparison to Global Aging System

Based upon SHRP recommendations, two, four, and eight days of laboratory aging
are approximately equal to four, eight, and sixteen years of service life (Bell and
Sosnovske, 1994). Figures 1 and 2 show the master curves predicted from the GAS
using the extracted and recovered binder properties for the respective RAP content as
dashed lines. The measured dynamic modulus using laboratory aging was less stiff than
the GAS predictions, with the greatest difference seen for the virgin mixture. Overall,
the data from the GAS model predictions suggest that the laboratory aging on this mix
simulated a shorter period of service life than predicted by the GAS model. The GAS
340
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

was then used to approximate an equivalent service life to correspond with each level of
laboratory aging for each RAP mixture. A master curve was fit to the unaged dynamic
modulus data for each RAP content. These master curves were then shifted using the
GAS until the best overlap over the entire frequency range was achieved between the
measured and GAS predicted master curves for each aging level. Table 3 summarizes the
service life in months each laboratory aging level simulated for each RAP content. The
mixtures are all very similar, with a large jump with the 8 day aging condition.
The unaged master curves for each RAP content (Fig 1b-1d) were compared to the
unaged master curve of the virgin mix (Fig 1a). The GAS was used to predict an amount
of aging that would create stiffness in the virgin mix equivalent to each unaged RAP
mix. These results are summarized in the last row of Table 4. It suggests that adding
20% or 30% RAP causes an increase in stiffness in the dynamic modulus curve that is
equivalent to field aging of less than a year predicted by the GAS. The 40% RAP causes
an increase in stiffness that is equivalent to field aging of almost 3.5 years.
Table 3. GAS predictions of the equivalent number of months of aging for the measured
dynamic modulus curves.
Equivalent GAS Aging Time in Months

0 % RAP 2 0 % RAP 3 0 % RAP 4 0 % RAP
2 days 4 4 2 3
Lab Aging
4 days 4 6 2 4
Level
8 days 20 10 20 30
Effect of RAP:
Comparison of Unaged n/a 4 8 40
to 0% RAP Unaged
4.1.3 Fatigue
The number of cycles to failure at different applied strain and stress levels can be
predicted using the S-VECD approach. In this study, fatigue predictions are made using
a loading frequency of 10 Hz at temperatures of 4oC and 25oC. Figure 5a shows the
number of cycles to failure as a function of applied strain level at 25oC. The unaged 20%
RAP mixture shows the best performance with the 30% eight day aged and 40% eight
day aged mixtures showing the worst. At a temperature of 4oC, the fatigue curves are
closer together with the 20% RAP mixture performing the best. The rankings of all the
mixtures at a range of strain values are summarized in Table 4. Within each
temperature, a few of the rankings change by one place with the different strain levels.
This can be seen visually in the figures when the lines cross. Comparing performance
between temperatures, up to two rank order changes are observed for some mixtures.
341
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Overall, the aged and unaged 20% RAP mixtures are the best performers and the aged
30% RAP and 40% RAP mixtures are the worst performers under strain control.
The predicted Nf for various stress levels at 25oC are shown in Figure 5b. Under
these conditions, the higher RAP, aged mixtures perform better. All mixtures have
better performance at the lower temperature which is expected due to the increase in
stiffness. Only two of the mixtures change rankings (Table 4) due to stress level at 25oC
and no rank changes are observed at 4oC. There are only a few rank changes with
temperature. Overall, the 40% RAP mixtures and the aged 30% RAP mixture are the
best performers and the unaged virgin and 20% RAP mixtures are the worst performers
in stress control.
The fatigue performance rankings for some of the mixtures change dramatically
from strain controlled to stress controlled. This highlights the importance of the
pavement structure in which these mixtures are placed. Some mixtures will show very
different performance in relatively thick or thin pavements (i.e., the unaged 20% RAP
mixture). Therefore, to design long lasting pavements, there must be a link between
mixture design and thickness design. The right mixtures must be designed and placed in
the right location within the pavement structure.
The predicted endurance limit values for each of the mixtures as a function of
temperature are shown in Figure 6. The endurance limit increases with temperature,
though several mixtures do change ranking with a change in temperature. The ranking of
the endurance limit is the same as the strain controlled fatigue. Also, the endurance limit
for the stiffer, aged mixtures changes less with temperature.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
342

Figure 5. Predicted Nf at 25oC for (a)Various strain levels (b) Various stress levels.
Figure 6. Endurance limit at various temperatures.
343
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 4. Fatigue ranking of mixtures.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
25C 4C Overall
Strain Amplitude Stress Amplitude Strain Amplitude Stress Amplitude
Mix 100 500 1000 100 500 1000 100 500 1000 100 500 1000 Strain Stress
0 0 D0 3 3 4 8 8 8 5 4 4 8 8 8 3 8
0 0 D8 6 6 6 5 7 7 6 6 7 5 5 5 6 5
2 0 D0 1 1 1 7 6 6 2 1 1 7 7 7 1 7
2 0 D8 2 2 2 4 4 4 1 2 2 4 4 4 2 4
3 0 D0 4 4 3 6 5 5 4 5 5 6 6 6 5 6
3 0 D8 7 7 7 1 1 1 7 7 6 2 2 2 7 1
4 0 D0 5 5 5 2 2 2 3 3 3 1 1 1 4 2
4 0 D8 8 8 8 3 3 3 8 8 8 3 3 3 8 3
344

4.2 Binder
4.2.1 Complex Shear Modulus
The modulus fits very well to the extended CAM model and the extended CAM
model phase angle formulation reflects the trends in the modulus-shifted phase angle as
well, although there is some variation in the measured data above and below the
numerical model for phase angle.
The master curve coefficients are provided in Table 5 along with the rheological
index (R) and cross over frequency determined from the master curves. Table 6 provides
the summary of all of the extracted binder properties. When these data were plotted
graphically to analyze the effects of RAP content and aging on the material properties,
there were mixed results without any clear increasing trend or decreasing trend
associated with either increasing RAP content or days of aging. However, there is a
general trend where the properties of the 20% RAP extracted binder do not exhibit a
significant change due to aging as compared to the other RAP contents. Also, when
properties were changed due to aging they tended to have the greatest change from four
to eight days of aging and were less affected by aging from zero to four days.
Complex shear modulus (Figure 7) responds as expected with increasing days of
aging but without a consistent, proportional trend. For example, with 0% RAP the
modulus master curves essentially overlapped except for eight days of aging, which was
stiffer.
Phase angle master curves (Figure 8) exhibited behavior like the complex shear
modulus where the decrease in phase angle did occur with increased aging and RAP but
was sporadic and not proportional to either RAP or days of aging. Coincidently, the
phase angle master curves tended to intersect at the 45 degree phase angle cross over
frequency and the behavior was mostly flat with respect to days of aging except for 0%
RAP which showed a noticeable trend of decreasing cross over frequency with days of
aging, as shown in Figure 9a. In other words, the binder generally responded the way
one would expect but not over the whole range of the measured master curve.
The temperature sensitivity of the rheological response was evaluated by analyzing
the change in the C2 shifting parameter in Figure 9b. The curves exhibit a slightly
increasing slope but for the most part overlap on top of one another for the different
RAP contents.
Glassy modulus was not fixed and was free to shift in the master curve fitting
optimization, which resulted in a scattered response ranging mostly between 2 GPa and
6 GPa and essentially flat with days of aging. The data points from the different RAP
contents overlapped with one another.
Rheological Index (R), which is the difference between the modulus at the cross
over frequency and the glassy modulus, provides a measure shape of the master curve
345
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

such as relatively gentle or relatively abrupt transition from glassy modulus to purely
viscous slope. No trend was observed for the effect of RAP and days of aging. The
rheological index data points for RAP and aging time were essentially flat and overlap at
a value of 3.
4.2.2 Performance Grade
High temperature continuous PG was fairly consistent where the 0% RAP mix had
the lowest continuous PG and continually increased with aging. The 40% RAP mixture
had magnitudes and behavior similar to the 0% RAP mix. 20% RAP and 30% RAP had
the higher initial continuous PG grade than 0% and 40% RAP and were less affected by
aging. Intermediate continuous PG grade was slightly scattered and mostly flat with
respect to days of aging and overlapped for the different RAP grades. Low temperature
continuous PG grade for all of the extracted binders was mostly scattered in a single -28
grade with a very slight trend of warming in the low PG grade with days of aging. RAP
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
content did not seem to influence the low temperature PG grade curves.
4.2.3 Brittleness and Ductility Parameters
CTOD remained unchanged for the 20% and 30% RAP. At 0% and 40% RAP the
parameter decreased with aging as expected because it is calculated from tensile energy
of failure that is divided by an increasing yield strength caused by aging and
embrittlement. The individual CTOD curves were not proportional to the RAP content.
G/(/G) exhibited a consistent increase. The values for 20% RAP were unchanged.
346

Table 5. Summary of binder master curve parameters using CAM model and Kaelbel shifting.
RAP Da y s Cross-over R
TK (oC) C1 C2 C me k Gg (Pa)
(%) Aging freq. (rad/sec) (log Pa)
0 0 2 1 .0 1 3 .8 1 0 9 .8 8.004E-02 1 .4 0 5 0.0957 4.65E+09 3 9 .4 2 2 .8 1
0 2 1 7 .7 1 4 .5 1 0 9 .0 6.809E-02 1 .4 0 1 0.0960 4.00E+09 3 1 .4 5 2 .8 0
0 4 1 7 .7 1 5 .0 1 1 3 .7 4.966E-02 1 .3 9 1 0.0981 2.97E+09 1 7 .9 1 2 .7 4
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0 8 2 1 .2 1 5 .8 1 2 3 .6 2.888E-06 1 .8 8 9 0.0737 9.97E+09 3 .0 3 3 .4 2
20 0 2 2 .0 1 4 .6 1 1 4 .7 4.056E-04 1 .6 4 6 0.0833 6.45E+09 8 .5 5 3 .1 1
20 2 2 0 .8 1 5 .0 1 1 7 .0 2.196E-03 1 .5 4 5 0.0881 5.25E+09 9 .4 2 2 .9 8
20 4 2 1 .6 1 4 .6 1 1 3 .6 4.907E-04 1 .6 3 3 0.0837 6.22E+09 8 .6 5 3 .1 0
20 8 1 9 .3 1 5 .2 1 1 6 .3 4.320E-04 1 .6 2 5 0.0837 5.41E+09 6 .9 1 3 .1 0
30 0 2 1 .0 1 4 .6 1 1 5 .0 5.179E-04 1 .6 4 9 0.0833 6.85E+09 1 1 .3 5 3 .1 1
30 2 1 9 .1 1 5 .2 1 1 3 .3 4.411E-04 1 .6 4 6 0.0849 5.48E+09 7 .7 2 3 .0 5
30 4 2 3 .9 1 4 .2 1 1 1 .5 3.908E-04 1 .6 2 6 0.0849 4.92E+09 5 .5 5 3 .0 6
30 8 1 9 .5 1 6 .5 1 3 0 .9 3.553E-04 1 .6 0 0 0.0859 3.95E+09 3 .4 6 3 .0 3
40 0 2 3 .9 1 3 .2 1 0 1 .4 4.801E-04 1 .6 7 1 0.0825 7.69E+09 1 4 .4 5 3 .1 3
40 2 1 9 .9 1 5 .3 1 1 2 .9 1.765E-04 1 .7 0 2 0.0875 2.60E+09 3 .9 9 2 .9 4
40 4 1 7 .9 1 5 .1 1 1 6 .0 3.285E-04 1 .7 0 2 0.0836 5.00E+09 1 1 .8 3 3 .0 8
40 8 1 8 .3 1 7 .3 1 3 5 .8 2.691E-04 1 .6 0 8 0.0872 3.12E+09 2 .4 7 2 .9 8
347

Table 6. Summary of extracted binder properties.
MSCR
RAP Da y s HTPG [oC] ITPG [oC] LTPG Jnr [kPa-1] CTOD G/(/G)
%Recovered
(%) Aging [oC] [mm] [Pa/s]
SL MC SL MC 100 Pa 3 2 0 0 Pa 100 Pa 3 2 0 0 Pa
0 0 7 2 .0 7 2 .3 1 4 .2 1 5 .2 -32.7 1 .3 5 1 .9 3 1 3 .8 4 .4 2 2 .7 384
0 2 7 1 .9 7 2 .2 1 7 .3 1 5 .4 -32.3 1 .3 3 1 .6 2 1 0 .8 0 .2 2 3 .3 391
0 4 7 3 .4 7 4 .0 1 8 .0 1 6 .0 -34.3 1 .0 7 1 .3 1 1 4 .0 2 .3 2 3 .0 540
0 8 8 2 .3 8 2 .1 1 8 .2 1 9 .1 -30.8 0 .3 6 0 .4 0 3 2 .2 2 4 .3 1 7 .1 1518
20 0 7 7 .6 7 7 .5 1 6 .0 1 7 .8 -31.2 0 .6 7 0 .8 0 2 2 .3 1 2 .1 1 8 .8 799
20 2 7 7 .7 7 7 .6 1 8 .9 1 8 .3 -30.7 0 .6 9 0 .8 0 2 0 .8 1 1 .9 1 8 .0 842
20 4 7 7 .5 7 7 .4 1 5 .5 1 7 .7 -31.6 0 .7 1 0 .8 4 2 1 .8 1 1 .9 1 8 .5 781
20 8 7 7 .5 7 8 .3 1 6 .1 1 7 .4 -31.1 0 .7 0 0 .8 4 2 3 .1 1 3 .1 1 8 .3 926
30 0 7 6 .9 7 7 .0 1 6 .1 1 7 .3 -31.6 0 .8 0 0 .9 4 2 0 .0 1 0 .3 1 7 .0 763
30 2 7 6 .4 7 7 .1 1 9 .7 1 8 .1 -31.8 0 .6 9 0 .8 1 1 7 .8 6 .7 1 8 .9 761
30 4 7 7 .4 7 7 .9 1 6 .6 1 8 .0 -31.8 0 .7 8 0 .9 4 2 1 .7 1 2 .3 1 8 .1 800
30 8 8 1 .5 8 1 .5 1 6 .0 1 8 .4 -30.6 0 .4 1 0 .4 6 2 8 .8 2 0 .5 1 7 .5 1688
40 0 7 3 .4 7 3 .8 1 7 .5 1 6 .9 -34.1 0 .9 7 1 .1 7 1 2 .0 1 .3 2 2 .6 395
40 2 7 4 .5 7 5 .3 1 9 .1 1 6 .2 -33.9 0 .9 4 1 .1 5 1 4 .5 4 .7 2 2 .4 614
40 4 7 5 .9 7 5 .8 1 6 .6 1 5 .3 -33.2 0 .7 0 0 .8 3 1 7 .7 6 .5 2 1 .2 735
40 8 8 1 .9 8 2 .0 1 7 .2 1 8 .5 -30.1 0 .4 0 0 .4 6 2 8 .8 2 0 .9 1 5 .6 1935
SL = Semi-log Interpolation, MC = Master Curve
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
348

0% RAP Binder Shear Modulus 20% RAP Binder Shear Modulus

100000000 100000000
1000000 1000000
|G*| (Pa)
|G*| (Pa)
0 Days 0 Days
10000 10000
2 Days 2 Days
100 4 Days 100 4 Days
1 8 Days 1 8 Days
1.E-07 1.E-03 1.E+01 1.E+05 1.E-07 1.E-03 1.E+01 1.E+05
Reduced Frequency (rad/sec) Reduced Frequency (rad/sec)
30% RAP Binder Shear Modulus 40% RAP Binder Shear Modulus
100000000 100000000
1000000 1000000
|G*| (Pa)
|G*| (Pa)
0 Days 0 Days
10000 10000
2 Days 2 Days
100 4 Days 100 4 Days
1 8 Days 1 8 Days
1.E-07 1.E-03 1.E+01 1.E+05 1.E-07 1.E-03 1.E+01 1.E+05
Figure 7. Binder complex shear modulus master curves for 0%, 20%, 30%, and 40% RAP.
349
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

0% RAP Binder Phase Angle 20% RAP Binder Phase Angle

90 90
Phase Angle (degrees)

75 75
60 60
0 Days 0 Days
45 45
2 Days 2 Days
30 30
15 4 Days 15 4 Days
0 8 Days 0 8 Days
1.E-07 1.E-03 1.E+01 1.E+05 1.E-07 1.E-03 1.E+01 1.E+05
30% RAP Binder Phase Angle 40% RAP Binder Phase Angle
90 90
Phae Angle (degrees)

75 75
60 60
0 Days 0 Days
45 45
2 Days 2 Days
30 30
15 4 Days 15 4 Days
0 8 Days 0 8 Days
1.E-07 1.E-03 1.E+01 1.E+05 1.E-07 1.E-03 1.E+01 1.E+05
Figure 8. Binder phase angle master curves for 0%, 20%, 30%, and 40% RAP.
350
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

45.00 140.0
40.00
Temperature Shifting C2
130.0
Cross Over Frequency
(reduced, rad/sec) 35.00
30.00 120.0
Parameter
25.00 0% RAP 0% RAP
110.0
20.00 20% RAP 20% RAP
15.00 30% RAP 100.0 30% RAP
10.00 40% RAP 40% RAP
90.0
5.00
0.00 80.0
0 5 10 0 5 10
Days of Aging Days of Aging
Figure 9. Binder shear modulus fit CAM model parameters (a) Cross over frequency as a function of RAP content
(b)Shift factor curve Kaelble modified WLF C2 parameter.
Figure 10. G/(/G) binder ductility parameter.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
351

4.3 Mixture and Binder Comparisons

The characteristics of the binder and mixture master curves are compared to
determine the degree to which the viscoelasticity of the fully blended virgin and RAP
binder may be reflected in the mixture where there is uncertainty in the amount of
blending between the RAP and virgin binder. The recovered binders cross over
frequency is compared to inflection point frequency of the mix because both are an
indicator of the transition from the more elastic regime to the more viscous regime. The
data are cross plotted in Figure 11a which shows a definite, proportional trend where the
mixture inflection frequency is about one order of magnitude slower. More importantly,
the 0% RAP mixture shows a wider range in both frequencies than the higher RAP
contents indicating the 0% RAP mix was more affected by aging.
The reference temperature of the binder and mixture master curves was 20oC and
allows a comparison of the C2 parameter which controls the steepness of the time
temperature superposition curves and, thus, is a measure of the temperature
susceptibility. The data are plotted in Figure 11b. There does not appear to be a
significant relationship between the binder and mixture for this property other than the
mixture having a consistently lower C2 value than the binder which is sensible given that
the volume of aggregate in the mixture decreases the temperature sensitivity from just
the binder.
The binder performance and mixture performance are further compared and
contrasted by analyzing the relationship between the mixtures fatigue life and three of
the binder parameters related to cracking and embrittlement: intermediate PG grade,
CTOD, and G/(/G) . A cross plot of the G/(/G) parameter against the 25oC
fatigue lives at 500 is provided in Figure 11c. The amount of agreement is quantified for
the different cross comparisons in Table 7. The familiar coefficient of correlation, R2, is
provided although this parameter is likely not useful in this particular scenario. A more
relevant parameter is the Kendalls Tau Measure of Association which is a non-
parametric, distribution free parameter that depends upon the number of data points and
quantifies the degree of the rank-order. The parameter can range between -1 for a perfect
inverse ranking of paired data (such as for number of cycles to failure and strain
amplitude) and +1 for perfect proportional raking of paired data (such as number of
cycles to failure in binder vs. mix). A score of 0 indicates no association at all. In
addition, a statistical indication of the likelihood of an underlying association between
the paired sets of data can be determined by considering the possible permutations for a
given rank scenario and total scenarios depending on the number points.
The analysis has revealed that there is naturally very low correlation by means of R2
given 8 data points between the binder parameters and the mixture logarithmic cycles to
failure. However, the Kendalls Tau Measure of Association (Kendall and Gibbons,
1996; Washington et al, 2001) is able to provide an assessment of the rank-order in spite
352
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of the low correlation. The interpretation of the scores and likelihood statistics is that all
three binder parameters do provide some indication of the mixtures fatigue resistance
with G/(/G) being slightly stronger than CTOD which is similarly as strong as
intermediate PG .
Table 7. Statistical parameters for binder mixture fatigue comparisons.
Mix Nf, 500 , Mix Nf, 500 kPa,

Strain Control Stress Control
25oC 4oC 25oC 4oC
R2 0 .6 5 0 .5 4 0 .1 9 0 .1 9
G/(/G) Kendall Tau -0.50 -0.50 0 .2 9 0 .3 6
K-T Significance 95% 95% 80% 86%
R2 0 .2 3 0 .2 2 0 .0 4 0 .0 3
CTOD Kendall Tau 0 .4 3 0 .4 3 -0.21 -0.14
R2 0 .3 4 0 .1 8 0 .2 1 0 .2 8
I TP G Kendall Tau -0.43 -0.29 0 .2 1 0 .4 3
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
353

0.1 200
Mixture Inflection Point Frequency (Hz)

180
160
0% RAP
Mixture C2
0% RAP 140
0.01 20% RAP
20% RAP
120 30% RAP
30% RAP
100 40% RAP
40% RAP
L OE
80
0.001
60
0.10 1.00 10.00
60.0 110.0 160.0
Binder Crossover Frequency (Hz) Binder C2
100,000,000
Mixture Nf at 500 mixrostrain 25C

10,000,000
1,000,000
0% RAP
20% RAP
100,000
30% RAP
40% RAP
10,000
1,000
0 500 1000 1500 2000 2500
Binder G/(/G)
Figure 11. Binder and mixture parameter comparisons (a) Inflection point and crossover frequency (b) Kaelble modified
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
WLF C2 shift parameter (c) Binder G/(/G) and mixture Nf.
354


In this study, the effect of long-term oven aging on RAP mixtures was evaluated.
Four plant-produced mixtures containing 0%, 20%, 30%, and 40% RAP were used to
fabricate specimens that were long-term oven aged at 2, 4, and 8 days at 85oC. Mixture
testing included complex modulus and fatigue in uniaxial tension. Binders were
extracted and recovered and characterized for rheological properties and PG grade. The
analysis included use of the Global Aging System model, evaluation of master curve
fitting parameters and comparisons between the binder and mixture properties and
parameters. The following observations have been made based on the testing and
analysis performed in this study:
Mixtures
The stiffening effect of long term oven aging on RAP mixtures is less than that of
virgin mixtures. This is likely due to the inclusion of already aged binder in the
RAP mixtures that ages at a much slower rate than virgin binder under laboratory
conditioning.
For New Hampshire conditions, analysis of the various oven-aged |E*| curves using
the GAS model suggests that the GAS model predicts more aging than equivalent
aging times recommended by the SHRP research. Two and four days of laboratory
aging appear equivalent to 2 to 4 months and 2 to 6 months of field aging as
predicted by the GAS model rather than 4 and 8 years of field aging. Eight days of
laboratory aging appears equivalent to 10 to 30 months of field aging as predicted
by the GAS model rather than 16 years of field aging. Adding 20% or 30% RAP
causes an increase in stiffness in the dynamic modulus curve that is equivalent to
field aging of less than one year while 40% RAP causes an increase equivalent to
field aging of almost 3.5 years.
The phase angle master curves shift to lower frequencies with RAP content and
aging time; aging also causes the peak value of the phase angle to decrease. The
shift in the phase angle master curve is quantified by the decrease in inflection point
frequency determined from the fitting parameters. The shapes of the inflection point
frequency curves indicate that the impact of the RAP materials decreases with aging
time.
Blackspace diagrams show convergence of the aging levels at higher stiffness values
while the aged mixtures have lower phase angles at lower stiffness values. This
indicates that the aged materials will be more susceptible to cracking and more
resistant to rutting in the low stiffness ranges.
The fatigue performance rankings for some of the mixtures change dramatically
under strain controlled or stress controlled conditions consistent with the changes in
stiffness. In strain control, the aged and unaged 20% RAP mixtures were the best
performers and the aged 30% RAP and aged 40% RAP mixtures were the worst
performers. In stress control, both 40% RAP mixtures and the aged 30% RAP
355
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

mixture were the best performers and the unaged virgin and unaged 20% RAP
mixtures were the worst performers. This highlights the importance of considering
mixture design and pavement structure together for the design of long lasting
pavements. The right mixtures must be designed and placed in the right location
within the pavement structure and there are suitable locations for high RAP
mixtures in a pavement structure.
Aging time had a very small impact on the temperature susceptibility of the
mixtures as evaluated by the Kaelble modified WLF C2 parameter. However, RAP
mixtures have higher C2 values than the virgin mixture indicating they are less
temperature susceptible.
Binders
Complex shear modulus responds as expected with increasing days of aging but
without a consistent, proportional trend. For example, with 0% RAP the modulus
master curves for each aging condition essentially overlapped except for 8 days of
aging which was stiffer.
Phase angle master curves exhibited behavior like the complex shear modulus where
the decrease in phase angle did occur with increased aging and RAP content but was
sporadic and not proportional to either RAP content or days of aging.
The temperature sensitivity of the rheological response evaluated using the C2
shifting parameter indicates a slight increase in temperature susceptibility with
aging level, but not a clear difference between RAP contents.
The high temperature PG grade stiffens at the high temperature, had minimal effect
at the intermediate temperature, and showed a slight warming trend at the lo w
temperature.
The CTOD decreased with aging time for some mixtures and the G/(/G)
parameter increases with aging time, indicating a reduced resistance to cracking.
Binder Mixture Comparison
Both the binder complex modulus and mixture complex modulus show similar
overall trends with respect to aging level and RAP content, however there are
differences in the magnitudes of changes observed between the binder data and
mixture data.
There is a clear relationship between the binder crossover frequency and the mixture
inflection point frequency which means characterization of the binder is linked with
the characterization of mixtures with varying RAP contents.
In terms of fatigue, the intermediate PG grade, intermediate PG, CTOD, and
G/(/G) all show a relationship with the mixture fatigue life where the
G/(/G) had a stronger association than CTOD and intermediate PG.
Finally, it is very important to emphasize that the mixtures evaluated in this study are
from plant production rather than laboratory fabricated specimens. Plant production
356
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

introduces variables that are outside of laboratory control; solvent extraction and
recovery also introduce more variables and destroy some of the in situ molecular
configurations and interactions with the aggregate. Regardless, consistent trends were
observed at both the binder and mixture scale. In addition, while the influence of RAP
on stiffness and fatigue resistance was clearly demonstrated with this study, the
magnitude of the effects of increasing RAP content were likely overshadowed by natural
variables of plant production as evidenced in the scatter of some of the relationships.
This is an important observation as it indicates that it might be possible to influence the
performance of mixtures with up to 40% RAP with certain proper material handling,
production and construction practices. This study only evaluated one set of RAP
mixtures; additional work needs to be performed on other types of mixtures. Field
evaluation of RAP mixtures over time and testing of field aged RAP mixtures is also
needed to fully understand the behavior of RAP mixtures over time.
6. Acknowledgements
The authors would like to acknowledge the Recycled Materials Resource Center for
funding this work and Pike Industries, Inc., for supplying the mixture. Geoff Rowe
provided timely ideas, advice, and guidance. Ashton Congalton helped with mixture
analysis. LaKesha Perry, Sue Needham, and Monte Simpson at the FHWA Turner-
Fairbank Highway Research Center are thanked for providing careful attention to the
asphalt binder recovery and testing.
7. References
Al-Khateeb, G., X. Qi, A. Shenoy, K. Stuart, and T. Mitchell. Assessment of Aging at
FHWAs Pavement Testing Facility, Transportation Research Record: Journal of
the Transportation Research Board, No. 1940, Transportation Research Board of
the National Academies, Washington, D.C., 2005, pp. 146155.
Anderson, D.A., Christensen, D.W., Bahia, H.U., Dongre, R., Sharma, M.G., Antle, C.E
and Button, J., Binder Characterization and Evaluation, Volume 3: Physical
Characterization, SHRP-A-369, Strategic Highway Research Program, National
Research Council, Washington, DC 1994.
Anderson, R.M., G. King, D. Hanson, and P. Blankenship, Evaluation of the
Relationship between Asphalt Binder Properties and Non-Load Related Cracking,
Journal of the Association of Asphalt Paving Technologists, Vol. 80, 2011, pp.,
615649.
357
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Bell, C.A., AbWahab, Y., Cristi, M.E., Sosnovske, D., Selection of Laboratory Aging
Procedures for Asphalt-Aggregate Mixtures, SHRP-A-383, Strategic Highway
Research Program, National Research Council, Washington, D.C., 1994.
Bell, C.A., Sosnovske, D., Aging: Binder Validation, SHRP-A-384, Strategic Highway
Research Program, National Research Council, Washington, D.C., 1994.
Daniel, J.S., Y.R. Kim, and H.J. Lee, Effects of Aging on Viscoelastic Properties of
Asphalt-Aggregate Mixtures, Transportation Research Record: Journal of the
National Academies, Washington, D.C.,1998, pp. 2127.
Gibson, N., Qi, X., Shenoy, A., Al-Khateeb, G., Kutay, M. E., Andriescu, A., et al,
Performance Testing for Superpave and Structural Validation, Final Report FHWA-
HRT-11-045, Federal Highway Administration, Washington, D.C. (Forthcoming),
2012
Glover, C., A. Epps Martin, A. Chowdhury, R. Han, N. Prapiatrakul, X. Jin, and J.
Lawrence, Evaluation of Binder Aging and Its Influence in Aging of Hot Mix
Asphalt Concrete: Literature Review and Experimental Design, Report
FHWA/TX-08/0-6009-1,Texas Transportation Institute, College Station, TX, 2009.
Harvey, J. and B. Tsai, Long-Term Oven-Aging Effects on Fatigue and Initial Stiffness
of Asphalt Concrete, Transportation Research Record: Journal of the
National Academies, Washington, D.C.,1997, pp. 8998.
Kendall, M. and Gibbons, J. (1996). Rank Correlation Methods, Fifth Edition, Oxford
University Press, Oxford, England.
Mirza M.W. and Witczak M.W., Development of a Global Aging System for Short and
Long Term Aging of Asphalt Cements, Journal of Association of Asphalt Paving
Molenaar, A., E. Hagos, and M. Van de Ven, Effects of Aging on the Mechanical
Characteristics of Bitumious Binders in PAC, Journal of Materials in Civil
Engineering, Vol. 22, No. 8, American Society of Civil Engineers, 2010, pp. 779
787.
Morian, N., E. Hajj, C. Glover and P. Sebaaly Oxidative Aging of Asphalt Binders in
Hot Mix Asphalt Mixtures, Transportation Research Board of the National
Academies, Washington, D.C., 2011.
Raad, L., S. Saboundjian, and G. Minassian, Field Aging Effects on Fatigue of Asphalt
Concrete and Asphalt-Rubber Concrete, Transportation Research Record: Journal
358
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of the Transportation Research Board, No. 1767, Transportation Research Board of

the National Academies, Washington, D.C.,1999, pp. 126134.
Rowe, G. and M. Sharrock, Alternate Shift Factor Relationship for Describing the
Temperature Dependency of the Visco-Elastic Behaviour of Asphalt Materials,
No. 2207, pp. 125135, 2011.
Said, S. Aging Effect on Mechanical Characteristics of Bituminous Mixtures,
No. 1901, Transportation Research Board of the National Academies, Washington,
D.C., 2005, pp. 19.
Tarbox, S. and J. Daniel, Long Term Oven Aging of RAP Mixtures, Transportation
Research Record: Journal of the Transportation Research Board, 2012, in press.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Walubita, L., A. Epps Martin, S. Jung, C. Glover, E. Park, A. Chowdhury, and R. Lytton
Comparison of Fatigue Analysis Approaches for Two Hot Mix Asphalt Concrete
Mixtures, FHWA/TX-05/0-4468-2, Texas Transportation Institute, College Station,
TX, 2005.
Walubita, L., A. Epps Martin, S. Jung, C. Glover, E. Park Application of Calibrated
Mechanistic Fatigue Analysis with Aging Effects, Report FHWA/TX-06/0-4468-3,
Texas Transportation Institute, College Station, TX, 2006.
Washington, S., Leonard, J., Manning, D.G., Roberts, C., Williams, B., and Bacchus,
A.R. (2001). NCHRP 20-45Statistics for Transportation Researchers, Scientific
Approaches to Transportation Research, National Cooperative Highway Research
Program. Accessed online: November 7, 2011.
(http://onlinepubs.trb.org/onlinepubs/nchrp/cd-22/readme.html)
Woo, W., E. Ofori-Abebresse, A. Chowdhury, J. Hilbrich, Z. Kraus, A. Epps Martin, C.
Glover Polymer Modified Asphalt Durability in Pavements, Report FHWA/TX-
07/0-4468-1, Texas Transportation Institute, College Station, TX, 2007.
Woo, W.J., A. Chowdhury, and C. Glover Field Aging of Unmodified Asphalt Binder
in Three Texas Long-Term Performance Pavements, Transportation Research
Record: Journal of the Transportation Research Board, No. 2051, Transportation
Research Board of the National Academies, Washington, D.C., 2008, pp. 15-22.
Underwood, B. S. and Y. R. Kim Improved Calculation Method of Damage Parameter
in Viscoelastic Continuum Damage Model, International Journal of Pavement
Engineering, Vol. 11, No. 6 2010, pp. 459476.
359

8. Discussion
DR. M. EMIN KUTAY: Very good work, Jo. Based on your conclusions, can we
conclude that high-RAP mixtures are prone to less aging in the field, as compared to a
less-RAP mixture?
PROFESSOR JO DANIEL: Based on what we saw, yes. You can see that definitely just
looking at the comparison of the dynamic modulus curves and phase angle curves, they
get closer together with the higher RAP contents, so there is less of a difference or less
of an impact of aging on those higher RAP contents.
DR. KUTAY: So can we say that when we design a high-RAP mix our predictions over
time will be a little more accurate than a virgin mix? Because, you know, a virgin mix
ages more and we really dont know how aging takes place in the field, and prediction of
that aging is much more difficult. So if you know that the high-RAP mixture ages less
and that kind of eliminates a little bit of an unknown from our predictions of the
performance in the future if you are in the business of, I dont know, modeling or
PROF. DANIEL: Actually, I hadnt thought about it from that perspective, but, yes, I
think so. I mean from what were seeing here in terms of the stiffness, yes.
DR. KUTAY: One more question about your aging protocol. I believe you aged the
actual mixture samples compacted and cut?
PROF. DANIEL: Yes.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. KUTAY: What do you think about aging gradients in the sample? Do you think that
has anything to do with any conclusions that you make in here?
PROF. DANIEL: Its a variable. We dont know exactly what those aging gradients are.
Then when we extracted and recovered, obviously, with the binder one of the
complications with the binder testing is we dont have a consistently aged specimen
there. Thats an issue with the mixture testing as well. But what we did, we decided to
follow the SHRP protocols in the absence of anything else at the time when we were
doing the work. So theres a lot of work I mean, I think everybody recognizes the fact
that thats an open question right now, and maybe well have a better answer sometime
in the future. But when we did this work, we followed what was existing in terms of the
recommendations for the mixture aging for the compacted specimens.
MR. KENNETH CRAWFORD: Great work. It really provokes a lot of thinking because
in our line of industry, were always trying to get it back to the original asphalt to get the
right quality, and this says well maybe thats not always the right thing. But I was
wondering for future work if using a rejuvenator and testing it at 40% and 60% RAP
levels would make sense because what were trying to do there usually is trying to
improve fatigue and cracking type properties and, you know, get the whole sweet spot
360

getting back to the Superpave specifications. So I was wondering how you felt about
future work in that area.
PROF. DANIEL: Well, I think the rejuvenators are something that a lot of folks are
looking at. I think its something we do need to consider in terms of trying to bring
those properties back, especially as were starting to get into the higher RAP contents.
That might be an appropriate use of bringing those materials back. I think one of the
things that need to be looked at with those is once you bring those materials back, then
we still need to know whats happening over time and how those materials are going to
age because we know they are going to age once theyre back out on the road.
MR. CRAWFORD: Thank you.
MR. FRANK FEE: A real quick one. As I understand it, this was plant mixed and then
reheated and lab compacted.
PROF. DANIEL: Yes.
MR. FEE: Jo, are you going to try to compare it to a plant mixed, plant compacted to see
if the reheating has any effect on the whole aging?
PROF. DANIEL: Actually, as part of the Pooled Fund Study, weve done a lot of that
work in comparing the specimens that are compacted without reheating, so immediately
theyre sampled and compacted immediately at the plant location. There is an impact of
the reheating, and what we have seen in that Pooled Fund study is that impact changes
with the amount of RAP that is in the mixture.
MR. FEE: Thank you.
PROFESSOR REYNALDO ROQUE: Jo, how were you able or were you able to control
closely the gradation between the different percent RAP mixtures? If not, did you see a
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
pattern for this particular RAP where there were differences in gradation between
mixtures with different RAP content that might impact your results?
PROF. DANIEL: For these studies, the asphalt content was targeted and the gradation
was targeted. So there are some differences in the gradation just because of the way the
stockpiles work, but we did the work and adjusted the mix designs so that the gradations
are very similar. That information is in the paper, but there is not a lot of difference
terms of gradation from one RAP mixture to the next.
PROF. ROQUE: Okay. And the design maintained the same total asphalt?
PROF. DANIEL: Total asphalt.
PROF. ROQUE: Assuming that all of the asphalt from the RAP was involved.
PROF. DANIEL: Right. Yes, that was the assumption.
PROF ROQUE: Okay. Thank you.
361

DR. GEOFFREY ROWE: Jo, very nice presentation. From what I saw in your
presentation, I can understand better what youre doing, so very nice work. I was very
interested to see that what we are calling the Glover-Rowe parameter correlate nicely
with the fatigue. If you look at that parameter in the black space, it sort of makes sense
and we can actually still relate that to things like the R value with the rheology, the
aging, and whats going on, and its very nice to do that. Ive got a comment and a
question for you. One of the things that we do with the testing of dynamic modulus,
quite often we have what I would call ill-posed data in that we have not equidistant
spaces in the data were capturing on a log scale for isotherms of stiffness. So when you
develop master curves from those, depending upon the shifting techniques youre using,
you can mischaracterize sometimes the temperature susceptibility and the shape of the
master curve because the shift, depending on how you do the shift, is dependent upon the
way that you're doing the shift. Ideally, I like to have equipoints, maybe five or six
points per decade of frequency so you can do a free shift so you can really fully
understand the temperature susceptibility of the binder from shifting and lining up those
isotherms without the assumption of any underlying model. I know weve sort of been
going back to the procedures that were laid down in TP 62 and those types of procedures
for the development of this work. One of the things Id like to suggest for researchers
like yourself and other people in the community who are trying to do more fundamental
research work is that we add some additional data points into the testing, which in terms
of research context wont add very much time to the actual testing. But I think when you
look at the datasets that you're trying to develop between things like the shift parameters,
you know, the C1, the C2s and some of these other parameters, I think you might start
to see some better relationships. I pose that as a comment and as a question for you as to
whether youve got any thoughts with regard to that sort of issue there?
PROF. DANIEL: You and I have discussed this several times, and just as a way of
explanation, the points that we collected for this particular data were equidistant in terms
of the frequencies that were sampled. We are looking at this for future work and
additional work going ahead and doing the testing at additional frequencies to get more
information so that we could do a better job with shifting. The shifting that we did, the
master curve construction for this particular project, we shifted to the generalized
logistic function, but we let the shift factors fall where they may. So we didnt constrain
the shift factors to a particular functional form, but we did do the master curve shifting to
the generalized logistic form. The WLF modified cable, we fit the shift factor curves
afterwards, so we let them free shift and then fit that functional form to them after they
were shifted. But thats some of what were kind of playing around with now is some of
the different shifting and using some of your free shifting techniques and looking at what
some of those differences might be.
DR. ROWE: Thanks, Jo, for that clarification. I guess its still a little bit of a model shift
if youre still shifting to model, so its not quite free.
362
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

PROF. DANIEL: Right, it wasnt a completely free shift. But at the time we were doing
the work, we didnt have the techniques to do the complete free shift, so we did it to the
modified logistic function.
DR. ROWE: This is excellent work. Im really delighted and compliment you on your
results youve generated. Thank you very much.
PROF. DANIEL: Thank you, Geoff.
DR. ERVIN DUKATZ: Excellent presentation, wonderful work. Its wonderful to see
past research corroborated, that it was correct. A similar project was conducted at
Clemson back in the1980s on the same type topic that showed that if you wanted to
predict your pavement response in five years, you put more RAP into it. Its nice to see
that as the tools improve, the same theories get repeated. The question I particularly have
is did you get a chance to take a second look at the modulus on the 30% RAP mix or
was that just a one truckload sample taken at the plant?
PROF. DANIEL: This was a one-time deal. So the mixtures were produced and we
sampled them.
DR. DUKATZ: One suggestion, in reviewing the data, is to look carefully at the
gradation data, because theres a shift to the coarser side both on the coarse side of the
gradation as well as on the fine side of the gradation thats not shown for the other three
mixes. That shift in gradation could account for a couple of those strange values in your
data.
PROF. DANIEL: It absolutely could.

DR. DUKATZ: The final two questions I have, which are both basically the same
question are you going to look or do you have plans in the future to look at different
gyration levels? The N-design of 50 mixes for most of the states that I work in would be
considered for very low volume road type mixes. So, I would be interested to see what
would happen as you go to the higher ESAL mixes where there is less effective asphalt
in the mix. Then the follow-up to that is to look at different PG grades.
PROF. DANIEL: Short answer, yes. The New Hampshire 50 gyration mix is actually the
standard mix for the New Hampshire traffic level, so thats why we have that particular
mixture. There are some 75-gyration mixtures that are part of the Pooled Fund Study that
weve been doing, weve been looking at, so we have some different gyration mixtures.
We have some different PG grades. We havent looked at the aging on those, but we are
looking at the impact of the different RAP contents and looking at the impact of a lot of
the different production parameters on whats happening with these mixtures. So there
are still a lot of questions out there and a lot of the work that we should be doing. The
other thing I should mention that weve been doing, New Hampshire DOT, they had
some 35% RAP mixtures that were down for close to 20 years, and so weve been doing
363
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

some forensic testing on those field cores. I have some information on doing binder
extraction/recovery, basically shaving off different levels and doing some mixture
testing that goes along with that. Actually that is a thesis thats being defended in a
couple weeks, so youll probably see that sometime next year in terms of a paper.
DR. DUKATZ: That will be very exciting. The other thing to take a look at is the rate of
oxidation. Youre going to have aging, but what is the rate of aging as the delta
between the high temperature grade and that low temperature grade increases.
PROF. DANIEL: That will be interesting to look at too, absolutely.

DR. DUKATZ: But you havent done anything on that yet.
PROF. DANIEL: I have not.

DR.DUKATZ: Okay. Thank you.
PROF. DANIEL: A lot of data, not as much time.
MR. DALE DECKER: Good morning, Jo. Nice presentation, as usual. I guess maybe a
little higher level question, up at 10,000 feet perhaps. As part of the recycling center, are
you guys looking at any sort of broad national review of RAP mixes and how theyre
performing? I mean, weve got millions and millions and millions of tons of mix down,
and we kind of all say, Oh, yeah, its performing equally or better, but we really dont
have any data to support that. Is anybody really looking at a major national study to
verify the performance of RAP?
PROF. DANIEL: So some of that work was being done in the RAP ETG, and, I kno w
we discussed that a lot and I dont know where that stands in terms of looking at. I kno w
part of the RAP ETG, we were tracking a lot of the RAP projects and that might be a
question for Matt.
MR. DECKER: Since it no longer exists.
PROF. DANIEL: The Recycled Material Resource Center no longer exists. RMRC
hasnt existed for the last two years, so RMRC is not in existence anymore. So with the
RAP ETG, I know that was something that we discussed doing, but I dont know where
it stands.
MR. DECKER: I guess the thing Im leading to is the kind Ray Brown did in the early
1990s with that rutting study, which was kind of a seminal study looking nationally at --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
whats going on. I was just curious. Not really related to your paper but it is.
PROF. DANIEL: No, but you're right. Its something that should be done.
MR. DECKER: Thank you very much.
364

DR. REBECCA MCDANIEL: This isnt exactly an answer to your question, Dale, but
the Long-Term Pavement Performance Program is adding the first new experiment in 20
years. It will be SPS10, which is looking at warm mix but warm mix with RAP and
warm mix with shingles, and there will be control sections. The control sections, I
believe, will also have RAP. So we will be collecting that data on a national level. It is
just getting started. Thank you.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
365

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluation of Use of Recycled Asphalt
Shingles in HMA
Fujie Zhoua*, Hongsheng Lia, Sheng Hua, Robert Leeb,
Tom Sculliona, German Clarosb, Jon Eppsa, and Joe Buttona
a
Texas A&M Transportation Institute, College Station, TX 77843-3135;
b
Texas Department of Transportation, Austin, TX 78705
ABSTRACT: This comprehensive study focused on best use of recycled asphalt shingles (RAS) in hot
mix asphalt (HMA). Consistency of processed RAS is discussed. It was found that, in Texas, the
processed RAS, including both tear-off asphalt shingles (TOAS) and manufacturing waste asphalt
shingles (MWAS), have very low variability in terms of gradation and RAS binder content.
Furthermore, the authors characterized RAS binder performance grade (PG). RAS binders are
very stiff, and TOAS binders, with an average high-temperature grade of 178C, are much stiffer
than MWAS binders that have an average high-temperature grade of 131C. Properties of
blended virgin and extracted/recovered RAS binders were investigated. Generally, the properties
of blended virgin and RAS binders are non-linear. However, for practical applications, the linear
blending chart can still be used if the RAS binder percentage is below 30%. Further, compared to
TOAS binders, MWAS binders have less impact on PG temperatures of virgin binders. Thus, it is
important to consider differentiating MWAS from TOAS when used in asphalt mixes. This study
also evaluated the impact of TOAS and MWAS on engineering properties of asphalt mixes. The use
of RAS has no significant influence on dynamic moduli of HMA mixes, but improves their
resistance to rutting/moisture damage. Meanwhile, adding RAS generally increases optimum
asphalt content (OAC) of HMA mixes and higher OAC corresponds to higher RAS content.
However, RAS mixes typically exhibit very poor cracking resistance compared to the 0% RAS
mixes with either PG 64-22 or PG 70-22, even though the RAS mixes have higher OAC. Therefore,
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
cracking resistance can be a significant concern for RAS mixes. This paper explored two
approaches for improving cracking resistance of RAS mixes in the laboratory and the field.
Laboratory test results clearly indicated that both using soft binder and increasing design density
can improve cracking resistance of RAS mixes. When considering rutting resistance of RAS mixes,
using a soft binder is superior to decreasing design air voids. The effectiveness of decreasing
The oral presentation was made by Dr. Zhou.
367

ZHOU, LI, HU, LEE, SCULLION, CLAROS, EPPS, BUTTON
design air voids was confirmed through two field test pavements on US 87 near Amarillo, Texas;
the four soft binder test pavements on FM 973, near Austin, are still being monitored
KEYWORDS: RAS, Binder Blending, Overlay Test, Rutting, Cracking, Dynamic Modulus
1. Introduction
Reclaimed asphalt pavement (RAP) has been one of the most often used recycling
materials in the asphalt industry. With increases in the price of asphalt cement and
subsequent price fluctuations, the industry has further amplified its recycling efforts.
Most recently, with the drastic increases in the cost of asphalt cement, the use of
recycled asphalt shingles (RAS) in hot-mix asphalt (HMA) has become another black
gold to the asphalt paving industry, since RAS contains approximately 20-30% asphalt
binder by its total mass. In addition to conserving energy and protecting the
environment, the use of RAS can significantly reduce the cost of HMA production and
paving. There are two basic types of RAS scraps: tear-off asphalt shingles (TOAS) and
manufacturing waste asphalt shingles (MWAS). In United States, around 10 million tons
of TOAS and 1 million tons of MWAS are available for recycling (Newcomb et al.
1993). Specifically, there are several nationwide, large roofing shingle manufacturers in
Texas, such as Owens Corning, GAF, TAMKO, Certain Teed, etc. Significant markets
exist for both the recycling and paving industries.
More than 30 years ago, some of the original pioneers established the first shingle
recycling plants, investigated mix designs containing RAS, and then published the first
technical literature in the late 1980s (Epps and Paulsen, 1986; Paulsen et al., 1986; and
Shepherd et al., 1989). More recently, HMA producers, Departments of Transportation
(DOTs), and researchers further expanded the expertise in shingle recycling in HMA
(Grzybowski, 1993; Newcomb et al., 1993; Button et al., 1996; Janisch and Turgeon,
1996; Foo et al., 1999; NAHB, 1999; Dykes, 2002; Lum, 2006; Brock 2007; McGraw et
al., 2007; Schroer, 2007, Johnson et al., 2010; Williams et al. 2011, and more on the
website: www.shinglerecycling.org). All these efforts paved the way for more DOTs to
use RAS in HMA.
In February 2009, the Texas Commission on Environmental Quality (TCEQ) issued
an Authorization Memo to allow HMA plants to include either MWAS or TOAS under
the TCEQ air quality standard permit for permanent HMA plants. Since then, RAS has
been widely used in Texas. Meanwhile, some concerns on consistency of processed
RAS, stiffness of RAS binder, and durability of RAS pavements were raised. To address
these concerns, in 2010, the Texas Department of Transportation (TxDOT) initiated a
368
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluation of Use of Recycled Asphalt Shingles in HMA
research study at the Texas A&M Transportation Institute (TTI) with an overall
objective of improving the use of RAS in HMA. To achieve this main objective, the
following steps were undertaken by the researchers:
Investigate consistency of processed RAS,
Characterize RAS binder properties,
Produce virgin binder/RAS binder blending charts,
Identify the impact of RAS on engineering properties of paving mixtures,
Evaluate approaches for improving durability of RAS mixes in the lab, and
Validate the approaches for improving durability of RAS mixes in the field.
Details of each of these are presented below. This is followed by the summary and
conclusions at the end of the paper.
2. Consistency of Processed RAS in Texas
Researchers visited different contractors and recyclers around Texas, and sampled ten
different types of processed RAS stockpiles: Four MWAS and six TOAS. For
simplicity, these are named MWAS-A, MWAS-B, MWAS-C, MWAS-D, TOAS-A,
TOAS-B, TOAS-C, TOAS-D, TOAS-E, and TOAS-F. For each processed RAS
stockpile, seven replicates of the processed RAS were collected and brought back to TTI
for laboratory testing.
2.1 Dry Sieve Analysis Results
Currently, the TxDOT specification requires 100% passing the in. sieve and 95%
passing a in. sieve. All ten processed RAS stockpiles tested met the specification and
they are very consistent in terms of gradation. Due to space limitation, Table 1 only
presents the dry sieve analysis test results for 4 processed RAS materials sampled in this
study. In fact, three processed RAS (TOAS-C, TOAS-F, and MWAS-C) had 100%
passing the in. sieve.
369
--`,`,,,``,,```,,`,,,,,`,``,``

Table 1. Dry sieve analysis results of four processed RAS materials.
RAS Standard
Sieve No. #1 #2 #3 #4 #5 #6 #7 Average
deviation
1 /2 " 100 100 100 100 100 100 100 100 0 .1
TOAS-C 3 /8 " 99 100 99 100 100 100 99 100 0 .2
#4 81 86 82 84 88 86 84 84 2 .3
1 /2 " 100 100 100 100 100 100 99 100 0 .4
TOAS-F 3 /8 " 100 100 100 100 100 100 98 100 0 .9
#4 94 93 95 94 93 94 89 93 2 .2
1 /2 " 100 100 98 99 100 100 100 100 0 .6
MWAS-A 3 /8 " 99 98 96 99 99 99 99 98 1 .1
#4 91 82 87 91 88 88 90 88 3 .1
1 /2 " 100 100 100 100 100 100 100 100 0 .0
MWAS-C 3 /8 " 100 100 100 100 100 100 100 100 0 .1
#4 85 80 89 93 87 89 90 87 4 .1
2.2 Ignition Oven Test and Associated Results
Normally, 1300 g of representative material of HMA or RAP is required for an ignition

oven test to determine the asphalt content and a washed sieve analysis. However, since
RAS has very a high asphalt content (more than 20%), the binder in 1300 g of RAS
could not be completely burned even if the specimen was burned more than three times.
After many trials, the researchers found that approximately 500-700 grams of RAS
material provides complete burning and consistent results in terms of RAS binder
content and RAS aggregate (solids) gradation. Following this practice, the researchers
performed the ignition oven test on all seven selected RAS materials and then a wet
sieve analysis on the RAS aggregates. For each selected RAS, seven replicates were
tested. Due to limited space, only two representative test results are presented below: one
for TOAS (see Table 2) and one for MWAS (Table 3).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The results listed in Tables 2 and 3 indicate that TOAS has higher binder content
than MWAS. MWAS generally has a consistent 20 percent binder content; TOAS has
various binder contents, ranging from 23% to 28%. Overall, RAS variability, in terms of
asphalt binder content and gradation, is low for both MWAS and TOAS. MWAS
exhibited a slightly lower variability.
370

Table 2. Gradations and binder contents from ignition tests: TOAS-B.

Cumulative % Passing of RAS Samples Standard
Sieve Size Av e r a g e
#1 #2 #3 #4 #5 #6 #7 Deviation
1/2" 100 100 100 100 100 100 100 100 0 .0
3/8" 100 100 100 100 100 100 100 100 0 .1
#4 98 91 98 99 94 98 99 97 2 .7
#8 95 86 96 96 86 93 95 92 4 .1
#16 75 65 76 80 66 72 74 73 5 .0
#30 53 44 53 64 48 49 52 52 5 .8
#50 48 38 47 60 43 43 46 46 6 .2
#100 41 31 39 54 36 35 38 39 6 .6
#200 30 22 28 46 28 25 27 30 7 .2
Binder content
25 24 28 28 28 28 28 27 1 .5
(%)
Table 3. Gradations and binder contents from ignition tests: MWAS-A.
Cumulative % Passing of RAS Samples Standard
Sieve Size Average
#1 #2 #3 #4 #5 #6 #7 deviation
1/2" 100 100 100 100 100 100 100 100 0 .0
3/8" 100 100 100 100 100 100 100 100 0 .0
#4 100 100 100 100 99 98 99 99 0 .5
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
#8 99 98 99 99 98 95 97 98 1 .5
#16 85 83 85 86 83 80 83 83 2 .0
#30 66 63 64 65 63 59 61 63 2 .1
#50 57 53 54 55 52 50 51 53 2 .3
#100 45 42 42 43 37 37 37 40 3 .1
#200 33 30 30 32 27 27 28 30 2 .2
Binder
21 19 21 19 20 20 19 20 0 .7
content (%)
3. RAS Binder Characterization and Virgin/RAS Binder Blending Charts

It is a known fact that the extracted/recovered RAS binders, regardless of MWAS or
TOAS, are very stiff, and they are far stiffer than any PG 76-22 binder. But their true
PG is not well known due to the temperature limitations of commonly used dynamic
shear rheometers (DSR). Since RAS binder properties have significant influence on
virgin binder selection and the allowable maximum amount of RAS used in asphalt
371

mixes, it is critical to determine the true performance grade of RAS binder. This study
sampled, extracted/recovered, and characterized a variety of RAS binders. Detailed
information is provided below.
3.1 Selection of RAS Samples
The same ten processed RAS products described previously were used: four MWAS
(MWAS-A, MWAS-B, MWAS-C, and MWAS-D) and six TOAS (TOAS-A, TOAS-B,
TOAS-C, TOAS-D, TOAS-E, and TOAS-F).
3.2 RAS Binder Extraction and Recovery
The following two methods to extract and recover RAS binder were used with
trichloroethylene as the solvent.
Tex-210-F - Determining Asphalt Content of Bituminous Mixtures by
Extraction: Part I-Centrifuge Extraction Method Using Chlorinated Solvent
ASTM D5404 - Standard Practice for Recovery of Asphalt from Solution Using
the Rotary Evaporator
The authors validated these two methods, first, through comparing both rheological
properties (measured by DSR and BBR) and chemical components (measured by Fourier
transform infrared spectroscopy) of one original shingle binder before and after the
extraction and recovery process (Zhou et al., 2012). Then, the ten RAS binders were
extracted and recovered following the validated extraction and recovery methods. Upon
completion of the extraction process, there was some difficulty in pouring out the
recovered TOAS binders, because they were so stiff, and just did not flow out of the
beaker even at 165C. In one case, the oven temperature was raised to 200C to
successfully drain out the TOAS binder.
3.3 RAS Binder Characterization
Both DSR and BBR were used to determine the PG grade of the ten extracted/recovered
RAS binders. Difficulties were encountered in grading the RAS binders using the BBR.
There are two criteria (S and m) for determining the low-temperature grade of asphalt
binders. RAS binders easily met the S (<300 MPa) criteria, but the measured m values
were always less than 0.3. The authors even tried to perform the BBR test at higher
372
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

temperatures (i.e., 18C and even 24C), but the measured m values were still less than
0.3, and in some cases, the beam deformation reached the limit of the BBR machine
within a very short of period of time. The reason for such a small m value is that RAS
binders have much less capability to relax. Therefore, no reliable results from BBR test
were obtained for any of the ten recovered RAS binders. Alternative tests (such as the
Asphalt Binder Cracking Device test) should be explored.
To measure the high-temperature grades of those extremely stiff binders, a high-
temperature DSR was purchased by TTI. Nine of the ten RAS binders were successfully
graded following the Superpave PG system. The high-temperature grade of one TOAS
binder was beyond the upper limit of the purchased DSR (200C), thus extrapolation
was used to estimate its high-temperature grade. For each extracted/recovered RAS
binder, both original and rolling thin-film oven (RTFO) aged residue were evaluated.
The high-temperature grades of the ten RAS binders are shown in Figure 1. At least two
observations can be made from Figure 1:
TOAS binders, with an average high-temperature grade of 178C, are much stiffer
than MWAS binders, with an average high-temperature grade of 131C.
MWAS binder exhibited smaller variation in the high-temperature grade, compared
with TOAS binder, varying from 159C to 214C.
These two observations may indicate that MWAS is different from TOAS so it is
necessary and important to differentiate MWAS from TOAS when using shingles in
asphalt mixes. For example, DOTs may desire to allow smaller amount of TOAS in the
specification when compared with MWAS.
In summary, the RAS binders are very stiff. Thus, it is critical to investigate the
impact of these extremely stiff binders on rheological properties of the combined binder
after blending with virgin binders, which is discussed next.
230
Te ar-offs
210 Ave rage PG = 178
CV= 10.8%
Manufacture Waste
190
Ave rage PG = 131
High Temp. Grade ( C )
170 CV= 5.5%
150
130
110
90
70
50
MWAS-A MWAS-B MWAS-C MWAS-D TOAS-A TOAS-B TOAS-C TOAS-D TOAS-E TOAS-F
Figure 1. High PG temperatures of both manufacture waste and tear-off shingles.
373
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

3.4 Virgin/RAS Binder Blending
Compared to virgin/RAP binder blending, very little work has been reported on
virgin/RAS binder blending in the literature. Nevertheless, AASHTO PP 53, Standard
Practice for Design Considerations When Using Reclaimed Asphalt Shingle (RAS) in
New Hot Mix Asphalt (HMA), recommends that the linear blending used for virgin/RAP
binders also be used with virgin/RAS binders. This study selected three virgin binders
(PG 64-22-A, PG 64-22-B, and PG 64-28) and four RAS binders (TOAS-A, TOAS-E,
MWAS-A, and MWAS-C) for investigating virgin/RAS binder blending. With these
binders, a total of four combinations of virgin/RAS binders, as listed below, were tested.
These four combinations were used in field test pavements that will be described later.
Virgin Binder: PG 64-22-A and RAS Binder: TOAS-E
Virgin Binder: PG 64-28 and RAS Binder: TOAS-A
Virgin Binder: PG 64-22-B and RAS Binder: MWAS-A
Virgin Binder: PG 64-22-B and RAS Binder: MWAS-C
For each combination, different percentages of virgin binder and RAS binder were
blended and then evaluated through DSR and BBR testing in terms of the high and low
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
PG temperatures. Test results for these four combinations are presented in Figures 2, 3,
4, and 5, respectively. The following observations are made from these figures.
Generally, virgin and RAS binder blending is non-linear.
For practical application, the linear blending chart can still be used if the RAS
binder percentage is below 30%. Below 30% RAS binder, not only is the linear
blending chart applicable, but the regular DSR and BBR can also be used to
evaluate the high and low ends of the PG grade of the blended binder.
Increasing RAS binder content will improve the high-temperature grade of virgin
binder and warm up its low-temperature grade, which is good for rutting resistance
but causes concerns about cracking resistance of the blended binder. Adding 20%
RAS binder can make a PG xx-22 binder become a PG xx-16 (or even a PG xx-10
[Figure 2]) binder after blending. Additionally, use of a PG xx-28 virgin binder to
obtain a PG xx-22 when 20% RAS binder is added is achievable. Note that 20%
RAS binder corresponds to 5% RAS by mass of the total mix with the assumptions
that the optimum asphalt content of a RAS mix is 5% and RAS contains 20%
asphalt binder.
The impact of MWAS binder on the high and low PG temperatures of virgin binder
is different from that of TOAS binder. Compared to the TOAS binders (Figures 2
and 3), the MWAS binders (Figures 4 and 5) have less impact on PG temperatures
374

of virgin binders, which makes sense, since MWAS binders are significantly softer
than TOAS binders (see Figure 1).
Binder Blending: PG64-22-A/TOAS-E Binder
180
160
140
120
100
80
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
RAS Binder (%)
Binder Blending: PG64-22-A/TOAS-E Binder

0
-5
Low Temp. Grade ( C )
-1 0
-1 5
-2 0
-2 5
0 10 20 30 40 50 60 70 80 90 100
RAS Binder (%)
Figure 2. Binder blending between PG 64-22-A and TOAS-E binder.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
375

Binder Blending: PG64-28 and TOAS-A Binder

180
160
140
High Temp. Grade ( C ) 120
100
80
60
40
20
0
0 10 20 30 40 50 60 70 80 90 100
RAS Binder (%)
Binder Blending: PG64-28 and TOAS-A Binder

0
-5
-1 0
-1 5
-2 0
-2 5
-3 0
-3 5
0 10 20 30 40 50 60 70 80 90 100
RAS Binder (%)
Figure 3. Binder blending between PG 64-28 and TOAS-A binder.
376
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Binder Blending: PG64-22-B/MWAS-C Binder

140
120
High Te m p. Grade (C)

100
80
60
40
20
0
0 20 40 60 80 100
RAS Binde r (% )
Binder Blending: PG64-22-B/MWAS-C Binder

0
-5
L ow Te m p. Grade (C)
-1 0
-1 5
-2 0
-2 5
-3 0
0 20 40 60 80 100
RAS Binde r (% )
Figure 4. Binder blending between PG 64-22-B and MWAS-C binder.
377
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Binder Blending: PG64-22-B and MWAS-A Binder

160
140

120
100
80
60
40
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
20
0
0 20 40 60 80 100
RAS Binder (%)
Binder Blending: PG64-22-B and MWAS-A Binder

0
-5
-1 0
-1 5
-2 0
-2 5
0 20 40 60 80 100
RAS Binder (%)
Figure 5. Binder blending between PG 64-22-B and MWAS-A binder.
In addition to the influence of RAS binders on blended binders, RAS itself has
significant impact on the engineering properties of HMA mixes containing RAS, which
is discussed in the next section.
4. Impact of RAS Content on Laboratory Mix Engineering Properties
Incorporation of RAS materials into asphalt mixes stiffens them and thus improves the
resistance to rutting but may jeopardize resistance to cracking. Additionally, TOAS and
MWAS may have different impacts; since these two are so different in terms of high-
temperature grade (see Figure 1). Therefore, it is necessary to thoroughly investigate the
impact of RAS content and RAS type (TOAS/MWAS) on mixture engineering
properties in terms of dynamic modulus, rutting, and cracking. In this study, two RAS
types (TOAS-E and MWAS-C) and three RAS percentages (0%, 3%, and 5%) were
considered. Note that TOAS-E is very similar to MWAS-C in terms of RAS aggregate
gradation and RAS binder content, and the main difference between them was RAS
378

binder high-temperature PG: TOAS-E = 166C vs. MWAS-C = 122C. A total of six
mixtures with the same aggregates and similar gradations, as listed in Table 4, were
evaluated using the dynamic modulus test, Hamburg wheel tracking test (HWTT), and
Texas Overlay test (OT). A 0% RAS/PG 64-22 mix was the control mix. A 0% RAS/PG
70-22 mix was added to compare with the mixes containing 5% RAS/PG 64-22, since
many DOTs allow a one grade dump of the virgin binder when 5% RAS is used.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Table 4. Ignition test results: MWAS-A.

RAS RAS Percentage1 and Virgin Binder
Type 0% RAS/PG 70-22 0% RAS/PG 64-22 3% RAS/PG 64-22 5% RAS/PG 64-22
TOAS-E X X
X X
MWAS-C X X
Note: 1- The percentage here is referred as to RAS rather than RAS binder.
The following steps were followed:

Step 1: Fix the RAS content (i.e., 5%) and adjust virgin aggregate percentage to
make the total aggregate gradation for each RAS mix as close to each other as
possible (Figure 6).
Figure 6. Gradations of HMA mixes with varying RAS contents.

Step 2: Design the RAS mixes and select an optimum asphalt content (OAC)
following standard TxDOT mix design procedure (Tex-204-F) for dense graded
mixes, which are widely used in Texas (75% of all the HMA used in Texas).
Step 3: Evaluate the dynamic modulus (or stiffness), rutting/moisture resistance, and
cracking resistance of mixes with varying RAS contents at its specific OAC.
379

o Dynamic modulus of each mix was measured following AASHTO TP 79,

Determining the Dynamic Modulus and Flow Number for Hot Mix Asphalt
(HMA) Using the Asphalt Mixture Performance Tester (AMPT). The 4 in.
(100 mm) diameter by 6 in. (150 mm) tall specimens with 7 0.5% air voids
were fabricated in accordance with AASHTO PP 60, Preparation of
Cylindrical Performance Test Specimens Using the Superpave Gyratory
Compactor (SGC). For each mix, two replicates were tested, and the average
value is reported here.
o Rutting/moisture resistance of RAS mixes was characterized using the
Hamburg Wheel Tracking Test (Tex-242-F). The specimen size of the HWTT
is 6 in. (150 mm) diameter by 2.5 in. (62 mm) height and its target air voids
content is 71 percent. The HWTT is conducted in a water bath at a constant
temperature of 122F (50C). The specimens are tested under a rolling 1.85 in.
(47 mm) wide steel wheel using a 158-lb (705-N) force. An average rut depth,
measured at several locations including the center of the wheel travel path, is
reported at end of the test.
o Cracking resistance of RAS mixes was determined using the Texas Overlay
Tester (Tex-248-F). The standard specimen size for OT is 6 in. (150 mm) long
by 3 in. (75 mm) wide by 1.5 in. (38 mm) high, and its target air voids content
is 71% after cutting. The OT is conducted in a displacement controlled mode
with a maximum opening displacement of 0.025 in. (0.63 mm) at test
temperature of 77F (25C). The number of cycles to failure (failure defined as
93% reduction of the cyclic maximum load from the one measured at the first
load cycle) is used as an indicator for cracking resistance. Note that five
replicates of OT specimens were tested for each mix, and the average value of
OT cycles was used for comparison. The correlation between OT result and
field cracking performance has been well documented (Zhou and Scullion,
2005, Zhou et al., 2007) and the OT has been used for evaluating both reflective
and fatigue cracking by different researchers (Bennert et al., 2008, Zhou et al.,
2009, Zhou et al., 2010, Mogawer et al., 2011).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
4.1. Impact of RAS on OAC of HMA mixes
Three mix designs were performed for mixes with TOAS-E: 0% TOAS-E/PG 64-22, 3%
TOAS-E/PG 64-22, and 5% TOAS-E/PG 64-22. The OAC for each mix was determined
based on 97% density (or 3% air voids) and is presented in Table 5. Mix designs for
mixes containing 3% MWAS-C/PG 64-22, and 5% MWAS-C/PG 64-22 were conducted
as well. It was found that the OACs for the mixes containing 3% and 5% MWAS-C are
very close to those of the mixes containing 3% and 5% TOAS-E. Thus, the same OAC
was selected for the mixes with the same amount of RAS, regardless of TOAS-E or
380

MWAS-C. Additionally, the OAC of the mix containing 0% RAS/PG 70-22 was kept
the same as that of the mix with 0% RAS/PG 64-22, since these two mixes have exactly
the same raw aggregates and gradation and the influence of binder type is considered
through mixing and compaction temperatures. In summary, Table 5 lists the OAC of
each RAS mix evaluated in this section.
It can be seen clearly that the OAC increases with the higher RAS content. The
reason for this is that the increasing RAS content increases the composite PG grade of
the blended RAS/virgin binder. (Note that it may also be due to incomplete use of all the
binder in the RAS as binder in the mix. That is, the RAS binder is extracted and counted
as binder, but it is not really active as binder in the mix). Therefore, with the higher
composite PG grade, the mixing and compaction temperatures should be increased for
high RAS mixes. When the mixing and compaction temperatures are kept the same for
each RAS mix, higher RAS mixes will need more asphalt binder to achieve the same
density. The higher OAC somehow offsets the impact of higher RAS content on
engineering properties of RAS mixes, as discussed next.
Table 5. OAC of each RAS mix.

RAS Optimum Asphalt Content (%)
Type 0% RAS/PG 70-22 0% RAS/PG 64-22 3% RAS/PG 64-22 5% RAS/PG 64-22
TOAS-E
4 .7 4 .7 4 .9 5 .2
MWAS-C
4.2. Impact of RAS on Dynamic Modulus
Dynamic moduli of each mix measured at different temperatures are shifted following
the time-temperature superposition principle and presented in master curve format.
Figure 7 shows master curves for all six mixes. Overall, the dynamic modulus master
curves of these six mixes are similar, except that the 0% RAS/PG 70-22 mix has a little
bit higher moduli. Higher RAS content does not always mean higher stiffness.
Additionally, in terms of dynamic moduli of HMA mixes, the TOAS-E and MWAS-C
have very similar impact.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
381

HMA Mixes with TOAS-E

100000
5% RAS/PG64-22 3% RAS/PG64-22
0% RAS/PG64-22 0%RAS/PG70-22
10000
Dynamic Modulus (MPa)
1000
100
10
1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2 1E+3
Reduced Frequency (Hz)
HMA Mixes with MWAS-C

100000
5%RAS/PG64-22 3%RAS/PG64-22
0%RAS/PG64-22 0%RAS/PG70-22
10000
Dynamic Modulus (MPa)
1000
100
10
1E-5 1E-4 1E-3 1E-2 1E-1 1E+0 1E+1 1E+2 1E+3
Reduced Frequency (Hz)
Figure 7. Impact of RAS on dfynamic moduli of HMA mixes.
4.3. Impact of RAS on Rutting Resistance

Figure 8 presents the HWTT test results. For those mixes containing PG 64-22, adding
RAS improved rutting/moisture damage, whether TOAS-E or MWAS-C. However, such
improvement is not significant enough to match the impact of the 0% RAS/PG 70-22
binder. This observation implies that the degree of blending between PG 64-22 virgin
binder and RAS binder in the mixing and curing process before compaction is different
from the blending of pure PG 64-22 binder and extracted/recovered RAS binders (see
Figures 2 and 4). One may take the mix with 5% RAS (TOAS-E)/PG 64-22 as an
example. If complete (or 100%) blending occurs, like those shown Figure 2, the blended
binder becomes PG 80-15, which is far stiffer than PG 70-22 binder, and accordingly,
the mix containing 5% TOAS-E should exhibit less rutting. Surely, complete (100%)
382

--`,`,,,``,,```

blending in the mixing, curing, and compaction processes did not occur, since the high-
temperature grade of the TOAS-E binder is 166C.
Additionally, Figure 8 shows that the mixes containing TOAS-E have lower rut
depths than those containing MWAS-C, which is reasonable, since the TOAS-E binder
is stiffer than the MWAS-C binder (see Figure 1).
Impact of RAS Content on Rutting/Moisture Damage

0%RAS/PG70-22 0%RAS/PG64-22
14
3%RAS/PG64-22 5%RAS/PG64-22
HWTT Rut Depth @10000
12
10
Passes (mm)
4
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0
TOAS-E MWAS-C
Figure 8. Impact of RAS on rutting/moisture damage of HMA mixes.
4.4. Impact of RAS on Cracking Resistance
Figure 9 shows the impact of RAS on cracking resistance of HMA mixes. For those
mixes containing PG 64-22, the use of RAS decreases cracking resistance of the HMA
mixes, whether TOAS-E or MWAS-C. The TOAS-E has a worse effect than the
MWAS-C, which implies that blending between the virgin PG 64-22 and the TOAS-E
binder (or the MWAS-C binder) actually occurred. If there is no blending at all, both of
the mixes containing TOAS-E should have the same OT cycles as those containing
MWAS-C. Although it is unknown how much blending actually occurred, one known
thing is that the mixes containing MWAS-C had much better cracking resistance than
those containing TOAS-E. This finding is consistent with the binder blending charts
presented previously, in which the impact of the MWAS-C on the low-temperature grade
of blended binder is much less than that of the TOAS-E (see low-temperature grades of
Figures 2b and 4b).
Compared with the 0% RAS/PG 70-22 mix, those PG 64-22 mixes with either 3% or
5% RAS have worse cracking resistance. This is not a surprise because, if blending
between virgin PG 64-22 and RAS binder occurs, the low-temperature grade would
become warmer (i.e., PG xx-15 shown in Figure 2); Conversely, if there is no or very
383

limited blending, the total effective asphalt binder amounts within those RAS mixes
would be much less than the 0% RAS mixes with PG 70-22 or PG 64-22 binder.
Regardless, the cracking resistance is a big concern for RAS mixes.
Impact of RAS Content on Cracking

0%RAS/PG70-22 0%RAS/PG64-22
50
3%RAS/PG64-22 5%RAS/PG64-22
40
OT Cycles
30
20
10
0
TOAS-E MWAS-C
Figure 9. Impact of RAS on cracking resistance of HMA mixes.
4.5. Summary and Discussion
Findings and discussions from the results presented above are provided below:
RAS has no significant influence on dynamic moduli of HMA mixtures, but
improves their rutting resistance and moisture damage. RAS generally increases
OAC of HMA mixes. However, RAS mixes exhibit very poor cracking resistance,
compared to the mixes containing only PG 64-22 or PG 70-22, even though the
RAS mixes have higher OAC. Therefore, cracking resistance is a big concern for the
RAS mixes.
Impact of the TOAS-E is different from that of the MWAS-C in terms of cracking
resistance, as the MWAS-C mixes offer much better cracking resistance. This
finding is consistent with results from binder blending discussed previously.
Both the HWTT and OT results indicated that some blending occurs between virgin
binder and RAS binder during the HMA mixing and curing (or short-term aging)
processes. But the blending is not 100%, and this is no surprise, since the high-
temperature grades of the TOAS-E and the MWAS-C binders are 166C and 122C,
respectively. It is well known that the mixing temperature for a PG 64-22 binder is
around 143C (290F). Extremely, impractically high temperature is required in
order to make the RAS binder flow and comingle with virgin binder, whether using
TOAS or MWAS. Although the degree of blending is still unknown, one known
384
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

fact is that the use of RAS increases cracking of HMA mixes, and some remedies
need to be explored, as discussed in next section.
5. Laboratory Evaluation of Approaches for Improving Cracking Resistance of

RAS Mixes
RAS has significant impact on cracking resistance and, consequently, on the durability
of HMA mixes. Some approaches need to be pursued to balance the performance of
RAS mixes. In general, there are at least four potential approaches:
Reduce RAS content (i.e., from 5% to 3%),
Rejuvenate RAS binder in the mix design process,
Use soft virgin binders, especially on the low-temperature grade (i.e., PG XX-28,
PG XX-34), and
Decrease design air voids.
Naturally, the first choice is to use less RAS. However, the previous results shown in
Figure 9 indicated that reducing RAS from 5% to 3% does not have significant
improvement on cracking resistance. Further reducing RAS usage quantities will
minimize the economic and environmental benefit. The second choice is to rejuvenate
RAS binder by using a rejuvenating agent. This may sound like a good idea and should
improve cracking resistance of RAS mixes (Tran et al., 2012), but there are practical and
technical issues when applied to normal asphalt plant operations. Apparently, more
research is needed in the area of rejuvenating agents and their practical application.
Thus, this paper focused on the last two approaches: using soft binder and decreasing
design air voids. The researchers evaluated the effectiveness of these two approaches in
improving cracking resistance of RAS mixes.
5.1. Use of Soft Binders
The same 5% RAS/PG 64-22 mixes containing TOAS-E and MWAS-C that were
previously designed were used as control mixes. Two soft binders were selected: PG
64-28 and PG 64-34. A total of six mixes (two RAS using three virgin binders), listed in
Table 6, were evaluated using the dynamic modulus test (AASHTO TP 79), HWTT
(Tex-242-F), and OT (Tex-248-F). The same 5.2% OAC was used for all six mixes,
since the purpose is to investigate the influence of soft binders. Figures 10, 11, and 12,
show the test results.
385
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 6. RAS mixes with soft virgin binders.

RAS 5%RAS/PG 64-22 5%RAS/PG 64-28 5%RAS/PG 64-34
TOAS-E X X X
MWAS-C X X X
It can be seen from Figure 10 that RAS mixes with softer binders have slightly lower
moduli, but the differences among these six mixes is very small. Meanwhile, compared
with the 5% RAS/PG 64-22 mix, the use of softer binders improved rutting/moisture
damage, as indicated in Figure 11. One reason for the improvement is that both PG 64-
28 and PG 64-34 are polymer-modified binders. As expected, the mixes containing
MWAS-C have deeper rut depth than those containing TOAS-E. Figure 12 clearly
indicates that using soft (modified) binders is an effective method to improve cracking
resistance of RAS mixes. For the cases presented here, one grade (-6C) lower can triple
the OT cycles of RAS mixes. Additionally, the mixes with the MWAS-C always have
better cracking life than those with the TOAS-E. In summary, the use of soft (modified)
binders has little impact on dynamic moduli of RAS mixes, whereas, it can improve both
rutting and cracking resistance of RAS mixes, particularly cracking resistance.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
386

HMA Mixes with 5% TOAS-E

100000
PG6 4 -2 2 PG6 4 -2 8 PG6 4 -3 4
10000
Dynam ic Modulus (MPa)
1000
100
10
1 E-5 1 E-4 1 E-3 1 E-2 1 E-1 1 E+ 0 1 E+ 1 1 E+ 2 1 E+ 3
Re duce d Fr e que ncy (Hz)
HMA Mixes with 5% MWAS-C

100000
PG6 4 -2 2 PG6 4 -2 8 PG6 4 -3 4
10000
Dynam ic Modulus (MPa)
1000
100
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
10
1 E-5 1 E-4 1 E-3 1 E-2 1 E-1 1 E+ 0 1 E+ 1 1 E+ 2 1 E+ 3
Re duce d Fr e que ncy (Hz)
Figure 10. Impact of soft binders on dynamic modulus of 5% RAS mixes.
387

Impact of Soft Binder on Rutting

12
PG64-22 PG64-28 PG64-34
HWTT Rut Depth @10000

10
Passes (mm) 8
0
TOAS-E MWAS-C
Figure 11. Impact of soft binders on rutting/moisture damage of 5% RAS mixes.
Impact of Soft Binder on Cracking

350
PG64-22 PG64-28 PG64-34
300
250
OT Cycles
200
150
100
50
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
TOAS-E MWAS-C
Figure 12. Impact of soft binder on cracking resistance of 5% RAS mixes.
5.2. Decreasing Design Air Voids
Another simple way to improve cracking resistance of RAS mixes is to add more virgin
binder with a lower low-temperature grade into the mixes through decreasing design air
voids when selecting OAC. Currently, the TxDOT design density for selecting OAC of
RAP/RAS mixes is 97%. To avoid bleeding problems, the maximum design density
should be less than 98%. Again, the same 5% RAS/PG 64-22 mixes with the TOAS-E
and MWAS-C, that were previously designed, were used here. Two design densities,
97% and 97.7%, were used, and accordingly, the corresponding OACs were 5.2% and
5.7%, respectively. Only the HWTT (Tex-242-F), and OT (Tex-248-F) testing was
performed. The dynamic modulus test was omitted, since the previous results showed
little difference among different RAS mixes. Figure 13 shows the test results.
388

It can be seen from Figure 13 that the higher OAC corresponding to decreased design
air voids significantly improves cracking resistance. Conversely, the higher OAC makes
the RAS mixtures more susceptible to potential rutting. Therefore, one must exercise
caution when improving cracking resistance of RAS mixes through decreasing design air
voids.
Impact of Decreasing Design Air Voids on

Rutting /Moisture Damage
16
HW TT Rut Depth @10000 Passes
5.2% AC w ith 3.0% air voids

14 5.7% AC w ith 2.3% air voids
12
10
(mm)
0
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
TO A S-E MW A S-C
Impact of Decreasing Design Air Voids on Cracking

90
80
70
60
OT C ycles
50
40
30
20
10
0
TO A S-E MW A S-C
Figure 13. Impact of decreasing design air voids on rutting and cracking resistance of
RAS mixes.
5.3. Discussion
The OT results presented above clearly indicated that both using soft binder and
decreasing design air voids can improve cracking resistance of RAS mixes. When
considering rutting of RAS mixes, using soft binder is superior to decreasing design air
voids (see Figures 11 and 13). In order to validate these laboratory test results and these
389

two approaches, field test pavements were constructed in Texas, which are discussed in
next section.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
6. Field Validation of Approaches for Improving Cracking Performance of RAS
Mixes
Field test pavements were constructed to validate the two approaches for improving
cracking performance of RAS mixes: using soft binder and decreasing design air voids.
A total of six asphalt overlay test pavements using RAS mixes were built in two
different areas of Texas. Field performance, to date, of these test pavements is presented
below:
6.1. RAS Test Pavements on US 87, Amarillo, Texas
Two 3 in. thick asphalt overlay test pavements were constructed end to end in the same
lane and travelling direction on US 87, Amarillo, Texas, in late October 2010. The main
objective of these two test pavements was to validate the effectiveness of decreasing
design air voids on improving cracking resistance of RAS mixes. The RAS mixes used
on the two test pavements are exactly the same (aggregates, gradation, virgin binder, and
RAS) except for the OAC; OAC for the control section was 4.6% while the other is
5.2%. Amarillos climate is a temperate, semi-arid climate characterized by numerous
freeze-thaw cycles and occasional blizzards during the winter season. Average daily
high temperatures for Amarillo range from 48F (9C) in January to 92F (33C) in July.
US 87 in Amarillo has medium traffic with around 5 million ESALs in 20 years. The
existing asphalt pavement exhibited severe transverse cracking. Cold weather and severe
existing pavement cracking plus high traffic make these two pavements a good case
study to rapidly validate the effectiveness of decreasing design air voids on improving
cracking resistance of RAS mixes.
After completion of construction of these two RAS test pavements, three field
surveys were conducted on Apr. 5, 2011, Dec. 15, 2011, and May 30, 2012. So far, no
rutting has been observed, but reflective cracking occurred in both test pavements
(Figure 14). The development history of the observed reflective cracking is shown in
Figure 15. Prior to placing the overlay, the number of pre-existing cracks in each
pavement was documented and mapped. The reflective cracking rate is therefore defined
as the ratio of the number of observed reflective cracks to the original number of cracks
before the 3 in. overlay. Apparently, decreasing design air voids significantly improved
reflective cracking performance of the RAS mix on US 87, which is clearly shown in
Figure 15.
390

Section with 96.5% design density Section with 97.7% design density
Figure 14. Observed reflective cracking of RAS test pavements on US 87, Amarillo,
Texas.
Reflective Cracking Development of RAS Test Sections

50
Highe r O A C-9 7 .7 %De nsity

40
Re fle ct ive Cracking Rat e (% )
O A C-9 6 .5 % De nsity (c ontr ol se c tion)
30
20
10
0
0 5 10 15 20
Mont hs ope n t o t r affic
Figure 15. Reflective cracking development history of RAS test pavements on US 87,
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Amarillo, Texas.
391

6.2. Field Test Pavements on FM 973, Austin, Texas
A comprehensive series of experimental asphalt overlay test pavements were constructed

on FM 973 near the Austin Bergstrom International Airport. Compared to the cold
weather in Amarillo, the weather in the Austin area is fairly warm. This roadway
experiences very heavy truck traffic, as it carries traffic from several aggregate quarries
and concrete batch plants. A total of nine test pavements were built between December
2011 and January 2012. One objective of these test pavements was to evaluate the
effectiveness of using soft binder on improving cracking resistance of RAP/RAS mixes.
Table 7 lists four HMA test sections related to this study. In addition to the virgin PG
70-22 mix (control mix), three other mixes were evaluated. Prior to the 2 in. asphalt
overlay, the overall pavement condition was good. Some areas exhibited longitudinal
cracking along the wheel paths. After completion of construction, these four RAP/RAS
test pavements were trafficked for six months. One survey was conducted in July 2012,
and neither rutting nor cracking was observed on any test section. Figure 16 shows
current conditions of Pavements 3 and 6. Apparently, more time is needed for these test
sections to show differences in rutting or cracking. TTI researchers will continue to
monitor performance of these pavements.
Table 7. Four test pavements on FM 973, Austin, Texas.
RAP (%)
Virgin Binder RAS
Mix
Grade (%)
0 15
PG 70-22 0 Section 1
3 Section 3
HM A PG 64-22
5 Section 4
PG 58-28 3 Section 6
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
392

Section 3: Section 6:
3% RAS/15% RAP/PG 64-22 3% RAS/15% RAP/PG 58-28
Figure 16. Conditions of RAP/RAS test pavements 3 and 6 on FM 973, Austin, Texas, in
July 2012.
This paper presents a comprehensive study on HMA mixes containing RAS, including
RAS consistency, RAS binder characterization and blending charts for virgin/RAS
binders; impact of RAS content on OAC and engineering properties of RAS mixes; and
approaches for improving cracking resistance of RAS mixes. Six RAS field test
pavements were constructed to validate the approaches for improving cracking resistance
of RAS mixes. Based on the research presented in this paper, the following conclusions
are offered:
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
All ten processed RAS materials tested in this study, whether TOAS or MWAS, met
the specification and are consistent in terms of gradation. TOAS had a higher binder
content than MWAS. MWAS had a consistent 20% binder content, while TOAS
yielded binder contents, ranging from 23% to 28%. Overall, RAS variability in
terms of binder content and gradation are low for both MWAS and TOAS. MWAS
exhibited slightly lower variability.
RAS binders are very stiff. TOAS binders, with an average high-temperature grade
of 178C, are much stiffer than MWAS binders, which had an average high
temperature grade of 131C. MWAS has smaller variation in the high-temperature
grades, compared to the TOAS, which varied from 159C to 214C.
393

Generally, virgin and RAS binder blending is non-linear. For practical application,
the linear blending chart can still be used if the RAS binder percentage is below
30%. Below 30% RAS binder, not only is the linear blending chart applicable, but
the regular DSR and BBR can be used to evaluate the high and low ends of the PG
grade of the blended binder.
Compared with the TOAS binders, the MWAS binders have less impact on PG
temperatures of virgin binders, which is reasonable, since MWAS binders are much
softer than TOAS binders. Thus, it is important to consider differentiating MWAS
from TOAS when used in asphalt mixes.
RAS has no significant influence on dynamic moduli of HMA mixtures, but
improves their rutting resistance.
Adding RAS generally increases OAC of HMA mixes, depending on compaction
temperature, curing time, and other factors.
RAS mixes exhibited very poor cracking resistance, compared with the 0% RAS
mixes containing PG 64-22 or PG 70-22, even though the RAS mixes have higher
OAC. Therefore, cracking resistance is of concern for the RAS mixes.
Two approaches for improving cracking resistance of RAS mixes were explored in
the laboratory. Test results clearly indicated that both using soft binder and
decreasing design air voids can improve cracking resistance of RAS mixes. When
considering rutting of RAS mixes, using soft (modified) binder is superior to
decreasing design air voids. Six field test pavements were built to validate these
two approaches. The effectiveness of increasing design density was confirmed
through field test pavements on US 87, Amarillo, Texas; the soft binder test
pavements on FM 973, in Austin, are being monitored.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
When using the softer binder and low air void approaches, one should be aware that,
if the RAS is not well blended into the mixture or, if segregation occurs during
mixing and/or placement, there will be spots on the pavement with "softer" mix,
which may fail due to rutting.
The findings from this study need to be further validated with field performance data
from different test pavements under various scenarios. More work is also needed to
investigate the performance of RAS/RAP mixes produced at warm mix temperatures.
Additionally, more research should be performed in two more areas:
Degree of blending between virgin binder and RAS binder in laboratory mixing and
compaction as well as plant production and, possibly, even after field compaction:
the degree of blending has significant influence on volumetric properties of asphalt
mixes and, thus the resulting OAC.
394

Intermediate temperature binder property: This study focused on high- and low-
temperature binder properties. Fatigue cracking is influenced more by the
intermediate temperature binder property.
8. Acknowledgements
This study was made possible by the Texas Department of Transportation in cooperation
with the Federal Highway Administration. The authors thank the pavement engineers of
Amarillo, Austin, and Fort Worth Districts of TxDOT for their support. Recognitions
are due to all TTI McNew lab staff.
9. Disclaimer
The contents and opinions of this paper reflect the views of the authors, who are solely
responsible for the facts and the accuracy of the data presented herein. The contents of
this paper do not necessarily reflect the official views or the policies of any agencies.
10. References
Bennert, T. and A. Maher, Field and Laboratory Evaluation of a Reflective Crack

Interlayer in New Jersey, Transportation Research Record, No. 2084, 2008, pp 112
123
Brock, J. D., From Roofing Shingles to Roads, Technical Paper T-120, Astec
Industries, Inc., Chattanooga, Tennessee, 2007.
Button, J.W., D. Williams, and J. Scherocman, Roofing Shingles and Toner in Asphalt
Mixtures, Research Report 1344-2F, Texas Transportation Institute, Texas A&M
University, College Station, Texas, 1996.
Dykes, J., Recycled Asphalt Shingle Use in Hot Mix Asphalt, presentation at seminar
program panel session TH-26Recycled Asphalt Shingle Use in Hot Mix Asphalt,
ConExpo-Con/Agg Show and Exposition, Las Vegas, Nevada, 2002.
Epps, J.A. and G.L. Paulsen, Use of Roofing Wastes in Asphalt Paving Mixtures
Economic Considerations, Report 709-2, Center for Construction Materials
Research, University of Nevada-Reno, 1986.
Foo, K. Y., D. I. Hanson, and T. A. Lynn, Evaluation of Roofing Shingles in Hot Mix
Asphalt, ASCE Journal of Materials in Civil Engineering, Vol. 11, No. 1, 1999.
Grzybowski, K.F., Recycled Asphalt Roofing MaterialsA Multi-Functional, Low-
Cost Hot Mix Asphalt Pavement Additive, Use of Waste Materials in HMA
395
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Pavement, ASTM Special Publication 1193, American Society for Testing and
Materials, Philadelphia, Pennsylvania, 1993.
Janisch, D.W. and C.M. Turgeon, Minnesotas Experience with Scrap Shingles in
Bituminous Pavements, Report MN/PR-96/34, Office of Materials and Road
Research, Minnesota Department of Transportation, Maplewood, Minnesota, 1996.
Johnson, E., G. Johnson, S. Dai, D. Linell, J. McGraw, and M. Watson, Incorporation of
Recycled Asphalt Shingles in Hot Mixed Asphalt Pavement Mixtures, Minnesota
Department of Transportation, Maplewood, MN, February 2010.
Lum, P., Roofing Shingles in HMA, 2006 National Conference, Materials and Energy
Conservation for HMA, Indianapolis, Indiana, 2006.
McGraw, J., A. Zofka, D. Krivit, J. Schroer, R. Olson, and M. Marasteanu, Recycled
Asphalt Shingles in Hot Mix Asphalt, Journal of Association of Asphalt Paving
Technologists, Vol. 76, San Antonio, TX, March 2007.
Mogawer, W., A. Austerman, M. E. Kutay, F. Zhou, Evaluation of Binder Elastic
Recovery on HMA Fatigue Cracking using Continuum Damage and Overlay Test
Based Analyses, International Journal of Road Materials and Pavement Design,
Vol. 12, No. 2, 2011, pp345376.
Newcomb, D., M. Stroup-Gardiner, B. Weikle and A. Drescher, Properties of Dense-
Graded and Stone Mastic Asphalt Mixtures Containing Roofing Shingles, Use of
Waste Materials in Hot Mix Asphalt, ASTM Special Technical Publication 1193,
ASTM, Conshohocken, Pennsylvania, 1993.
Paulsen, G., M. Stroup-Gardiner, and J.A. Epps, Roofing Waste in Asphalt Paving
Mixtures Economic Considerations, Report No. 6-331-709-1, Center for
Construction Materials Research, University of Nevada-Reno, 1986.
Schroer, J., Asphalt Shingles in HMA: Missouri DOT Experience, North Central
Users and Producers Group (NCAUPG) Conference, Minneapolis, Minnesota, 2007.
Shepherd, P.B., T.J. Powers, J. Hardy, R. Maloof, and C. Patenaude, Energy
Conservation through Recycling of Factory Asphalt Roofing Waste, prepared for
the U.S. Department of Energy, Washington, D.C., 1989.
Tran, N., A. Taylor, and J. R. Willis, Effect of Rejuvenator on Performance Properties of
HMA Mixtures with High RAP and RAS Contents, NCAT Report 12-05, National
Center for Asphalt Technology, Auburn, AL, 2012.
Williams R. C., Cascione A., Haugen D. S., Buttlar W. G., Bentsen R. A., and Behake J.,
Characterization of Hot Mix Asphalt Containing Post-Consumer Recycled Asphalt
Shingles and Fractionated Reclaimed Asphalt Pavement, Iowa State University,
Department of Civil, Construction and Environmental Engineering, Ames, IA,
January 2011.
396
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Zhou F. and Scullion T., Overlay tester: a simple performance test for thermal reflective
cracking, Journal of the Association of Asphalt Paving Technologists, Vol. 74, 2005,
pp 443484.
Zhou F., Hu S., Scullion T., Chen D., The Overlay Tester: A Simple Performance Test
for Fatigue Cracking, Transportation Research Record, No. 2001, 2007, pp 18.
Zhou F., Hu S., Scullion T., Mikhail M., Walubita L. F., A Balanced HMA Mix Design
Procedure for Overlays, Journal of the Association of Asphalt Paving Technologists,
Vol. 76, 2007, pp. 823850.
Zhou F., Hu S., Hu X., Scullion T., Mechanistic-Empirical Asphalt Overlay Thickness
Design and Analysis System, Research Report FHWA/TX-09/0-5123-3, Texas
Transportation Institute, College Station, TX, October 2009.
Zhou F., Fernando E.G., Scullion T., Development, Calibration, and Validation of
Performance Prediction Models for the Texas M-E Flexible Pavement Design
System, Research Report FHWA/TX-10/0-5798-2, Texas Transportation Institute,
College Station, TX, August 2010.
Zhou F., Li H., Hu S., Button J., and Epps J., Characterization and Best Use of
Recycled Asphalt Shingles in Hot-Mix Asphalt, Research Report FHWA/TX-13/0-
6614-2, Texas A&M Transportation Institute, College Station, TX, October 2012.
12. Discussion
PROFESSOR IMAD-AL-QADI: Thanks for a good presentation. There is one comment

that you mentioned at the end that I would like to comment on, and I have also a
question. The comment is after 30% of binder replacement, you showed nonlinear
blending. The question Is there really blending?
DR. FUJIE ZHOU: Thats a really good question. I dont know.
PROF. AL-QADI: Then there is no linear or nonlinear, right?
DR. ZHOU: Yes.
PROF. AL-QADI: So thats the first question.
DR. ZHOU: I dont know. Because if you look at how stiff the shingles binder is we
showed, you know, we are talking about PG131 or PG178. I really dont know ho w
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
much blending occurs. There may be blending at a later time during the life of the
pavement through diffusion between the shingles and the virgin asphalt binder. But
when you look at it and ask is there any blending? Even for the RAP, how much
blending, is there any blending? I just cant answer.
397

PROF. AL-QADI: I think we force blending when we do extraction for these materials.
My other question or comment here is that we have been doing some work on this topic
and we went up to 64 binder replacement from 7.5% RAS. The major thing that we
noticed was related to fatigue behavior as well as thermal cracking. When you run the
overlay tester on the different materials using the same binder almost, you compared the
number of cycles. I always have a concern that stress level is different and a stiffer
binder or a stiffer mix would experience greater stress. In that case, you are not
comparing the number of cycles at the same level of stress. So what are your thoughts on
that? Because you have more experience than any of us here on the overlay tester.
DR. ZHOU: First of all, we talk about overlays. For overlays the main problem we saw
is reflection cracking. When we are talking about fatigue, were really talking about the
new construction. Thats my understanding. Im talking about Texas; Im not talking
about Illinois. In Texas, most of the work we are doing is overlays. Were talking about
2 in. overlays or 3 in. overlays. Were not talking about 10 in. or 6 in. new construction.
So when we are talking about overlays, most of the time you see the crack already on
your pavement. Thats the big difference between the new construction and the overlays.
Most of the crack has a problem. The crack causes the stress concentration at the bottom
of your overlay. That is completely different from the new construction. I think youll
agree on that.
PROF. AL-QADI: Not really, but continue.
DR. ZHOU: So were really not talking about small strains first of all. Were talking
about overlays. You see the daily thermal effect and how traffic affects, and thats the
overlay. Thats why we see the overlay test is the best test for the overlays. I think I
looked at your paper, probably from the past TRB, you did compare the push-pull test
results versus the overlay tests.
PROF. AL-QADI: No, no, no. We didnt compare that, absolutely. These are two
different animals.
DR. ZHOU: You didnt compare, but I looked at your results, which provide exactly the
same ranking. The push-pull test, the fatigue analysis shows you the higher cracking
resistance. Overlay test will give you the same cracking resistance. I am talking about
ranking. For most studies we see, they probably give you pretty consistent ranking in
terms of the push-pull fatigue analysis versus the overlay test. To address another
question, you know, how can we look at overlay tests or cycles to directly compare that
thing? We have discussed these things several times. Youre not looking at the same
strain, say, different RAP, but we really think we need to integrate the overlay test result
with your pavement structure. Thats the whole process and not just looks at the mix
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
only, not only look at the cycles. Thats what were doing in Texas and not just the
cycles only.
398

PROF. AL-QADI: So I guess to get to the point so Ill leave time for my colleagues here
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
to question, do I need to control the stress or the displacement when I run a test like the
overlay tester? Its just a simple question. What do you think?
DR. ZHOU: I think for the reflection cracking it should be strain-controlled.
PROF. AL-QADI: Okay. Thanks.
PROFESSOR ELIE HAJJ: In your overlay tester, did you observe a difference in the
crack path between the different RAS mixes versus the 0% RAS? Also, was the
variability of the overlay tester higher when you introduced the RAS?
DR. ZHOU: We didnt see much difference. Because, you see, the total cycles even with
the virgin mix, we only got around to 40 cycles. I know your mix is lasting beyond 1200
but all the mixes in Texas we tested, most of them, not all of them, most of the dense
grade mixes are below 100. We really didnt see much difference in terms of
PROF. HAJJ: So the crack path was more or less the same?
DR. ZHOU: The cracking path we didnt see a difference.
PROF. HAJJ. Did you ever look also at the cost savings due to the addition of RAS
versus the additional cost associated with the use of polymer-modified asphalt binder,
what is the difference and what is the real net savings?
DR. ZHOU: Thats a really good question. We have not reached that point yet. Thats
the thing were going to do, in terms of cost effectiveness. We have to look at the whole
thing, not just the lab test or field test section. But thats the next step were going to do.
PROF. HAJJ: Thank you.
DR. BRIAN PROWELL: Nice presentation.
DR. ZHOU: Thank you.
DR. PROWELL: I was interested in the fact that your optimum asphalt contents increase
with increasing RAS percentage. My recollection is that TxDOT specifies a constant
compaction temperature for a grade of binder. Is that correct?
DR. ZHOU: Thats correct.
DR. PROWELL: So you did all three of those mixtures at one compaction temperature.
Is that correct?
DR. ZHOU: Thats correct.
DR. PROWELL: What was that compaction temperature?
DR. ZHOU: That is 250F.
399

DR. PROWELL: Do you think that constant and maybe somewhat lower compaction
temperature than we might see contributed to the increased binder content?
DR. ZHOU: I think thats one of the potential contributions or reasons that lead to higher
binder content. Because if its assumed there is total blending, youre going to have the
total PG binder with a higher grade, so that you need a little bit higher compaction
temperature. Thats one of them. The other thing, potentially, is there is not much
blending going on. So there are two potential reasons, so thats one of them, we kept
using the same compaction temperature. If it were higher, I think for sure because it
lubricates maybe more, it gets easier to compact.
DR. PROWELL: So do you think that possibly when were working with RAP or RAS
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
mixes that keeping a constant compaction temperature might actually provide you a way
of seeing the degree of blending that occurs in the lab? I think that were thinking along
the same lines or maybe were not thinking along the same lines. But youre seeing some
of the same things. Obviously, if you raise the temperature very high, you can artificially
create maybe more blending than might occur in the field, but if you keep a constant
reasonable compaction temperature, you might be able to self-correct for some of that.
Does that seem reasonable?
DR. ZHOU: It does for me, yes. Because thats what you see in the results. Thats why
you see higher RAS but at the same time youve got a higher binder content, total binder
content.
DR. PROWELL: Thank you very much.
DR. STACEY GLIDDEN: Very nice presentation. Im glad to see that other people are
looking at the RAS. My concern would be that you're suggesting that a soft binder is
going to be better, but I didnt see any data on just modified binder. So using a 76-22
versus using a 64-34, which is a pretty highly modified binder. Soft grades are getting
harder and harder to come by, and I think thats a case across the industry. So Im just
wondering why or if maybe something we should pursue is looking at modified binders
that arent necessarily softer and see how that compares, especially in your overlay
testing.
DR. ZHOU: Thats a great comment. We thought the shingles or RAP is already pretty
stiff. We dont want to add PG 76-22 even though they have modifier. Our main concern
is the lower end, we can improve the cracking. But at the same time, if you go to PG 76-
22 on the higher end, although you have modifier, I dont know, maybe improve, maybe
not.
DR. GLIDDEN: Its just something worth looking at. Its something were going to
pursue, as well.
400

DR. ZHOU: Okay. Thank you.

MR. GALE PAGE: I guess I find it interesting, as Brian said, that as you increase the
RAS, youve got to increase the additional binder content to obtain optimum asphalt
content. Even with the higher optimum asphalt contents, those RAS mixes still had
some lower performance in cracking and rutting. Im going to make a philosophical
point rather than dwelling on technical details. I think its important that agencies allow
these recycled materials and let the contractor figure out the economics. Apparently
some contractors have figured out the economics because they still have to meet the
performance criteria even at slightly higher asphalt content in some cases for these RAS
mixes. But owners of contracting firms and agency administrators, when you're looking
at a material such as RAS with 20 to 30% asphalt content, think it needs to be allowed to
be used because their perception is that you get 100% replacement with the asphalt thats
in the RAS. Having this technical information allows the contractors and the agencies to
determine what they need to do in order to meet the performance criteria and work out
the economics. But let the contractor work out the economics. Researchers or agency
people for that matter dont do well working out economics.
DR. ZHOU: Thank you, Gale, for the comments.
MR. PHILLIP BLANKENSHIP: Really good, comprehensive work. I have a comment
and a question. The Asphalt Institute has performed much work recently with RAP and
RAS. Weve been using the Disk-Shaped Compact Tension [DC(t)] developed by Bill
Buttlar. Its a mixture ductility test. We have obtained very confusing results with RAP
and RAS. Normally we see the same response as you would see. If you use a softer
binder, we should see an increase in fracture energy, but we saw something different. I
am not sure were getting blending of the binder and RAP in some cases. This lack of
blending seems more evident with RAS. The DC(t) test seemed to pick up on that well
and maybe showing a lack of cohesion as we overlaid the results with the peak loads
from the same test.
Now the question. When you performed the extraction and the recovery and you turned
that flask upside down, what temperature did you have to use to get that RAS out of the
flask?
DR. ZHOU: When you talk about the manufactured waste?
MR. BLANKENSHIP: Yes.
DR. ZHOU: I think that was 175C.
MR. BLANKENSHIP: Yes, were seeing about that too.
DR. ZHOU: When youre talking about tear-offs, we have to raise it up to 200C.
MR. BLANKENSHIP: Right. Okay.
401
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DR. ZHOU: Otherwise, the recovered tear-offs shingle binder will not flow out the flask.
MR. BLANKENSHIP: Excellent. Then my point back to you would be that I would love
to see this work expanded and maybe to challenge the other folks here to look at this
issue dynamically. As we move more toward warm mix and lowering our mixing
temperatures while adding softer binders at lower temperatures, how is all of this
working together? So I leave that with you. Thank you, Fujie.
MR. FRANK FEE: Just two comments. The way we look at the overlay tester, its an
empirical test much like an APA or a Hamburg; therefore, we use it to rank mixtures.
Im not looking for fundamental properties out of it. Number 2, as far as the cost issue
with regard to polymer-modified materials, if you are looking for the properties of
polymer-modified material and you want it to be equivalent of the virgin material, as a
manufacturer, we are going to have to keep the polymer at the same level. Therefore, if
you want to put a 40% replacement in there, Ive got to replace that polymer. That
material is going to cost you more money, so there is going to be a line. Theres going to
be a delta here, if you want that kind of performance. Thank you.
DR. ZHOU: Thank you, Frank.
402
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Evaluating the Effect of Rejuvenators on the
Degree of Blending and Performance of High
RAP, RAS, RAP/RAS Mixtures
Walaa S. Mogawera*, Abbas Booshehriana, Siavash Vahidia ,
and Alexander Austermana
a
University of Massachusetts - Highway Sustainability Research Center, Fall River, MA
02723
ABSTRACT: The objective of this study was to examine if asphalt rejuvenators can offset the
stiffness attributed by the hardened binder from RAP and RAS in mixtures that incorporate high
RAP and RAS content without adverse impact on the performance of the mixtures. Also, to assess,
if rejuvenators can help the hardened binder from the RAP/RAS comingle with the virgin binder.
Overall, the results showed that asphalt rejuvenators can mitigate the stiffness of the resultant
binder. The cracking characteristics of the mixture improved by the addition of the rejuvenators,
however, the rutting and moisture susceptibility were adversely impacted at the dosage and the
testing conditions used. Also, the tests results at 4C generally showed that there was blending of
the rejuvenated and virgin binder, however, no conclusion could be made at the higher
temperatures.
KEYWORDS: rejuvenators, blending, softening, RAP, RAS
The oral presentation was made by Professor Mogawer. --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
http://dx.doi.org/ 10.1080/14680629.2013.812836
403

MOGAW ER, BOOSHEHRIAN, VAHIDI, AUSTERMAN
1. Introduction
Since the 1970s, Reclaimed Asphalt Pavement (RAP) has been used in asphalt
mixtures at percentages ranging from 10 to 20% in the top lift (Mogawer et al., 2012;
Swiertz et al., 2011; Mallick et al., 2010). The resulting pavements have generally
performed as well as pavements made solely with virgin materials. Because of the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
increasing price of asphalt binder, there has been a continuous increase in the price of
Hot Mix Asphalt (HMA). Correspondingly, the HMA industry has been searching for
methods to decrease the costs of materials without compromising performance. One
method is to use higher percentages of RAP. A survey conducted as part of a Federal
Highway Administration (FHWA) sponsored state of the practice report for RAP in hot
mix asphalt showed that many state transportation agencies specifications allow up to
30% RAP in the surface layers, nevertheless, the majority of these states are only using
RAP percentages of 10 to 20% (Copeland, 2011).
Another valuable source of recycled materials is Reclaimed Asphalt Shingles (RAS).
Approximately, 11 million tons of asphalt shingle waste is generated in the U.S. each
year (Hansen, 2009). This waste is comprised of approximately one million tons of
scraps from asphalt shingle manufacturers and ten million tons of construction scraps
from installations and tear-offs from re-roofing (U.S. EPA, 2012). It has been estimated
that recycling one ton of shingles is equivalent to avoiding the use of one barrel of
asphalt binder (Owens Corning Forms, 2012). Heavily aged, hard binder in RAS does
not allow the engineers to manufacture a high percentage RAS mixture without
justification. Most states that specify RAS in new pavements allow up to 5% RAS b y
weight because the previous studies indicated that mixes that contained up to 5% RAS
by weight performed equally as well as traditional HMA pavements in the field (Swiertz
et al., 2011).
In addition to the possible variability and inconsistency of RAP and RAS resources
(Mallick et al., 2010), one reason for the reluctance of state transportation agencies to
use more RAP and RAS is the concern that the resultant mixtures will be too stiff and
consequently less workable, difficult to compact, and may lead to mixtures more prone
to field failures (cracking, rutting, etc.). Another reason is that it is unknown if adequate
blending occurs between the reclaimed and new materials (Al-Qadi et al., 2007; Huang
et al., 2005). However, even with the reluctance to use more RAP and RAS and
questions concerning mixture stiffness and blending, the recent increases in the cost of
asphalt binder as well as diminishing supplies of quality aggregates have made using
higher contents of these materials in HMA mixtures a priority for the industry as a
method to optimize the use of available resources (Mogawer et al., 2012; Copeland,
2011). The industry has been addressing the stiffness and blending concerns by using
softer binders when using higher RAP/RAS contents. However, several studies have
404

Evaluating the Effect of Rejuvenators on the Degree of Blending and Performance
shown that asphalt rejuvenator agents can allow more reclaimed materials to be
incorporated in asphalt mixtures than using softer binders (Shen et al., 2007a;
OSullivan, 2011; Al-Qadi et al., 2007).
Significant weathering and oxidation of reclaimed materials, reduces the ratio of
maltenes to asphaltenes resulting in a dry and brittle pavement (Karlsson, 2002; Boyer,
2000). Asphalt rejuvenating agents are manufactured to restore the rheological properties
of the RAP and RAS binder by diffusing into it and restoring its colloidal structure and
reconstituting its chemical components (Shen et al., 2005; Brownidge, 2010).
Rejuvenators have been extensively used in pavement preservation to revive the hard
and oxidized top layer by penetrating into the pavement and fluxing with the aged binder
to balance the maltenes to asphaltenes ratio (Brownidge, 2010). However, for recycling
purposes, rejuvenators are not encouraged by some of the state agencies in USA because
of the potential rutting-related concerns while dealing with recycled mixtures containing
a rejuvenator (Shen et al., 2007b).
In an attempt to address the mentioned concern of the industry, this study focused on
evaluating the effects of multiple rejuvenators with different chemical bases and
compositions on the rheological properties of a neat binder and how these properties
might impact the performance of high RAP/RAS mixtures. The use of rejuvenators can
eliminate the need for a softer binder, which is typically used with the high RAP/RAS
contents, and allows the engineers to aim for higher RAP/RAS percentages. The main
focus of the study presented herein was to evaluate the effects of rejuvenators on the
performance of binders and high RAP/RAS content mixtures.
Besides, to address one important concern in dealing with high RAP/RAS content,
the effect of each rejuvenator on the degree of blending between the RAP, RAS, and
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
virgin binders was evaluated by comparing the dynamic modulus of mixtures to a

dynamic modulus predicted using the complex modulus of the recovered binders and the
Hirsch model (Bonaquist, 2005).
2. Objectives
The main objective of this research was to study if asphalt rejuvenators can be used with
high RAP and RAS mixtures to offset the stiffness contributed by the aged binder from
the RAP and RAS without negatively impacting the performance of the mixtures,
especially in terms of rutting, and to assess if rejuvenators can help the hardened binder
from the recycled materials comingle with the virgin binder.
405

In order to achieve the objectives of the study, an experimental plan was developed as
shown in Figure 1. Three different rejuvenators were added to a virgin binder and
performance grading and viscosity tests were conducted. Furthermore, the effect of these
rejuvenators on the performance of the virgin binder was evaluated in terms of fatigue
using the Linear Amplitude Sweep (LAS) test and rutting in terms of the Multiple Stress
Creep Recovery (MSCR) test.
A typical Superpave 9.5 mm was designed with virgin materials. This mixture was
the designated control mixture. Three additional mixtures were designed: the control
mixture plus 40% RAP, the control mixture plus 5% RAS, and the control mixture with
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
35% RAP plus 5% RAS. Each rejuvenator was added to the virgin asphalt binder, a PG
58-28, at a dosage recommended by the manufacturers of the rejuvenators. The
rejuvenator-modified asphalt binders were used to fabricate the control mixture with
40% RAP, 5% RAS, and 35%RAP plus 5% RAS. In total, 13 mixtures were investigated
in this study including the control mixture. In order to evaluate the restoration of the
rheological properties of RAP and RAS binders, the performance characteristics of the
mixtures were evaluated. The stiffness of the mixtures was obtained by measuring the
dynamic modulus; fatigue characteristics using the Texas Overlay Tester (OT); low
temperature cracking using the Thermal Stress Restrained Test Specimen (TSRST), and
rutting and moisture damage using the Hamburg Wheel Tracking Device (HWTD).
To have a more comprehensive understanding of the rejuvenator impact, the mixtures
with 35% RAP and 5% RAS, as the stiffest mixtures, were selected and rheological
properties of the extracted and recovered binders were measured by constructing a
master curve for each binder using the Christensen-Anderson model (Christensen and
Anderson, 1992). MSCR and LAS were also conducted to study the rutting and fatigue
properties of the extracted/recovered binders.
Furthermore, the effect of each rejuvenator on the degree of blending between the
RAP, RAS, and virgin binders in the mixtures with 35% RAP plus 5% RAS was
evaluated. The blending was evaluated by comparing the dynamic modulus of each
mixture with the predicted dynamic modulus using the complex modulus of the
recovered binders and the Hirsch model (Bonaquist, 2005).
406

Reclaimed Materials: Neat Binder Rejuvenating Agents:

1. Reclaimed Asphalt Pavement (RAP) PG 58-28 1. BituTech RAP
2. Reclaimed Asphalt Shingles (RAS) 2. SonneWarmix RJT
3. SonneWarmix RJ
9.5 mm
SuperPave Binder + Rejuvenators
Mixture
Control RAP 40% RAP 0% RAP 35% Binder

RAS 0% RAS 5% RAS 5% Performance
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Performance
Mixture Degree of Blending Gr a d e
Performance Between Virgin
and RAP/RAS Kinematic
Binders Viscosity
Reflective Cracking
Overlay Tester Extract/Recovered Fatigue
Binder from Mixture Linear Amplitude
Moisture Susceptibility Sweep (LAS)
& Rutting Construct Binder
Hamburg Wheel Master Curve Rutting
Tracking Device Multiple Stress
(HWTD) Creep Recovery
Low Temperature Compare R, Td, c (MSCR)
Cracking
Thermal Stress Predict |E*| Using
Restrained Specimen Hirsch Model
Test (TSRST)
Mixture Stiffness Compare Measured Determine

Dynamic Modulus |E*| |E*| and Predicted Quality of
|E*| Blending
Figure 1. Experimental plan.
407

4. Materials
4.1. Asphalt Binder
A PG 58-28 was selected to be used in this study since it is one of the readily available
binders. Based on the viscosity of the binder, the mixture mixing temperature was 150C
(302F) and the compaction temperature was 137C (279F).
4.2. Rejuvenators
Three locally used rejuvenators with different chemical compositions were selected.
1. BituTech RAP is manufactured by Engineered Additives.
2. SonneWarmix RJT is produced by Sonneborn Inc.
3. SonneWarmix RJ is another product manufactured by Sonneborn Inc. It is a
combination of wax and rejuvenator which allows for lower mixing and
compacting temperatures, and at the same time serves as a rejuvenator.
Accordingly, for mixtures containing SonneWarmix RJ, the mixture mixing
temperature was reduced to 130C (265F) and the compaction temperature was
119C (245F).
The dosage used for each rejuvenator was the one recommended by the manufacturer
of each rejuvenator. The manufacturers recommendation for 40% RAP mixtures was
0.5% by weight of total RAP. Since the main purpose of rejuvenators is to revive the
oxidized binder in recycled materials, the percentage of rejuvenators by weight of
RAP/RAS binder (oxidized binder introduced to the mixture) was selected as the base
ratio which equals to 9.28% [0.5/5.39*100=9.28%] for the RAP mixtures. As shown in
Table 1, this ratio was held constant for mixtures containing 5% RAS, and 35% RAP
plus 5% RAS. Thus, proportionally to the weight of the aged binder in each mixture, the
dosage of rejuvenator in RAP/RAS and total mixture could be calculated.
408
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 1. Dosage of rejuvenators in different mixtures.
Rej by Rej by Rej by

Total Replaced
Replaced Recycled Total
Mixture Binder Binder
Binder Material Mix
(%) (%) (%) (%) (%)
4 0 % RAP * 6 2 .1 6 9 .2 8 0 .5 0 0 .2 0 0
5% RAS** 6 0 .8 9 9 .2 8 1 .6 4 0 .0 8 2
5% RAS+35%
6 2 .7 7 9 .2 8 0 .6 4 0 .2 5 7
RAP
*RAP Binder Content: 5.39%
**RAS Binder Content: 17.7%
4.3. Aggregates, RAP, RAS, and Mix Gradation
The aggregates used were from a crushed stone source in Wrentham, Massachusetts.
Three aggregate stockpiles were used: 9.5 mm crushed stone, stone sand, and natural
sand. The RAP was obtained from the same contractor.
The RAS was provided by Recycle America Enterprise. RAS from manufacturing
waste was utilized in this study. Using tear-off shingles presents several potential
challenges that do not exist with the use of manufacturing waste shingles. Tear-off
shingles have aged because of weathering exposure, possibly causing more brittleness
that could affect the durability of the pavement. In addition, asbestos was used in
domestic shingles in small amounts prior to the mid-1980s (Maupin, 2010).
Each aggregate stockpile and the RAP and RAS were tested to determine their
properties, which are shown in Table 2. To be able to generate the same aggregate
skeleton for different mixes, selective sieving was used to separate the 9.5 mm crushed
stone and stone sand into different sieve sizes. As a result, a similar gradation for all
mixtures was obtained as indicated in Table 2.
409
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 2. Average virgin aggregate, RAP, RAS stockpile properties and mixture gradation.
9.5 mm
Sieve Size 9 .5 Stone Natural Mixture
RAS RAP Superpave
(mm) mm Sand Sand Gradation
Specification
19 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 -
1 2 .5 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .7 100 min.
9 .5 1 0 0 .0 1 0 .0 1 0 0 .0 1 0 0 .0 1 0 0 .0 9 7 .1 90-100
4 .7 5 7 4 .1 9 9 .8 9 9 .7 1 0 0 .0 7 4 .1 6 6 .8 90 max.
2 .3 6 5 7 .8 8 3 .7 9 8 .3 9 9 .1 5 7 .8 4 7 .8 32-67
1 .1 8 4 5 .5 5 4 .3 9 3 .3 8 4 .6 4 5 .5 3 3 .5 -
0 .6 3 4 .4 3 3 .8 7 3 .3 6 5 .9 3 4 .4 2 3 .0 -
0 .3 2 2 .4 1 9 .0 2 9 .7 5 8 .7 2 2 .4 1 3 .3 -
0 .1 5 1 3 .5 9 .4 4 .8 4 3 .8 1 3 .5 7 .1 -
0 .0 7 5 9 .1 4 .3 0 .9 2 6 .4 9 .1 4 .4 -
Bulk
Specific
2 .6 3 8 2 .6 4 4 2 .6 2 4 2 .6 2 9 2 .6 3 8 - -
Gravity,
Gsb
Absorption,
0 .7 6 0 .5 3 0 .4 5 0 .6 0 0 .7 6 - -
%
- Not Applicable
5. Evaluation of Binder Properties
In order to study the effect of the rejuvenators on the virgin PG 58-28 binder,
rejuvenators were added directly into the binder. To calculate the additive dosage, as
mentioned in section 4.2., the mixture incorporating 40% RAP was selected as the
reference mixture. The authors assumed that the binder in the mixtures with additives
consisted of two components: (1) the oxidized stiff binder from the RAP, and (2) the
combination of the virgin binder and added rejuvenator. Then, the two binders would
commingle and produce the resultant balanced binder in the mixture. Accordingly, on
one hand, by having an understanding about the properties of the aged binder, and on the
other hand, by evaluating the properties of the combination of virgin binder and
rejuvenator, a rough prediction of the performance of the resultant binder in the mixture
would be achievable.
Based on this assumption, the rejuvenator dosage added to the virgin binder was
calculated by assuming an arbitrary weight of mixture with 40% RAP. For this study
410
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

RAP, the percent of rejuvenator by weight of the virgin binder equals to 5.218%
[9.276%*2.16%/(6-2.16)%=5.218%]. The virgin binder was heated to the mixing
temperature of 150C (302F), and then rejuvenator was added to the binder and mixed
for 5 minutes. Next, the following tests were conducted on the binders.
5.1 Performance Grading
The binders were tested to determine their continuous and performance grade in
accordance with AASHTO R 29, Grading or Verifying the Performance Grade of an
Asphalt Binder, and AASHTO M 320, Standard Specification for Performance-
Graded Asphalt Binder (AASHTO, 2011). Kinematic viscosity was also conducted in
accordance with AASHTO T 316, Standard Method of Test for Viscosity
Determination of Asphalt Binder Using Rotational Viscometer, to investigate the
influence of rejuvenators on the viscosity of virgin binder. The results of the PG grading
and kinematic viscosity test are shown in Table 3.
Table 3. Binder performance grading and kinematic viscosity results.

Continuous Grade
Performance Grading
Binder PG-Grade High-Low

Low Mid High
PG 58-28 -28.5 1 6 .6 5 9 .7 58-28 8 8 .1 1
PG 58-28 + BituTech -34.9 9 .2 5 2 .5 52-34 8 7 .3 3
PG 58-28 + Sonne
-34.4 1 0 .1 5 3 .6 52-34 8 7 .9 8
RJT
PG 58-28 + Sonne RJ -33.4 1 1 .6 5 4 .2 52-28 87.60
Viscosity Viscosity Mixing Compaction
Binder @135 @150 Temperature Temperature
Kinematic Viscosity
cP cP C C
PG 58-28 3 2 4 .7 1 6 8 .3 1 5 0 .0 1 3 7 .0
PG 58-28 + BituTech 2 2 1 .5 1 1 8 .2 1 4 2 .5 1 2 7 .5
PG 58-28 + Sonne
2 2 4 .5 1 2 0 .8 1 4 3 .0 1 2 7 .0
RJT
PG 58-28 + Sonne RJ 2 1 6 .9 1 1 6 .4 1 4 2 .0 1 2 6 .0
Based on the data presented in Table 3, the rejuvenators did drop the high end
continuous PG and intermediate temperature of the control binder by approximately 6
C, and the low end continuous PG increased by 6 C. Furthermore, the viscosity of
411
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

the different rejuvenators plus the neat binder dropped similarly relative to the neat
binder. These data prove the considerable softening effect of the rejuvenators to the
virgin binder, which is a required phenomenon when high RAP/RAS mixtures are used
in order to substitute the need to use softer binders.
5.2 Performance Testing
5.2.1. Resistance to Rutting-MSCR
The Multiple Stress Creep Recovery (MSCR) test is designed to more accurately
characterize the resistance to permanent deformation (rutting) of asphalt binders
(FHWA, 2011). Testing was conducted in accordance with AASHTO TP 70-10
Multiple Stress Creep Recovery (MSCR) Test of Asphalt Binder Using a Dynamic
Shear Rheometer (DSR) (AASHTO, 2011).
The test was run on the binders aged by Rolling Thin Film Oven Test (RTFOT) to
simulate the rutting that happens at the beginning of the pavement service life. The DSR
25-mm parallel plate geometry with 1-mm gap was utilized to test the samples at the
continuous high grade of each binder. Ten cycles of creep (1 second) and recovery (9
seconds) at two different stress levels (0.1 kPa and 3.2 kPa) were applied to the
sandwiched disk. Equations 1.a and 1.b were used to calculate percent recovery and non-
recoverable creep compliance for each cycle at different stress levels.
[ 1 a]
[1 b ]
where: r = Percent recovery for each cycle

c = The strain value at the end of the creep portion of each cycle.
0 = The strain value at the beginning of the creep portion of each cycle.
10= The strain value at the end of the recovery portion of each cycle.
Jnr = Non-recoverable creep compliance for each cycle.
c= The stress value.
412
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

r and Jnr were used to calculate the average percent recovery (R0.1 and R3.2) and
average non-recoverable creep compliance (Jnr0.1 or Jnr3.2) at stress levels of 0.1 kPa and
3.2 kPa. The results are shown in Table 4.
The relationship of the MSCR test specification parameter Jnr to actual rutting has
been extensively evaluated. The Jnr is the non-Recoverable Creep Compliance (Jnr) it is
a measure of the amount of residual strain left in the specimen after repeated creep and
recovery, relative to the amount of stress applied. The lower the Jnr the better resistant to
rutting the binder will be. The MSCR test was tested at the continuous PG high
temperature. At that temperature the G*/sin value is constant and equal to 2.2 KPa.
Table 4. Multiple Stress Creep Recovery test results.

Continuous PG
0.1 3.2 diff Jnr0.1 Jnr3.2
Binder Temperature
C (%) (%) (%) (%) (%)
PG 58-28 5 9 .7 4 .1 9 1 .7 6 5 7 .8 9 2 .4 1 2 .1 2
PG 58-28 + BituTech 5 2 .5 2 .7 4 0 .9 3 6 5 .8 4 3 .5 9 3 .9 1
PG 58-28 + Sonne RJT 5 3 .6 4 .5 6 1 .2 0 7 3 .5 2 3 .2 9 3 .6 9
PG 58-28 + Sonne RJ 5 5 .9 4 .4 4 1 .4 0 6 8 .5 8 3 .1 1 3 .5 2
Based on the results in Table 4, the Jnr at the two stress levels was higher for the
virgin binder with each rejuvenator than the virgin binder used. This was an indication
that, as expected, the rejuvenators softened the virgin binder. The higher the percent
recovery, , the more resistant to rutting the binder will be. The results for 3.2 confirms
the softening effect of the rejuvenators, whereas it was decreased for rejuvenator
modified binders. The results for 0.1 illustrated a similar trend. Softening a binder is a
function of rejuvenators and different rejuvenators show various softening impact.
However, to determine whether this increase in the Jnr and decrease in might cause any
rutting problems, a rutting performance test on the mixtures containing rejuvenators was
performed, which is explained later.
5.2.2. Resistance to Fatigue - LAS
The Linear Amplitude Sweep test indicates the resistance of the asphalt binders to
fatigue damage by cyclic loading employing a linearly ramping amplitude sweep test. It
is conducted using the DSR at the continuous intermediate temperature of the
performance grade of the asphalt binder (MARC, 2011).
413
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The binders were initially short-term aged by using the RTFOT, and then aged using
the Pressure Aging Vessel (PAV) to simulate the aging for in-service asphalt pavements.
The sample was tested using the DSR 8 mm parallel plate geometry with a 2 mm gap.
First, a frequency sweep test was conducted on the sample to measure the undamaged
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
rheological properties of the binder. Next, at a constant frequency of 10Hz, a series of
systematically increasing oscillatory load cycles was applied on the sample to accelerate
the damage (Hintz et al., 2011). Using a regression model, the parameters A and B for
the binder fatigue performance model can be calculated. Equation 2 can be used to
obtain the number of cycles to failure (Nf) at maximum strain amplitude.
[2 ]
where: A, B = regression parameters

max = the maximum expected binder strain for a given pavement structure
The results for LAS are shown in Table 5.
Table 5. Linear Amplitude Sweep test results.
Continuous Nf
Nf
Binder intermediate A B @
@ 2 .5 %
Temp. C 5 .0 %
PG 58-28 1 6 .7 9.58E+06 -5.98 39,805 629
PG 58-28 + BituTech 9 .2 1.55E+07 -5.53 98,136 2 ,1 3 0
PG 58-28 + Sonne RJT 1 0 .1 5.10E+07 -6.38 147,138 1 ,7 6 4
PG 58-28 + Sonne RJ 1 1 .6 1.16E+08 -6.57 283,087 2 ,9 8 1
The number of cycles improved at both strain levels for all the rejuvenators. This was
expected as the rejuvenators did soften the virgin binder as illustrated by the MSCR test.
It should be noted that the improvement with respect to neat binder varied depending on
the type of rejuvenator.
5.3. Performance Grading of the Extracted Binders
In order to investigate the direct effect of rejuvenators on the aged binders, the mixtures
containing 35% RAP plus 5% RAS were extracted. These mixtures were selected
because they contain the highest amounts of oxidized binders. For comparison purposes,
the control mixture binder was extracted as well. Besides, the RAP and RAS binders
414

were extracted to specify the stiffness of binders introduced into the mixtures while
using these two recycled materials.
To do so, appropriate masses of mixtures were used to extract and then recover the
binders (combination of new and aged binder) according to AASHTO T 164, Standard
Method of Test for Quantitative Extraction of Asphalt Binder from Hot Mix Asphalt
(HMA), and AASHTO T 170Standard Method of Test for Recovery of Asphalt Binder
from Solution by Abson Method (AASHTO, 2011).
Next, the performance grading was conducted on all the above-mentioned extracted
and recovered binders in accordance with AASHTO R 29, Grading or Verifying the
Performance Grade of an Asphalt Binder, and AASHTO M 320, Standard
Specification for Performance-Graded Asphalt Binder (AASHTO, 2011). The
recovered binders were assumed to be already short-term aged due to the aging
encountered by the binder during the mixture mixing and aging process. Table 6 outlines
the test results for the extracted and recovered binders.
Table 6 illustrates that the RAP and RAS stiffened the virgin binder as evident by the
increase in the high temperature continuous PG from 60.6C to 74.2C and changing the
low temperature continuous PG from -31.1C to -25.2C. The rejuvenators helped
mitigate the stiffening effect due to the presence of RAP and RAS by changing the
continuous PG of the RAP/RAS mixture from a PG 74.2-25.2 to a PG 68.5-26.8 (based
on averaging the continuous PG of the three extracted binders with rejuvenators). These
temperatures illustrated that the three rejuvenators at the dosage selected reduced the
stiffness of the resultant binder in the RAP/RAS mixture.
A worthy observation is the difference for low temperature grading based on the
stiffness versus m-value of the binders containing RAP and RAS binders. The m-value is
an indicator of the relaxation capacity of the binder at low temperatures. The control
binder indicated that there was a slight difference between the two temperatures (-32.49
for stiffness and -31.13 for m-value). For binders containing RAP and RAS, the m-value
showed lower relaxation capacity while binder stiffness gave fairly similar low
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
temperature grade to the control mixture. The reason might be related to the ability of the
rejuvenators to merely soften the binders. It might be concluded that another additive
would be helpful to revive the relaxation capacity of the binder.
5.4 Binder Master Curve
The master curve is a relationship between the binder stiffness (G*) and reduced
frequency using the Christensen-Anderson Model (CAM) (Christensen and Anderson,
1992). To construct the master curves, G* of the as-recovered binders was measured at
multiple positive temperatures (above 0C) and frequencies using DSR device. Stiffness
415

of the binders at the negative temperatures (below 0C) was obtained by the running the
Bending Beam Rheometer (BBR) and the values were converted into G* and frequency.
The collected points were combined by shifting them to a reference temperature
(Anderson et al., 1991; Christensen and Anderson, 1992).As a result, these master curves
make it possible to predict viscoelastic properties over a wide frequency range and also
to predict viscoelastic properties at any temperature. For comparison purposes, all the
binder master curves were shifted to a single reference temperature. In this study a
reference temperature of 25C was used.
R
log 2
log 2
R

G * ( ) = G g 1 + c [3]
r

where: G*( ) = complex shear modulus

Gg = glass modulus assumed equal to 1GPa
r = reduced frequency at the defining temperature, rad/sec
c = cross over frequency at the defining temperature, rad/sec
= frequency, rad/sec
R = rheological index
The Christensen-Anderson model is a very useful tool because the master curve
parameters ( c, R, and Td) have specific physical significance. The cross-over
frequency, c, is a measure of the overall hardness of the binder. As the cross-over
frequency increases, the hardness of the binder decreases, which is desirable for
rejuvenated binders. The rheological index, R, is an indicator of the rheological type. It
is defined as the difference between the log of the glassy modulus and the log of the
dynamic modulus at the cross-over frequency. As the value of R increases, the master
curve becomes flatter, indicating a more gradual transition from elastic behavior to
steady-state flow. Normally, R is higher for oxidized asphalt. The defining temperature,
Td, is related to the glass transition temperature of the binder, and is an indicator of the
temperature dependency of the material. The temperature dependency increases as Td
increases (Anderson et al., 1991; Christensen and Anderson, 1992). Table 6 shows the
rheological properties of extracted/recovered binders.
Generally, the rheological properties agreed with the data from the MSCR and LAS
tests. The recovered binder from the RAP/RAS mixtures had a higher c relative to the
same mixtures with incorporation of the rejuvenators. This indicated that the
rejuvenators reduced the hardness of the binders. The R-value increased as the oxidation
level increased. R for the Control mixture is about 0.68109 smaller in comparison with
416
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

mixture with RAP and RAS. The addition of rejuvenators reduced the difference
depending on the type of the rejuvenator.
In order to make the physical parameters more meaningful, Figure 2 schematically
compares the master curves of the extracted binders. As depicted in this figure, adding
the rejuvenators softened the RAP/RAS binder and moved the curve closer to the control
binder. It is worthy of mention that these extracted/recovered binders master curves are
the main requirement to evaluate the degree of blending described in Section 8.
Cont r ol 58-28 35%RAP+5%RAS

35%RAP+5%RAS+Bit uTech 35%RAP+5%RAS+Sonne RJ T
35%RAP+5%RAS+Sonne RJ Ext r act ed RAP Binder
1 .0 E+0 9
1 .0 E+0 8
1 .0 E+0 7
1 .0 E+0 6
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
10
1 .0 E+0 5 8
G* (Pa)
6
Lo g Shif t Fa cto r
1 .0 E+0 4 4
2
1 .0 E+0 3 0
-2
1 .0 E+0 2
-4
-6
1 .0 E+0 1 -3 0 -1 0 10 30 50 70
Tempera ture ( C)
1 .0 E+0 0
1 .0 E-0 6 1 .0 E-0 4 1 .0 E-0 2 1 .0 E+0 0 1 .0 E+0 2 1 .0 E+0 4 1 .0 E+0 6 1 .0 E+0 8
Reduced Frequency at 25 C (rad/sec)
Figure 2. Comparison of extracted/recovered binder master curves.
417

Table 6. Performance Grading of the extracted/recovered binders, and RAP/RAS

properties, and Christensen-Anderson Master Curve Parameters.
35%RAP + 5%RAS 100%

Properties Unit Control No BituTe Sonne Sonne RAP
Add. ch RJT RJ Binder
Delta 84.51 80.78 81.66 80.28 80.55 81.44
High Grade
Temp C 60.91 74.21 68.35 68.87 69.06 77.38
Delta 41.69 36.88 38.53 34.18 35.35 41.16
Int. Temp.
Temp C 16.20 21.11 17.48 15.89 17.83 23.69
m-value Temp C -31.13 -25.21 -27.53 -26.50 -26.20 -23.49
Stiffness Temp C -32.49 -31.54 -31.99 -31.71 -31.68 -28.78
Continuous 60.9 74.2 68.4 68.9 69.1 77.4
Grade -31.1 -25.2 -27.5 -26.5 -26.2 -23.5
PG Grade 58-28 70-22 64-22 64-22 64-22 76-22
R 1 0 9 1.990 2.682 2.122 2.521 2.524 2.545
c@ C rad/sec 1310.4 12.3 361.3 88.8 72.8 13.9
Td Temp C -9.46 2.38 -8.99 -3.72 -1.71 2.90
Phase
Temp G*/sin Temp Stiffness m-value
Recovered RAS
angle
C kPa C Mpa
Binder
118 8.329 54.91 -12 123.0 0.244

124 5.765 57.04 -6 74.4 0.260
130 3.616 59.63 0 42.7 0.282
5.5 MSCR on PAV-Aged Extracted/Recovered Binders
In order to study the effects of the rejuvenator on the performance of extracted/recovered

binders in terms of rutting, MSCR was conducted on the PAV-aged binders at the same
temperature (58C). Table 7 outlines the MSCR test results. Based on the non-
recoverable creep compliance (Jnr) and in conformance with the previous binder data, the
binder extracted from the RAP/RAS mixture exhibited the best resistance to rutting and
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
addition of rejuvenators slightly impacted the binders in terms of rutting. Obviously, the
control mixture showed the least rutting resistance.
418

Table 7. PAV-aged extracted/recovered MSCR test results.
PAV-aged Extracted/Recovered 0.1 3.2 diff Jnr0.1 Jnr3.2

Binder (%) (%) (%) (%) (%)
Control 1 8 .2 6 1 3 .0 6 2 8 .4 7 0 .4 7 3 0 .5 0 8
35%RAP + 5%RAS 5 1 .1 5 5 0 .3 4 1 .5 8 8 0 .0 4 3 0 .0 4 2
35%RAP + 5%RAS + BituTech 4 0 .1 3 3 7 .2 1 7 .2 8 4 0 .1 0 8 0 .1 1 2
35%RAP + 5%RAS + Sonne RJT 4 8 .0 5 4 5 .3 4 5 .6 5 9 0 .0 8 0 0 .0 8 3
35%RAP + 5%RAS + Sonne RJ 5 2 .2 7 4 9 .3 8 5 .5 3 2 0 .0 6 4 0 .0 6 7
6. Mix Design Procedure
A control mixture and control mixtures incorporating 40% RAP, 5% RAS, and 35%
RAP plus 5% RAS were designed using the PG 58-28. Each mixture had the same
aggregate gradation (Table 2) and was developed to meet the requirements for a 9.5 mm
Superpave mixture in accordance with AASHTO M 323, Superpave Volumetric Mix
Design, and AASHTO R 35, Superpave Volumetric Design for Hot Mix Asphalt
(AASHTO, 2011). The volumetrics of the mixtures with recycled materials and
rejuvenators were verified, however, only the 40% RAP mixture data is presented in the
paper as the 35% RAP, the 5% RAS, and the 35% RAP and 5% RAS showed similar
results.
The design Equivalent Single Axle Loads (ESALs) for this project was selected as
0.3 to <3 million which is consistent with surface course mixtures in New England. All
specimens were fabricated using the Superpave Gyratory Compactor (SGC).The design
Superpave gyratory compactive effort for this ESALs level was Ndesign = 75 gyrations.
The RAP was air dried until a constant mass was achieved. To minimize the
hardening of the aged binder in RAP and heating the high RAP percentage up to the
mixing temperature, two hours prior to mixing, it was added on top of the heated
aggregate. Generally, the RAS binder is too stiff and care must be taken to have minimal
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
extra oxidation while dealing with this material. Therefore, RAS was put in 5 minutes
prior to mixing on the heated aggregate, and since the RAS dosage is limited to 5% by
total mix, it can rapidly achieve the mixing temperature. The binder was added to the
aggregates and RAP/RAS, if applicable, and at this stage, the rejuvenating agent was
added to the binder in the mixing bowl.
The volumetric properties for each mixture are shown in Table 8. Note that all
mixtures met the Superpave specification volumetric targets except for the percent air
voids. The percent air voids for the control mixture was 3.5% - it increased to 4.3%
when RAP was incorporated, 4.6% when RAS was incorporated, and 4.7% when RAP
419

and RAS were used. This increase in the air voids might be an indication that the
hardened binders of the RAP and RAS are not blending completely with the virgin
binder resulting in a mixture with less total binder content. The use of rejuvenators
appeared to mitigate the increase in the air voids. This might be attributed to the
diffusion of the rejuvenator in the hardened binder and allowing it to commingle with
the virgin binder leading to mixtures meeting the total binder content.
Table 8. Mixture Volumetric Properties.
35% RAP 5%
Properties Control 40% RAP 5% RAS
RAS
Total Binder Content, % 6 6 6 6
Virgin Binder Added, % 6 3 .8 4 5 .1 2 3 .2 3
Air Voids,% 3 .5 4 .3 4 .6 4 .7
VMA, % 1 6 .2 1 6 .7 1 7 .1 17
VFA, % 7 8 .4 7 4 .3 7 3 .1 7 2 .4
Dust to Binder Ratio 0 .7 5 0 .7 6 0 .7 5 0 .7 6
4 0 % RAP 9.5 mm
Properties Superpave
Bitutech Sonne RJT Sonne RJ Specification
Total Binder Content, % 6 6 6 -
Virgin Binder Added, % 3 .8 4 3 .8 4 3 .8 4 -
Air Voids,% 4 .1 3 .6 3 .7 4 .0 0 %
VMA, % 1 6 .6 1 6 .2 1 6 .6 15% min.
VFA, % 7 5 .3 7 7 .8 7 7 .7 65-78
Dust to Binder Ratio 0 .7 5 0 .7 5 0 .7 5 0.6 -1.2
- Not Applicable
VMA = Voids in Mineral Aggregate
VFA = Voids Filled with Asphalt
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
420

7. Mixture Performance Testing
7.1. Mixture Stiffness
As the main stiffness indicator, complex dynamic modulus |E*| testing was conducted to
determine changes in mixture stiffness due to the incorporation of rejuvenators with
RAP and/or RAS. Three gyratory specimens were fabricated for each mixture. Each
specimen was subsequently prepared for dynamic modulus testing in accordance with
AASHTO PP 60 (AASHTO, 2011). The final test specimens had air void contents of 7.0
1.0%. Test specimens were placed in the Asphalt Mixture Performance Test (AMPT)
device and subjected to a sinusoidal (haversine) axial compressive stress at the different
temperatures and frequencies. The resultant recoverable axial strain (peak-to-peak) was
measured, and consequently the dynamic modulus was calculated.
Dynamic modulus testing was conducted in accordance with TP 62 (AASHTO,
2011). All the specimens were tested at temperatures of 4C, 20C, and 35C (39F,
68F, and 95F) and loading frequencies of 25 Hz, 10 Hz, 5 Hz, 1 Hz, 0.5 Hz, 0.1 Hz,
and 0.01 Hz (35C only) in accordance with AASHTO PP 61(AASHTO, 2011). Three
frequencies were added to the AASHTO PP 61 to increase the precision of the resultant
master curve.
In order to develop the mixture master curve, 15C (59F) was selected as the
reference temperature to model the intermediate temperature of the Northeast. Figures 3-
5 illustrate a comparison between different mixtures.
Figure 3 illustrates the master curves for the control mixture, control mixture with
40% RAP, and the control mixture with 40% RAP and each rejuvenator. The control
mixture with 40% RAP was stiffer than the control mixture with no RAP. Incorporating
the rejuvenators dropped the stiffness of the RAP mixture closer to the control mixture.
This is an indication that the rejuvenators did soften the resultant binder of the 40% RAP
mixture causing the overall stiffness of the mixture to get closer to the stiffness of the
control mixture. This agrees with the binder testing, in particular, the binder master
curves. Similar observations apply to Figures 4 and 5, however, the drop in stiffness for
the RAS and RAP/RAS mixtures was not as much as it was for the RAP mixtures. This
might be attributed to the RAS binder being stiffer than the RAP binder suggesting that
the dosage of rejuvenators for mixtures containing RAS might need some adjustment.
421
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

1,000
Dynamic Modulus E*, ksi
100
Control
40% RAP + BituTech
10
40% RAP + SonneWarmix RJT
40% RAP + SonneWarmix RJ
40% RAP
1
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04
Figure 3. Mixture master curve comparison of PG 58-28 and 40% RAP mixtures.
1,000
100
Control
5% RAS + BituTech
10
5% RAS + SonneWarmix RJT
5% RAS + SonneWarmix RJ
5%RAS
1
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04
Figure 4. Mixture master curve comparison of PG 58-28 and 5% RAS mixtures.
422
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

1,000
100
Control
35%RAP 5%RAS + BituTech
10 35%RAP 5%RAS + SonneWarmix RJT
35%RAP 5%RAS + SonneWarmix RJ
35%RAP 5%RAS
1
1.0E-06 1.0E-04 1.0E-02 1.0E+00 1.0E+02 1.0E+04
Figure 5. Mixture master curve comparison of PG 58-28 and 35% RAP plus 5% RAS
mixtures.
7.2. Rutting and Moisture Susceptibility
Since the rejuvenators did soften the binder based on the binder testing of the virgin and
extracted binder, it was essential to determine the effect of these rejuvenators on the
cohesiveness of the mixtures. Testing was conducted using the Hamburg Wheel-
Tracking Device (HWTD) in accordance with AASHTO T 324 (AASHTO, 2011). The
rut depth versus numbers of passes of the wheel is plotted to determine the Stripping
Inflection Point (SIP) at which the test specimen begins to exhibit stripping (moisture
damage). The compacted and trimmed specimens had air void contents of 7.01.0% as
required by AASHTO T 324.
The test was conducted at a temperature of 45C (113F). The termination criteria
was either 20,000 passes or until visible stripping was noted. In addition to SIP, the rut
depths at 10,000 and 20,000 passes (if applicable) were measured.
Table 9 presents the data from the HWTD tests. The control mixture performed
poorly in terms of rutting and moisture damage. The 40% RAP, and 35% RAP plus 5%
RAS mitigated the poor performance. These mixtures had minimal rutting and no SIP
while the 5% RAS showed relatively less improvement. The reason might be related to
423
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

the stiff RAS binder not blending sufficiently with the virgin binder and, consequently,
the influence is not as evident. The incorporation of the rejuvenators to the 35% RAP
plus 5% RAS slightly reduced the mixtures resistance to rutting and moisture damage
(although no SIP was observed prior to the 20,000 passes). Contrary to the RAP/RAS
mixture, the incorporation of the rejuvenators to the 40% RAP and 5% RAS mixtures
increased the mixtures susceptibility to rutting and moisture damage. The rut depths
were greater than 20 mm and a SIP was observed prior to the completion of the tests.
This might indicate that these two mixtures did not have enough stiff binder to balance
the impact of the rejuvenators.
Table 9. HWTD test results.
SIP Rut Depth @ Rut Depth @

Mixture 10,000 Passes 20,000 Passes
# Passes (mm) (mm)
Control 5590 N/ A N/ A
No Additive None 1 .3 7 2 .7 9
5%RAS 40% RAP
BituTech 12270 6 .2 5 N/ A
SonneWarmix RJT 11550 3 .4 5 N/ A
SonneWarmix RJ 10550 4 .7 3 N/ A
No Additive 8230 7 .2 7 N/ A
BituTech 7210 1 1 .4 3 N/ A
SonneWarmix RJT 5190 1 6 .9 0 N/ A
SonneWarmix RJ 6690 1 2 .5 3 N/ A
No Additive None 0 .8 0 1 .0 3
35%RAP +
5%RAS
BituTech None 1 .5 8 3 .2 9
SonneWarmix RJT None 1 .8 0 3 .7 4
SonneWarmix RJ None 1 .5 1 2 .8 5
N/A: Test termination happened before 10,000 or 20,000 passes.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
7.3. Reflective Cracking
The OT device is designed to evaluate the fatigue and reflective cracking potential of
asphalt mixtures (TxDOT, 2009). The device applies tension loading to test specimens
while recording load, displacement, temperature and time (TxDOT, 2009). Research
studies have been conducted that outline the use of this device for evaluation of asphalt
mixture cracking susceptibility (Zhou et al., 2007a, 2007b).
For this study, the Texas Department of Transportation specification (Tex-248-
F) for testing bituminous mixtures with the OT (TxDOT, 2009) was followed. Five
specimens were fabricated, trimmed and tested to cope with the OT potential high
424

variability. The air void level of the trimmed specimens was 7.01.0%. All mixtures for
this study were tested with a joint opening (displacement) of 0.6 mm (0.025 in.), test
temperature of 15C (59F), and a failure criteria of 93% reduction in the load measured
during the first cycle or 2000 cycles (whichever occurred first). The 15C (59F) was
selected because it represents a typical intermediate temperature for the northeast. It was
also used as the reference temperature for the mixtures master curves.
The average and standard deviation results of the five specimens are shown in Figure
6. Generally, mixtures exhibiting more cycles to failure exhibit more cracking
resistance. The RAP, RAS, and RAP/RAS mixtures exhibited a significant drop in the
number of cycles to failure relative to the control mixture. This is expected because of
the hardened binder in the RAP and RAS. The incorporation of the rejuvenators
improved the cracking performance of the RAP, RAS, and RAP/RAS mixtures.
However, this improvement was relatively lower for SonneWarmix RJ, although it was
shown in the binder section that this rejuvenator can enhance the fatigue properties of the
virgin binder. This observation is possibly related to the lower mixing and compaction
temperature. This lower amount of energy introduced to the mixture might not be
appropriate for the mixtures containing stiff recycled materials (Mogawer et al. 2012)
resulting in less available binder content and more brittle mixtures. The other two
rejuvenators performed very similarly.
7.4 Low Temperature Cracking
The low temperature cracking of each mixture was assessed by the Thermal Stress
Restrained Specimen Test (TSRST) device in accordance with AASHTO TP 10-93
(AASHTO, 1993).In the TSRST test, the asphalt specimen is placed in an environmental
chamber and pre-cooled at 4C for 30 minutes. Next, the chamber is cooled down at a
constant rate of -10C/hour. In the meanwhile, the specimen length is held constant by
the TSRST device. As the test specimen cools down, it tries to contract but cannot.
Gradually, the resultant thermal stress is accumulated until it exceeds the tensile capacity
of the specimen resulting in specimen fracture and test termination. The temperature at
which this fracture occurs is reported as the low temperature cracking of the mixture.
Three gyratory specimens (185 mm (7.3 in.) tall by 150 mm (5.9 in.) in diameter)
were fabricated for each mixture. TSRST specimens were cored and cut to reach the
final height of 160 mm tall (6.3 in) by 54 mm (2.1 in) in diameter and the air voids of
71%. The results for this test are depicted in Figure 6.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
425

OT TSRST
40% RAP 5% RAS 35% RAP + 5% RAS

2000 -30.00
-27.40 -27.07 -27.36 -27.90 -27.26
-26.28 -26.19
1800 -25.24 -25.25
-24.45
-23.54 -23.43 -23.26 -25.00
OT Number of Cycles to Failure
1600
TSRST Cracking Temperature ( C)

1400
-20.00
1200
1000 -15.00
800
-10.00
600
400
-5.00
200
0 0.00
Cont rol
No Ad d it ive
No Ad d it ive
No Ad d it ive
SonneWarmixRJ
SonneWarmixRJ
SonneWarmixRJ
SonneWarmixRJT
SonneWarmixRJT
SonneWarmixRJT
Bit uTech
Bit uTech
Bit uTech
Figure 6. Overlay Tester and TSRST results.
The low temperature cracking of the control mixture lost approximately -1.7C when
RAP was incorporated and -2C when RAS or RAP plus RAS were incorporated in the
mixture. The three rejuvenators helped mitigate the loss in the low temperature
cracking. The incorporation of the rejuvenators in the mixtures with the recycled
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
materials interestingly improved the low temperature cracking relative to the control
mixture (almost -2C). These data agreed with the rheological properties obtained using
the CAM model, particularly the R parameter.
8. Degree of blending
The degree of blending/mixing between the RAP/RAS and the virgin binders will have a
significant impact on the volumetric and performance of HMA mixtures (Copeland,
2011). The degree of blending was assessed by comparing the measured dynamic
modulus (E*) of the mixtures with predicted dynamic modulus from binder testing of as-
426

recovered binders (Bonaquist, 2005; Bonaquist, 2007). Since the |E*| is highly sensitive
to the stiffness of the binder (G*) in the mixture, it represents the real blending of the
virgin binder with RAP/RAS, and the extracted binder is assumed to show the fully
blended condition.
The degree of blending between the RAP and RAS and virgin binders was evaluated
using the binder master curves mentioned in section 5.4. The estimated G* from the
master curve is then substituted into the Hirsch model (Equations 4 and 5) to calculate a
predicted dynamic modulus E*. The dynamic modulus of a mixture is highly sensitive
to the stiffness of the binder in the mixture (Christensen et al., 2003)
VMA VFA VMA
| E* |mix = Pc 4,200,0001 + 3 | G* |binder
100 10,000
1 Pc
+
VMA [4]
1 100 VMA
+
4,200,000 3 VFA | G* |binder

0.58
VFA 3 | G* | binder
20 +
VMA
Pc = 0.58
VFA 3 | G* | binder [5]
650 +
VMA
where: |E*|mix = mixture dynamic modulus, psi

VMA = Voids in mineral aggregates, %
VFA= Voids filled with asphalt, %
The confidence intervals at a level of significance = 0.05 are calculated for the
measured and the predicted E*. If the two confidence intervals overlap, it is concluded
that a good degree of blending exists. Figures 7 to 11 illustrate the degree of blending
bar charts for selective frequencies (1 Hz for slow traffic, and 10 Hz for high speed) at
the |E*| test temperature range (4C, 20C, and 35C).
The figures show that all predicted E* were greater than the measured E*. Even so,
the tests results at 4C generally showed that there was blending of the rejuvenated and
virgin binder in the 35% RAP plus 5% RAS mixture. However, as the temperature
increased, the predicted E* for the control mixture became significantly greater than that
of the measured, especially at 35C. Therefore, no conclusion regarding the degree of
427
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

blending on E* at the higher temperatures could be made. Why the predicted E*values
were greater than the measured E* needs to be determined.
The control mixture and its bars were considered as the reference for full blending,
and then the other mixtures were compared respectively. The figures illustrate that the
addition of the rejuvenators to the 35% RAP and 5% RAS mixtures adjusted the
predicted and measured values of the RAP/RAS mixture similar to the control mixture.
Therefore, it can be concluded that the addition of the rejuvenators can help to alleviate
the high stiffness caused by the addition of RAP and RAS binder.
Temperat ure = 4C & Frequency = 10Hz
1000
Dynamic Modulus (ksi)
100
10
1
Control RAP+RAS RAP+RAS + RAP+RAS + RAP+RAS +
BituTech SonneWarmix SonneWarmix
Measured E* Predicted E* RJT RJ
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 7. Degree of blending at a temperature 4C and frequency of 10 Hz.
428

Temperature = 4C & Frequency = 1Hz
1000
100
10
1
Figure 8. Degree of blending at a temperature 4C and frequency of 1.0 Hz.
1000
100
10
1
429
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

1000
100
10
1
Figure 10. Degree of blending at a temperature 20C and frequency of 1.0 Hz.
1000
100
10
1

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,
430

This study focused on evaluating the effects of three different asphalt rejuvenators on
mitigating the increase in stiffness of laboratory mixtures having high RAP and RAS
contents without adversely impacting the performance of the mixtures. The study also
assessed if rejuvenators can help hardened binder from RAP and RAS comingle with
virgin binder. Based on the data and the analysis of the data, the following conclusions
were made:
1. The addition of the rejuvenators to the virgin binder, a PG 58-28, decreased the
viscosity of the virgin binder. This showed the softening effect of the rejuvenators
on the virgin binder.
2. The MSCR test illustrated, at the two stress levels used, that the non-recoverable
creep compliance of the virgin binder increased due to the addition of the
rejuvenators. Again, this was an indication that these rejuvenators had a softening
effect on the virgin binder.
3. The results from the LAS test illustrated that the rejuvenators increased the number
of cycles to failure of the virgin binder at the two strain levels used for testing. This
also showed the softening effect.
4. The rheological properties of the master curves of the extracted and recovered
binders agreed with the data from the MSCR and LAS tests. The recovered binder
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
from the RAP/RAS mixture had a lower c relative to the recovered binders from
same mixtures with rejuvenators. This indicated that the rejuvenators reduced the
hardness of the binders since c is a measure of the overall hardness of the binder.
5. The percent air voids for the control mixture increased by 0.8 to 1.2% when RAP
and/or RAS was incorporated. Since the binder content for all mixtures was the
same, it was assumed that the increase in the air voids indicated that the hardened
binders of the RAP and RAS were not blending completely with the virgin binder,
resulting in a mixture with less total effective binder content. The use of
rejuvenators appeared to mitigate the increase in the air voids. This might be
attributed to the rejuvenators diffusing into the hardened binder and allowing it to
commingle with the virgin binder.
6. Based on the dynamic modulus data, the rejuvenators did soften the resultant binder
of the 40% RAP mixture causing the overall stiffness of the mixture to get closer to
the stiffness of the control mixture. However, the drop in stiffness for the RAS and
RAP/RAS mixtures was not as great as for the 40% RAP mixture. This might be
attributed to the RAS binder being stiffer than the RAP binder.
7. The HWTD showed that the rejuvenators increased the rutting and moisture
susceptibility of the 40% RAP and 5% RAS.
8. The data from the Texas Overlay Tester showed that the RAP, RAS, and RAP/RAS
mixtures exhibited a significant drop in the number of cycles to failure relative to
431

the control mixture. The incorporation of the rejuvenators improved the cracking
performance of the RAP, RAS, and RAP/RAS mixtures.
9. The TSRST data showed that the rejuvenators helped mitigate the loss in the low
temperature cracking of the mixtures due to the incorporation of RAP and RAS.
These data agreed with the rheological properties obtained using the CAM model,
particularly the R parameter.
10. The test results at 4C generally showed that there was blending of the rejuvenated
and virgin binder in the 35% RAP plus 5% RAS mixture. However, no conclusion
regarding the degree of blending on E* at the higher temperatures could be made.
11. Overall, the data and data analysis showed that asphalt rejuvenators can be used to
mitigate the stiffness of the resultant binder in a high RAP content and RAS
mixtures. Based on the results from the HWTD, it is recommended to revisit the
dosage used and develop a rational procedure for determining the proper dosage of
asphalt rejuvenators. Furthermore, the use of a polymer in addition to a rejuvenator
should be investigated as a way to balance the softening effect of an asphalt
rejuvenator.
10. References
AASHTO, American Association of State Highway and Transportation Officials.
AASHTO Provisional Standards. American Association of State Highway and
Transportation Officials (AASHTO), Washington, D.C., May 1993.
AASHTO, Standard Specifications for Transportation Materials and Methods of
Sampling and Testing, American Association of State Highway and Transportation
Officials, Washington, D.C., 30th Edition. 2010.
Anderson, D., Christensen, D., and Bahia H. Physical Properties of Asphalt Cement and
the Development of Performance Related Specifications. Journal of the
Association of Asphalt Paving Technologists, Vol. 60, 1991, pp. 437532.
Al-Qadi, L.I., Elseifi, M., and Carpenter, H.S. Reclaimed Asphalt Pavement A
Literature Review, FHWA-ICT-07-001. Illinois Center of Transportation, March
2007.
Bonaquist, R. Laboratory Evaluation of Hot Mix Asphalt (HMA) Mixtures Containing
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Recycled or Waste Product Materials Using Performance Testing. Publication

FHWA-PA-2005-006+98-32(19), Pennsylvania Department of Transportation,
Office of Planning and Research, 2005.
Bonaquist, R. Can I Run More RAP? Hot Mix Asphalt Technology. Sep/Oct2007, pp.
1113.
432

Boyer, E.R. Asphalt Rejuvenators-Fact or Fable. Asphalt Institute, Prepared for

presentation at TS2K workshop, August 2000.
Brownidge, J. The role of an asphalt rejuvenator in pavement preservation: Use and
need for asphalt rejuvenation. Compendium of papers from First International
conference on pavement preservation, 2010, chapter 5, pp. 351364.
Christensen, D., and Anderson, D. Interpretation of Dynamic Mechanical Test Data for
Paving Grade Asphalt. Journal of the Association of Asphalt Paving Technologists,
Vol. 61, 1992, pp. 67116.
Christensen, D., Pellinen, T., and Bonaquist, R. Hirsch Model for Estimating the
Modulus of Asphalt Concrete. Journal of the Association of Asphalt Paving
Copeland, A. Reclaimed Asphalt Pavement in Asphalt Mixtures: State of the Practice.
Publication FHWA-HRT-11-021, Turner-Fairbank Highway Research Center,
Federal Highway Administration, McLean, VA, April 2011.
FHWA-HIF-11-038. The Multiple Stress Recovery (MSCR) Procedure.
http://www.fhwa.dot.gov/pavement/materials/pubs/hif11038/tb00.cfm, Accessed
July 2012.
Hansen, K.R. Guidelines for Use of Reclaimed Asphalt Shingles in Asphalt
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Pavements. Information Series 136. National Asphalt Pavement Association,
Lanham, Md., 2009.
Hintz, C., Velasquez, R., Johnson, C., and Bahia, H. Modification and Validation of the
Linear Amplitude Sweep Test for Binder Fatigue Specification. Journal of the
Transportation Research Board, 2011, pp. 99106.
Huang, B., Li, G., Vukosavljevic, D., Shu, X., and Egan, K.B. Laboratory Investigation
of Mixing Hot-Mix Asphalt with Reclaimed Asphalt Pavement. Transportation
Research Record: Journal of the Transportation Research Board, No. 1929,
Transportation Research Board of the National Academies, Washington, D.C., 2005,
pp. 3745.
Karlsson, R. Investigations of Binder Rejuvenation Related to Asphalt Recycling.
Doctoral thesis. Royal Institute of Technology, Stockholm, Sweden, 2002
Mallick, R., Tao, M., OSullivan, K., and Frank, R. Why Not (Use Rejuvenator For)
100% RAP Recycling? Proceedings of the 89th TRB Annual Meeting, TRB,
Washington, DC, 2010.
MARC (Modified Asphalt Research Center), Draft Standard Method of Test for
Estimating Fatigue Resistance of Asphalt Binders Using The Linear Amplitude
433

sweep. University of Wisconsin, 2011, http://uwmarc.wisc.edu/linear-amplitude-

sweep.php, Accessed July 2012
Maupin, G.W. Investigation of the Use of Tear-Off Shingles in Asphalt Concrete.
FHWA/VTRC 10-R23, Virginia Transportation Research Council, May 2010.
Mogawer, W.S., Bennert, T., Daniel, J., Bonaquist, R., Austerman, A.J., and
Booshehrian A. Performance Characteristics of Plant Produced High RAP
Mixtures. Journal of the Association of Asphalt Paving Technologists (AAPT),
2012, pp. 403440.
OSullivan, K. Rejuvenation of Reclaimed Asphalt Pavement (RAP) in Hot Mix
Asphalt Recycling with High RAP Content. A Thesis Submitted to the Faculty of
the WPI in Partial Fulfillment of the requirement for the M.Sc. in Civil Engineering,
April 2011.
Owens Corning Forms, Owens Corning Forms Alliance with Earth911 to Expand
Footprint of Nation's Leading Shingle Recycling Program.
http://owenscorning.mediaroom.com/index.php?s=2370&item=66527, Accessed
July 2012.
Shen, J.A., Amirkhanian, S.J., and Lee, S.J. The Effects of Rejuvenating Agents on
Recycled Aged CRM Binders. International Journal of Pavement Engineering,
2005, pp. 273279.
Shen, J., Amirkhanian, S., and Miller, J. Effects of Rejuvenating Agents on Superpave
Mixtures Containing Reclaimed Asphalt Pavement. Journal of Materials in Civil
Engineering, May 2007a, pp. 376384.
Shen, J.A., Amirkhanian, S.J., and Tang, B.M. Effect of Rejuvenator on Performance-
based Properties of Rejuvenated Bitumen Binder and Mixtures. Construction and
Building Materials, 2007b, Vol21, No. 5, pp. 958964.
Swiertz, D., Mahmoud, E., and Bahia, H. Estimating the Effect of Recycled Asphalt
Pavements and Asphalt Shingles on Fresh Binder, Low-Temperature Properties
Without Extraction and Recovery, Journal of the Transportation Research Board,
2011, pp. 4855.
TxDOT, Test Procedure for Overlay Test. Texas Department of Transportation,
TxDOT Designation Tex-248-F, January 2009.
U.S. EPA Office of Solid Waste and Emergency Response, Tear-off Asphalt Shingles
Recycling. --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
http://www.epa.gov/oswer/iwg/pilots/docs/2005_0520_asphalt_formatted_final.pdf,
Accessed July 2012.
434

Zhou, F., Hu, S., Scullion, T., Chen, D., Qi, X., and Carlos, G. Development and
Verification of the Overlay Tester Based Fatigue Cracking Prediction Approach.
Journal of the Association of Asphalt Paving Technologists, Vol. 76, 2007a, pp.
627662.
Zhou, F.S., Chen, D., and Scullion, T. Overlay Tester: A Simple Performance Test for
Fatigue Cracking. Transportation Research Record: Journal of the Transportation
Research Board, No.2001, Transportation Research Board of the National
Academies, Washington, D.C., 2007b, pp. 18.
11. Discussion
DR. M. EMIN KUTAY: Im wondering about that degree blending analysis based on
Hirsch model. Once you extract the binder, you do the extraction (E*), instead of doing
that, once you extract it, take it back and actually remix with the aggregate to make a
new mix and run the E* again and see if you get a match. I wonder if anybody has done
it just to see if in doing the extraction process, you are not artificially evaluating the
combination of RAP binder and virgin binder. Do you know?
PROFESSOR WALAA MOGAWER: This is a very good point. However, to fabricate
E* samples using extracted and recovered binder would require a lot of binder to be
extracted and recovered. This would be a monumental effort. But its a brilliant idea, it
really is.
DR. KUTAY: Just to check if that method works
PROF. MOGAWER: Thanks.
PROFESSOR MIHAI MARASTEANU: Very nice presentation. Can you go back to the
slide that shows predicted versus measured?
PROF. MOGAWER: The good blending one or the bad blending one?
PROF. MARASTEANU: Yes, that one. We have done similar work using Hirsch model
and a method that was originally developed by Herve di Benedetto and his group. Our
analysis was done for low temperature only and we compared the extracted and the
back-calculated binder. Your slide shows that the measured E* is lower than the
predicted E*. The predicted E* is based on 100% blending, while in the measured you
do not see the full contribution of the aged since your modulus is lower. In other words,
there is not enough blending, which leads to a softer mix, which may be better for lo w
temperature. We have also seen similar results. Your low temperature strength results
were not that good, so Im wondering if some of this old binder that didnt blend acted
as a weakening interface in your mixture when you do your fracture test.
435
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

PROF. MOGAWER: It could be. The method we used is an empirical method that relies
on a macro scale testing. I would like a method where were looking in micro scale
testing. This year at TRB, there were a couple of papers from Europe that presented very
good work looking at the degree of blending. What I heard at TRB is only extracted and
recovered binders were used to look at the zone where the two binders touch each other
using an AFM. Were looking at that method as well.
PROF. MARASTEANU: Yeah. That also tells you that you really need to fracture these
specimens; stiffness is a bulk property, but this is not a homogeneous mixture to begin
with. Very nice work.
PROF. MOGAWER: Thank you very much.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
436

Effect of Rubber Characteristics on Asphalt
Binder Properties
J. Richard Willis*a, Pamela Turnera, Clayton Plemmonsa,

Carolina Rodeznoa, Tom Rosenmayerb, Codrin Darangac,
and Doug Carlsond
a
National Center for Asphalt Technology, Auburn University, Auburn, AL 36849
b
Lehigh Technologies Inc., Tucker, GA 30084
c
Blacklidge Emulsions, Gulfport, MS 39503
d
Liberty Tire Recycling, Pittsburgh, PA 15222
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
ABSTRACT: The use of scrap tires in ground tire rubber (GTR) modified binders has continued to
evolve since its introduction in the early 1960s. Currently, many states have developed recipe
specifications requiring contractors to blend GTR with asphalt binders using specified rubber
sizes, percentages, and grinding methods. While these specifications were developed based on
early research, the GTR industry has developed new methods and techniques which might improve
the quality of GTR in asphalt binders. The objective of this research was to assess how rubber
properties affect the properties of an asphalt binder. This objective was completed by blending 12
unique GTR samples with a singular asphalt binder at a loading of 10% rubber. Two of the
selected rubbers were additionally tested at 15% loading. These 14 GTR-modified asphalt binders
were then tested using the performance grade (PG), multiple stress creep recovery (MSCR), cigar
tube separation test, and softening point methodologies. Statistical analyses were conducted to
determine how particle size, grinding temperature, rubber chemistry, and surface area affected the
four test results. GTR particle size was the most influential parameter on the majority of the test
results. The smaller particle sizes improved the high and low temperature performance grade and
particle separation as tested by cigar separation tubes and the softening point.
KEYWORDS: ground tire rubber, asphalt binder, separation, binder grading
The oral presentation was made by Dr. Willis.
http://dx.doi.org/ 10.1080/14680629.2013.812845
437

WILLIS, TURNER, PLEMMONS, RODEZNO, ROSENMAYER, DARANGA, CARLSON
1. Introduction
The use of scrap tires in ground tire rubber (GTR) modified binders has continued to
evolve since its introduction in the early 1960s due to the enhancement in
performance including resistance to rutting, thermal, reflective and fatigue cracking
(Lee et al., 2006; Huang et al., 2002). Some other benefits reported include reduction in
maintenance, smooth riding, good skid resistance, and noise reduction (Kaloush et al.,
2002; Way, 2000).
The concept of using GTR in asphalt mixtures can be divided into two broad
categories: a dry process or a wet process. In the dry process, GTR is blended with the
aggregate before the asphalt binder is added in the mix. The GTR particles in this
process are generally coarser than those in the wet process and are considered as part of
the aggregate gradation (Huang et al., 2002). In the wet process, a rubberized asphalt
binder or GTR-modified asphalt binder is produced. The GTR is pre-blended with the
asphalt binder at high temperature and is then mixed with the aggregate.
The two main GTR-modified binders can also be divided into asphalt rubber and
terminally blended binder. For asphalt rubber, the range of GTR added typically ranges
from 15 to 22% by weight of the asphalt. Terminal blends use finely ground (No. 40
mesh or smaller) GTR and the amount of GTR may vary from 5-20% depending on the
application.
For asphalt rubber, during the blending process a reaction occurs when the GTR
particles absorb a portion of the oils in the asphalt binder and the particles swell (Shen
and Amirkharian, 2009). This reaction does not melt the GTR into the binder. This
reaction is compared to a compressed dry sponge in a water bath. As the sponge absorbs
the water, it swells and softens. For GTR, when the reaction with asphalt cement occurs,
it also swells and softens (Heitzman, 1992). This swelling increases the viscosity that
results in a thicker film coating on the aggregate particles in asphalt mixtures. The
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
increase in film thickness provides for a more durable HMA mixture showing increased
resistance to oxidative aging (Arizona Department of Transportation, 1977). For the
terminal blended binders, the blending takes place in the refinery or stationary asphalt
terminal and the component materials are heated over an extended period of time. This
results in dissolving of the rubber particles or complete digestion. This process allows
the product to be engineered specifically to meet the climatic conditions specified by the
Superpave performance graded system. (Hicks et al., 2010). In some cases, additives
are added to terminally blended binders in addition to the GTR.
GTR can be produced by three different processes. The cracker mill process is the
most commonly used. This process produces irregularly shaped particle sized from No.
4 to No. 40 sieve. The granulator process produces cubical, uniformly shaped particles
438

Effect of Rubber Characteristics on Asphalt Binder Properties
from 3/8 in. to No. 40 sieve. The micro-mill process produces a very fine GTR from No.
40 to No. 200 sieve. In practice, there are also two temperatures to produce GTR:
ambient temperature grinding and cryogenic grinding. The cryogenic grinding takes
place by fracturing rubber particles that have been frozen with liquid nitrogen.
Cryogenic fracturing of the rubber particles leaves the rubber angular, but can reduce the
particle size much finer than ambient grinding. Ambient grinding, however, leaves the
particles with a rougher texture which increases the particle surface area; however, the
particles are larger in size due to the tearing of the rubber.
The improvement in performance has been attributed to the interaction of GTR with
the base binders (Shen and Amirkharian, 2009). This interaction is influenced b y
different factors such as the amount of aromatic fraction in the binder, the temperature
and time duration of the blending reaction, the production method, particle size, specific
surface area and chemical composition of the GTR (Heitzman, 1992).
A study was conducted to evaluate the GTR-binder interaction effect and the GTR
particle effect (Putman and Amirkharian, 2009). They measured the rheological
properties of the modified binder with three sizes of GTR, two GTR processing methods
(ambient and cryogenic grinding), two GTR contents, and three binder sources. They
found that the interaction effect was significantly affected by the crude source and the
amount of GTR in the binder. The particle effect was most significantly affected by the
GTR content, followed by the GTR particle size.
A follow up laboratory study was conducted on the properties of GTR binders as a
function of GTR method and percentages (Lee et al., 2008). A total of 24 GTR binders,
(three binder sources, two processing methods and four GTR percentages) were used in
this study. Superpave binder tests were conducted using the rotational viscometer, DSR
and BBR to evaluate the viscosity and rutting and cracking properties of the modified
binders. They found that the ambient GTR resulted in higher viscosity values than the
cryogenic GTR at the same percentage usage, this was attributed to the increased surface
area and irregular shape of the ambient GTR. The study also concluded that, in general,
the higher GTR percentages appeared to lead to better fatigue cracking properties and
that the higher GTR percentage resulted in the lower stiffness. GTR produced with
ambient gridding showed a better resistance to low temperature cracking than those with
the cryogenic process. Finally it was found that the binder source influence on the binder
properties at low temperature was not as significant as the GTR percentage increased.
While previous studies have reported significant disadvantages to using crygogenic
GTR, using smaller particle sizes which can be produced using the cryogenic process
can increase the surface area of the cryogenic particles and thus might increase the
viscosity of the asphalt binder modified with cryogenic rubber. To assess this theory, a
439
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
study was developed to determine the most significant rubber properties which affect
asphalt binder properties.
1.1. Binder Content and Air Voids
The objective of this research was to assess how GTR properties affect the properties of
an asphalt binder. This objective was completed by blending 12 unique GTR samples
with a singular asphalt binder at a loading of 10% rubber. Two of the selected rubbers
were additionally tested at 15% loading. These 14 GTR-modified asphalt binders were
then tested using the performance grade (PG), multiple stress creep recovery (MSCR),
cigar tube separation test, and softening point methodologies. Both ambiently and
cryogenically ground rubbers were evaluated.
2. Methodology
The following section details the testing methodology used to quantify the properties of
the 14 GTR-modified binders developed for this study. Each of the 12 GTR products
was originally characterized for particle size and chemical compositions before it was
blended with a standard performance graded PG 67-22 asphalt binder at either 10 or
15% loading. The GTR binders were then tested for performance grade, creep recovery,
particle separation, and softening point.
2.1. Ro-Tap Particle Size Analysis
Ro-Tap testing was conducted to determine the particle size of the twelve GTR products
used in the study. The testing was conducted according to Lehigh Technologies testing
methodology using approximately 100 grams of GTR. This testing methodology is
similar to ASTM D5644. The GTR was mixed with Flow-aid, a product used to ease the
movement of the GTR through the sieves. Additionally, the Flow-aid is -#200 material;
therefore, it did not affect the test results on the +No. 200 sieves.
A set of sieves was chosen based on the estimated particle size of the sample. The
zero screen was designated as the sieve where 99.9% of the material should pass. The
designation sieve corresponded to the sieve where at least 90% of the materials pass.
Additional sieves were placed below the estimated designation sieve to further
characterize the size of the rubber materials. Two rubber balls were placed on each sieve
to aid the material in passing through the sieves without accruing material degradation.
440

The GTR and Flow-aid mixture was then poured into the top sieve and placed with
the stack of sieves in the Ro-Tap machine (Figure 1). The samples were agitated for at
least 10 minutes and for a maximum of 20 minutes based on the estimated particle size
of the product. The smaller the particle size, the longer the material was agitated in the
Ro-Tap.
Upon completion of agitation, each sieve was checked for fiber. Any accrued fiber
was removed from the sieve and weighed. Additionally, any rubber material adhering to
the rubber balls on each sieve was brushed from the balls back to the sieve. The mass of
the material on each sieve was quantified. The weights on each sieve and weight in the
bottom pan were then determined. The particle size was ultimately determined by
quantifying the last sieve where 90% or more GTR particles pass through the sieve.
This mesh size designation was consistent with ASTM D5603-01; however, this
specification has no provision for No. 16 mesh material or specific cuts of rubber such as
the -80/+140. Therefore, the standard classification has limitation. The Ro-Tap data
were also assessed for mean particle size.
Figure 1. Ro-Tap device.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`
441

2.2. Thermogravimetric Analysis (TGA)
TGA testing was conducted to analyze the GTR for its makeup using ASTM E1131-03.
This test determines the percent extractible, polymer, carbon black, and ash in each GTR
source. Testing was completed in a Perkin-Elmer TGA device. A sample was placed in
the testing device and the temperature ramped to 325C at 10C per minute in a nitrogen
environment. The percent of material lost at this temperature was calculated and deemed
to be extractibles. The temperature in the chamber was increased at a ramp of 10C per
minute in nitrogen until it reached the target temperature of 530C where the
temperature was held constant for 10 minutes. After the temperature was held constant,
the percent material lost was calculated. The material lost in this temperature range was
classified as polymer. The temperature was then increased again at 10C per minute to
850C in an oxygen environment where the percent loss quantified the percent carbon
black in the GTR sample. The remaining material was considered ash.
2.3. Specific Surface of GTR Samples
Quantachrome was contracted by the project sponsors to quantify the surface area of
each GTR material incorporated in the study. The methodology quantified the surface
area of each rubber source by determining the physical absorption on Krypton for each
GTR sample. To conduct this test, an approximately 1 to 3 gram sample of rubber was
transferred to a testing cell which was connected to a degassing unit. In ambient
conditions, a vacuum was used to remove the gas from the cell for 16 hours. The cell
was removed from the degassing unit and the sample was weighed. Multipoint
Brunauer-Emmett-Teller (BET) testing was then conducted using an ultra-high purity
Krypton gas at relative pressures of 0.05, 0.075, 0.10, 0.15, 0.20, 0.25, and 0.30 at a bath
temperature of 77.35K. The surface area was calculated by assessing the volumetric
changes between the gas adsorbed or desorbed and the amount of gas required to fill the
space around the sample in conjunction with Equation 1.
/ [1 ]
where: S = surface area (m2/g)

Wm = monolayer capacity
N = Avogadros number
A = cross-sectional area of Krypton
M = molecular weight of Krypton
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
442

2.4. GTR-Modified Binder Blending
Prior to blending, the virgin binder was heated in a 135oC oven until fluid enough to
pour (approximately 3 hours). While heating, the binder was stirred with a glass rod
every 30 minutes. Once heated, the binder was proportioned into pre-weighed cans and
set aside until needed for blending. The amount of GTR needed for each sample was
calculated based on the final weight of the asphalt binder and GTR blend to ensure a
representative sample; a significant portion of GTR was poured onto a tray and the
material was manually blended to ensure homogeneity. The required weight of GTR
was then split from this homogeneous pile.
Each can of binder was heated for 30 minutes at 135oC prior to being blended. A
heating mantle was used to ensure the temperature of the binder and rubber sample
remained constant during blending. The target temperature of the GTR and binder
mixture during blending was 163C. This was monitored through the use of a
thermometer probe. The GTR sample for each blend was added at a constant rate during
the first two minutes of blending. While adding the GTR, the stirring paddle was set at
700 RPM to help prevent rubber particles from being blown outside the system. Once all
the GTR was added, the blender was set to 1000 RPM and the material blended for 30
minutes. To ensure that each blend received the same style of blending, the blender
propeller was placed at a depth to ensure a one inch vortex during blending.
Immediately after blending, a small portion of the GTR binder was poured into an
aluminum tube for separation testing according to ASTM D7173-11 using inch tubes
instead of the specified 1 inch tubes due to their lack of availability. The rest of the
material was covered and set aside at room temperature for further testing.
2.5. Performance Grading
Prior to testing, the pre-blended cans of GTR binder were placed in a 163C oven for 30
minutes and then re-blended at 1000 RPM for 10 minutes. Again, the temperatures of the
samples were monitored with a lab thermometer and the target mixing temperature of
163C was held constant through the use of the heating mantle.
The 14 GTR binders were then tested and graded according to AASHTO M 320-10.
The standard 1 mm Dynamic Shear Rheometer (DSR) gap was used initially used on all
14 blends. Past research (Bahia and Davies, 1994; Bahia and Davies, 1995) has
suggested using a 2 mm DSR gap to improve the repeatability of PG testing on DSR
results. However, only the binder modified with the largest particle sizes was
inconsistent in its critical temperature analyses; therefore, a 2 mm gap was used to test
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
that blend. The variability of the testing was reduced using this procedure for that one
binder.
443

2.6. Multiple Stress Creep Recovery
In addition to the PG grading of the GTR binders according to AASHTO M 320-10, the
performance grades of the 14 GTR binders were also determined in accordance with
AASHTO MP 19-10. The multiple stress creep recovery (MSCR) test was performed on
the rolling thin film oven (RTFO) aged material in accordance with AASHTO TP 70-09.
Testing was conducted at 64C which is the average 7-day maximum pavement design
temperature in Auburn, Alabama. The same RTFO aged specimen used in the DSR was
used for MSCR testing. The MSCR results measure the non-recoverable creep
compliance. The acceptable non-recoverable creep compliance at 3.2 kPa and percent
differences for varying levels of traffic as specified in AASHTO MP 19-10 are given in
Table 1.
Table 1. Requirements for non-recoverable creep compliance.
Max Jnr3.2 Max Jnrdiff
Traffic Level
(kPa-1) (%)
Standard Traffic S Grade 4 .0 75
Heavy Traffic H Grade 2 .0 75
Very Heavy Traffic V Grade 1 .0 75
Extremely Heavy Traffic E Grade 0 .5 75
2.7. Separation Tubes
The separation tube test procedure is performed to determine the tendency of an asphalt
modifier to separate from the asphalt binder during static heated storage. If a modified
asphalt binder shows a tendency to separate during storage, this must be taken into
account either by providing some sort of agitation or stirring, or by re-formulating the
binder. Testing was conducted in accordance with ASTM D7173-11, Standard Practice
for Determining the Separation Tendency of Polymer from Polymer Modified Asphalt
with the exception of the tube size.
Immediately after the blending was performed, 50 0.5g of the hot asphalt binder
was poured into a cylindrical aluminum tube which was closed on one end and
supported vertically in a rack. Once filled, the open end of the tube was sealed to
prevent air from reaching the sample. The tube and rack were placed in an oven at 163
5C and allowed to condition for 48 1 hr.
After the conditioning was completed, the tube assembly was removed from the oven
and immediately placed in a freezer for at least 4 hours. Once frozen, the tube was
removed from the freezer and cut into three equal portions. The middle portion was
444
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

discarded and the top and bottom portions were placed in separate 3 oz. tins and heated
until sufficiently fluid that asphalt binder flows from the metal tubing. The two asphalt
binder samples were stirred thoroughly and used for comparative testing to determine if
the asphalt binder in the top and bottom portions of the tube was significantly different
in terms of critical high temperature grading in the DSR.
2.8. Softening Point
The Ring-and-Ball softening point test (ASTM D36-09) measures the temperature at
which an asphalt binder begins to flow at elevated service temperatures. It was used
here in conjunction with the separation tube procedure to determine if there was a
difference in properties of the asphalt binder from the top and bottom portions of the
separation tube. A difference in the softening point values of the upper and lower
portions of the tube would indicate that the modifier had separated from the asphalt
binder during the storage period.
The softening point test was performed by heating the asphalt binders obtained fro m
the top and bottom portions of the aluminum separation tube and pouring them into two
small brass rings. The asphalt binders and rings were allowed to cool at room
temperature for at least 30 minutes. The asphalt binder was then trimmed flush with the
top of the rings.
A ring holder assembly was placed in a fluid bath along with two steel balls where
they were allowed to condition to a starting temperature of 5 1C for 15 minutes. A
bath of distilled water was chosen for this study based on the expected softening point
(between 30 and 80C) of the material being tested. Other fluid options exist for
materials with higher softening points.
After the conditioning period, the steel balls were placed on top of the asphalt binder
in the rings. The bath was heated at a rate of 5C/min until the asphalt binder softened to
the point that the balls dropped and touched the bottom plate of the assembly. The
temperature at which each ball touched the plate was recorded as the softening point for
the asphalt binder in that ring. Softening points were considered different if they vary by
more than 1.2C between samples tested by the same operator.
3. Rubber Particle Characterization
Twelve unique rubber sources were included in this research study. To determine the
effects of the rubber characteristics on the GTR-modified binder, each rubber product
was characterized by particle size, chemical make-up, surface area, tire type, and
grinding temperature.
445
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

3.1. Ro-Tap Particle Size Analysis
Each rubber product initially was characterized for particle size using the Ro-Tap
procedure. It was found that the coarsest GTR was rubber number 12 with 30% retained
on the No. 20 sieve. The finest rubber particle was rubber No. 5 where 77.3% of the
rubber particles passed the No. 200 sieve. Most of the products ranged in size from No.
30 to No. 60 mesh materials with few exceptions. It should also be noticed that three
products had trace amounts of fiber which were removed from their particle size
distribution.
While engineers commonly refer to rubber particles using the mesh with 90%
passing, there is a current trend in the rubber industry to refer to particles by the mean
particle size which is more representative of the actual particle sizes included in the
product. Table 2 provides both the mesh and mean particle size for the products
evaluated in this study.
Table 2. Ro-Tap Results.
Grinding
GTR Temperature Mesh Size Mean Particle Size, microns
1 Cryogenic -40 180
2 Cryogenic -40 180
3 Ambient -40 180
4 Cryogenic -80 cut* 105
5 Cryogenic -140 50
6 Ambient -30 250
Cryogenic
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
7 -30 250
8 Cryogenic -80 125
9 Ambient -30 250
10 Ambient -30 250
11 Ambient -20 250
12 Ambient -16 600
*-80 cut is a specific cut of rubber particles that are designed to have a required range
of particle sizes between two sieves.
446

3.2. Thermogravimetric Analysis
TGA testing was conducted to quantify the makeup of each ground rubber product
according to ASTM E1131-03. In this test, the percent extractibles, polymer, carbon
black, and ash were calculated for each of the twelve GTR sources. The results are
given in Table 3. As can be seen, most of the GTR products had extractibles which fell
within a 4% band, polymer contents which varied by no more than 6.5%, carbon black
differences of 4.7% maximum, and ash contents within 5 % of each other. Therefore,
while there was some deviation in the test results of the ground rubber sources, most of
the particles have similar makeups. The 6.5% difference in polymer content of some
rubber sources could be inflated when used at 10 to 15% loading. This would
theoretically give one modified binder an additional 0.65% polymer than another which
could influence both high and low temperature critical temperatures.
Table 3. TGA results.

GTR Percent Material, %
Extractibles Polymer Carbon Black Ash
1 6 .7 8 5 8 .1 1 3 0 .1 3 4 .8 0
2 7 .3 5 5 7 .9 5 2 9 .8 6 4 .7 0
3 1 0 .1 7 5 6 .1 0 2 8 .3 6 5 .1 8
4 8 .7 9 5 4 .4 8 3 0 .8 8 5 .6 7
5 1 0 .9 7 5 2 .4 3 3 0 .1 7 6 .2 6
6 1 0 .3 7 5 1 .8 9 3 1 .4 0 6 .2 0
7 8 .1 1 5 2 .7 8 2 9 .3 5 9 .6 3
8 8 .6 7 5 3 .7 3 3 2 .3 1 5 .3 1
9 7 .7 3 5 8 .4 6 2 7 .5 6 5 .9 6
10 9 .6 9 5 2 .2 9 3 1 .5 4 6 .3 6
11 9 .8 6 5 4 .3 6 3 0 .6 5 5 .0 3
12 9 .4 9 5 5 .5 3 2 8 .4 3 6 .4 4
3.3. Specific Surface of GTR Results
The surface area of each rubber product was characterized by Quantichrome using the
BET procedure previously described in conjunction with Krypton gas. The surface area
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
for each rubber product is given in Table 4. As seen, the product with the smallest
surface area was rubber No. 9. The material with the highest surface area was rubber No.
5. This material was cryogenically ground and had the smallest particle size.
447

Table 4. Specific surface results.
GT R Mesh Grinding Temperature Surface Area (m2/g)

1 -40 Cryogenic 0 .4 0 7
2 -40 Cryogenic 0 .0 7 9
3 -40 Ambient 0 .4 0 0
4 -80 cut Cryogenic 0 .2 7 5
5 -140 Cryogenic 0 .7 5 1
6 -30 Ambient 0 .1 0 4
7 -30 Cryogenic 0 .0 4 4
8 -80 Cryogenic 0 .1 0 4
9 -30 Ambient 0 .0 5 6
10 -30 Ambient 0 .1 1 4
11 -20 Ambient 0 .0 9 2
12 -16 Ambient 0 .0 7 9
4. GTR-Modified Binder Characterization
While it was important to characterize the particle sizes and makeup for each GTR
product, it is even more critical to understand how each GTR product will affect the
properties of the asphalt binder it is modifying. Therefore, all 12 GTR products were
blended with a standard PG 67-22 asphalt binder at a loading rate of 10%. Two of the
GTR products (No. 6 and No. 11) were blended at loading rates of 15% to determine
how the additional 5% rubber affected the binder properties. These two products were
chosen at the discretion of the sponsors. Each of the 14 GTR-modified binders was then
analyzed for performance grade, multiple stress creep recovery grade, and separation
potential via the separation tubes and softening point procedure.
4.1. Performance Grade
The GTR-modified asphalt binders were tested and graded according to AASHTO M
320-10. Table 5 summarizes the true grade and performance grade of each blended
binder. All of the rubber products were tested using a 1 mm gap in the DSR with the
exception of rubber No. 12. This product was the largest in size and showed signs of the
448
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

GTR particles affecting the DSR test results. In order to get a more accurate reading of
the GTR-modified binder properties, a 2 mm gap was used to characterize this product in
addition to the 1 mm gap.
The base binder used in this study had a true grade of PG 70.0-25.5. A dosage rate of
10% was used for each rubber product in an attempt to modify the base binder to a PG
76-22. All 14 GTR-modified binders had a PG critical high temperature of at least 76C.
Additionally, the two rubber blends which were tested at both 10 and 15% loading had
higher critical temperatures when loaded at 15%.
Both grinding method and particle size seemed to make no difference on the critical
high temperature of the blended binder. Figure 2 graphically shows the relationship
between GTR particle size and critical high temperature of the blended binders. The
best relationship that could be developed has a low R2 value of 0.19 suggesting particle
size did not really influence the true grade of the binder; however, outliers were not
removed from the data set. Using an average value of the GTR-modified asphalt binders
with ambient No. 30 mesh material did not improve the goodness-of-fit. These results
were based on limited particle sizes and one binder.
Table 5. Performance Grade results.
GTR Grinding Mesh Loading, True Performance
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Temp Size % Grade Grade
Base NA NA 0% 70.0 25.4 70 22
1 Cryo -40 10% 79.0 23.0 76 22
2 Cryo -40 10% 80.4 24.2 76 22
3 Ambient -40 10% 82.1 20.8 82 - 16
4 Cryo -80 cut 10% 81.2 25.4 76 22
5 Cryo -140 10% 77.9 25.6 76 22
6 Cryo -30 10% 82.2 23.2 82 22
6 Cryo -30 15% 86.7 19.3 82 16
7 Ambient -30 10% 82.8 23.1 82 22
8 Cryo -80 10% 83.6 24.9 82 22
9 Ambient -30 10% 79.8 20.4 76 16
10 Ambient -30 10% 80.7 23.6 76 22
11 Ambient -20 10% 83.1 24.6 82 22
11 Ambient -20 15% 87.9 21.3 82 16
12 (2 mm gap) Ambient -16 10% 84.7 21.8 82 16
12 (1mm gap) Ambient -16 10% 76.3 21.8 76 16
449

160 700
140 600
Mean Particle Size, microns

120
500
Particle Size, mesh
Mesh Mean Particle Size, micron
100
400
80
300
60
200
40
20 100
0 0
75.0 80.0 85.0 90.0
Critical High Temperature, C
Figure 2. Effect of particle size on critical high temperature.
The grinding temperature seemed to have little effect as both cryogenically and
ambiently ground particles were able to achieve the desired PG 76-22 performance grade
when blended with a PG 67-22 binder. Additionally, both ambient and cryogenic GTR
products were able to modify the PG 67-22 binder to a PG 82-22 binder at 10% loading
in some cases. To further support this point, a one-sided t-test was used to statistically
compare the high temperature grade of the four GTR-modified asphalt binders using
ambiently ground No. 30 mesh material to the GTR-modified asphalt binders using No.
30 mesh cryogenically ground rubber products. The ambient No. 30 mesh GTR-
modified binders had an average critical high temperature grade of 81.6C while the
GTR-modified binder with No. 30 mesh cryogenically ground rubber had a critical high
temperature grade of 82.2C. There was no evidence of a statistical difference (p-value
= 0.509) between the GTR-modified asphalt binders using ambiently or cryogenically
ground No. 30 mesh material.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
450

The General Linear Model (GMA, = 0.05) was attempted to statistically assess if
particle size, grinding temperature, tire type, surface area, polymer content, and loading
rate statistically affected the critical high temperature grade of the rubber modified
binders (Table 6). Using surface area and polymer content as covariates due to the
nature of the data, none of the factors were statistically significant in explaining the high
temperature grade of the modified binder. This could be due to the unbalanced nature of
the dataset. While none of the variables were statistically significant, the model only
explained 68% of the total variability. The two factors which explained the most
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
variability in the model were mean particle size (28.5%) and surface area (21.2%).
Four of the 14 GTR-modified binders failed to meet the critical low temperature
specification of -22C. Of the four materials which did not meet this specification, two
were rubber-binder blends which used 15% rubber, and one of the other rubber modified
asphalts used the GTR with the largest particle sizes. The fourth product which graded
at -16C was the -30 with a high amount of fines. In each of the four cases, the m-value
caused the rubber blend to fail the low temperature criterion. A possible solution for
three of the GTR blends could be to reduce the amount of GTR added to the binder.
Three of the GTR-modified binders are relatively close to passing the -22C grade and
have high temperature true grades substantially higher than the 76C target. A slight
reduction in the GTR for these materials should achieve a passing low temperature grade
while still maintaining the desired PG 76. It should only take a small decrease in this
GTR to achieve the target grade. The fourth GTR-modified binder which did not pass
may also benefit from a decrease in GTR percentage, but it may prove difficult to obtain
the -22C grade without dropping below the PG 76C for the high grade.
Figure 7 shows the relationship between particle size and critical low temperature.
While the relationship was stronger than the relationship for critical high temperature (R2
= 0.3615), the relationship is not robust enough to draw any strong conclusions about the
relationship between particle size and critical low temperature. The goodness-of-fit for
this relationship was reduced due to the number of No. 30 GTR sources with differing
low temperatures. Using the mean particle size did not improve the relationship (R2 =
0.3414).
While it is difficult to determine if there is a difference in the effect of grinding
method on low temperature binder performance, the only blends which failed the -22C
criterion at 10% loading used an ambient grind. All the binder blends using
cryogenically ground materials passed this specification.
451

Table 6. ANOVA Critical High Temperature Results.

Source DF Seq SS Adj SS Adj MS F P
Surface Area* 1 2 0 .1 0 0 .9 5 0 .9 5 0 .0 6 0 .8 2 6
Polymer* 1 0 .7 9 0 .2 2 0 .2 2 0 .0 1 0 .9 1 5
Loading Rate 1 0 .5 3 1 .2 2 1 .2 2 0 .0 8 0 .8 0 3
Mean Particle Size 5 2 7 .0 4 1 7 .8 2 3 .5 6 0 .2 4 0 .9 1 6
Tire Type 2 1 2 .5 3 1 2 .9 3 6 .4 7 0 .4 3 0 .7 0 0
Temperature 1 3 .5 9 3 .5 9 3 .5 9 0 .2 4 0 .6 7 4
Erro r 2 3 0 .2 5 3 0 .2 5 1 5 .1 3
Total 13 9 4 .8 4
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The GLM ( = 0.05) was used to assess if particle size, grinding temperature, tire
type, surface area, polymer content, and loading rate statistically affected the critical low
temperature grade of the GTR-modified binders (Table 7). Using surface area and
polymer content as covariates due to the nature of the data, again, none of the factors
were statistically significant in explaining the low temperature grade of the modified
binder. However, unlike the high temperature critical temperature, approximately 92%
of the model variability was explained by two factors. The two factors which had the
most influence on the low temperature grade were the mean particle size (34.6%) and
loading rate (21.3%). The effect of loading rate was evident as changing from 10 to 15%
loading always increased the critical high temperature by at least one binder grade.
452

Figure 3. Critical low temperature vs. particle size.
Table 7. Critical low temperature ANOVA.

Surface Area* 1 4 .9 9 6 0 .4 3 0 0 .4 3 0 0.21 0.690
Polymer* 1 3 .6 6 6 .5 6 7 6 .5 6 7 3.24 0.214
Loading Rate 1 10.323 9 .2 0 9 9 .2 0 9 4.54 0.167
Mean Particle Size 5 16.766 14.563 2 .9 1 3 1.44 0.459
Tire Type 2 3 .5 6 7 8 .8 2 5 4 .4 1 2 2.17 0.315
Temperature 1 5 .3 5 3 5 .3 5 3 5 .3 5 3 2.64 0.246
Er r o r 2 4 .0 5 8 4 .0 5 8 2 .0 2 9
Total 13 48.472
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
*Covariate
4.2. MSCR
To determine the performance grade in accordance with AASHTO MP 19-10, the

MSCR test (AASHTO TP 70-09) was conducted at 64oC (the average 7 day maximu m
453

pavement design temperature for Auburn, Alabama) to determine the non-recoverable

creep compliance for all the binders. The same RTFO aged specimen used in the
dynamic shear rheometer (DSR) test according to AASHTO T 315-09 was used in the
MSCR test. Table 8 summarizes the MSCR testing results. The base binder was graded
as an H for heavy trafficking.
Ten of the GTR-modified binders met the requirements for the highest trafficking
level E. The E stands for extremely high traffic or trafficking of greater than 30
million equivalent single axle loads (ESALs) and standing traffic. When compared to
the PG grades, all the GTR-modified binders had critical high temperatures greater than
76C.
Four of the rubber modified binders exceeded the maximum allowable Jnr difference
of 75%. This is most likely due to extremely low Jnr values at 0.1 kPa. However, while
these blends did not meet the E specification, the Jnr0.1 and Jnr3.2 results suggest they
should have adequate resistance to permanent deformation under those conditions.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Extremely low Jnr values suggest high resistance to permanent deformation.

Table 8. MSCR results.
GTR Loading, Jnr % Re c o v e r y Traffic
% 0 .1 3 .2 % 0.1 kPa-1 3.2 kPa-1 Level
kPa-1 kPa-1 Diff
Base 0% 1 .1 5 0 1 .3 5 3 1 7 .6 8 1 2 .9 9 5 .6 1 6 H
1 10% 0 .1 7 8 0 .2 0 2 1 3 .5 2 4 2 .9 0 3 6 .6 9 E
2 10% 0 .1 3 9 0 .1 6 6 1 9 .1 9 5 1 .2 3 4 3 .5 5 E
3 10% 0 .1 2 3 0 .1 6 0 5 7 .1 8 4 6 .8 8 3 0 .5 3 E
4 10% 0 .1 7 5 0 .2 0 1 1 4 .9 0 4 4 .6 6 3 8 .0 2 E
5 10% 0 .2 7 3 0 .3 4 4 2 6 .0 8 3 6 .8 6 2 4 .5 0 E
6 10% 0 .1 5 0 0 .2 0 1 3 4 .2 1 5 0 .5 9 3 6 .6 8 E
6 15% 0 .0 6 2 0.1949 2 1 6 .2 6 8 .8 7 4 4 .7 2 E *
7 10% 0 .1 2 2 0 .1 8 3 5 0 .8 8 5 9 .4 6 4 2 .7 3 E
8 10% 0 .1 2 3 0 .1 9 0 4 6 .3 9 5 8 .4 4 4 1 .8 7 E
9 10% 0 .0 9 2 0 .1 2 7 3 7 .6 8 6 1 .6 2 4 9 .5 8 E
10 10% 0 .2 0 1 0 .2 3 3 1 5 .9 5 4 3 .5 6 3 6 .4 2 E
11 10% 0 .0 8 6 0 .1 5 9 8 5 .8 1 6 9 .1 6 4 6 .4 5 E *
11 15% 0 .0 3 0 0 .1 9 3 5 5 4 .0 8 5 .9 4 5 0 .4 3 E *
12 10% 0 .0 8 8 0 .1 2 2 3 9 .4 1 6 3 .0 3 5 0 .4 4 E
12 10% 0 .0 9 6 0 .7 2 0 6 5 2 .5 9 5 .2 0 6 6 .4 5 E *
454

4.3. Separation Tubes
Separation tubes were used in conjunction with the DSR to determine if particle size and
grinding method affected the overall separation of the GTR particles from the asphalt
binder. The amount of separation was quantified by the difference in the critical high
temperature of the modified binder removed from the top half of the separation tube
compared to the critical temperature grade of the modified binder removed from the
bottom half of the separation tube. The results are shown in Table 9.
There currently is no consensus pass/fail criterion for this testing methodology;
therefore, one can only rank the performance of the modified binders. The binder which
had the greatest difference was the GTR-modified binder which used the GTR product
with the largest sized particles. No trend was noticed when comparing the mesh rubber
size to the difference in critical high temperatures; however, there was a slight trend
between separation and mean particle size (Figure 4). The finer the mean particle size,
the less separation occurred.
Table 9. Separation tube results.

GTR Loading, % Critical High Temperature, C
Top Bottom Absolute % Difference
Difference
1 10% 8 2 .7 5 1 0 2 .7 1 9 .9 5 2 4 .1 1
2 10% 8 1 .7 9 1 0 2 .6 2 0 .8 1 2 5 .4 4
3 10% 8 6 .0 8 9 9 .9 2 1 3 .8 4 1 6 .0 8
4 10% 7 4 .9 8 8 1 .5 6 6 .5 8 8 .7 8
5 10% 7 9 .9 3 9 2 .5 3 1 2 .6 0 1 6 .6 1
6 10% 7 7 .7 1 1 0 2 .1 2 4 .3 9 2 3 .8 9
6 15% 7 9 .8 5 1 1 0 .5 3 0 .6 5 3 8 .4 0
7 10% 7 7 .0 4 1 0 2 .4 2 5 .3 6 3 2 .9 0
8 10% 7 8 .8 7 9 7 .5 2 1 8 .6 5 2 3 .6 5
9 10% 8 5 .6 3 9 7 .0 7 1 1 .4 4 1 1 .7 9
10 10% 7 6 .6 7 1 0 3 .6 2 6 .9 3 3 5 .1 2
11 10% 7 5 .8 4 1 0 1 .7 2 5 .8 6 3 4 .1 0
11 15% 9 4 .3 1 0 7 .4 1 3 .1 0 1 2 .2 0
12 (2 mm) 10% 7 9 .7 8 1 9 7 .5 117.72 147.56
455
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 4. Particle size vs. DSR separation.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
To statistically assess this hypothesis, the GLM ( = 0.05) was used to assess the
variability of the test results in terms of particle size, grinding temperature, loading rate,
surface area, and polymer content (Table 10). The only factor which statistically
influenced the separation results was particle size. Smaller particles showed less
separation. Over 86% of the model variability was explained using this term alone. The
second most influential factor was surface area; however, it only accounted for 7.8% of
the data. Overall, the model had less than 1% error.
Table 10. Separation Test ANOVA Results.

Surface Area* 1 753.73 0 .1 3 0 .1 3 0 .0 0 0 .9 6 2
Polymer* 1 108.26 116.96 116.96 2 .5 8 0 .2 5 0
Loading Rate 1 281.86 1 4 .9 7 1 4 .9 7 0 .3 3 0 .6 2 4
Mean Particle Size 5 8343.08 7608.22 1521.64 33.55 0 .0 2 9
Tire Type 2 6 .5 1 2 5 .8 0 1 2 .9 0 0 .2 8 0 .7 7 9
Temperature 1 5 4 .3 5 5 4 .3 5 5 4 .3 5 1 .2 0 0 .3 8 8
Erro r 2 9 0 .7 2 9 0 .7 2 4 5 .3 6
Total 13 9638.51
*Covariate
456

4.4. Softening Point
The softening point test was conducted according to ASTM D36-09 to determine if the
rubber particles separated when mixed with the asphalt binder. The results are presented
in Table 11. Since the separation tube procedure does not specify a value at which
separation can be said to occur, the value used as an indication of separation for this
testing was 1C. The GTR-modified binder which showed the least amount of
separation used the GTR with the smallest particle size. All of the other GTR-modified
binders showed at least 4C difference.
While most of the rubber particles seemed to settle in the asphalt binder, the research
team wanted to determine if the particle size or grinding technique influenced the
amount of separation. It should be noted that the one particle size which did not separate
in the separation tubes was the smallest particle size. Figure 5 shows the relationship
between particle size and softening point differences. Using the developed relationship
and a maximum allowable temperature difference of 1C, the data suggest that particle
sizes larger than No. 100 mesh may have a tendency to separate from the asphalt binder.
It should also be noted that only one product had a particle size less than No. 80 mesh.
This suggests that contractors using larger particle sizes will need agitation systems to
ensure the rubber and binder stay blended.
To statistically assess this hypothesis, the GLM ( = 0.05) was used to assess the
factors which had the most influence on the softening point results (Table 12). None of
the variables were statistically significant; however, mean particle size (30.5%) and
surface area (22.9%) explained the most variability in the model. The smaller particles
with larger surface areas were less susceptible to settling than the larger particles with
smaller surface areas.
457
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 11. Softening point results.

GTR Loading, % Softening Point, C
Top Bottom Absolute Difference % Difference
1 10% 5 9 .7 7 7 .2 1 7 .5 2 5 .6
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
2 10% 6 1 .7 7 3 .3 1 1 .6 1 7 .2
3 10% 6 2 .2 7 1 .4 9 .2 1 4 .8
4 10% 5 8 .6 6 3 .0 4 .4 7 .2
5 10% 6 0 .6 6 1 .1 0 .5 0 .8
6 10% 5 9 .7 7 0 .7 1 0 .3 1 5 .9
6 15% 6 1 .9 7 5 .0 1 3 .1 1 9 .1
7 10% 5 8 .9 6 6 .7 7 .8 1 2 .4
8 10% 6 1 .4 7 8 .0 1 6 .6 2 3 .8
9 10% 6 2 .8 7 2 .8 1 0 .0 1 4 .7
10 10% 5 8 .9 7 2 .5 1 3 .6 2 0 .7
11 10% 5 8 .9 6 9 .4 1 0 .5 1 6 .4
11 15% 6 6 .4 7 3 .3 6 .9 9 .9
12 10% 5 9 .2 6 9 .7 1 0 .5 1 6 .3
Figure 5. Softening point vs particle size.
458

Table 12. Softening point ANOVA results.

Surface Area* 1 5 9 .9 1 1 6 .1 7 1 6 .1 7 1 .5 2 0 .3 4 3
Polymer* 1 4 9 .3 3 6 .6 9 6 .6 9 0 .6 3 0 .5 1 0
Loading Rate 1 0 .5 7 4 .2 2 4 .2 2 0 .4 0 0 .5 9 3
Mean Particle Size 5 7 9 .8 0 123.75 2 4 .7 5 2 .3 3 0 .3 2 7
Tire Type 2 1 3 .4 7 2 2 .4 8 1 1 .2 4 1 .0 6 0 .4 8 6
Temperature 1 3 7 .4 7 3 7 .4 7 3 7 .4 7 3 .5 3 0 .2 0 1
Erro r 2 2 1 .2 4 2 1 .2 4 1 0 .6 2
Total 13 261.78
*Covariate
5. Conclusions and Recommendations
Based on the results of this research, the following conclusions can be drawn.
All 14 modified binders met the criterion for high temperature grade classification
of PG 76. Additionally, five of the 12 blends loaded at 10% and both blends using
15% loading met the criterion of a PG 82 binder.
Surface area and particle size of the rubbers had the most influence on increasing
the critical high temperature grade of the modified binder. Grinding temperature
(ambient versus cryogenic) had little to no influence on the results due to additional
surface area of the minus No. 40 cryogenically ground material. This is contrary to
some other reported studies; however, these other reported results were based
primarily on viscosity testing.
Four of the 14 asphalt blends using rubber particles did not meet the -22C PG
specification. Both of the mixtures using 15% rubber loading were graded as -16C
binders.
Polymer content and grinding method had little influence on the low temperature
properties of the modified binders. Loading rate and particle size had the most
influence; however, strong relationships were not developed due to the scatter of the
data.
While all of the GTR-modified binders met the MSCR Jnr requirements for the
highest traffic level at 64C, four of them did not meet the percent difference Jnr
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
requirements. The low Jnr0.1 values for these binders may have artificially inflated
the percent difference value; however, given the low Jnr values it would be expected
that these binders would have adequate rutting resistance in the field.
459

87% of the variability in DSR data from the separation tubes was explained by the
particle size of the rubber. The larger particle sizes showed greater discrepancies
between the high temperature grades of the binder from the top of the tube
compared to the bottom of the tube.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The GTR products with smaller particle sizes separated less in the softening point
test. Particle size and surface area explained the most variability in the model.
The loading rate (10%) was arbitrarily chosen for this study and using less than 10%
rubber could modify binder properties enough to meet a PG 76 criterion. These data
support the idea of states moving to a performance grade specification for binders
instead of specifying raw materials.
Using the previous stated conclusions, the following recommendations are made:
Ground tire rubber should be considered an appropriate asphalt binder modifier to
achieve critical high temperature performance in mixtures.
Ambient and cryogenically ground GTR performed equivalently in terms of binder
modification and separation. Specifications should not distinguish between the two
types of materials when the GTR is -#30 or smaller.
A rubber content of 10% is an appropriate level of loading for increasing a PG 67-
22 binder to a PG 76-22. Increasing the rubber content of the binders increased the
critical low temperature grade of the modified material. To achieve a similar lo w
temperature performance grade (i.e. -22C), one might need to use a binder which
has a lower virgin critical temperature (i.e. -28C).
When using GTR particle sizes larger than No. 100 mesh, continuous agitation
systems should be used to prevent separation of the rubber particles and asphalt
binder.
6. References
Arizona Department of Transportation, The Chemical and Physical Properties of

Asphalt-Rubber Mixtures, ADOT-RS-14(162), 1977.
Bahia, H., Davies, R. Effect of Crumb Rubber Modifiers on Performance Related
Properties of Asphalt Binders, Journal of the Association of Asphalt Paving
Technologists, Vol. 63, 1994, pp.414449.
460

Bahia, H., Davies, R. Factors Controlling the Effects of Crumb Rubber on Critical
Properties of Asphalt Binders, Journal of the Association of Asphalt Paving
Heitzman, M. State of the Practice Design and Construction of Asphalt Paving
Materials with Crumb Rubber Modifier, FHWA-SA-92-022, Federal Highway
Administration, 1992.
Hicks, R., Cheng, D., Duffy, T. Evaluation of Terminal Blend Rubberized Asphalt in
Paving Applications, Report No. CP2C-2010-102TM, California Pavement
Preservation Center, 2010.
Kaloush, K., Witczak, M., Way, G., Zborowski, A. Abojaradeh, M., Sotil, A.
Performance Evaluation of Arizona Asphalt Rubber Mixtures Using Advanced
Dynamic Material Characterization Tests: Final Report, Arizona State University,
Tempe, Arizona, 2002.
Lee, S., Akisetty, C.K., Amirkhanian, S. The effect of crumb rubber modifier on the
performance properties of rubberized binders in HMA pavements, Construction
and Building Materials, Vol. 22, Issue 7, 2008, pp. 13681376.
Lee, S., Amirkhanian, S., Shatanawi, K. Effects of crumb rubber on aging of asphalt
binders, Proceedings of the Asphalt Rubber 2006 Conference, Palm Springs,
California, 2006, pp. 779795.
Huang, B., Mohammad, L., Graves, P.S., Abadie, C. Louisiana Experience with Crumb
Rubber-Modified Hot-Mix Asphalt Pavement, Transportation Research Record,
No. 1789, 2002, pp. 113.
Putman, B., Amikrhanian, S. Crumb Rubber Modification of Binders: Interaction and
Particle Effects, Proceedings of the Asphalt Rubber 2009 Conference, Nanjing,
China, 2009, pp. 437448.
Shen, J. Amirkharian, S. Comparison of the Properties of Laboratory and Field
Prepared CRM Binders, Proceedings of the Asphalt Rubber 2009 Conference,
Nanjing, China, 2009, pp. 437448.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Way, G. Flagstaff I-40 Asphalt Rubber Overlay Project Ten Years of Success,
Transportation Research Record, No. 1723, 2000, pp. 4552.
461

7. Discussion
MR. SEBASTIAN PUCHALSKI: Very nice presentation. Have you had a chance to
take a look at storage stability in terms of your PG grading for long-term storing at high
temperatures?
DR. J. RICHARD WILLIS: You have to hold the separation tubes at elevated
temperatures. I dont remember the exact time, but 48 hours is kind of what is ringing a
bell with me in terms of the specification. So youre holding it at a heated temperature
for two days and then seeing how much of everything falls out. So that would kind of
give you an idea in terms of long-term storage of how you could hold it in a tank at an
elevated temperature without causing the separation issues. I think that kind of helps us
answer those questions a little bit.
MR. PUCHALSKI: I see. So you havent seen any deterioration in terms of PG grade?
DR. WILLIS: Well, what you're going to see is if the binder is not being agitated, you're
going to see a lower PG grade at the top of the tank than at the bottom. In terms of the
entire tank losing PG grade, I havent seen any data that would support that or that
would even suggest that.
MR. PUCHALSKI: Okay. Thank you.
DR. WILLIS: You are welcome.
PROFESSOR CARL MONISMITH: Do you have only one asphalt source in the
Southeast?
DR. WILLIS: No, we do have multiple sources, but due to funding on the project, we
only really wanted to look at one.
PROF. MONISMITH: Different asphalts have different chemistry, and I am wondering
if the materials behave a little differently.
DR. WILLIS: There have been studies that show the binder source does play a part in
the overall performance grade of the binder.
PROF. MONISMITH: Amy Epps did a study a number of years ago with asphalt rubber
binders, using two different asphalt sources. The low temperature parameter was
affected by these two binder sources sufficiently to change the grade of the binder in the
low temperature range.
DR. WILLIS: I completely agree with you that the binder source is going to play a
critical component in how the GTR and the binder interact.
DR. HUA QIN LIU: Maybe I missed it during the presentation, what was the
temperature you used when you were preparing the crumb rubber asphalt? How long did
you keep the crumb rubber asphalt at that temperature? And Im surprised that the lo w
462
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

temperature grading at -22C got worse comparing with the base asphalt because there is
some extender oil in the crumb rubber. When the crumb rubber is mixed in hot asphalt,
the extender oil will migrate to the asphalt. You would observe that the low temperature
performance was improved if there is no oxidation or evaporation of the extender oil.
DR. WILLIS: What we did is the base binder was heated for Im trying to remember
the specifics but I think about three hours at a temperature just to make sure it was
completely fluid. I dont remember the exact mixing temperature. Im thinking it was
165F; I know its written in the paper. But that is the temperature where the mixing
occurred. And then all the binders were left to let the rubber digest into the asphalt
binder for... All the testing was conducted the next day. Everything was re-blended to
make sure that if any separation had occurred, the rubber particles were back into
suspension. So it wasnt like we were going in as soon as we mixed it, testing it, putting
the binder through the performance grade. We did give the rubber time to digest into the
asphalt binder.
DR. LIU: So the material was kept in the oven overnight?
DR. WILLIS: No, they were not kept in the oven overnight. Most of the binder blending
was conducted first thing in the morning, and then overnight they were not left in the
oven. They were reheated the next morning, reblended. We used a method developed by
the Clemson University and so thats why Im a little hazy on all the details on that right
now.
DR. LIU: Okay. Thanks.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
MR. GALE PAGE: A question and a comment. First is the question. When you did the
PG grading, did you fix the intermediate temperature that you tested all the binders at? It
makes a difference because if the high temperature characteristic of a binder increases,
the intermediate temperature at which you will test will increase correspondingly. So I
guess thats my question. Did you fix the intermediate temperature during your testing?
DR. WILLIS: Ill be honest. I did not look at the intermediate temperature data.
MR. PAGE: Well you need to test at the intermediate temperature to determine PG
grading.
DR. WILLIS: Yes. I do not believe it was fixed. I believe that.
MR. PAGE: The intermediate test temperature will make a difference in some of your
PG grading results.
DR. WILLIS: I believe that Pamela Turner went through the specification to determine
it, but it was not a fixed parameter.
MR. PAGE: Some agencies using the current PG grading system do fix the intermediate
temperature using the environmental temperatures and then test polymer-modified or
463

modified asphalts to determine their high temperature grading, but use a fixed
intermediate temperature. Because as John DAngelo said with the MSCR test, its the
environmental temperatures that make a difference. Using the MSCR test you dont want
to be testing high temperature characteristics of a binder at 76 or 82C. Another
comment that I have is with the MSCR test, I always felt that testing temperatures could
certainly be simplified. Agencies went through this whole process with neat binders.
Using the LTPPBind program, I believe that in Florida we would have had at least three
different PG neat binder grades. As an agency, we simplified it by specifying one PG
grade, and we called it our neat environmental temperature grade. Now we have the
MSCR test and were back to using LTPPBind again to determine the environmental
temperature PG grade again. Within our state we have terminals that will deliver from
the north part of the state to the south part of the state. And if you implement a MSCR
test using the LTPPBind program to determine the PG grade, you're really complicating
things, and simplifying it by choosing a single (or just a few) PG grades certainly would
help agencies implement the MSCR test.
DR. WILLIS: I agree. I know Alabama has gone to one temperature for the entire state.
We did this testing before they set it up at 67F. So the test that they require is a little bit
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
higher in terms of temperature than what we tested at, but we did use LTPPBind. But I
agree with you.
PROFESSOR SHADI SAADEH: Very nice presentation. I have just one curious
question. When you did your statistical analysis, you had a lot of parameters. Have you
considered those fixed factors or random factors or a mix of this and that?
DR. WILLIS: There were things like tire type, grinding temperature. Those we set as
covariants. Things like the measured properties, we did have as variable results. So it
was a mix of test results and covariants in that statistical analysis. So we did try to take
some parameters out, move them around and see if interactions played. But these
consistently gave us the best results in terms of explaining the most variability, and the
model is when we put these six in together.
PROF. SAADEH: Thank you.
464

Understanding Mechanisms Leading to
Asphalt Binder Fatigue in the Dynamic
Shear Rheometer (DSR)
Cassie Hintza* and Hussain Bahiab
a
Department of Civil, Construction, and Environmental Engineering, North Carolina
State University, Raleigh, NC 27695;
b
Department of Civil and Environmental Engineering, University of Wisconsin -
Madison, Madison, WI 53706
ABSTRACT: Asphalt binder fatigue resistance is important in determining overall pavement fatigue
performance. One procedure commonly used to characterize asphalt binder fatigue resistance is
the time sweep test, which consists of repeated cyclic loading in the Dynamic Shear Rheometer
(DSR). Generally, apparent changes in complex modulus and phase angle with respect to number
of loading cycles are used to define fatigue performance. Although results from DSR testing have
been shown to correlate well with asphalt mixture fatigue performance, the mechanisms
responsible for changes in binder properties in the DSR were previously not well understood.
Results in this study demonstrate that fracture can explain the changes in loading resistance
during fatigue testing in the DSR. Under cyclic torsional loading of cylindrical specimens,
fracture manifests as a circumferential crack that starts at the periphery of a sample and
propagates inward, reducing the effective sample radius. Digital visualization of binder
specimens following testing allows for determination of the fractured and intact sample radius.
Predictions of fracture propagation based on measurements of loading resistance and fracture
mechanics concepts agreed favorably with direct measurements based on visualization.
Furthermore, it is shown in this study that fracture morphology and progression of crack growth
of asphalt binders under time sweep testing match those observed for other materials under
similar loading conditions. Based on these findings, fracture mechanics concepts are used to
present an analysis framework that allows for predicting binder fatigue life at any loading
amplitude using the results of a single time sweep test.
KEYWORDS: asphalt binder, fatigue, fracture, failure mechanism
The oral presentation was made by Dr. Hintz.
http://dx.doi.org/ 10.1080/14680629.2013.818818
465
--`,`,,,``,,```,,`,,,,,`,``,````-`-

HINTZ, BAHIA
1. Introduction
Fatigue cracking is one of the primary modes of failure in asphalt pavements. Asphalt
mixtures used as a pavement wearing surface are composite materials consisting of
aggregate, asphalt binder, and air voids. Cracking typically occurs either within the
asphalt binder or at the asphalt binder - aggregate interface. Thus, asphalt binder fatigue
resistance plays a critical role in overall pavement performance.
One of the most widely used procedures to evaluate asphalt binder fatigue resistance
is the time sweep test, which consists of repeated cyclic torsional loading at constant
amplitude in the Dynamic Shear Rheometer (DSR). In the DSR, a cylindrical asphalt
binder specimen is inserted between two parallel plates. The top plate rotates whereas
the bottom plate is fixed. A depiction of the DSR test set-up is provided in Figure 1.
Generally, 8 mm diameter specimens with 2 mm height are used in time sweep testing.
Results of time sweep testing of asphalt binder have been shown to correlate well with
asphalt mixture fatigue performance, indicating the procedure captures asphalt binder
contribution to mixture fatigue (Bahia et al., 2001).
Figure 1. Schematic of DSR.
Generally, apparent changes in viscoelastic properties (e.g., |G*|, |G*|sin) with

number of cycles of loading are used to define fatigue performance of the asphalt binder.
However, the mechanisms responsible for apparent changes in these viscoelastic
properties were previously not well understood. Fatigue testing of binders is conducted
at intermediate temperatures for which asphalt binders exhibit complex viscoelastic
behavior. Thus, multiple phenomena could potentially contribute to changes in
properties (e.g., complex moduli and phase angle) during fatigue testing, including
thixotropy, plastic flow, micro-defect/void formation, and macro fracture.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
466

Understanding Mechanisms Leading to Asphalt Binder Fatigue in the DSR
A fundamental understanding of the damage mechanisms leading to reductions in

loading resistance during time sweep testing is needed to determine if the DSR is an
appropriate method for fatigue characterization. Additionally, there is a concern that
analysis of fatigue measurements using Viscoelastic Continuum Damage (VECD)
concepts requires empirical assumptions of the damage progression rate power factor
(i.e., ). For asphalt binders, the damage progression rates are estimated from the slope
of the log storage modulus versus log frequency local relationship (i.e., m value) (Hintz
et al., 2011) measured in linear viscoelastic range. Understanding the damage
mechanism(s) leading to fatigue, and a better estimate of damage or crack propagation
rate, can improve the current analysis methods and enable mechanistic fatigue
characterization of asphalt binders and mixtures.
2. Hypothesis
It is hypothesized that changes in sample radius due to cracking, rather than changes in
viscoelastic properties, are responsible for the apparent changes in load carrying capacity
during fatigue testing of asphalt binders in the DSR. If this hypothesis is confirmed,
fracture mechanics concepts can be applied for mechanistic characterization of fatigue
resistance of asphalt binders.
Under cyclic torsional loading of cylindrical specimens, fracture generally manifests
as an edge crack. An edge crack is a circumferential crack starting at the periphery of a
sample that propagates inward as loading is applied (Aboutorabi et al., 1998), effectively
reducing the sample size. A depiction of edge fracture is provided in Figure 2 where ri is
the initial sample radius, a is the crack length and r is the intact sample radius.
Figure 2. Fracture schematic.
It should be noted that apparent edge fracture in DSR testing can be caused by flow
instability resulting from the development of normal stresses (Keentok and Xue, 1999).
467
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
This phenomenon is more properly termed "edge flow" as it is not an actual fracture
phenomenon. Previous researchers have attributed edge fracture in DSR testing of
asphalt binders to flow instability (Anderson et al., 2001). In a previous study by the
authors, a detailed analysis of normal forces during time sweep testing was conducted
(Hintz et al., 2011). Results demonstrated that at the testing temperatures used for
binder fatigue testing, normal stresses were negligible compared to the torsional stresses.
In this study, normal forces were not observed and thus flow instability was ruled out
as a cause of damage. Furthermore, the morphology of the specimen cracked surfaces
following testing indicated fracture planes rather than flow was responsible for geometry
changes during time sweep testing.
3. Research Methodology
Several measures were taken to evaluate if fracture can account for the changes in
loading resistance during fatigue testing of asphalt binders in the DSR. An experimental
study was conducted comparing changes in loading resistance during testing with direct
measurements of crack lengths. Additionally, the fracture morphology and trends in
crack growth of asphalt binders under time sweep testing were compared with other
materials under similar loading conditions.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
3.1. Prediction of Crack Length from Changes in Loading Resistance
DSRs are only capable of measuring / controlling torque (T) and deflection angle ()
directly. Hence, strain () and stress () are calculated using these measurements coupled
with the sample geometry (height (h) and radius (r)) as follows:

[1 ]

2
[2 ]
3
Equations 1 and 2 can be combined to arrive at the following equation for complex
modulus, |G*|:
2
| | [3 ]
4
468

In calculating |G*|, rheometers assume constant sample geometry. Thus, changes in

effective sample radius due to fracture could result in apparent changes in |G*| based on
DSR output. Following this logic, if one assumes a constant |G*|0, relative changes in T
and can be used to predict the change in sample radius due to fracture using the
following equation derived from rearranging Equation 3:
2
4 [4 ]
| |0
3.2. Fracture Morphology
Observations of fracture morphology provide information regarding fracture

mechanisms and can explain trends in crack growth. If fatigue fracture is the cause of
fatigue degradation in the DSR, the fracture morphology of asphalt binders following
testing should resemble the fracture morphology of other materials under similar loading
conditions.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Under torsion loading, fracture can initiate and propagate in two modes: shear (Mode
III) and tension (Mode I). If fracture under torsional loading is caused by tension,
cracking will occur +/- 45 degrees relative to the specimen axis (i.e., along the principal
plane) whereas in Mode III fracture will occur along axial or radial shear planes.
The fracture morphology of cylindrical specimens with an edge crack subjected to
cyclic torsion has been studied in steel (Tschegg et al., 1983) and polymeric materials
(Keller et al., 2008). A depiction of steel and poly-dimethylsiloxane (an elastomeric
polymer) with an edge crack under cyclic torsion is provided in Figure 3. Experimental
observations demonstrate the initial fracture surface, corresponding to the outer edge of
the sample, appears macroscopically flat, indicative of Mode III crack growth. However,
as the crack propagates inward, surface interactions between crack faces become
significant, leading to a rough fracture surface comprised of radial peaks and valleys.
This type of fracture morphology is referred to as "factory roof"(Tschegg et al., 1983).
Factory roof morphology can lead to mixed mode fracture due to interaction between
crack faces.
469

HINTZ, BAHIA
( a) (b)
Figure 3. Fracture morphology of (a) steel (Tschegg et al., 1983) and (b) poly-
dimethylsiloxane (Keller et al., 2008) under cyclic, torsional loading.
3.3. Trends in Crack Growth
Materials with an edge crack under cyclic torsional loading demonstrate distinct stages
of crack propagation, commonly referred to as shallow and deep crack growth.
Hence, if asphalt binder is undergoing fatigue fracture during time sweep testing, similar
trends are expected. These general trends in crack growth are depicted in Figure 4 where
a is the crack length and da/dN is the crack growth rate (length / cycle). Aboutorabi et al.
(1998), demonstrated in the initial stage of crack propagation both crack growth rate and
energy release rate increase as crack length increases, termed shallow crack growth.
However, once a critical crack length is exceeded, both energy release rate and crack
growth rate decrease with increasing crack length, which Gent and Yeoh (2003) referred
to as deep crack growth.
Figure 4. Depiction of evolution of crack growth.
470
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Two explanations exist for the trends in crack growth rate. The first is concerned
with the energy source driving fracture and the second involves fracture surface
interactions. Initially, the energy driving crack growth is derived from the material near
the sample edge in the immediate vicinity of the crack (Gent and Yeoh, 2003). As the
crack grows inward, the available energy source increases and consequently crack
growth rate increases, until limited by the physical boundaries of the specimen. At this
point, the crack becomes deep and energy for crack growth is derived from the interior
of specimen. The interior specimen size decreases with increasing crack depth. Hence,
energy release rate and crack growth rate decrease with increasing crack depth.
Additionally, as the crack becomes deep, interaction between fractured surfaces increase
which leads to mixed mode fracture and consequently reduces the crack growth rate
(Tschegg et al., 1983).
To determine if fracture is responsible for fatigue damage in the DSR, deflection angle
controlled tests were conducted and terminated at different loading durations in order to
attain visual measurements of crack length at various stages of the fracture process.
Visual observations of edge fracture can only be obtained through destruction of samples
following test termination. Hence, in order to obtain visual images of fracture at different
stages of crack growth, numerous tests with varying duration are needed.
Note that if fracture is occurring as anticipated, it is very difficult to conduct a true
strain controlled test due to continuous changes in sample geometry. Thus, what is
currently referred to as a strain controlled time sweep test in the DSR is more accurately
described as a deflection angle or displacement-controlled test. In this study, time sweep
tests used for hypothesis verification were conducted using sinusoidal loading with
deflection angle (i.e., angle of rotation) amplitude of 0.02 radians and 10 Hz loading
frequency at 25C. Testing was conducted using standard specimen geometry, consisting
of cylindrical specimens 8mm in diameter with 2mm height.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Following each test, samples were chilled to 0C and yellow paint was applied
around the sample perimeter using a foam paintbrush. The paint used was an acrylic
polymer emulsion, produced by Liquitex. The paint was fluid at room temperature,
which allowed it to seep into cracks to the crack tip. Additionally, it is believed that due
to very narrow crack depth at the end of the crack, capillary action forces the paint to
reach very close to the actual end of the crack.
Following paint application, the paint was allowed to dry and then the upper portion
of the sample adhering to the DSR spindle is detached from the lower part of the sample
adhering to the lower plate. Photos of paint application and the remaining specimen after
471

HINTZ, BAHIA
spindle detachment are provided in Figure 5. The purpose of paint application was to
enhance contrast between the fractured and intact portions of samples for easy detection
using digital image processing. After detachment of spindle, samples were photographed
and digital imaging analysis was used to determine the change in effective sample area
due to fracture.
( a) (b )
Figure 5. Procedure for attaining image of fractured surface: (a) paint application and
(b) sample after spindle detachment.
Two binders were tested using this procedure: an unmodified binder and a binder
modified with cross linked sytrene-butadiene-styrene (SBS). Both binders were tested
after Rolling Thin Film Oven (RTFO) laboratory aging. Tests were terminated in five
minute intervals beginning once a significant decrease in torque was achieved and
ending once torque measurements became relatively constant with time. Table 1
provides a summary of the experimental plan for testing the hypothesis and verification.
Table 1. Experimental plan for verification of hypothesis.
Asphalt Binder Times of Test
Termination (min)
Ne a t 25, 30, 35, 40, 45,50, 55
PG 64-22
SBS modified 40, 45, 50, 55, 60, 65, 70,
PG 70-22 75, 80
472
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

To validate findings of the study based on digital image analysis and to allow for the
development of a mechanistic fatigue analysis framework, additional binders were tested
using deflection angle controlled testing at varying temperatures and loading amplitudes.
A loading frequency of 10 Hz was used for all tests. Two base, unmodified asphalts were
used. These base binders were tested without modification and also with polymer
modification. Two modifiers were used: polyethylene (PE), a plastomeric modifier; and
styrene-butadiene-styrene (SBS), the elastomeric modifier. SBS was used in two forms:
cross linked (x-linked) and linear. All binders were evaluated using time sweep tests at a
minimum of two temperatures. To eliminate effects of initial stiffness, iso-stiffness
conditions of |G*|sin at 10 Hz loading equal to 30 MPa, 6.5 MPa and 2 MPa were used
to determine test temperatures. Master curves were constructed using frequency sweep
tests at multiple temperatures to determine the temperatures corresponding to the iso-
stiffness conditions for the individual binders. These iso-stiffness conditions led to
testing temperatures ranging from approximately 10C to 30C. A summary of the
binders tested and corresponding testing conditions is provided in Table 2.
Table 2. Experimental plan for verification and model development.
Undamaged Strain (%)

Base Binder Modification
|G*|sin=6.5MPa |G*|sin=2MPa |G*|sin=30MPa
none 2 , 3 , 3 .5 2 ,3
2% SBS
2, 3 3, 5
A (x-linked)
(PG 64-22) 2% SBS
2, 3, 4 2, 3
(linear)
2% 2
2, 3, 4 2, 3
Polyethylene
none 2 , 3 , 3 .5 2, 3 2
B
(PG 64-22) 2% SBS
2, 3, 4 2, 3
(x-linked)
473
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
5. Analysis of Results
5.1. Comparison between Predicted and Measured Crack Length
The hypothesis that fracture is the primary mechanism of material degradation in fatigue
testing of asphalt binders in the DSR was tested by using constant deflection angle
testing and comparing predictions of crack length based on torque measurements using
Equation 4 with direct measurements of intact sample radii obtained from image
analysis. A similar analysis could be conducted by comparing the change in deflection
angle with visual observations in a constant torque amplitude test. However, failure
occurs rapidly in controlled torque tests. Therefore, controlled deflection angle tests
were used in this study because failure occurs at a slower rate and there is no
accumulated deformation.
Torque values with normalization by the initial torque for the unmodified asphalt
binder versus number of loading cycles for each of the samples that were loaded at a
frequency of 10 Hz are shown. The first sample was terminated at 25 minutes of loading
and then aborted, fractured and imaged. Six more samples were tested each with five
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
minutes more loading, with the last one imaged after 55 minutes of continuous loading.
The torque measurements for all samples are overlapped and presented in Figure 6.
Results demonstrate a reduction in torque of roughly 75% at the maximum loading time
of 55 minutes (75% loss of effective complex modulus). To simplify the plot, only the
additional loss in normalized torque is shown for each sample. The results show
significant change in torque between each sample as the time interval is increased and
the tests were aborted. Each sample was imaged after the test was aborted to measure the
crack length and compare to the reduction in torque.
Figure 7 presents the normalized torque results of modified asphalt binder for tests of
nine samples loaded to various loading times. Roughly 75% reduction in torque is
reached at the maximum loading time of 80 minutes. It can be seen that the modified
binder exhibits higher fatigue resistance than the unmodified binder as the torque
decreases at a slower rate. The enhanced resistance should result in reduced crack
growth rate, as will be shown later.
474

Figure 6. Normalized torque results for the unmodified binder, based on averaged data.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 7. Normalized torque results for the modified binder, based on averaged data.
475

HINTZ, BAHIA
Equation 4 was used to predict the intact sample radius based on changes in torque
with number of loading cycles. Additionally, photographed samples following
termination of tests were used for direct measurement of the intact sample radius. An
example of a photographed sample following testing is provided in Figure 8. It can be
seen the fractured edge is covered in yellow paint, allowing for easy determination of the
crack length.
Figure 8. Fractured sample with paint application.
Microscope imaging of fractured samples after paint application was conducted to

verify that the paint reached the crack tip and did not diffuse into the asphalt binder. An
image of the interface between the painted and un-painted portions of the samples
obtained from an optical microscope is provided in Figure 9. At the inner edge of the
paint, the fracture front can be seen. The fracture morphology at the crack front is
characterized by radial peaks and valleys. It can be seen that these peaks contain yello w
paint but there is no paint in front of these features. Thus, the microscope imaging
demonstrates the paint is sufficient for crack detection.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 9. Microscope image of fractured sample with paint application.
476

Images of fractured samples were processed using MATLABTM. A program was

written using built-in MATLABTM image analysis functions in order to detect the
fractured portion of the sample. Based on Gonzalez et al. (2004) the most appropriate
image processing functions were identified. Upon identification of the suitable image
filters, a variety of functions and sequences of these functions were tried to determine
the optimal combination and sequence for crack identification. Figure 10 provides a
depiction of the image processing steps. Once the fractured portion of the sample was
detected, the outer area including cracked and un-cracked portions of the sample (i.e.,
initial sample area) and intact sample area (i.e., black area inside white fractured portion)
can be calculated in terms of number of pixels. Using the areas, measured in pixels, radii
are calculated assuming the area is circular (i.e., assuming area equals r2). Noting that
the initial sample radius equals 4 mm, one can convert the radii measured in pixels to
millimeters. The difference between the radius calculated for the initial and intact radius
is used to estimate crack length. Thus, crack length is taken to be the difference between
the initial an intact sample radius.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
477

HINTZ, BAHIA
Figure 10. Image processing steps.
Figure 11 depicts the progression of edge fracture based on visual observation for the
unmodified binder. It can be seen that as loading time increases, the intact sample size
decreases, indicating fracture propagation. Visual observations of edge fracture in the
modified binder at different loading times are shown in Figure 12. Similar trends are
observed in the modified binder; fracture progresses as loading time increases and
consequently the intact sample area decreases.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
478

(a) 30 minutes (b) 40 minutes (c) 55 minutes
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 11. Visualization of edge fracture of unmodified binder at varying loading
durations.
40 minutes (b) 50 minutes (c) 60 minutes

Figure 12. Visualization of edge fracture of modified binder at varying loading
durations.
Crack lengths determined from image analysis were compared with crack lengths
predicted from torque measurements. The crack lengths predicted from torque
measurements and visualization of edge fracture (i.e., "measured") for the unmodified
and modified binders are shown in Figures 13 and 14, respectively. Additionally, the
comparisons between measured and predicted crack lengths for the unmodified and
modified binders are presented in Tables 3 and 4, respectively. Results indicate good
agreement between visual observations and torque predictions of crack length.
Furthermore, results match the trends reported in the literature (Fleischman et al., 2001):
the torque predicted values of crack length are slightly less than the measured fracture,
particularly at higher crack lengths. This is thought to be a result of the fractured edge
carrying a small portion of the load due to surface interactions at the crack front.
Therefore, the comparison between crack length measurements using digital image
analysis and predictions based on torque measurements indicates a change in sample
geometry due to fracture is the cause for apparent changes in material properties during
time sweep testing of asphalt binders in the DSR.
479

HINTZ, BAHIA
Figure 13. Comparison between crack lengths predicted from torque measurements and
measured based on digital image analysis for unmodified binder.
Figure 14. Comparison between crack lengths predicted from torque measurements and
measured based on digital image analysis for modified binder.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,
480

Table 3. Comparison between measured and torque predictions of sample radii

for the unmodified binder.
Intact Radius (mm)

Loading Time (min) Measured Predicted from Torque
25 3 .4 4 3 .4 1
30 3 .3 5 3 .4 0
35 3 .2 2 3 .2 4
40 2 .9 0 3 .0 7
50 2 .7 8 2 .8 7
55 2 .6 2 2 .7 7
60 2 .6 1 2 .7 2
Table 4. Comparison between measured and torque predictions of sample radii

for the modified binder.
Intact Radius (mm)

Loading Time (min) Measured Predicted from Torque
40 3 .5 4 3 .5 8
45 3 .3 7 3 .4 8
50 3 .3 0 3 .3 9
55 3 .1 6 3 .2 0
60 3 .1 0 3 .1 4
65 2 .9 8 2 .9 8
70 2 .9 2 2 .9 6
75 2 .8 6 2 .9 4
80 2 .8 6 2 .9 4
481
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
5.2. Analysis of Fracture Morphology
Figure 15 is a photograph of a fatigue fractured asphalt binder specimen (without paint

application). The fracture morphology matches experimental observations in steel,
rubber, and elastomers under torsional fatigue loading where crack growth originates in
Mode III and radial peaks and valleys (factory roof) morphology develops due to
surface interactions (Fleischman et al., 2001; Tschegg et al., 1983; Keller et al., 2008).
Examples are shown in Figure 3. Fracture does not occur at a 45 angle relative to
specimen axis, indicating fracture is not caused by tension and hence fracture is
occurring primarily in Mode III.
Figure 15. Fatigue fractured sample.
5.3. Analysis of Crack Growth Stages
Given that the fracture morphology of asphalt binder under fatigue loading in the DSR
resembles other materials, it is also expected that similar trends in fatigue crack growth
rate will be observed in asphalt binders and other materials under similar loading
conditions. Crack growth rate (da/dN) with respect to number of loading cycles, based
on torque predictions of crack length, for the unmodified and modified binders are
shown in Figures 16 and 17, respectively. Initially there is a decrease in da/dN with
increasing crack length. This contradicts the crack growth trends for torsional Mode III
fracture reported in the literature, which demonstrates an increase in the rate of energy
release and hence, crack propagation with increasing crack length for shallow cracks.
However, the results reported in the literature are based on crack propagation of notched
482
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

specimens. Crack growth can be divided into two phases: crack initiation and crack
propagation. Trends in crack propagation may be invalid during crack initiation due to
complex geometry dependence of initiation (Hertzberg, 1989), which could explain the
decreasing magnitude of da/dN during the initial portion of crack growth. Excluding
initiation, results match expected trends. There is an increase in crack growth rate with
crack length, assumed to be shallow crack growth, followed by a decrease in crack
growth rate as the crack becomes deep. These results provide promising evidence edge
fracture is responsible for fatigue in asphalt binders tested in the DSR.
Figure 16. Predicted crack growth rate as a function of number of loading cycles for the
unmodified binder.
483
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
Figure 17. Predicted crack growth rate as a function of number of loading cycles for the
modified binder.
6. Fracture Mechanics Based Analysis Framework for Fatigue Characterization
Based on the finding that cracking is the main cause of damage in asphalt binders during
time sweep testing, an analysis framework for characterization of asphalt binder fatigue
resistance based on fracture mechanics concepts was developed. The analysis framework
allows for predicting fatigue life at any loading amplitude using the results of a single
time sweep test (Hintz, 2012).
In this framework, fatigue failure is defined as the peak in energy release rate (Gf)
and crack growth rate (da/dN), corresponding to the transition from shallow to deep
crack growth, denoted by the arrow in Figure 18. Although the definition is
phenomenological, it is considered easier to quantify and a better approach than
arbitrarily selecting a percentage reduction in effective stiffness as fatigue failure.
Energy release and crack growth rates were found to peak at a crack length of
approximately 0.3 mm regardless of binder type and test temperature based on results of
experimental plan shown in Table 2. This is demonstrated for unmodified binder B,
which was not used in the comparison between torque predictions and digital image
measurements of crack length, as shown in Figure 19. Furthermore, as demonstrated in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 20, which shows fatigue life results at the |G*|sin = 6.5 MPa iso-stiffness
484

condition at two loading amplitudes, clearly demonstrates that the time sweep test with
the proposed failure definition is clearly able to distinguish between the relative fatigue
performance of different binders and clearly captures the benefit of polymer
modification.
Figure 18. Failure definition.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(a)
485

HINTZ, BAHIA
(b)
(c)
(d)
Figure 19. Trends in crack growth and energy release rates for binder B at (a) |G*|sin
= 6.5MPa, i = 3%, (b) |G*|sin = 6.5MPa, i = 4%, (c) |G*|sin = 2MPa, and (d) i =
2%, |G*|sin = 30MPa, i = 2%.
486
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 20. Fatigue life results at |G*|sin = 6.5 MPa iso-stiffness condition.
It should be noted that the failure definition depicted in Figure 18 is only valid for
standard 8 mm parallel plate geometry with a 2 mm sample height. Fracture is dependent
on sample size and geometry when there are significant effects of a fracture process
zone. Thus, it is expected that the crack length at the peaks in energy release and crack
growth rates depends on the sample geometry. In order to verify this hypothesis, two
asphalt binders were tested using a sample height of 3 mm as opposed to 3 mm at an
initial, undamaged strain amplitude of 3%. Crack lengths were approximated based on
torque measurements. Figure 21 depicts crack growth rate as a function of crack length
using a sample height of 3 mm. As expected, results demonstrate trends in crack growth
rate are dependent on the sample geometry. Results indicate that when the sample height
is 3 mm the crack growth rate peaks at a crack length of approximately 0.6 mm.
487
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
Figure 21. Crack growth rate as crack length (a) function of a when sample height = 3
mm.
Details on the development of the fracture mechanics based analysis framework are
presented in Hintz (2012). A brief overview is presented here. The general form of the
model is presented in Equation 5.
= [5 ]
where: i is the undamaged, initial strain amplitude,
=+ [6 ]
and
= 2 [7 ]
The parameters I and P in Equation 6 represent contributions from crack initiation

and crack propagation, respectively. No precise definition for distinguishing between
crack initiation and propagation exists (Hertzberg, 1989). However, crack initiation and
propagation can be separated through observation of the relationship between crack
488
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

growth rate (da/dN) and energy release rate (Gf). Schapery (1975) proposed that
viscoelastic materials follow a power law relationship between Gf and da/dN during
crack propagation, which was validated for asphalt binders under time sweep testing.
Results of time sweep testing of a typical asphalt binder at various stiffness conditions,
specified in terms of |G*|sin values (corresponding to different test temperatures) and
loading amplitudes, specified in terms of undamaged, initial strain are shown in Figure
22. It can be seen that a power law relationship between da/dN and Gf exists with the
exception of data at low Gf values, corresponding to small crack lengths associated with
crack initiation. Based on these findings, the transition from initiation to propagation
was found to occur at a crack length of 0.05mm, (for standard 8mm parallel plate
geometry), regardless of binder type and testing temperature.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 22. Relationship between energy release rate and crack growth rate.
These findings led to the development of a crack propagation model based on the
relationship between Gf and da/dN during shallow crack propagation as follows:

= [8 ]

where k and are curve fit parameters and Gf is defined as follows:

| |0 2
= ( + (1 ))3 (1 ) [9]
2 2 ( )
489

HINTZ, BAHIA
where | |0 is the undamaged complex modulus, h is the sample height, ri is the initial
sample radius (4mm), a is the crack length, and c is a numerical factor equal to 0.1 for
asphalt binders tested using 8mm parallel plate geometry. A depiction of crack
propagation curves for all binders tested is provided in Figure 23. It can be seen that the
relationship between Gf and da/dN is independent of loading amplitude, indicating
model parameters k and can be determined using the results of a single time sweep test.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
( a)
(b )
490

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
( c)
(d )
491

HINTZ, BAHIA
( e)
(f)
Figure 23. Crack propagation curve fitting for (a) Binder A, (b) Binder A + SBSX, (c)
Binder A + SBSL, (d) Binder A + PE, (e) Binder B, (f) Binder B + SBSX.
Numerical integration of Equation 9 yields the following model relating number of

cycles of crack propagation to failure (Np) and undamaged, initial strain amplitude (i):
= [1 0 ]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
492

where B is defined in Equation 7, and P in Equations 8 and 6 is defined as:

1 2 2 ( )
(| 3 [11]
|0 )

+ 1 1
where ai is the crack length at the start of crack propagation (0.05 mm) and af is the
crack length at failure (0.3 mm). Because model parameters k and are independent of
loading amplitude, parameters P and B can be determined using the results of a single
time sweep test.
As demonstrated in Figure 22, there is no clear trend in the relationship between Gf
and da/dN during crack initiation. Therefore, a rigorous means to characterize crack
initiation could not be developed. However, it was determined that the number of cycles
of crack initiation, Ni follows the same loading amplitude dependency as Np. Thus, a
crack initiation model can be defined as follows:
[1 2 ]
where the parameter B is defined as in Equation 7 and can be determined using crack
propagation data. Note that since B can be determined using crack propagation data, I
can be determined using the results of a single test.
Combining Equations 10 and 12 yields the comprehensive fatigue model presented in
Equation 5, which allows for predicting fatigue life at any loading amplitude using the
results of a single fatigue (time sweep) test. Equation 5 was validated through a
comparison between measured fatigue lives (Nf) and those predicted using the proposed
model. As demonstrated in Figure 24, measured and predicted fatigue lives are in
agreement, indicating fatigue life can be accurately predicted at any loading amplitude
using the results of a fatigue test.
493
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
Figure 24. Comparison between measured and predicted fatigue lives.
7. Summary
This study demonstrates that fracture is the cause of fatigue damage during time sweep
testing of asphalt binders in the DSR. Changes in sample geometry caused by fracture
rather than changes in viscoelastic material properties lead to reduction in loading
resistance. Changes in loading resistance (i.e., torque) of asphalt binders under time
sweep testing are directly related to crack propagation, thus crack lengths can be
determined using torque measurements.
Visual observations of fracture morphology of asphalt binders show similar trends to
other materials under similar loading conditions, validating the damage mechanism. The
fracture surface of asphalt binders following time sweep testing is initially
macroscopically flat and demonstrates factory roof morphology as the crack
propagates inward. These findings indicate fracture initiates in Mode III but once a
substantial crack has developed interactions between crack faces leads to fixed mode
fracture. Additionally, trends in crack propagation match those observed in other
materials; initially the rate of crack growth increases with increasing crack depth until a
critical crack length is exceeded at which point the crack growth rate decreases. It should
be noted that the results of this study are based on tests over a limited range of
494
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

temperatures. Future research is needed to determine the temperature range over which
the findings of this study are valid. It is speculated that if the temperature is increased
substantially, flow instability rather than fracture will occur during time sweep testing in
the DSR.
This study validates the time sweep test is an appropriate test method for fatigue
characterization of asphalt binders. Additionally, these results allow for improved
asphalt binder fatigue characterization using fracture mechanics. A model was presented
that applies fracture mechanics concepts for prediction of fatigue life at any loading
amplitude using the results of a single time sweep test. It is proposed that results of a
time sweep test can be used very effectively to estimate crack growth rate and derive the
coefficients for binder fatigue law that can account for variation in loading frequency
and temperature.
8. References
Anderson, D.A., Hir, Y., Marasteanu, M., Planche, J.P., Martin, D., and Gauthier, G.
Evaluation of Fatigue Criteria for Asphalt Binders, Transportation Research
Record, No. 1766, p. 4856, 2001.
Aboutorabi, H., Ebbot, T., Gent, A.N. "Crack Growth in Twisted Rubber Disks. Part I:
Fracture Energy Calculations," Rubber Chemistry and Technology, Vol. 71, No. 1, p.
7683 1998.
Bahia, Hussain U., D.I. Zeng, H., Khatri, M. A. Zhai, and R. M Anderson.
Characterization of Modified Asphalt Binders in Superpave Mix Design, NCHRP
Report 459. Washington D.C., National Cooperative Highway Research Program,
2001.
Fleischman, T.S., Kerchman, V., Ebbott, T.G. "Torsional Crack Growth Test To Simulate
Belt Edge Deformation," Tire Science and Technology, Vol. 29, No. 2, p. 91107,
2001.
Gent, A.N., and Yeoh, O.H. "Crack Growth in Twisted Rubber Disks. Part 3. Effects of
Crack Depth and Location." Rubber Chemistry and Technology, Vol. 76, No. 5, p.
91107 2003.
Gonzalez, R.C., Woods, R.E., Eddins, S.L. Digital Image Processing Using MATLAB,
Upper Saddle River, New Jersey: Pearson Education, 2004.
Hertzberg, R.W. Deformation and Fracture Mechanics of Engineering Materials, New
York, New York, John Wiley & Sons, 1989.
495
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

HINTZ, BAHIA
Hintz, C. Understanding Mechanisms Leading to Asphalt Binder Fatigue, PhD

Dissertation, University of Wisconsin Madison, Madison, WI, 2012.
Hintz, C., Velasquez, R., Bahia, H.B., and Stimilli, A. Normal Stresses in Dynamic
Shear Rheometer Testing for Binder Damage, University of Wisconsin Madison
Modified Asphalt Research Center (MARC) white paper, 2011.
Johnson, C.M. Estimating Asphalt Binder Fatigue Resistance Using an Accelerated
Test Method, PhD Dissertation, University of Wisconsin Madison, Madison, WI,
2010.
Keentok, M., and Xue, S. Edge fracture in cone-plate and parallel plate flows,
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Rheologica Acta, Vol. 38, No. 4, p.321348, 1999.
Keller, M.W., White, S.R., Sottos, N.R.. "Torsion fatigue response of self-healing
poly(dimethylsiloxane) elastomers," Polymer, Vol. 49, , p. 3136-3145, 2008.
Schapery, R.A. "A theory of crack initiation and growth in viscoelastic media. III.
Analysis of continuous growth," International Journal of Fracture, Vol. 11, No. 19,
p. 549562, 1975.
Tschegg, E., Ritchie, R,. McClintock, F. On the Influence of Rubbing Fracture Surface
on Fatigue Crack Propagation in Mode III, International Journal of Fatigue, Vol.
5, No. 1, p. 2935 1983.
9. Discussion
DR. ALAEDDIN MOHSENI: Cassie, very nice presentation. I really enjoyed it. How do you
determine the borderline between crack initiation and when the crack propagation starts?
DR. CASSIE HINTZ: Very good question. The distinction between crack initiation and
propagation was made by looking at the relationship between the crack growth rate and
energy release rate for a lot of binders under varying conditions. As discussed in the
presentation, it was found that a linear relationship exists between crack growth and
energy release rates in log space during crack propagation. However, no clear trend
exists during the crack initiation phase as damage is not localized. It was found that the
start of crack propagation occurred at roughly the same effective crack length predicted
from torque measurements for all binders and conditions evaluated. This effective crack
length, estimated based on loading resistance measurements, was really very small, (less
than 0.1 mm).
DR. MOHSENI: Okay. What do you call the shallow and deep cracking? The shallow
cracking seems to be not even called cracking in other disciplines. It seems to be we
call it microcrack here in asphalt, but other people maybe call it crazing, which is not
496

necessarily a bad thing and its fully recoverable. So it looks like it changes the material,
but is it really cracking?
DR. HINTZ: Actually, during the shallow crack growth, this is a stage of crack
propagation. We have localization of the damage in the form of a macrocrack at the
sample periphery. So when I show this slide that shows this transition, our crack has
reached sufficient length that its not microcracking anymore. There is a macrocrack
present in the material. The shallow and deep crack growth shouldnt be confused with
microcracking versus macrocracking. Shallow and deep crack growth are both stages of
macrocrack propagation.
DR. MOHSENI: So essentially when you have a crack is when this sample starts to fail?
DR. HINTZ: Yes. And the localization and damage in this test occurs fairly early on,
which kind of makes sense if you think about it because in the DSR theres a radial
distribution of the stress and strain with the maximum stress and strain occurring at the
sample periphery. Thus, it is reasonable that damage becomes localized fairly early on in
our test compared to other tests with a more uniform stress/strain distribution.
DR. HOHSENI: Okay, thank you.
PROFESSOR DAVID ANDERSON: First I want to ask you, at what region of stiffness
are you running this test?
DR. HINTZ: Weve run it on a fair range of conditions. So all the way for some binders
its found to work and we can prevent flow all the way to roughly a few megapascals
upward of 50 MPa.
PROF. ANDERSON: From 2 to 50 MPa?
DR. HINTZ: Approximately over that range.
PROF. ANDERSON: Okay. Do we have instability flow in this test?
DR. HINTZ: In some cases we have seen flow instability. We neglect the test results in
these instances.
PROF. ANDERSON: Okay.
DR. HINTZ: Thats something we are looking into. If we do use really low frequencies
of loading or relatively high temperatures, we do see this type of failure mechanism and
future research will aim to provide guidance for the range of conditions the test can
reasonably be used.
PROF. ANDERSON: Is this test applicable when we have instability flow?
DR. HINTZ: In my opinion for fatigue characterization, no.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
497

HINTZ, BAHIA
PROF. ANDERSON: Okay. So thats something that should be discussed very, very
emphatically. Most people dont know what were talking about. But I have real
difficulty with your analysis because you are assuming a linear viscoelastic analysis of
the material. You're assuming that the R changes. Youve got an equation that has D0
in it, okay? So you're assuming then basically that this is a linear viscoelastic behavior
throughout all the way through the whole test. Thats the only way. Thats implicit in
the way youve done your analysis. I dont think thats correct. The other thing that I
think you really need to do before this test goes too much farther is to give us a good
understanding of the stress and strain distribution within the test specimen. We have
equations of compatibility and equilibrium that need to be satisfied. So we have
boundary conditions in this test that have not been explained adequately, certainly not to
the asphalt literature. I mean the equations youve got for fatigue growth and all that,
theyre all nice and they fit in with what everybody is accustomed to looking at, but we
dont know whats going on this specimen. We dont know what initiates the crack, if
its really a crack. I believe its a flow problem, not a cracking problem. So I think were
a long, long way from understanding the test from the very bottom and we should be
there before we start doing an analysis. If we can make it work, thats fine, but I think
were a long way Right now, this is a torture test, thats all it really is. It may be
cloaked in a lot of nice fatigue curves and whatnot, but its a torture test.
DR. HINTZ: Thank you for your comments. Your points are well taken. One of the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
challenges in dealing with the DSR is it is very hard to measure the types of things that
you are referring to. So for this analysis, we acknowledged that some of this is probably
going on, but we found that this change in geometry was the dominant mechanism
responsible for the changes in loading resistance.
PROF. ANDERSON: I mean if you watch this test run very often, youre eyebrows start
to go up and you say, Whats really going on here? Why is this thing flowing at the
center of the specimen? Is that really plastic flow? I mean this material is not elastic.
Were in a region of behavior where theres a lot of viscoelastic flow in the specimen,
and I dont think that is properly taken into account.
DR. HINTZ: Thank you for your comments.
DR. NELSON GIBSON: Forgive me if I start off with a simplistic understanding of
mixture fatigue or damage, but I guess I, you know, use initiation as kind of the same
thing as localization, all right? So a simplistic understanding of mixture damages: You
have microcracking so you can apply the math that comes out of continuum damage,
which I guess you're going to quote me because its AAPT, but I guess the definition of
continuum damage means you can still treat it like a solid, no change in geometry, no
macrocrack. Then once a macrocrack happens, then youve got to switch to fracture
mechanics. So a question I have in what you're proposing, you're suggesting there is
initiation and then propagation with binders, okay? So if thats correct, in any of your
methodology, do you calculate a reduction in modulus before there is a macrocrack? In
498

other words, does any of your methodology follow the philosophy of continuum
damage?
DR. HINTZ: No, but were currently looking into that. So this was just an initial take on
it, focused more on the propagation of the fracture since that was what we were looking
at primarily.
DR. GIBSON: Okay. So anything thats not macrocracked, you're assuming its not
microcracked, I guess.
DR. HINTZ: No, there is some microcracking but as damage localization and hence,
macrocracking occurs fairly early on in the test because of the radial distribution in stress
in strain, at which point the macrocrack propagation becomes more significant than
microcracking. Were not applying fracture mechanics principles to the crack initiation
phase of damage. Its incorporated empirically, as a simplistic kind of means to
incorporate it into the fatigue crack propagation portion of the model.
DR. GIBSON: Thank you.
PROF. HERVE DI BENEDETTO: If I understand correctly, you have two key
hypotheses for your fundamental or theoretical analysis. One is that inside your sample
no modulus reduction is considered and once the crack is created, no shear stress is
transmitted through this crack. The question is whether these hypotheses are physically
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
founded or not. This is a key question. For the first one, when you suppose that there is
no decrease of the modulus, no damage, dont you think its questionable?
DR. HINTZ: I agree that its questionable. Ill address that one first.
PROF. DI BENEDETTO: The second one is also questionable because if you imagine
your rheometer, even if you have a crack, then stress by friction can go from one side to
the other. My question is what do you think about these two hypotheses?
DR. HINTZ: In reference to your first question regarding the modulus reduction in the
intact portion of the specimen, I agree some of that is undoubtedly occurring, but its
just not dominant compared to how fast this crack is going inward. Thus, for our analysis
to make that underlying assumption, worked fairly well.
PROF. DI BENEDETTO: It seems to be working, so this is the point.
DR. HINTZ: Yes.
PROF. DI BENEDETTO: Okay.
DR. HINTZ: So for that at least to gauge if this cracking could account for the changes
in loading resistance, we thought that was reasonable.
PROF. DI BENEDETTO: Okay.
499

HINTZ, BAHIA
DR. HINTZ: Your second question about stress redistribution at the outer edge of the
specimen retaining some load carrying capacity, honestly, in what we saw and whats
been reported in some other materials fatigued under similar conditions, they saw similar
things to what we saw. We do have questions about why this is happening, and your
points are well taken. But, again, from what weve seen so far, the cracking that is
occurring during this test can be interpreted as an effective reduction in the specimen
radius. Thank you.
PROF. DI BENEDETTO: Okay, to come back to the modulus reduction, I do think you
may have some more modulus reduction at least due to heating, I think.
DR. HINTZ: Okay. Thank you. Thats something well look into.
PROFESSOR HAIFANG WEN: Very nice presentation. Two questions. Number one is
related to what has been asked. When you have cracks and calculate the modulus, did
you just take the modulus from the DSR or did you consider the reduction of area?
DR. HINTZ: Im sorry. I dont understand your question.
PROF. WEN: When you're doing the test, you have cracks and that point when you have
crack, the sample area is smaller, right? So did you consider the reduction of area?
DR. HINTZ: We considered the reduction in effective sample radius due to the cracking.
We did not consider changes in modulus of the intact portion of the specimen.
PROF. WEN: But it would affect the modulus calculation, right? In the beginning, there
are no cracks. Then later on, you have crack halfway through the area, then your area
becomes much smaller.
DR. HINTZ: Were taking that into account. Were taking into account the change in
specimen geometry.
PROF. WEN: You do, okay.
DR. HINTZ: Yes, we are taking that into account.
PROF. WEN: Good. The second one is that because the crack is starting from outside,
stress distribution is also different. So you use the shear strength and outside?
DR. HINTZ: Again. Were accounting for the effective reduction in sample radius
during testing. So the effective boundary of the specimen is moving inward.
PROF. WEN: So when you're using this relationship of fatigue of life related
DR. HINTZ: Its the initial strain amplitude of the test. The test is, in fact, a
displacement controlled, not strain controlled test.
PROF. WEN: Initial strain and where?
DR. HINTZ: On the outer edge of the specimen, corresponding to the maximum strain.
500
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

PROF. WEN: And the crack propagated probably inside. Are you using also the shear
strain and crack tape?
DR. HINTZ: No, because the model is in terms of the initial strain amplitude.
PROF. WEN: Okay, thank you.
DR. HINTZ: Thanks.
501
--`,`,,,``,,```,,`,,,,,`,

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Fatigue Endurance Limit for HMA
Based on Healing
Mena I. Soulimana*, Waleed Zeiadab, Michael Mamloukb,

and Kamil Kaloushb
a
Department of Civil and Environmental Engineering, University of Nevada-Reno, Reno,
NV 89557
b
Civil, Environmental and Sustainable Engineering Program, Arizona State University,
Tempe, AZ 85287-5306
ABSTRACT: A long lasting flexible pavement has been the goal of the pavement community for
many years. Designing perpetual pavements requires knowledge of the stresses or strains below
which damage does not occur or can be healed during unloading. The endurance limit, as
applied to hot mix asphalt (HMA) and flexible pavement design, is the strain or stress level below
which the HMA would endure indefinite load repetitions and the pavement would not experience
bottom-up fatigue cracking. The purpose of this study is to determine the endurance limit for
HMA under different conditions using laboratory beam fatigue tests. The approach used in this
study assumes that if stresses or strains are kept below a certain level, fatigue damage may not
occur or can be healed during unloading. Relating healing to the endurance limit makes this
approach unique compared to previous studies that investigated these concepts separately. An
extensive laboratory beam fatigue testing program covering a wide range of asphalt mixtures was
conducted according to the AASHTO T 321-03 test procedure as a part of the NCHRP 9-44A
project. The endurance limit was determined when no accumulated damage occurred indicating
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
complete healing during the rest period after each load application. The endurance limit varied
between 22 and 264 micro-strains (s) depending on binder grade, binder content, air void,
temperature and the rest period between load applications when 0.1 sec loading cycles are used.
The oral presentation was made by Dr. Souliman.
503

SOULIMAN, ZEIADA, MAMLOUK, KALOUSH
The results of this study can be used to design perpetual pavements that can sustain a large
number of load applications if traffic volume and vehicle weights are controlled.
KEYWORDS: hot mix asphalt, fatigue, healing, endurance limit, rest period, stiffness, volumetric
1. Introduction
Bottom-up fatigue cracking is one of the main distress types in flexible pavement.
Current design methods of flexible pavement assume that cumulative damage occurs
where each load cycle uses up a portion of the finite fatigue life of the asphalt layer
regardless of traffic load or volume. The concept of endurance limit (EL) assumes that
there is a strain or stress value below which fatigue damage may not occur or can be
healed during unloading. The fact that traffic loads are separated by rest periods allows
for healing some damage that might occur, which in turn increases the endurance limit.
Therefore, if the pavement is thick enough to keep strains or stresses below the
endurance limit, the fatigue life of the pavement can be considerably extended. This
concept has significant design and economic implications.
The concept of endurance limit has been known for many years for metallic materials
(Whler, 1867). A hyperbolic function between stress ratio and number of load cycles
was developed. The asymptote of this curve parallel to the load cycle axis indicates that
there is a stress level below which the number of cycles to failure does not
proportionally increase with decreasing stress, thus the material tends to have unlimited
fatigue life (Prot, 1948). This asymptote represents the fatigue endurance limit.
In 1972, Monismith and McLean (1972) first proposed an endurance limit of 70
micro-strains for asphalt pavements. More recently, Nishizawa et al. (1997) analysed in-
service pavements in Japan and reported an endurance limit of 200 micro-strains. Wu et
al. back-calculated Falling Weight Deflectometer (FWD) data and reported strains at the
bottom of the asphalt layer between 96 and 158 micro-strains for a long-life pavement in
Kansas (Wu et al., 2004). Bhattacharjee et al. (2009) employed the elasticviscoelastic
correspondence principle and identified the strain at which hysteresis loops form in a
stress-pseudo strain relationship, indicating that damage is occurring. The approach
requires the linear viscoelastic characterization of the mixture through dynamic modulus
testing. The endurance limit values obtained through uniaxial testing ranged from 115 to
250 micro-strains, which showed comparable magnitudes as those obtained from beam
fatigue tests.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
504

Fatigue Endurance Limit for HMA Based on Healing
Studies performed at the University of Illinois (Carpenter et al., 2003; Marshall and
Carpenter, 2006) and the National Center for Asphalt Technology (NCAT) (Prowell,
2008) provided clear evidence that the fatigue behavior of HMA is much different if the
strain is decreased. Their lab tests showed that fatigue life becomes significantly longer
if the strain is kept below approximately 100 micro-strains.
Another major concept that has been investigated by researchers recently was the
healing phenomenon of HMA. Healing of micro damage was proposed as the primary
reason for the increased fatigue life at low strain levels (Advanced Asphalt
Technologies, 2007; Carpenter and Shen, 2006; Kim and Roque, 2006). Healing is
generally considered as the capability of a material to self-recover its mechanical
properties (stiffness or strength) to some extent upon resting due to the closure of micro-
cracks. Phillips (1998) proposed that the healing of asphalt binders is a three-step
process consisting of: 1) The closure of micro-cracks due to wetting (adhesion of two
crack surfaces together driven by surface energy), 2) The closure of macro-cracks due to
consolidating stresses and binder flow, and 3) The complete recovery of mechanical
properties due to diffusion of asphaltene structures. Kim and Little (1989) developed a
mechanical approach to identify the healing potential of asphalt concrete. They
performed cyclic loading tests with varying rest periods on notched beam specimens of
sand asphalt. They obtained a consensus that rest periods enhance the fatigue life
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
through healing and relaxation mechanisms.

There is mounting evidence that healing and endurance limit of HMA are related to
each other. It has been observed in laboratory studies of fatigue at low strain levels with
rest periods and in properly constructed thick pavements that bottom initiated fatigue
cracking is almost non-existent. The HMA endurance limit, however, does not reflect an
absence of load induced damage in the material. It is the result of a balance of damage
caused by loading and healing or damage recovery that occurs during rest periods
(Marshall and Carpenter, 2006). The endurance limit of HMA is, therefore, not a single
value, but varies depending on the load pattern and environmental conditions applied to
the HMA. Considering an endurance limit in flexible pavement design requires the
consideration of the effects of loading, environment and material properties on both
damage accumulation and healing. These findings concerning the endurance limit for
HMA served as the research hypothesis upon which this research project (NCHRP 9-44)
was formulated.
In conclusion, the literature provides endurance limits at certain conditions but there
is no general predictive model currently available to estimate their values under different
conditions based on the healing that occurs during the rest period. Also, the literature
does not provide a clear relation between endurance limit, rest period and healing, which
is one of the main contributions of this study.
505

2. Factors Affecting Fatigue Life and Endurance Limit
2.1. Binder Content and Air Voids
Results from the SHRP A-003A project indicated that low asphalt contents and low air
voids lead to higher stiffness, while high asphalt contents and low air voids lead to
higher fatigue lives (SHRP A-404, 2008). Harvey and Tsai (1996) produced similar
results for a typical California mix. By using the models for stiffness and fatigue lives
obtained from laboratory test results, their simulation showed an increase in pavement
predicted fatigue life with higher asphalt contents and lower air voids. Tayebali et al.
(1994) also found that air voids have a large effect on fatigue life. As air voids
increased, fatigue life decreased for both controlled strain and controlled stress tests. It
was concluded that stiffer mixes would perform better for thick pavements, while low
stiffness mixes would perform better for thin pavements.
2.2. Temperature and Binder Grade
Because the endurance limit of HMA is tied closely to the healing potential of the
binder, healing occurs more rapidly at higher temperatures and softer binder grades and
the strain level that can be tolerated with no damage accumulation is increased
(Carpenter and Shen, 2006).
Verstraeten et al. (1977) performed dynamic two-point bending beam tests in a
constant-stress mode. The loading conditions were maintained either until failure or
80% of stiffness reduction. The specimens were then stored for periods varying from 3
to 21 hours at different temperatures. The authors concluded that the longer the rest
periods and the higher the temperatures, the greater the beneficial effect, although their
effects on the susceptibility of mixtures to fatigue could not be quantified.
2.3. Rest Period
Over the last five decades, several researchers have studied the significance of rest
periods between load applications during fatigue testing of HMA. Different findings
have been presented in the literature showing different opinions on the effect of rest
period. Some researchers thought that the rest period only leads to a temporary modulus
recovery without actually extending fatigue life, while others found that the modulus
recovery did extend fatigue life by a certain amount.
506
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Van Dijk and Visser (1977) investigated the effect of rest period on the fatigue life of
a rolled asphalt base course mix. It was found that increased rest periods can increase
fatigue life by a factor of 1 to 10 times. Monismith et al. (1961) assessed the effect of
rest period by conducting repeated flexure tests on beam specimens supported by a
spring base. Under their assumptions and test conditions, the test results indicated that
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
increasing the rest period had no significant effect on fatigue performance.
In an attempt to investigate a rational value of a threshold rest period, Raithby and
Sterling (1972) applied a range of rest periods between null and 25 times the loading
time (i.e., 0.1 sec loading time and 2.5 sec rest period) on a rolled asphalt base course
using a dynamic push-pull test. It was found that fatigue life does not increase
significantly for rest periods greater than ten times the loading time (or 1 sec rest period)
and the waveform influence was less important than the duration of the rest periods.
Van Dijk and Visser (1977) also tested a rolled asphalt base course mixture in a
three-point bending beam apparatus in a constant strain mode (40 Hz frequency and
20C temperature) with loading ratios varying from 1 to 25 (0.025 sec loading time and
up to 0.625 sec rest period). The results showed increasing fatigue lives with increasing
rest periods. The maximum beneficial effect of the rest period on the fatigue life (load
ratio of about 10) was determined by means of extrapolation to be achieved at a loading
ratio of about 50.
Bonnaure et al. (1982) conducted a three point bending beam fatigue test on
rectangular beams. An intermittent loading was applied with various rest periods of 0, 3,
5, 10, and 25 times the length of the loading cycle (0, 0.075, 0.125, 0.25 and 0.625 sec).
The tests were done under both constant stress and constant strain modes of loading at
three temperatures and a 40 Hz frequency. The authors concluded that increasing the
rest period between the loading cycles increases fatigue life. The maximum beneficial
effect of rest periods on the fatigue life was at a rest period equals to 25 times the
loading cycle (or 0.625 sec).
3. Objectives
The main objective of this study was to determine the fatigue endurance limit for HMA
and relate it to healing that occurs during the rest period between load applications.
Another objective was to develop models that relate endurance limit of HMA to material
properties, loading conditions, and temperature. These objectives were achieved by
carrying out an extensive beam fatigue laboratory experiment to identify the mixture and
pavement design features related to endurance limit for bottom-initiated fatigue
cracking. Issues studied included incorporating rest periods between loading cycles and
the effect of rest period on the healing and endurance limit of HMA.
507

4. HMA Endurance Limit and Healing
A rational procedure was developed to relate the HMA healing phenomenon to the
endurance limit, which makes this approach unique compared to previous research
projects that studied these concepts separately. If the fatigue test is conducted with and
without rest period between load applications, the typical stiffness ratio (SR) versus
number of load cycles will be as shown in Figure 1. Both curves start at a stiffness ratio
of one (no damage). The curve for the test without rest period is steeper than the other
curve because of the continuous deterioration during the test. The test with rest period
shows higher SR values during the test because of healing that occurs during the rest
period after each load application. A larger separation between the two curves indicates
more healing, and vice versa. If the curve for the test with rest period remains
horizontal, it indicates that full healing occurs after each load cycle.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 1. Typical stiffness ratio versus number of loading cycles for tests with and
without rest period.
The approach used in this study is to develop a relation between stiffness ratio and
various factors in the form of Equation 1:
SR = f (BG, AC, Va, T, , N, RP) [1]
In Equation 1, SR is the stiffness ratio, BG is the binder grade, AC is the binder
content, Va is the air voids, T is the temperature, is the initial strain, N is the number of
load applications, and RP is the rest period between load applications. If the stiffness
508

ratio in Equation 1 is set to one, the strain will become the endurance limit, which
implies that full healing occurs after each loading cycle. Also, since it is assumed that
full healing occurs after each loading cycle at the endurance limit, the number of loading
cycles is redundant and may be removed from the equation without large effect.
Note that the rest period in the lab is related to the average annual daily truck traffic
(AADTT) in the field. If the AADTT is low, the spacing between trucks and/or axles is
large which corresponds to longer rest periods in the lab.
5. Materials, Mix Design and Fatigue Testing
A 19 mm Superpave mix design was prepared for the project according to the
requirements of typical paving materials used for paving arterial roads in Arizona. Three
asphalt concrete mixes were designed using three binder grades (PG 58-28, PG 64-22,
and PG 76-16) that cover a wide range of stiffness. The same aggregate gradation was
consistently used for all mixes. Figure 2 shows the designed aggregate gradation
distribution curve and the specification limits.
Figure 2. Designed aggregate gradation distribution.
Beams were prepared using vibratory loading applied by a servo-hydraulic loading

machine. A mold was used with inside dimensions larger than the required dimensions
of the beam to allow for sawing to achieve standardized geometric dimensions. The mix
was compacted using a stress-controlled sinusoidal load until the pre-determined density
509
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

was reached. The specimens were brought to the required dimensions for fatigue testing
by sawing 1/4 in. from each side. Air voids were measured using the saturated surface-
dry procedure (AASHTO T 166, Method A). Any specimen with an air void content
larger than 1 percent deviation from the target value of 7 percent was rejected. The
details of beam preparation and verification of air void uniformity within the specimen
are presented elsewhere (Souliman, 2012).
Beam fatigue tests were conducted according to the AASHTO T 321-03 test
procedure using displacement-controlled sinusoidal loading. Some tests were performed
without rest periods at a frequency of 10 Hz according to AASHTO T 321, while others
used a 0.1 sec sinusoidal loading followed by a rest period.
Tests without rest period were performed up to failure, which occurs when the
stiffness ratio reaches 0.5. Since the tests with rest periods take considerably longer than
tests without rest periods, it was decided to run all tests with rest period up to 20,000
cycles only. Therefore, extrapolation was used to predict the SR for the test with rest
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
period at Nf w/o RP (number of cycles to failure for the test without rest period). Figure 3
shows the extrapolation used to determine the stiffness ratio (SR) for tests with rest
period at Nf w/o RP. Several tests with rest period were run up to Nf w/o RP to verify the
accuracy of the extrapolation process and showed that extrapolation has high accuracy in
predicting SR at Nf w/o RP.
1.2 Extrapolate of test with Res t

1.1 Test w/ Rest Period
Test w/o Rest P eriod Period till N f w/o Res t Peri od
1
0.9
Stiffness Ratio
0.8
0.7 HI
0.6 Healing
0.5
0.4
0.3
0.2
0.1 N f w/o Res t Peri od
0
0 50000 100000 150000
Cycles
Figure 3. Typical extrapolation to estimate SR (with rest period) at Nf w/o RP (PG 64-22,
40F, 4.2 AC%, 4.5 Va%, 200 in/in).
510

In this study, two models were developed, namely the volumetric model and the
stiffness model. The purpose of the volumetric model was to determine the endurance
limit at different combinations of material volumetric properties, such as binder content
and air voids. The stiffness model was to determine the endurance limit at certain
stiffness instead of the volumetric properties.
6. Volumetric Endurance Limit Model
6.1. Experimental Design
The volumetric model considers the following factors:

1. Binder grade (three levels: PG 58-28, PG 64-22, PG 76-16)
2. Binder content (two levels: optimum 0.5 %)
3. Air voids (two levels: 4.5, 9.5 %)
4. Strain level (two levels: L, H)
5. Temperature (three levels: 40, 70, 100oF)
6. Rest period (two levels: 0, 5 sec)
The criterion of selecting the two strain levels at each temperature was to reach a
fatigue life in tests without rest period (Nf w/o RP) after approximately 20,000 cycles at the
high strain level and approximately 100,000 cycles at the low strain level. Pilot beam
fatigue tests were conducted at different strain levels at three temperatures of 40, 70 and
100oF, from which the strain levels were determined.
It was clear that a complete factorial design using the above-mentioned factors and
levels would be practically prohibited considering the time and resources available.
Added to the problem is the time it takes to run beam fatigue tests with rest periods. For
example, a fatigue test with a 0.1 sec loading cycle, 5 sec rest period and 20,000 load
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
applications takes 28.3 hours, in addition to the time needed for mix preparation,
specimen compaction and sawing, and air void determination. Therefore, it was decided
to use a 6-factor fractional factorial statistical design with partial randomization since it
would provide accurate results and requires fewer tests. This design allows for
determining the effects of the main factors and up to 3-factor interactions.
Table 1 shows the factor combinations used in the whole study. Highlighted cells
show factor combinations used to develop the volumetric model. The cells marked with
X show tests that were added to the database to develop the stiffness model as discussed
is a later section.
Three replicates were tested at each factor combination during the development of
the volumetric model requiring a total of 288 tests. A small statistical study using the
PG 64-22 data points was then performed to determine the minimum number of
511

replicates to maintain the required accuracy. The results concluded that the accuracy of
the model does not change when using two or three replicates for each factor
combination. Therefore, it was decided to use two replicates for the additional tests to
develop the stiffness model adding up to 152 tests (76 test combinations x two
replicates). Thus, 440 tests were used to develop the stiffness model.
Table 1. Factor combinations used in the study.
512
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

6.2. Development of volumetric model
The JMP software (2012) was used in developing the volumetric model by optimizing
the different combinations of factors. The significant factors are selected if the
individual p-values are less than the significant level of 0.05.
Because of the limitations of the fractional factorial statistical software, all factors
have to have the same number of levels. The three factors (strain, temperatures, and
binder grade) have three levels each whereas the other factors have two levels each.
Therefore, it was necessary to assume the strain, temperatures, and binder grade as
categorical factors. After developing the model, they have been converted fro m
categorical to numerical factors. The conversion was done by developing relationships
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
between the three levels of each factor and numerical coefficients (Souliman, 2012).
Each level of the strain was represented by the corresponding number of cycles at the
end of the test. To convert temperature from a categorical to a numerical variable,
relationships between the three levels of temperature (40, 70, and 100oF) and the
categorical coefficients were developed (Souliman, 2012). While converting the binder
type from categorical variable to numerical, it was decided to use the elastic modulus
values obtained from the E* test at 70oF and 10 Hz at the optimum design condition
(4.7% AC and 7% Va) for each mix. As a result, the finalized volumetric model was
developed as shown below.
SR = 0.1564774 + (0.00079*BG) + (0.070059744*AC) + (0.00393*Va) +

(0.10095*RP) - (1.268*10-7 *N) - (0.0024676 *T) - (0.0001677*BG*AC) +
(3.29961x10-5 *BG*RP) + (3.488*10-6 *BG*T) + (0.00794848*AC*RP) -
(0.0042225*Va*RP) + (0.0006044*AC*T) - (0.0001035*Va*T) - (2.889*10-8*RP*N)
+ (2.9191*10-9 *N*T) - (0.0025*RP*T) - (3.97*10-7 *BG2) - (1.20135*10-5*T2) +
(8.434*10-8 *BG2*AC) - (2.8756*10-8 *BG2*RP) + (1.9558*10-6 *AC*T2) +
(6.6137*10-7 *Va*T2) - (1.582*10-11 *N*T2) + (1.262x10-5 *RP*T2) - (1.176*10-6
*Va*RP*T2) + (3.124*10-12 *N*RP*T2) - (7.4*10-7 *BG*AC*T) + (3.92*10-7
*BG*RP*T) + (0.00013185 *Va*RP*T) + (2.19 * 10-9 *N*RP*T)
[2]
In Equation 2, SR is the stiffness ratio, BG is the binder grade, AC is the asphalt

content (%), Va is the air voids (%), RP is the rest period (sec), N is the number of cycles
to failure, and T is the temperature (oF). In order to evaluate the accuracy of the model,
the measured SR values were plotted versus the predicted SR values as shown in Figure
4. The figure shows good correlations with a high adjusted R2 value of 0.980, which is a
good indication of the accuracy of the model.
513

0 .9
y = 0 .9 4 5 5 x + 0 .0 2 9 6
0 .8 2
R = 0 .9 7 8 9
0 .7
Pred icted SR
0 .6
0 .5
0 .4
0 .3
0 .2
0 .1
0
0 0 .1 0 .2 0 .3 0 .4 0 .5 0 .6 0 .7 0 .8 0 .9 1
M e asu r e d SR
Figure 4. Measured versus predicted SR values based on the volumetric model for all
three mixtures.
6.3. Predicting Endurance Limit Using the Volumetric Model
Using the developed volumetric model, the SR values were plotted versus strain that
were used for each mixture at each temperature separately since it is expected to have
different endurance limit values for different mixtures and different temperatures.
Figure 5 shows examples of the relationship between the stiffness ratio and strain for the
64-22 mixture at the three temperatures. Since there are two strain levels for each factor
combination, a logarithmic relation was assumed and the curve was extended to reach a
SR value of 1 (complete healing). The corresponding strain in each case was determined
and reported as the endurance limit as discussed earlier. Since the actual relationship
between SR and strain cannot be exactly determined using two points only, the
logarithmic relationship was used as the best approximation.
Figure 6 shows an example of the estimated endurance limits for all possible factor
combinations based on the developed model at a 5-second rest period. The endurance
limit ranged from 22 s (at 40oF) to 264 s (at 100oF). As expected, increasing the
binder content increased the endurance limit, while increasing the air voids decreased the
endurance limit. It was also noted that the endurance limit increases by increasing the
temperature.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
514

Figure 5. Stiffness ratio versus strain for the PG 64-22 mixture.
515
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 6. Endurance limits for different factor combinations for a 5-second rest period
using the volumetric model.
7. Stiffness Endurance Limit Model
7.1. Model Simplification Using Initial Stiffness
Although the volumetric model developed in the previous section is very powerful, it is
rather extensive and there is a need for simplification. Since binder content, air voids,
binder grade and temperature affect stiffness, they were all replaced with the initial
stiffness of the mixture. This is a rather innovative approach that relates the endurance
limit to a basic material property such as material stiffness.
7.2. Introducing Other Rest Periods and Strain Levels
Since the volumetric model is based on two levels of rest period and two levels of strain,
the mathematical relationship between endurance limit and these two factors cannot be
based on the true nonlinear function. In the volumetric model, a logarithmic function
was arbitrarily used to incorporate nonlinearity without good justification. Previous
516

studies demonstrated that increasing the rest period above a certain threshold value
would not add a significant gain in the HMA healing, which means no changes would
occur to the endurance limit value. In order to check the true nature of the relationship
between endurance limit, rest period and strain level, two other rest period levels and
another strain level were introduced. Adding these extra levels (see Table 1) allows for
gaining more data points in developing accurate relationships between healing and
endurance limit. It would also allow for compiling all the data together to have a more
accurate stiffness regression model.
7.3. Healing and Rest Period
Several trials were made to determine the best mathematical form that relates the three
independent variables (rest period, strain level, and stiffness) to SR. It was found that
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
there is a need for a logarithmic transformation for both strain and stiffness values. It
was also concluded that the best mathematical form to relate SR to rest period was the
tangent hyperbolic (Tanh) function since it was noticed that there was no large healing
gained by applying 10 sec rest period compared to 5 sec observed during the laboratory
tests. This also supports the literature that showed a threshold rest period beyond which
no more healing is gained.
7.4. Adding More Data Points
As discussed earlier, the applied strain was pre-selected to reach failure for the test
without rest period at a certain number of cycles (Nf w/o RP). This issue resulted in co-
linearity between the strain and the number of cycles. To remove the co-linearity in the
stiffness model, SR data were collected at three different locations along the SR-N
relationship for tests with rest period. Two of these points were taken during the test,
while the third point was taken at Nf w/o RP. Figure 7 shows the typical SR-N
relationships for the tests with and without rest period and the locations where data
points were taken. Note that the test results with rest period are extrapolated to Nf w/o RP
as discussed earlier.
517

Healing
Figure 7. Selection of data point locations.
7.5. Development of Stiffness Model
A total of 946 data points were used to build the stiffness model. Two statistical
packages were used to build the regression model: STATISTICA (2012) and Excel.
STATISTICA was used to come up with the best initial values for the coefficients. An
optimization process was performed using Excel to minimize the sum of squared error
followed by setting the sum of error equal to zero. A statistical analysis (Montgomery,
2008; JMP, 2012) was then used to remove the outliers in order to improve the accuracy
of the model. Equation 3 shows the stiffness model after removing the outliers.
SR= 2.0844 - 0.1386*log(Eo) - 0.4846*log() - 0.2012*log(N) +
1.4103*tanh(0.8471*RP) + 0.0320*log(Eo)*log() -
0.0954*log(Eo)*tanh(0.7154*RP) - 0.4746*log()*tanh(0.6574*RP) +
0.0041*log(N)*log(Eo) + 0.0557*log(N)*log() + 0.0689*log(N)*tanh(0.259*RP)
[3]
In Equation 4, SR is the stiffness ratio, Eo is the initial flexural stiffness (ksi), is the
applied strain (micro-strains), RP is the rest period (sec), and N is the number of cycles.
The adjusted R2 value of the model was 0.891. Figure 8 shows the predicted versus
measured SR values of the stiffness model, which indicates accurate prediction. By
substituting the stiffness ratio in Equation 4 with 1.0 (complete healing condition), the
endurance limit can be predicted for different values of Eo, N, and RP.
518
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 8. Measured versus predicted SR for the stiffness model.
7.6. Effect of N on Endurance Limit
After N was included in the model, it was important to check the effect of changing the
value of N on the endurance limit. A sensitivity analysis study was performed, where
SR was plotted versus strain and rest period for different Eo values and three values of N
(20,000, 100,000, 200,000 cycles). It was concluded that the number of loading cycles
has little or no effect on the SR value, especially at large values of N (Souliman, 2012).
This conclusion validates the assumption that complete healing occurs during the rest
period after each load application at the endurance limit, which makes the number of
load applications redundant. As a result, the endurance limit was calculated at a
conservative value of 200,000 cycles in the stiffness model.
7.7. Predicting Endurance Limit Using the Stiffness Model
The endurance limit was estimated by plotting stiffness ratio versus strain at rest periods
of 1, 2, 5, 10 and 20 sec. In each case, the endurance limit was obtained as the strain
corresponding to an SR value of 1.0, indicating complete healing during the rest period.
Figure 9 demonstrates examples of stiffness ratio versus strain at rest periods of 1 and 5
sec.
519
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 9. Examples of strain versus SR for several initial stiffness values (at 1 and 5 sec
rest period).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 10 and Table 2 show a summary of the endurance limit values obtained fro m
the stiffness model for several rest periods and stiffness values. The results show that
decreasing the stiffness makes the material more ductile and, therefore, the endurance
limit increases. Also, increasing the rest period increases the endurance limit. This
520

means that the endurance limit at larger strains is obtained at longer rest periods in order
to allow for complete healing. The endurance limit increased from a range of 22-82 s
at one sec rest period to a range of 80-220 s at a 10 second rest period. The endurance
limit values at rest periods of 10 and 20 sec were almost the same. This indicates that
the threshold rest period to complete the healing is between 5 and 10 sec.
Figure 10. Summary of endurance limit values versus several rest periods and stiffness
values using the stiffness model.
The endurance limits of the stiffness model were slightly less than those obtained
from the volumetric model and found to be comparable. The volumetric model
produced endurance limit values of 22 to 264 s, while the stiffness model produced
endurance limit values of 22 to 223 s .
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`
521

Table 2. Endurance limits predicted from the stiffness model.
Rest Period, Sec Stiffness, ksi Predicted EL, s EL Range, s

3 ,0 0 0 22
1 ,0 0 0 32
1 22 - 82
200 56
50 82
3 ,0 0 0 45
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
1 ,0 0 0 62
2 45 - 138
200 96
50 138
3 ,0 0 0 66
1 ,0 0 0 90
55 66 - 187
200 134
50 187
3 ,0 0 0 80
1 ,0 0 0 106
10 80 - 220
200 161
50 220
3 ,0 0 0 81
1 ,0 0 0 108
20 81 - 223
200 164
50 223
8. Incorporating Endurance Limit in Fatigue Relationships
The fatigue relationships shown in the literature are typically parallel straight lines for
different Eo values on a log-log graph as shown in Figure 11. These lines show that
when the applied strain decreases the number of load applications to failure (Nf)
increases following a log-log relationship, without consideration to endurance limit. The
endurance limit obtained from the stiffness model (Equation 3) can be incorporated in
the fatigue relationships for a specific rest period as demonstrated in Figure 11. This
means that if the applied strain is above the endurance limit, failure will accumulate and
the layer will last up to a certain values of Nf. However, if the strain is below the
endurance limit value and is applied at a certain rate (for a specific rest period), complete
healing will occur and damage will not accumulate. This concept has a significant
522

implication on pavement design, where the load magnitude and/or the rate of loading
(rest period) can be controlled to extend pavement life.
Figure 11 also shows that when the material stiffness increases, Nf decreases if the
strain is above the endurance limit. It also shows that when the material stiffness
increases, the endurance limit decreases for the same rest period. The figure
demonstrates the endurance limit concept for a specific rest period that corresponds to a
specific AADTT in the field. If the rest period increases (smaller traffic volume) up to a
certain value (5-10 sec), it would allow for more healing at higher strains, which means
that the endurance limit would increase.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 11. Incorporating endurance limit in fatigue relationships.
9. Incorporating Endurance Limit in the MEPDG
The current MEPDG software (DARWin-ME) requires the designer to input a single
value of endurance limit. As shown earlier, the endurance limit values vary depending
on the material stiffness and the rest period between loading cycles. In the MEPDG
software, a simulation is performed every month during the pavement service life.
Therefore, the incorporation of the endurance limit into the MEPDG requires additional
software that calculates the endurance limit value for the given material stiffness and the
average rest period associated with each month MEPDG simulation and feeds it into the
MEPDG software during the analysis process. The average rest period is a function of
the average annual daily truck traffic (AADTT) during the one month simulation period.
523

The endurance limit can be obtained every one month of the MEPDG simulation
using the stiffness model developed in this study (Equation 3) for a stiffness ratio of 1.0
(complete healing) and a large value of N such as 200,000 cycles.
At the same time, the critical strain value of the HMA layer (or sublayer) for the
average truck axle for this period is calculated using the Jacob Uzan Layered Elastic
Analysis (JULEA) multilayer elastic computer program built in the software. If the
critical strain calculated from the JULEA program is less than the fatigue endurance
limit, the average axle is not counted in the analysis for this period, which means that
there is no fatigue damage during this time period. However, if the critical strain is
greater than the fatigue endurance limit, the average axle is counted as causing
cumulative fatigue damage during this period.
The concept of perpetual pavement requires the knowledge of the endurance limit of hot
mix asphalt. The main purpose of this study was to validate the endurance limit for
HMA using laboratory beam fatigue tests with rest periods between loading cycles. A
comprehensive study was performed to estimate the endurance limit of a wide range of
typical HMA due to healing that occurs during the rest periods. Six main factors were
selected for evaluation, which are binder grade, binder content, air voids, temperature,
duration of the rest period between loading cycles, and strain. A fractional factorial
statistical design was used in order to reduce the number of tests and still maintain the
necessary accuracy.
Extensive laboratory displacement-controlled flexure fatigue tests were performed
according to AASHTO T 321-03 test procedure. Hot mix asphalt was used with three
binder grades, two binder contents, two levels of air voids, three levels of applied strain,
three test temperatures and four values of rest periods between loading cycles. The
study assumes that the endurance limit is obtained due to healing that occurs during the
rest period between loading cycles. Volumetric and stiffness models were developed
that can predict the stiffness ratio at various test conditions which can be related to the
healing gained during the rest period. The strain level that allows for complete healing
at a certain rest period was obtained to estimate the endurance limit below which a very
large number of load repetitions can be applied to the pavement without fatigue damage
accumulation.
After developing the stiffness model, a concept of integrating the endurance limit in
the strain-Nf fatigue relationships was discussed as a step toward incorporating the
endurance limit in the MEPDG software.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
524

The following are the conclusions drawn from this study.

1. HMA exhibits endurance limits that vary with mixture properties and pavement
design conditions. There is no single value of the endurance limit for all
conditions. The endurance limit varies depending on the binder grade, binder
content, air voids, temperature and the rest period between load applications.
2. Mixtures using softer binders exhibit higher endurance limits than mixtures
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
using stiffer binders for the same rest period.
3. For the same rest period, high binder contents and low air voids produced the
highest endurance limit values compared to low binder contents and high air
voids, which showed the lowest endurance limit.
4. Endurance limit values were higher at high temperatures, which correspond to
soft mixtures, compared to low temperatures that correspond to stiff mixtures.
5. HMA stiffness is a good mixture property that represents other volumetric
parameters. It exhibits the same effect of four combined pavement mixture
variables: binder grade, binder content, air voids, and temperature.
6. The rest period that ensures complete healing ranges from 5 to 10 sec based on
pavement design conditions.
7. Number of loading cycles has little effect on endurance limit for tests with rest
period. This concludes that the endurance limit can be determined based on a
relatively short number of cycles in the laboratory since damage will always be
healed at the end of each loading cycle.
8. The predicted endurance limit values based on the volumetric and stiffness
models were comparable. The endurance limit ranged from 22 micro-strains to
264 micro-strains. Either model can be used to obtain the endurance limit of
typical hot mix asphalts.
9. The endurance limits obtained in this study can be incorporated in the strain-Nf
fatigue failure criteria. They can also be incorporated in the MEPDG that
would eventually lead to the design of perpetual pavements that can sustain a
large number of truck loads without accumulated fatigue damage. If pavement
thickness, traffic volumes and vehicle weights are controlled, a very large
number of vehicle repetitions can be applied without causing fatigue damage to
the HMA layer.
Filed validation is highly recommended as a prudent step in implementing the
developed models.
525

11. Acknowledgement
The paper was prepared as a part of NCHRP Project 9-44A, which is funded by the
National Cooperative Highway Research Program. The guidance and valuable input of
the project PI, Dr. Matthew Witczak, are greatly acknowledged and appreciated.
12. Disclaimer
The opinions and conclusions expressed or implied in the paper are those of the Arizona
State University research team. They are not necessarily those of the NCHRP.
13. References
Advanced Asphalt Technologies, LLC, Hot Mix Asphalt Endurance Limit Workshop:
Executive Summary, National Cooperative Highway Research Program Project 9-
44, November, 2007.
Bhattacharjee, S., Swamy, A., and Daniel, J., Application of ElasticViscoelastic
Correspondence Principle to Determine Fatigue Endurance Limit of Hot-Mix
Asphalt. Transportation Research Record: Journal of the Transportation Research
Board, No. 2126, Washington, D.C., 2009, pp. 1218.
Bonnaure, F., Huibbers, A.H.J.J., and Booders, A. "A Laboratory Investigation of the
Influence of Rest Periods on Fatigue Characteristics of Bituminous Mixes."
Proceedings of the Association of Asphalt Paving Technologists, Vol. 51, 104, 1982.
Carpenter S.H., Ghuzlan, K., and Shen, S., Fatigue Endurance Limit for Highway and
Airport Pavements, Transportation Research Record: Journal of the Transportation
Research Board, No. 1832, TRB, National Research Council, Washington, D.C., pp.
131138, 2003
Carpenter, S. H., and Shen, S., A Dissipated Energy Approach to Study HMA Healing
in Fatigue, Journal of the Transportation Research Board, 2006.
Harvey, J., Tsai, B., Effect of Asphalt Content and Air Void Content on Mix Fatigue
and Stiffness. Transportation Research Record No: 1543, Washington, D.C. pp.
3845, 1996.
JMP software, SAS Institute Inc, http://www.jmp.com/, 2012.
Kim, Y, R. and Little, D, N. Evaluation of Healing in asphalt concrete by means of the
theory of nonlinear viscoelasticity, Journal of Transportation Research Record:
Transportation Research Board, No. 1228, pp. 198210, 1989.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,
526

Kim, B. and Roque, R., Evaluation of Healing Property of Asphalt Mixtures,

Transportation Research Record No. 1970, Transportation Research Board,
Washington, D.C., 2006.
Thompson, M. R., Carpenter, S. H., Considering Hot-Mix-Asphalt Fatigue Endurance
Limit in Full-Depth Mechanistic-Empirical Pavement Design, International
Conference on Perpetual Pavement, Ohio 2006.
Monismith, C. L., and McLean, D. B., Structural Design Considerations, Proceedings
of the Association of Asphalt Paving Technologists, Vol. 41, 1972.
Monismith, C L, Secor, K E and Blackmer, W., Asphalt mixture behaviour in repeated
flexure, Proceedings of Association of Asphalt Paving Technologists, Vol. 30,
pp.188222, 1961.
Montgomery, D. C., Design and Analysis of Experiments, Wiley, 2008.
Nishizawa, T., Shimeno, S., and Sekiguchi, M.. Fatigue Analysis of Asphalt Pavements
with Thick Asphalt Mixture Layer, Proceedings of the 8th International
Conference on Asphalt Pavements, Vol. 2. University of Washington, Seattle, WA,.
Pp. 969976, August 1997.
Phillips, M.C. Multi-Step Models for Fatigue and Healing, and Binder Properties
Involved in Healing. Proceedings, Eurobitume Workshop on Performance Related
Properties for Bituminous Binders, Luxembourg, Paper No. 115, 1998.
Prot, E. M. "Rev Metallurgie." translated by Ward, E.J. WADC Tech Rep, pp. 52148,
1948.
Prowell, B. et al., Endurance Limit of Hot Mix Asphalt Mixtures to Prevent Fatigue
Cracking in Flexible Pavements, NCHRP 9-38, Updated Draft Final Report,
NCHRP, May 2008.
Raithby, K. D. and Sterling, A. B. Some Effects of Loading History on the Performance
of Rolled Asphalt, TRRL-LR 496, Crowthorne, England, 1972.
SHRP, A-404. "Fatigue Response of Asphalt-Aggregate Mixes". Strategic Highway
Research Program, National Research Council, 1994.
Souliman, M. Integrated Predictive Model for Healing and Fatigue Endurance Limit for
Asphalt Concrete. Ph.D. Dissertation, Arizona State University, AZ, 2012.
STATISTICA software, StateSoft, http://www.statsoft.com/, 2012
Tayebali, A. A., Rowe, G.M., and Sousa, J.B., Fatigue Response of Asphalt Aggregate
Mixtures. Journal of the Association of Asphalt Paving Technologists Vol: 61. pp.
333360, 1994.
527
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Van Dijk, W., and W. Visser. The Energy Approach to Fatigue for Pavement Design.
Proceedings Journal of the Association of Asphalt Paving Technologists Vol. 46.
pp. 140, 1977.
Verstraeten, J., Romain, J. E., and Veverka, V., The Belgian Road Research Centers
Overall Approach Structural Design, Fourth International Conference on The
Structural Design of Asphalt Pavements, Vol. 1, Proc., Ann Arbor, MI, Aug 1977.
Whler, A. "Versuche uber die Festigkeit der Eisenbahnwagenachsen"; English
summary (1867). Engineering, 4, 160161, 1860.
Wu, Z., Siddique, Z. Q., and Gisi, A. J., Kansas Turnpike An Example of Long
Lasting Asphalt Pavement. Proceedings International Symposium on Design and
Construction of Long Lasting Asphalt Pavements. National Center for Asphalt
Technology, Auburn, pp. 857876, AL 2004.
14. Discussion
PROFESSOR REYNALDO ROQUE: Thank you for your presentation. Am I correct

that you used only one gradation which looked like a coarse, dense graded mixture and
then all the variables were accomplished by essentially changing either asphalt content
or compaction?
DR. MENA SOULIMAN: Yes. We used one typical aggregate gradation that is used in
Arizona.
PROF. ROQUE: So do you think that if were to change that gradation somewhat, then
all of the conclusions that youve drawn would still be applicable?
DR. SOULIMAN: Thats a very good question. Aggregate gradation must be viewed as
a factor that possibly may have a potential impact upon the fatigue endurance limit.
However, AC base gradation specifications will not vary significantly between DOT
agencies. The NCHRP 9-44A research team felt that highway agencies typically use
standard base gradations and filler contents based on their previous experience. These
standard gradations and filler contents have been selected to optimize the properties of
their mixes and any changes in these factors might result in poor performance. Thus, the
selection of a typical gradation for the mixtures used in this experiment was believed to
suffice until more results are finalized from this research project and other studies. But
definitely if the gradation would change drastically from what we had, it may affect the
results. I trust that the developed endurance limit model is going to be sensitive to the
change in gradation, as well.
PROF. ROQUE: My experience is that depending on the mixture and depending on what
part of the gradation curve one changes, even small changes in gradation can have very
528
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

large changes in failure limits and damage rates, which are directly related to endurance
limit and healing. So I think that its important to qualify what youve done with respect
to that limitation. As an example, even with the stiffness I wanted to ask you about
that. I can stiffen a mixture in different ways. I can stiffen a mixture by aging it. So
aging will make the mixture stiffer. I can improve the gradation and also get a much
stiffer mixture. If I make a mixture finer, then I will have higher asphalt content than if I
have a coarser mixture, so in asphalt mixtures all factors are interrelated. So once you
consider the range in gradations and aggregate types used in paving, development of
these very general types of relationships is going to be quite challenging to implement
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
across the board.

DR. SOULIMAN: Well, yes, its a very good question. I think part of revealing the
effect of such factors would be the field study. So I would say that we need to perform a
field verification study to implement different gradations and different binder contents
and other volumetrics to see how accurate the developed EL model is.
PROF. ROQUE: I suggest you try in the lab first. Even in the lab, youll see some big
differences. I have no doubt in my mind that youll see huge differences if you just go to
a different gradation.
DR. SOULIMAN: Well, if we were to choose another one or two gradations we would
have needed to run another 800-1000 beams on top of the 500 beams that we tested in
three years. Therefore, if we have another three to four years and additional three-
quarters of a million dollars we would have added aggregate gradation into the
experimental design. Thanks.
DR. BRIAN PROWELL: Thanks, Rey. It appeared like all your binders are neat binders.
Is that correct?
DR. SOULIMAN: Yes.
DR. PROWELL: So do you have any thoughts on how your model might account for
polymer modification? Because there are changes that will change the stiffness and
the elasticity dont seem to go hand in hand.
DR. SOULIMAN: Yes, thats a very typical question that I get. For this project, our
main concentration was to develop a model for neat binders. So, definitely, I agree with
you. Once we have a polymer-modified binder, we need to make use of the procedure
that we came up with. So the good thing about this study is we are showing a procedure,
how you can come up with a new mix and apply the same procedure. Even WMA
mixtures, in another meeting, someone asked how about WMA, can I use the model? So
for now, you can use it because were accounting for the stiffness of the mixture. Were
relying on the change in the stiffness, and the EL model might be applicable. But what
Im saying is the procedure that were showing, how we can incorporate healing with the
endurance limit with certain pavement volumetrics that you have and with certain
529

polymer-modified materials, you can do it within a month or two and come up with your
own predictions or come up with your own endurance limit for that specific mix. But the
model that I am showing here is applicable for conventional neat binders.
DR. POWELL: I guess along the same lines, RAP and RAS, if you dont get complete
blending, would fall in. So maybe then, the procedure is good, but when you go to
implement this in the MEPDG, maybe not implement a model thats based on straight-
run binders, one gradation, but a way you can implement the procedure you might
consider.
DR. SOULIMAN: I agree. Thanks.
DR. RAMON BONAQUIST: My question is along the lines of how you implement it in
pavement design. I think that the methodology that youve put together so far for
implementation in pavement design isnt really correct. I think that with the healing-
based endurance limit that was the philosophy that we originally put in the work plan for
the project that preceded yours, you can design any pavement to basically be a perpetual
pavement and never have to deal with damage. What you would want to do is for the
traffic volume that you have, you have a given rest period. For the temperature that you
have, that rest period would then generate an endurance limit or strain level, and as long
as I keep my strain induced by traffic below that, I have a perpetual pavement. I think
that it can greatly simply pavement design, and I encourage you to think of that process
when you actually go to implement this in pavement design. You dont have to track
damage, you dont have to do millions and millions of runs of computer programs. What
you need to do is keep the induced strain in the pavement below the level of the
endurance limit. And I would like to hear your thoughts on that.
DR. SOULIMAN: Yes, I think thats a very good point. But the question is, youre
saying I need to keep the strain at the bottom of the HMA layer less than the endurance
limit. Is that right? So the question is how are we going to predict endurance limit, that
we need to be below that value?
DR. BONAQUIST: Youve done that.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. SOULIMAN: Exactly.
DR. BONAQUIST: The methodology that you're doing does that.
DR. SOULIMAN: To predict the endurance limit.
DR. BONAQUIST: Yes. What Im saying is that when I go to design a pavement
knowing this model, the methodology that we use in the mechanistic empirical design
guide where we run all of these simulations over and over again, its not necessary. We
have a given environment, we know the traffic level. The traffic level creates a rest
period. You can calculate rest periods for traffic. Theyre going to be shorter on higher
volume roads. Theyll be longer on low-volume roads. So from that rest period and
530

temperature calculate an endurance limit. Now all youve got to do is keep the strain
induced in the pavement for the design traffic below the endurance limit and you have a
perpetual pavement, right? This is what we want. We dont want any bottom-up fatigue
cracking, we only want to replace the surface. I guess the analogy that I have is I hope
the people who designed this building, designed it so that it doesnt fail. This isnt what
we do right now with pavements. We assume that they have finite life and that were
going to get fatigue cracking at some point in time. I think this healing-based endurance
limit actually gives you a reason why perpetual pavements ought to exist and that we
should be able to design every flexible pavement as a perpetual pavement.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
DR. SOULIMAN: I totally agree with what you mention. Again, the whole concept here
is I want to have a thick enough pavement that Im going to ensure within the design life
of the pavement, at the bottom of the HMA layer, Im having a low enough strain which
is going to be always less than the endurance limit. If I reach that, Im reaching my goal.
DR. BONAQUIST: Thats correct. So if thats case, you dont need to have the fatigue
curve. You dont need to have the high-strain part of the fatigue curve.
DR. SOULIMAN: Yes.
DR. BONAQUIST: You just need to know the endurance limit and keep the strain below
the endurance limit.
DR. SOULIMAN: I agree. And thats why here the way we developed the regular
fatigue lines is just if we are still interested to see how the old way of designing it. But
for me, I personally agree with you that what we should look at endurance limit value
only, were now going to have unlimited number of repetitions. Thank you.
531

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Investigation of Fracture Properties of
California Asphalt Mixtures Using Semi-
Circular Bending and Beam Fatigue Tests
Hamed Hakimelahia, Shadi Saadeha*, and John Harveyb
a
Department of Civil Engineering and Construction Engineering Management,
California State University, Long Beach, Long Beach, CA 90840
b
Department of Civil and Environmental Engineering, University of California, Davis,
CA 95616
ABSTRACT: Fatigue cracking is a primary distress in asphalt concrete due to repetitive stresses and
strains caused by both traffic loading and environmental factors. The fatigue resistance of AC is
investigated by a number of fatigue tests. The main objective of this study is to investigate the use
of the SCB test as a quality assurance/quality control (QA/QC) measure for field construction.
Comparison of fracture properties of seven AC mixtures from the semi-circular bending (SCB)
and Beam Fatigue test (BFT) is conducted. The Jc and K1c values for the seven mixtures were
determined. BFT was performed on the same mixtures and initial stiffness, Nf and PV were
determined. The coefficient of variation (CV) ranged from 0 to 38% for Jc and from 0 to 35% for
K1c.The coefficient of variation (CV) ranged from 10 to 93% for the initial stiffness, 2 to 83% for
Nf, and 8 to 167% for PV. The SCB Jc and BFT Nf and PV indicated lower fracture properties for
PG64-10RAP (LIME), AN-HMA and WMA-ADVERA mixtures than other mixtures. The BFT Nf
and PV achieved similar ranking for all mixtures. There is good correlation between Nf and PV
with Jc, and not a good correlation between initial stiffness with Jc, Nf and PV. This has indicated
that the initial stiffness is not a good representative for fracture properties of AC while Jc, Nf and
PV are better indicators. The results of this study indicate that the SCB test has great potential as
a QA/QC test of fracture properties of asphalt mixtures.
KEYWORDS: asphalt, semi-circular bending, beam fatigue, fracture, variability

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The oral presentation was made by Professor Saadeh.
http://dx.doi.org/ 10.1080/14680629.2013.812835
533

HAKIMELAHI, SAADEH, HARVEY
1. Introduction
Fatigue resistance of asphalt concrete (AC) is defined as the ability of the AC to resist
repeated traffic loading without significant cracking or failure (Harvey et al., 1995).
Fatigue cracking is a primary distress in asphalt concrete due to repetitive stresses and
strains caused by both traffic loading and environmental factors. The fatigue resistance
of AC is investigated by a number of fatigue tests. Fatigue cracking occurs where the
pavement has been stressed to the limit of its fatigue life by a repetitive axial load.
Fatigue cracking is often associated with loads that are too heavy for pavement structure
or more repetitions of a given load, which is provided for the design (Roberts et al.,
1996).
Wagoner et al. (2005) developed the disk-shaped compact tension (DC(T)) test as a
practical method for obtaining the fracture energy of asphalt concrete. Suitable specimen
geometry was developed using the ASTM E399 standard for compact tension testing of
metals. The fracture energy of the asphalt specimens can be calculated from this test.
Zhou et al. (2007) presented the Texas overlay test (OT). The researchers at the
Texas Transportation Institute developed this test to assess the fatigue cracking
prediction. The device tests AC specimens that are glued to two steel plates, with half of
its length resting on each plate. One of the steel plates is fixed, and the other moves
horizontally to simulate the opening and closing of the crack under overlays. The AC
mixture specimen has a standard dimension of 150 mm length by 75 mm width by 38
mm height. The test specimen can be fabricated from a Superpave gyratory compactor
(SGC) or from field cores.
Chiangmai (2010) examined the relation between fatigue and fracture parameters
obtained from the standardized laboratory tests. Two fatigue parameters, number of
cycles to failure and plateau value (PV), were obtained from flexural bending test. The
disk-shaped compact tension, DC(T), fracture test was used to obtain fracture energy
parameters, including: fracture strength, pre-peak fracture energy, post-peak fracture
energy, and total fracture energy. The correlations between fatigue and fracture
parameters indicated that the plateau value (PV) has a better relationship with fracture
parameters than the traditional fatigue criteria (Nf50 ). Also, results indicated an inverse
trend between fatigue and fracture parameters at low strain levels, such as 300 micro-
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
strains, in comparison with high strain levels, 700 and 1000 micro-strain.
534

Investigation of Fracture Properties of California Asphalt Mixtures
In California, the Asphalt Research Program performed a five-year study on the

fatigue performance of AC mixtures, (Harvey et al., 1995). To predict fatigue
performance of asphalt mixtures, the Beam Fatigue Test (BFT) was developed during
this study. The BFT is used as a quality control/quality assurance (QC/QA) test, however
the long testing period and the high variability of this test makes it impractical for the
QC/QA testing.
This study was motivated by the desire to investigate the feasibility of predicting the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
BFT test results with the SCB test results. The reason that SCB test is chosen in this
study is that it is simple to conduct (regular stabilometer can be used to conduct the test),
inexpensive (one compacted specimen makes four SCB specimens), simple to analyze
(the output parameter is indicative of the dissipated energy during crack propagation),
and the failure mode of SCB is due to tensile stress induced by bending.
During the 1990s a simple SCB test was developed to determine the crack resistance
and crack growth rate of bituminous mixtures. In recent years the SCB test has become
popular to determine the fracture toughness of AC. Hofman and his group found that the
cyclic SCB test is very promising to determine the constants for the Paris law for crack
propagation (Hofman et al., 2003).
Saadeh and Eljairi (2011) investigated the influence of moisture on fracture
properties of asphalt mixtures. Twenty-four SCB tests were conducted on two California
asphalt mixtures in dry and wet conditions at 20C. In addition, 16 beam fatigue tests
(BFT) were conducted on the same HMA mixtures in dry and wet conditions. The
results of these two tests were compared. The data analysis showed that there is a good
correlation between BFT and SCB tests in sample evaluation; however, they
recommended more tests should be done on more HMA mixtures in order to get more
definitive results.
Recently, the SCB test has been used in many pavement material studies to
investigate the fracture properties of HMA mixtures (Molenaar et al., 2002; Mull et al.,
2002; Mohammad et al., 2004; and Mohammad et al., 2003). The SCB test can be a
potential test to further explore the material fracture properties.
2. Objective
The objective of this study is to investigate the use of the semi-circular bend (SCB) test
as a quality assurance/quality control (QA/QC) measure for field construction and to
investigate the feasibility using SCB test results as a trigger for further investigation
using BFT. This objective was achieved by testing seven asphalt mixtures including
535

mixtures with recycled asphalt pavement (RAP), warm mix asphalt (WMA), lime, and
polymer modified asphalt.
3. Scope
Seven AC mixtures types (PG64-10RAP (LIME), PG64-28PM (LIME), 710P4-AR, AN-

HMA, AN-WMA, MnROAD, WMA-ADVERA) were investigated in this study.
BFT and SCB tests were conducted to investigate the performance of the AC
mixtures. The fracture properties of the AC mixtures were determined by these two tests.
Rolling wheel compaction (RWC) was used to compact the AC mixtures. The same
specifications and methodology were used to prepare cores for SCB testing and beams
for BFT. The SCB test was performed at California State University of Long Beach, and
BFT was conducted at University of California Pavement Research Center.
4. Methodology
In this study the use of the SCB test as a QA/QC test to trigger further investigation of
the fracture properties of asphalt mixtures using BFT is examined. To achieve this goal,
six different AC mixtures used in California and one from MnROAD are selected.
Different asphalt mixtures with different gradations and mixture designs were selected to
achieve a comprehensive comparison and evaluation.
4.1 Material Properties and Mixture Design
A total of 42 SCB tests and 21 BFT were conducted on seven asphalt mixtures (PG64-
10RAP (LIME), PG64-28PM (LIME), 710P4-AR, AN-HMA, AN-WMA, MnROAD,
WMA-ADVERA). For the SCB specimens, three notch depths 25.4, 31.8, and 38.0 mm
(1, 1.25 and 1.5 in.) were used in this study. Rolling wheel was the method of
compaction used. A similar compaction method was used to prepare the BFT specimens.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Table 1 provides additional information regarding the mixture properties used in this
study.
5. Semi-Circular Bend (SCB) Test
The SCB test is used to characterize the fracture resistance of asphalt mixtures based
on the fracture mechanics concept. The critical strain energy release rate, which is called
536

the critical value of J-integral or Jc was determined for each mixture tested. Another
parameter was determined called fracture toughness (K1c) according to Hofman et al.
(2003). In this test, semi-circular samples with three notch depths of 25.4, 31.75 and
38.1 mm (1, 1.25, and 1.5 in.) are used. The tests were conducted at 20C. A semi-
circular specimen was loaded monotonically until fracture under a constant cross-head
deformation rate of 0.508 mm/min (0.02 in/min) in a three-point bend load configuration
as shown in Figure 1. The maximum load (Fmax) at which the specimen fails was
determined. The load and deformation were continuously recorded and the critical value
of J-integral and K1c were determined using the following equation:
1 U
Jc b a [1]
where b is sample thickness, a is the notch depth, and U is the strain energy to
failure. In addition,
4.263.Fmax
max D.t
[2 ]
a
K1c max af W [3 ]
where
a a a 2 a 3
f 0.623 29.29 171.2 457.1
W W W W
a 4 a 5
561.2 265.54
W W
and max is the maximum stress, Fmax is the maximum load, K1c is fracture
toughness, a is the initial notch depth and W is the specimen height. Hofman et
al. (2003) determined the correlation factor f through 2D finite element analysis.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
537

Table 1. HMA mixture properties.

Specimen
Ag g r e g a t e
Mix Type Binder Type Mix Design Preparation
Source
Type
PG64-10RAP SHRP Level Red Bluff
PG64-10 (Valero) LMLC*
(lime treated) 1 mix design (District 2)
PG64-28PM PG64-28PM SHRP Level Red Bluff
LMLC
(lime treated) (Valero) 1 mix design (District 2)
San Gabriel
AR-8000 (PG64- Caltrans
710P4-AR River Valley at LMLC
16) Hveem
Azusa
PG64-10 (Valero- Caltrans Graniterock-
AN- H M A LMLC
Benicia) Hveem Wilson Quarry
PG64-10 (Valero- Caltrans Graniterock-
AN-WMA LMLC
Benicia) Hveem Wilson Quarry
SHRP Level
MnROAD FMFC**
1
PG64-16 (Valero-
WMA- Benicia) with Caltrans Graniterock-
FMFC
ADVERA Advera warm mix Hveem Wilson Quarry
treatment
*Lab-Mixed Lab-Compacted, **Field Mixed Field Compacted.
538
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 1. Set-up of semi-circular bend test.
6. Beam Fatigue Test (BFT)
This test was conducted according to AASHTO T 321. The test was conducted at a
strain level of 400 micro-strains and temperature of 20C. A total of 21 BFT on seven
asphalt mixtures were conducted based on controlled strain method.
In BFT, number of load cycles to failure (Nf50) is determined as the number of cycles
for a 50% reduction in initial stiffness (AASHTO T 321-03).
Ghuzlan and Carpenter (2000) used the ratio of dissipated energy (RDEC) as an
energy approach to describe fatigue damage. RDEC can be calculated based on Equation
4. The definition shows the change in damage from one cycle to another. In this equation
the typical number of cycles between b and a is 100, and this number can be higher
if the DE between 100 load cycles is low (Shen and Carpenter, 2007).
DE DE
RDECa = DE a(ba)
b
[4 ]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
a
where: a, b = loading cycle (LC) a and b, respectively (kPa);

RDECa = average ratio of dissipated energy change at cycle a, compared to the
next cycle b;
DEa, DEb = dissipated energy at load cycle a and b.
539

The damage curve which describes the relation between RDEC and number of cycles
to failure is shown in Figure 2. (Shen and Carpenter, 2007) divided the curve in three
stages: initial stage (stage I), damage growth stage (stage II) and the stage at which
damage grows rapidly (stage III). They proposed RDEC value, at the 50% stiffness
reduction point, is defined as the plateau value (PV), which is also shown in Figure 2.
Figure 2. Typical RDEC vs. Loading Cycles Plot and the Indication of PV and Nf50
(Shen and Carpenter, 2007).
Shen and Carpenter (2007) provided an equation to calculate PV, Equation 5. They
also proposed a simplified equation to calculate PV for low strain fatigue testing,
Equation 6.
100 k
1(1+ )
Nf 5 0
PV = 100
[5]
k
PV = N [6]
f5 0
where K= slope of the DE-LC curve.
540
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

7. Discussion of Results
7.1. Semi-Circular Bend (SCB) Test
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
A total of 42 SCB tests were performed on seven mixtures (PG64-10, PG64-28, 710P4-
AR, AN-HMA, AN-WMA, MnROAD and WMA-ADVERA) with three notch-depths
25.4, 31.8, and 38.0 mm (1, 1.25 and 1.5 in.), two replicates each.
Figure 3 presents the load-displacement curves for three notch depths of AC
mixtures; similar curves were constructed for all mixtures. The maximum load (Fmax)
was recorded and the area under the load displacement curve was used to determine the
strain energy (U). The calculated strain energy for each notch was used to determine the
critical strain energy Jc, Figure 4. By dividing the slope of the linear fitting line to the
specimens thickness the Jc value can be calculated. The calculated peak load and the
dissipated energy for each SCB test were used to determine the critical K1c and Jc values,
according to Equations 1 and 2.
The SCB Jc and K1c values for each AC mixture are presented in Table 2. The table
presents the dissipated energy (U), the coefficient of variation (CV%), the critical value
of the J-integral (Jc), and the ranking of the mixtures based on the Jc and K1c parameters.
The values of Jc ranged from 0.66 to 1.198 kNm (7.9 to14.4 lbin). The values for K1c
ranged from 1.7 to 4.4 N /m2. The presented values are the average of the calculated
data, and the coefficient of variation ranged from 0 to 38% for Jc and from 0 to 35% for
K1c.
In order to better compare the ranking of each parameter, analysis of variance
(ANOVA) was used. Using ANOVA, the p-value is used to determine whether the
model is significant. Typically the p-value is compared to an alpha value of 0.05. A p-
value that is lower than alpha indicates that the model is significant. The Tukey analysis,
which is a part of ANOVA, was used to examine the difference in the groups with
different rankings.
In order to categorize Jc and K1c values, Tukey analysis was conducted. As a result,
another basis for comparing results is provided. An example for statistical analysis is
provided in Tables 3 and 4. Tukey analysis categorized the seven mixtures into three
groups, group A (MnROAD and 710P4-AR), group AB (AN-WMA and PG64-28PM),
and Group B (WMA-ADVERA, AN-HMA, and PG64-10RAP). This grouping will be
using in the comparison of the ranking with BFT ranking in the following sections.
541

Figure 3. Load-deflection curve, PG64-10RAP.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 4. Strain energy curve, PG64-10RAP.
542

Table 2. SCB test results (Jc) and (K1c).

Notch U CV Jc K 1c
Mix Type Rank Rank
(mm) (kN.m) (%) (kN/m) (N/m2)
2 5 .4 1 .2 4 6 3
P G6 4 -
3 1 .7 5 1 .0 3 2 1 0 .6 6 2 7 4 .4 3 1
10RAP
3 8 .1 0 .7 9 7 10
2 5 .4 1 .0 5 1 28
P G6 4 -
3 1 .7 5 0 .7 4 9 18 0 .8 1 6 4 2 .8 8 4
28P M
3 8 .1 0 .4 9 8 26
2 5 .4 1 .5 4 3 11
710P4-
3 1 .7 5 1 .2 6 9 13 1 .1 3 2 2 4 .1 2 2
AR
3 8 .1 0 .7 7 6 22
2 5 .4 1 .1 1 3 37
AN-HMA 3 1 .7 5 0 .8 1 15 0 .6 9 6 6 3 .6 2 3
3 8 .1 0 .6 4 2 21
2 5 .4 1 .1 5 5 17
AN-
3 1 .7 5 0 .8 9 6 35 0 .8 6 6 3 2 .4 6 6
W MA
3 8 .1 0 .5 6 8 32
2 5 .4 1 .3 2 5 29
MnROAD 3 1 .7 5 1 .2 5 9 0 1 .1 9 8 1 1 .7 7
3 8 .1 0 .5 1 3 27
2 5 .4 1 .0 6 9 4
W M A-
3 1 .7 5 0 .8 6 38 0 .7 1 1 5 2 .6 9 5
ADVERA
3 8 .1 0 .5 8 7 12
Table 3. ANOVA for SCB, Jc value.

Seq Adj Adj
Source DF F P
SS SS MS
Mixture 6 8.00E+05 8.00E+05 1.30E+05 4 .3 9 0 .0 3 7
Erro r 7 2.10E+05 2.10E+05 3.00E+04
Total 13 1.00E+06
543
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 4. Tukey analysis for SCB, Jc value.

M ean
Mixture Group
*10^2
MnROAD 1 .2 A
710P4-AR 1 .1 A
AN-WMA 0 .9 AB
PG64-28PM 0 .8 AB
WMA-ADVERA 0 .7 B
AN-HMA 0 .7 B
PG64-10RAP 0 .7 B
7.2. Beam Fatigue Test (BFT)
A total of 21 BFTs were performed: seven mixtures (PG64-10, PG64-28, 710P4-AR,

AN-HMA, AN-WMA, MnROAD and WMA-ADVERA) and three replicate each. The
test was performed at 20C and a strain level of 400 micro-strains. The dissipated
energy, number of cycles to failure (Nf), initial stiffness, average ratio of dissipated
energy (RDEC) and plateau value (PV) were determined for all conducted BFT.
Tables 5, 6 and 7 present initial stiffness, Nf, and PV values for the conducted BFT
respectively. The tables presented the average of three replicates for each mixture except
710P4-AR, as BFT was conducted on two replicates for this mixture. In order to
compare the test results, the ranking for each test parameter was considered.
Furthermore, to check the variability of the test results, the coefficient of variation for
the test results was calculated. In Table 6 the results of the Nf parameter are presented. It
can be seen that the CV values were extremely high for three mixtures (PG 64-10RAP,
PG64-28PM, and AN-HMA). In order to reduce the coefficient of variation, it was
decided to split the results into two scenarios besides the original results.
In the first scenario the higher values are eliminated, and in the second one the lower
values are eliminated. The difference between the two modified groups is in the way of
eliminating outlier test results. The goal was to reduce the coefficient of variation.
Comparing the ranking of both scenarios it can be seen that the ranking did not change
very much, except for AN-HMA mixture. Therefore the analysis was continued based on
the first scenario. It can be noticed that the Nf and the PV gave similar ranking, Tables 6
and 7, whereas the initial stiffness, Table 5, was totally different.
544
--`,`,,,``,,```,,`,,,,,`,``,

ANOVA was done on the modified (first scenario) BFT results. The results for initial
stiffness, Nf and PV values with 95% degree of confidence are presented in Tables 8a, 9a
and 10a, and the results show that the p-value for all three parameters is lower than 0.05,
which indicts their significance. Tables 8b, 9b and 10b present the Tukey analysis results
for initial stiffness, Nf and PV values. Tables 9b and 10b categorize Nf and PV
parameters in four and six groups respectively. Table 11a presents ANOVA for initial
stiffness with 60% degree of confidence. The reason that ANOVA was conducted with
60% degree of confidence for this parameter is that the Tukey analysis with 95% degree
of confidence categorized initial stiffness values in one group, Table 8b. For sake of
ranking the mixtures, the degree of confidence was lowered, Table 11b; as a result three
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
groups were found.

The following section compares the ranking among the BFT parameters and against
the SCB parameters.
Table 5. BFT results (initial stiffness).
Modified Test Results Modified Test Results
Original Test Results
First Scenario Second Scenario
Mix Type
Ini S CV In i S In i Ini S
Ini S In i S Rank Rank
(GPa) (%) M1 S M2
5 .2 5 .2
PG64-
5 .0 5 .4 10 5 .2 5 5 .0 5 .5 3
10RAP
6 .1 6 .1
6 .1 6 .1
PG64-
2 .3 3 .6 61 6 .1 3 2 .3 2 .3 7
28PM
2 .4 2 .4
7 .3 7 .3 7 .3
710P4-AR 6 .7 13 6 .7 2 6 .7 2
6 .1 6 .1 6 .1
5 .9 5 .9
AN-HMA 5 .7 5 .4 11 5 .9 4 5 .7 5 .2 4
4 .7 4 .7
4 .9 4 .9 4 .9
AN-WMA 4 .2 4 .3 12 4 .2 4 .3 6 4 .2 4 .3 5
3 .8 3 .8 3 .8
3 .5 3 .5 3 .5
MnROAD 4 .3 3 .5 24 4 .3 3 .5 7 4 .3 3 .5 6
2 .6 2 .6 2 .6
82 82 82
WMA-
282 137 93 282 137 1 282 137 1
ADVERA
46 46 46
545

Table 6. BFT results (Nf).
Modified Test Modified Test Results

Original Test Results
Results First Scenario Second Scenario
Mix Type
Nf
*1003
Nf CV (%) Nf Nf M1 Rank Nf Nf M2 Rank
65 65
P G6 4 -
200 143 49 65 5 200 181 5
10RAP
163 163
163 163
P G6 4 -
5788 3703 83 163 3 5788 5473 1
28P M
5158 5158
710P4- 100 100 100
83 28 83 4 83 6
AR 67 67 67
45 45
AN-HMA 473 395 81 45 6 473 571 3
668 668
194 194 194
AN-
335 249 30 335 249 2 335 249 4
W MA
218 218 218
972 972 972
MnROAD 649 846 20 649 846 1 649 846 2
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
917 917 917
4 4 4
W M A-
4 4 2 4 4 7 4 4 7
ADVERA
4 4 4
546

Table 7. BFT results (PV).

Modified Test Modified Test
Original Test Results Results First Results Second
Mix
Scenario Scenario
Type
CV PV Ran PV Ran
PV*E10 PV PV PV
(%) M1 k M2 k
69 69
P G6 4 -
29 44 51 69 5 29 31 5
10RAP
33 33
33 33
P G6 4 -
0 .4 11 167 33 3 0 .4 0 1
28P M
0 .4 0 .4
710P4- 40 40 40
45 17 45 4 45 6
AR 51 51 51
74 74
AN-
5 .5 28 140 74 6 5 .5 5 3
HM A
5 .4 5 .4
27 27 27
AN-
15 23 30 15 23 2 15 23 4
W MA
28 28 28
3 .3 3 .3 3 .3
MnROA
3 .7 4 11 3 .7 4 1 3 .7 4 2
D
4 .1 4 .1 4 .1
154 154
1544 4 4
W M A-
169 182 182
ADVER 8 1696 7 1696 7
1824 6 4 4
A
171 171
1719 9 9
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
547

Table 8a. ANOVA for BFT, modified initial stiffness, 95.0% confidence.
Mixture 6 4.1E+10 4.1E+10 6.8E+09 1 .4 7 0 .3 1
Erro r 7 3.2E+10 3.2E+10 4.6E+09
Total 13 7.3E+10
Table 8b. Tukey analysis, modified initial stiffness values.

Mixture M ean Grouping
WMA-ADVERA 136520 A
710P4-AR 6670 A
PG64-28PM (LIME) 6050 A
AN-HMA 5855 A
PG64-10RAP (LIME) 5235 A
AN-WMA 4294 A
MnROAD 3488 A
Table 9a. ANOVA for BFT, modified Nf, 95.0% confidence.

Mixture 6 1.4E+12 1.4E+12 2.3E+11 2 2 .5 8 0
Erro r 7 7.1E+10 7.1E+10 1.0E+10
Total 13 1.5E+12
Table 9b. Tukey analysis, modified Nf values.

MnROAD 846093.3 A
AN-WMA 248782.3 B
PG64-28PM (LIME) 163116 BC
710P4-AR 83416 BC
PG64-10RAP (LIME) 64937 BC
AN-HMA 44667 BC
WMA-ADVERA 3699.5 C
548
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 10a. ANOVA for modified BFT PV, 95.0% confidence.

Mixture 6 6.1E+05 6.1E+05 1.2E+05 3 8 .4 8 0 .0 0 1
Er r o r 7 1.6E+04 1.6E+04 3.2E+03
Total 13 6.3E+05
Table 10b. Tukey analysis, modified PV values.

Mixture Mean *10^9 Grouping
MnROAD 3 7 .1 A
AN-WMA 2 3 3 .8 B
PG64-28PM (LIME) 3 3 1 .1 BC
710P4-AR 4 5 4 .4 CD
PG64-10RAP (LIME) 693 D
AN-HMA 7 3 8 .8 D
WMA-ADVERA 16955.7 E
Table 11a. ANOVA for BFT, modified initial stiffness, 60.0% confidence.
Mixture 6 4.1E+10 4.1E+10 6.8E+09 1 .4 7 0 .3 1
Erro r 7 3.2E+10 3.2E+10 4.6E+09
Total 13 7.3E+10
Table 11b. Tukey analysis, modified initial stiffness values, 60.0% confidence.
WMA-ADVERA 136519.7 A
710P4-AR 6670.2 AB
PG64-28PM (LIME) 6049.9 AB
AN-HMA 5854.5 AB
PG64-10RAP (LIME) 5234.9 AB
AN-WMA 4293.7 B
MnROAD 3487.5 B
549
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

8. Comparison between SCB Jc and BFT
The SCB test and BFT parameters ranking are provided in Table 12. The initial stiffness
did not have a good correlation with the other BFT parameters nor with the SCB
parameters. This is expected as it is consistent with previous studies that showed that the
initial stiffness is not a good indicator of fracture properties of AC mixtures.
The ranking of Nf and PV from the BFT is identical for these test results.
There is a good relation between Jc and Nf and PV. Tukey analysis of Jc grouped the
PG64-10RAP (LIME), AN-HMA and WMA-ADVERA mixtures in the lowest grouping
(B) and ranked 7, 6, and 5, respectively. Similarly the BFT Nf grouped these mixture in
the lowest groups (BC, BC, and A) and they ranked 5, 6, and 7.
ANOVA was used to compare the BFT and SCB test results. Table 13 shows the
direct results from Tukey analysis, which categorized mixtures based on BFT and SCB
test parameters. In order to better compare the Tukey analysis results, a constant scale
was needed for the sake of demonstration of the results on a figure. Therefore, it was
decided to convert the alphabetic results to numerical with a constant scale. The scale of
4 was chosen to convert the results. The converted results of the SCB test data as well as
the BFT data are presented in Table 13.
The converted test results from the BFT and SCB tests are used to correlate the
parameters of these two tests to each other. Figure 5 presents the correlation between the
BFT and SCB test results. In this figure Jc and K1c from the SCB test are plotted verses
PV and Nf values from the BFT. Figures 5a and 5b show that there is a good correlation
between the Nf and PV with the Jc values.
Figure 6 provides correlation between another SCB test parameter (K1c), and the BFT
parameters (Nf and PV). The correlation shows that the K1c has no correlation with the
Nf and PV, this correlation is almost reversed. This might be due to the fact that K1c is
calculated based on the peak load without consideration of the deformation.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
550

Table 12. Specimens ranking comparison.
Mix Type Ini S Nf PV Jc K 1c

PG64-10RAP 5 5 5 7 1
PG64-28PM 3 3 3 4 4
710P4-AR 2 4 4 2 2
AN-HMA 4 6 6 6 3
AN-WMA 6 2 2 3 6
MnROAD 7 1 1 1 7
WMA-ADVERA 1 7 7 5 5
Table 13. ANOVA for test results ranking.
Mix Type Nf PV Jc K 1c
PG64-10RAP BC 3 D 3 .3 3 B 4 A 1
PG64-28PM BC 3 BC 2 AB 2 .6 7 BC 2 .4
710P4-AR BC 3 CD 2 .6 7 A 1 A 1
AN-HMA BC 3 D 3 .3 3 B 4 AB 1 .6
AN-WMA B 2 B 1 .3 3 AB 2 .6 7 CD 3 .2
MnROAD A 1 A 1 A 1 D 4
W M A-
C 4 E 4 B 4 BC 2 .4
ADVERA
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
551

Nf M 1 9 5 %
5.00
WMA-
4.00 ADVERA
3.00 710P4-AR PG64-28PM AN-HMA

Nf
PG64-10RAP
2.00 AN-WMA
1.00 MnROAD
0.00
0.00 1.00 2.00 3.00 4.00 5.00
Jc
( a)
P V M1 9 5 % WMA-
5.00
ADVERA
4.00
AN-HMA
3.00
710P4-AR PG64-10RAP
PV
2.00 PG64-28PM
1.00 MnROAD AN-WMA
0.00
0.00 1.00 2.00 3.00 4.00 5.00
Jc
(b )
Figure 5. BFT and SCB test results correlation (a) Nf Jc (b) PV- Jc.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
552

Nf M 1 9 5 %
5.00
WMA-
4.00 AN-HMA ADVERA
3.00 PG64-28PM
Nf
2.00 710P4-AR AN-WMA

1.00 PG64-10RAP MnROAD
0.00
0.00 1.00 2.00 3.00 4.00 5.00
K1c
( a)
P V M1 9 5 %
5.00
4.00 PG64-10RAP WMA-
3.00 AN-HMA DVERA

A
PV
PG64-28PM
2.00 710P4-AR AN-WMA
1.00 MnROAD
0.00
0.00 1.00 2.00 3.00 4.00 5.00
K1c
(b )
Figure 6. BFT and modified SCB test results correlation (a) Nf K1c (b)PV- K1c.
9. Conclusion
The main objective of this study was to investigate the use of the SCB test as a quality
assurance/quality control (QA/QC) measure for field construction. Comparison of
fracture properties of seven HMA mixtures from the BFT and SCB test was conducted.
The Jc and K1c values for the mixtures were determined. BFT was performed on the
same mixtures and initial stiffness, Nf and PV for each mixture were determined. The
following conclusions were drawn:
The coefficient of variation (CV) ranged from 0 to 38% for Jc and from 0 to 35% for
K1c.
The coefficient of variation (CV) ranged from 10 to 93% for the initial stiffness, 2 to
83% for Nf, and 8 to 167% for PV.
553
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The SCB Jc and BFT Nf and PV indicated lower fracture properties for PG64-
10RAP (LIME), AN-HMA and WMA-ADVERA mixtures than other mixtures.
The BFT Nf and PV achieved similar ranking.
There is good correlation between Nf and PV with Jc, and not a good correlation
between initial stiffness with Jc, Nf and PV. This indicates that the initial stiffness is
not a good representative for fracture properties of AC while Jc, Nf and PV are better
indicators.
The results of this study indicate that the SCB test has great potential as a QA/QC
test of fracture properties of asphalt mixtures.
10. Recommendation
The results of this study indicated that the SCB test has great potential as a QA/QC test
of fracture properties of asphalt mixtures. However, further studies are recommended in
the following areas:
The BFT results need to be assessed for higher strain levels (700-1000 micro-strain).
Sensitivity analysis on different variables such as test temperature and binder
content, etc., should be evaluated.
The simplicity of performing the SCB test makes it the preferred test for the QA/QC
procedure on the fracture properties of a asphalt mixture. In addition, the SCB test
can lend itself easily to performance modeling and finite element analysis.
11. Acknowledgment
The authors would like to thank the United States Department of Transportation, the
California Department of Transportation and METRANS for their interest and provision
of grant support to make this study possible. We would also like to thank the University
of California Pavement Research Center and particularly the UC Berkeley laboratory for
their invaluable technical and operational support in this study.
12. Reference
Chiangmai C.N., Fracture Relation on Asphalt Concrete Mixtures, Thesis, University

Of Illinois at Urbana-Champaign, 2010.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
554

Ghuzlan, K. A., and Carpenter, S. H., Energy-Derived, Damage Based Failure Criterion
for Fatigue Testing, Transportation Research Record, 1723, Transportation
Research Board, Washington, D.C., 2000, 141149.
Harvey, J.T., Deacon, J.A., Tsai, B.W., and Monismith, C.L., Fatigue Performance of
Asphalt Concrete Mixes and Its Relationship to Asphalt Concrete Pavement
Performance In California, Asphalt Research Program: CAL/APT Program, Institute
of Transportation Studies, University of California, Berkeley, October 1995.
Hofman R., Oostcrbaan B., Erkens S., Kooji J., Semi-Circular Bending Test To Assess
The Resistance Against Crack Growth, Road and Hydraulic Engineering Institute,
Directorate Public Works and Water Management, Ministry Of Transport, Public
Works And Water Management, Delft, Netherlands, 2003
Mohammad, L. N., Negulescu, I, Wu, Z., Daranga, C., Daly, W.H and Abadie, C.,
Investigation of The Use of Recycled Polymer Modified Asphalt Binder In Asphalt
Concrete Pavements, Asphalt Paving Technology: Association of Asphalt Paving
Technologists, Vol. 72, Asphalt Paving Technology 2003, p 551594.
Mohammad, L. N., Wu, Z., & Aglan, M. A. Characterization of fracture and fatigue
resistance on recycled polymer-modified asphalt pavements, in C. Petit, I. L. Al-
Qadi, & A. Millien (Eds.), The 5th International RILEM conference on cracking in
pavements: Mitigation, risk assessment and prevention, RILEM Publications SARL,
Limoges, 2004, pp. 375382.
Molenaar, A.A.A.; Scarpas, A.; Liu, X.; Erkens, S.M.J.G. Semi-Circular Bending Test:
Simple but Useful? Association of Asphalt Paving Technologists, v 71, 2002, p
794815.
Mull, M. A., Stuart, K. and Yehia, A., Fracture Resistance Characterization of
Chemically Modified Crumb Rubber Asphalt Pavement, Journal of Materials
Science, Vol. 37, 2002, pp. 557566.
Roberts, F.L., Kandhal, P.S., Brown, E.R., Lee, D., Kennedy, T.W., Hot Mix Asphalt
Materials, Mixture Design and Construction, National Center for Asphalt
Technology, Third Edition.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Saadeh, S., Eljairi, O., Development of A Quality Control Test Procedure for
Characterizing Fracture Properties of Asphalt Mixtures, California State University
of Long Beach, Project No. 1024, June 2011.
Shen, S., and Carpenter, S. H., "An Energy Approach for Airport Pavement Low
Damage Fatigue Behavior," FAA Worldwide Airport Technology Transfer
Conference, Atlantic City, New Jersey, USA, 2007.
Wagoner, M. P., Buttlar, W. G., and Paulino, G. H., Disk-Shaped Compact Tension
Test For Asphalt Concrete, Journal of Experimental Mechanics, v 45, No. 3, June,
2005, pp. 270277.
555

Zhou, F., Hu, S., Scullion, T., Chen, C., Qi, X., Claros, G., Development and
Verification of the Overlay Tester Based Fatigue Cracking Prediction Approach,
Journal of the Association of Asphalt Paving Technologists, Volume 75, 2007, pp.
524577.
13. Discussion
PROFESSOR GABRIELE TEBALDI: Excellent presentation. Good job. Only one

question. If your goal is the quality control and to have simple specimen for quality
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
control, why not try to use Superpave IDT? You avoid cutting in the middle, you dont
have to do the notch. Is just the core and the cut or?
PROFESSOR SHADI SAADEH: Well, I guess this takes me back to my presentation.
Our goal was to have a test that is similar in inducing tensile stresses to the fatigue beam
that is used in California, in which the tensile stresses are induced by bending.
PROF. TEBALDI: Yes, but fracture energy is independent of distress; it is a
fundamental parameter. You use some fracture energy parameter and the fracture energy
is not related with the way that you apply the stress.
PROF. SAADEH: Well, I guess this is what we were thinking when we used this.
556

Development and Validation of a Rutting
Model for Asphalt Mixtures Based on the
Flow Number Test
Maria C. Rodeznoa* and Kamil E. Kaloushb

a
National Center for Asphalt Technology (NCAT) at Auburn University, Auburn, AL,
36830
b
Civil, Environmental and Sustainable Engineering Program, Arizona State University,
Tempe, AZ 85287-5306
ABSTRACT: This study presents the development and validation of a rutting prediction model based
on input of the Flow Number test. It was realized that such a model would require a very large
laboratory testing program along with the construction of field test sections. Therefore, a Flow
Number predictive equation, developed and published by the authors in a previous study, was one
of the important tools used in this paper together with the Mechanistic Empirical Pavement
Design Guide (MEPDG, now DARWin-ME). The MEPDG allowed the generation of hundreds of
rutting values for different pavement structures, climatic locations, and traffic levels. A total of
1440 asphalt pavement sections were used in this analysis. For this rutting Flow Number
relationship, a power model form was recommended that also included traffic and layer thickness
as input variables and had very good statistical goodness of fit. In addition, the model was
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
validated using actual laboratory Flow Number testing and field rutting data from 10 test sections
installed at the National Center Asphalt Technology (NCAT) Test Track in 2006. The results
showed that the model had very good accuracy in predicting the field rutting of the NCAT test
sections.
KEYWORDS: Flow Number Test, Rutting Model, MEPDG, DARWin-ME, NCAT Test Sections.
1. Introduction and Background
The Flow Number is defined as the load repetition value when shear deformation, under
constant volume, starts. This is equivalent to the start of tertiary flow within the asphalt
mixture. The Flow Number parameter was found to be a good indicator of the asphalt
The oral presentation was made by Dr. Rodezno.
557

RODEZNO, KALOUSH
mixture potential for permanent deformation in the field (Witczak et al., 2002).
However, there is no universal relationship developed to date relating the Flow Number
obtained in the laboratory and the rut depth measured in the field.
In 2002, an approach was published where a conceptual relationship between the
Flow Number obtained in the laboratory and the rut depth measured in the field for given
traffic levels was presented (Kaloush and Witczak, 2002). The relationship is presented
in Figure 1. For a given traffic level and a maximum desired rut depth, the required
number of load repetitions is determined in the laboratory and compared to the number
of cycles to failure for the particular mixture. This approach was presented using limited
data from three experimental sites with known field performance: MnROAD, WesTrack
and ALF. The test conditions used in this approach were unconfined and conducted at
54.4C (130F). The Flow Number data were shifted at the effective pavement
temperature and a stress level of 69 kPa (10 psi). This stress level was selected based on
the laboratory testing program conducted. The field rut depth measured for all three
experimental sites were only available at an ESAL of 175,000. (Kaloush and Witczak,
2002).
Rut Depth Versus Laboratory Flow Number of Repetitions at

Site Effective T emperature
(Actual Curve Developed at Stress Level of 10 psi and Traffic Level of
175,000 ESALs)
45
40
35 MnRoad
Rut Depth, mm
30 ALF
Increasing Traffic Lev el
25 W esTrack
20
15
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
10 Hypothetical
5
A c tu a l
0
0 20000 40000 60000 80000 100000
Flow Number of Repetitions
Figure 1. Relationship between field rut depth and laboratory Flow Number (Kaloush
and Witczak, 2002).
Another attempt to develop a relationship between the Flow Number and rutting was
presented in NCHRP Report 580 (Witczak, 2007). The study explored the relationship
558

Development and Validation of a Rutting Model for Asphalt Mixtures
between the Flow Number and field rut depth at different traffic levels. Using laboratory
temperature shift factors, laboratory measured Flow Numbers were reduced for
unconfined and confined tests. The analysis included laboratory and field mixtures from
the NCHRP 9-19 test sites, Arizona DOT, NCAT and Long Term Pavement
Performance (LTPP) database (Witczak, 2005). A wide range of temperatures and stress
levels were used in the laboratory-testing program to develop relationships for
temperature-reduced Flow Number. The results of the permanent deformation criteria
developed showed that there are very good correlations among laboratory measured
Flow Number, field rutting and traffic. However, these correlations were only specific
for each individual test site. The model used for predicting field rut depth (Rd) as a
function of Flow Number for the individual sites was presented as follows:
Rd = i1 * ( N ) m2 * ( Fn) m1*Log ( N ) l0 [1]
where: i1, m2, m1 and l0 are fitting parameters and N is the traffic in ESAL
In 2010, an effort was undertaken by the authors to develop a Flow Number
predictive model (Rodezno et al., 2010). The model developed used HMA mixture
volumetric properties and stress/temperature testing conditions as predictor variables.
The laboratory test data used was a combination of two major databases. The first one
included tests conducted at Arizona State University; the second one included tests
conducted by the Federal Highway Administration (FHWA) Mobile Asphalt Material
Testing Laboratory (MAMTL). Ninety-four mixtures were evaluated and a total of 1759
Flow Number test results were available. Since its publication in 2010, the model was
slightly revised to include a total of 1801 data points. The model has eight terms and a
coefficient of determination R2 of 0.64. The final form of the model is as follows
(Rodezno et al., 2010):
log FN = 0.485 + 0.644 LogV1 + 0.0874 P200 3.323Logp + 0.0129 R04 0.0803Va +
2.593 log q 0.0142 R34
[2 ]
where: FN = Flow Number
V1 = Binder Viscosity at testing temperature defined in terms of Ai and VTSi
P200 = Percentage passing No. 200 sieve
R04 = Percentage retained in No. 4 sieve
R34 = Percentage retained in in. sieve
Average Normal Stress p (psi), p = (1+3)/2
Maximum Shear Stress q (psi), q = (1-3)/2, 1 and 3 = principal stresses
559
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

RODEZNO, KALOUSH
The availability of the above Flow Number predictive equation was instrumental in
developing the Flow Number rutting relationship as explained in the subsequent
sections of this paper.
2. Study Objective
The main objective of this research study was to develop a rutting prediction model for
asphalt mixtures based on the Flow Number test and to validate the model using field
performance data from ten NCAT test sections constructed in 2006. In order to
accomplish this objective, a comprehensive analysis using the Flow Number predictive
equation in conjunction with the MEPDG (now DARWin-ME) was used to develop such
a model. The model was successfully validated with actual Flow Number and field
performance test data from the NCAT test sections.
3. Methodology
The conceptual approach presented earlier by Kaloush was evaluated and further
developed (Kaloush and Witczak 2002). The availability of the Flow Number predictive
equation and the MEPDG allowed the generation of the necessary data to develop such
correlations. This effort was undertaken in two parts as follows:
1. For a wide range of asphalt mixture properties, the Flow Number values were
generated using the Flow Number predictive Equation 2.
2. The MEPDG was used to generate HMA rutting values. This was accomplished by
using the same asphalt mixtures selected in part 1, a wide range of climatic
locations, and pavement structures. Therefore, the MEPDG was a necessary tool
that provided the predicted rutting for these mixtures. Otherwise, such an approach
would have required an extensive effort for field data collection of performance and
mixture data. The MEPDG also allowed the use of different traffic levels, which
was an important input for this approach.
The ultimate output of such analysis was that given the Flow Number test result of a
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
specific asphalt mixture at defined stress/temperature conditions, the field rut depth
could be estimated or predicted for the given traffic level and layer thickness.
3.1 Flow Number Values Generation
The Flow Number predictive equation was used to obtain Flow Number values for
different asphalt mixtures. The variables for these mixtures were selected based on the
560

variables in the predictive equation. The variables were: viscosity at testing temperature
(V1); air voids level; temperature; aggregate gradations of the mixture, %P200, %R04 and
%R34; and the average normal stress and maximum shear stresses (p and q). The deviator
stress (d=1-3) was fixed at 483 kPa (70 psi).
Table 1 summarizes the range of each variable used to generate the Flow Number
values corresponding to a wide range of asphalt mixture properties. For each climatic
location, two binder types were selected. This gave a total of 320 Flow Number values.
(20 locations x 2 binders x 2 air voids x 4 gradations).
Table 1. Variables used in the Flow Number prediction analysis.
Variable Levels Description
Climatic Location 20 Cold to Hot Climate (Alaska- Arizona)
Air Voids 2 4 and 10%

R(%) R3/8 (%) R4 (%) P200 (%)
1 0.4 5.8 27.0 5.4
Gradation 2 0.0 24.6 51.1 3.5
3 3.0 35.5 57.5 4.2
4 9.2 47.9 70.0 6.0
The selection of different climatic locations had an impact on the Flow Number
prediction. This is because the binder types and temperatures were selected individually
for each location. The binder selection was based on the LTPPBind program, version
3.1, developed by Pavement Systems LLC. The temperature selected for each location
corresponded to the effective pavement temperature for permanent deformation
calculated using the latest version of the effective pavement temperature equation (El-
Basyouny and Jeong, 2009). The equation is follows:
Teff = 14.62 3.361Ln(Freq ) 10.94 z + 1.121(MAAT )+ 1.718( MMAT ) 0.431(Wind ) +
0.333(Sunshine ) + 0.08(Rain )
[3]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
where: Teff = modified Witczak effective temperature for rutting, F

Freq = frequency loading, Hz
z = desired pavement depth, inches
MAAT = mean annual air temperature, F
MMAT = standard deviation of the mean monthly air temperature
561

RODEZNO, KALOUSH
Wind = mean annual wind speed, mph

Sunshine = mean annual sunshine percentage, %
Rain = cumulative annual rainfall depth, inches
Table 2 presents a summary of the 20 environmental locations selected with their

respective Mean Annual Air Temperature, Mean Annual Average Wind speed, Mean
Annual Sunshine Percentage, Cumulative Annual Precipitation, and Mean Monthly Air
Temperature Standard Deviation. These values were obtained from the MEPDG
Enhanced Integrated Climatic Model (EICM) Database. Table 2 also presents the final
effective temperatures for HMA rutting obtained using Equation 3. These values were
used in the Flow Number predictive equation.
3.2 HMA RuttingValues Generation
The HMA rutting values used in this study were generated using the MEPDG. A closed
form solution developed in 2008 was used in this approach as an alternative tool to
dramatically reduce the run time for each MEPDG simulation. (El-Basyouny and Jeong,
2009).
The HMA rutting values were obtained for the environmental locations identified in
Table 2. The basic structure of each pavement section in those locations consisted of a
variable HMA layer thickness, over 25.4 cm (10 in) of base course material with a
resilient modulus of 275,760 kPa (40,000 psi), and a subgrade with a resilient modulus
equal to 377,880 kPa (20,000 psi). Three traffic levels were used, together with the
gradations and air void levels defined. Table 3 shows a summary of the HMA thickness
and ESAL levels used. Table 4 also summarizes the different binders selected for each
location based on the traffic levels. As mentioned earlier, the binder selection was based
on LTPP Bind 3.1. This program recommends binders that are adjusted when the traffic
level and speed are defined.
For this analysis, two binder types were selected for each location; one for a traffic
level up to 3 million ESAL, and one for a traffic level between 10 and 30 million ESAL.
The traffic speed was defined as fast for all the binders selected. It is noted that for
Anchorage, Alaska, the same binder was specified at all traffic levels, since the MEPDG
does not include the low binder PG grade recommended for low traffic level (PG 40-16).
The above variable combinations generated a total of 1440 AC rutting predictions.
(20 locations x 2 air voids x 4 gradations x 3 traffic levels x 3 thicknesses). The
predictions of Flow Number values are a function of the gradation and volumetric
properties of the mixture only. This means that the same Flow Number value can be
related to AC layer structures with different thicknesses and traffic levels.
562
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 2. Summary of selected environmental locations.

Mean
Mean Mean Mean Annual
Monthly
No. Location Annual Annual Annual Cumulative T e ff
Air Temp
Air Temp Wind Speed Sunshine Rainfall C (F)
St Dev.
C (F) C (F) km/h(mph) (%) cm (in)
1 Phoenix, AZ 23.9 8.3 8.5 16.4 48.8
60.9
(75.0) (15.12) (5.3) (6.5) (119.9)
2 8.2 7.6 9.0 47.1 34.0
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Flagstaff, AZ 78.2
(46.8) (13.75) (5.6) (18.5) (93.2)
3 Intern. Falls, 3.6 11.5 10.9 74.7 32.6
52.5
MN (38.5) (21.0) (6.8) (29.4) (90.6)
4 Indianapolis, 12.0 9.2 13.3 106.5 37.1
37.6
IN (53.5) (16.7) (8.3) (41.9) (98.8)
5 Madison, WI 8.7 10.0 10.4 80.7 35.8
54.5
( 47.7) (18.2) (6.5) (31.8) (96.5)
6 Miami, FL 24.7 3.1 11.3 161.1 41.6
42.8
(76.5) (5.6) (7.0) (63.4) (106.8)
7 10.9 4.8 10.1 99.6 24.6
Seattle, WA 28.8
(51.6) (8.8) (6.3) (39.2) (76.3)
8 Grand Rapids, 9.3 9.4 13.3 82.6 33.8
36.0
MI (48.8) (17.2) (8.25) (32.5) (92.9)
9 Great Falls, 7.4 8.5 15.9 31.9 35.8
63.2
MT (45.4) (15.5) (9.9) (12.6) (96.5)
10 7.2 9.2 32.1
Redmond, OR 9.1 (48.4) 68.3 20.9 (8.2)
(13.0) (5.7) (89.8)
11 Corpus 22.2 5.4 15.7 74.1 46.2
55.7
Christi, TX (71.9) (9.8) (9.8) (29.1) (115.2)
12 Anchorage, 3.3 8.5 8.8 39.1 19.8
24.0
AK (38.0) (15.5) (5.5) (15.4) (67.7)
13 Los Angeles, 16.8 2.6 10.5 41.9 30.5
50.0
CA (62.3) (4.8) (6.5) (16.5) (86.9)
14 Houston, TX 20.6 6.2 9.5 131.3 39.9
38.9
(69.0) (11.3) (5.9) (51.7) (103.9)
15 Las Vegas, 20.7 9.1 11.3 49.1
65.4 12.7 (5.0)
NV (69.3) (16.6) (7.0) (120.3)
16 16.7 7.1 10.8 118.1 38.7
Atlanta, GA 44.8
(62.0) (12.9) (6.7) (46.5) (101.7)
17 Grand Forks, 4.7 12.0 14.3 50.8 36.2
52.3
ND (40.4) (21.9) (8.9) (20.0) (97.1)
18 13.9 9.7 15.8 87.1 43.5
Wichita, KS 49.9
(57.1) (17.6) (9.8) (34.3) (110.3)
19 10.0 8.9 8.2 125.0 35.8
Danbury, CT 60.4
(50.0) (16.1) (5.1) (49.2) (96.4)
20 16.3 8.9 12.7 98.0 43.1
Tulsa, OK 50.0
(61.3) (16.1) (7.9) (38.6) (109.6)
563

RODEZNO, KALOUSH
Table 3. Additional variables used in the AC rutting prediction analysis.
Variable Levels Description

AC thickness cm (in) 3 7.6, 15.2, 22.9 (3, 6, 9)
ESALS 3 2, 10 and 30x106
Table 4. PG Binder Grade Selected by Location.
No. PG Grading
Location
< 3 Millions ESALs 10-30 Millions ESALs
1 Phoenix, AZ PG 70-10 PG76-10
2 Flagstaff, AZ PG 52-22 PG 58-22
3 International Falls, MN PG 52-34 PG 64-34
4 Indianapolis, IN PG 58-22 PG70-22
5 Madison, WI PG 52-22 PG70-22
6 Miami, FL PG 64-10 PG 76-10
7
Seattle, WA PG 52-10 PG 64-10
8
Grand Rapids, MI PG 52-22 PG70-22
9
Great Falls, MT PG 52-28 PG 64-28
10 Redmond, OR PG 52-22 PG 64-22
11
Corpus Christi, TX PG 70-10 PG 76-10
12
Anchorage, AK PG 52-28 PG 52-28
13
Los Angeles, CA PG 52-10 PG 70-10
14
Houston, TX PG 70-10 PG 76-10
15
Las Vegas, NV PG 70-10 PG 76-10
16
Atlanta, GA PG 64-10 PG 76-10
17
Grand Forks, ND PG 52-28 PG70-28
18 Wichita, KS PG 64-16 PG 76-16
19 Danbury, CT PG 58-16 PG70-16
20 Tulsa, OK PG 64-16 PG 76-16
564
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

4. Initial Assessment of the Rutting Flow Number Relationship Approach
The first part of the analysis consisted of finding the relationship between the Flo w
Number values and AC rutting for a 3 in. AC layer, three traffic levels and only one
aggregate gradation. This step was considered important since the sensitivity analysis of
the Flow Number showed dependency on the gradation of the mixture. On the other
hand, the AC rutting prediction using the MEPDG was not sensitive to the AC mixture
gradation changes.
Table 5 shows the comparison between predicted Flow Number and predicted AC
rutting for a 7.6 cm (3 in.) AC layer structure, two traffic levels, all 20 locations and four
different gradations. As can be observed from these tables, the Flow Number values
predicted are clearly different for the four different gradations; on the other hand, the AC
rutting values are very similar. Because of this, the initial approach in this section was to
find the relationship between AC rutting, Flow Number, traffic level and ESALs using
Gradation 1 only, presented in Table 1.
For this initial analysis, Figure 2 shows the Flow Number-Rutting relationships for
the 7.6, 15.2, 22.9 cm (3, 6, 9 in.) AC layers, three traffic levels, and using mixture
Gradation 1. The graphs show that there is a power non-linear relationship that is shifted
on the vertical scale for the different traffic levels for all thicknesses.
A nonlinear regression analysis was performed on the data generated using Gradation
1. This included a total of 360 data points. The dependant variable was the AC rutting,
and the independent variables were the Flow Number, traffic level and thickness of the
AC layer. The form of the equation is as shown in Equation 4:
R = k1 * FN k 2 * ESALs k 3 * h k 4
[4]
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
where: R = AC Rutting (in)

FN = Flow Number (cycles)
ESALs = Traffic level in ESALs
H = Thickness of the AC layer (in)
k1, k2, k3, k4 = regression parameters
The regression analysis resulted in the following estimates of the coefficients:
R = 0.005813 * FN 0.3116 * ESALs 0.498 * h 1.108 [5]
R2adj=0.91; Se/Sy=0.28
565

RODEZNO, KALOUSH
Table 5. Predicted Flow Number and AC rutting for four gradations, 7.6 cm (3 in.) AC layer and 2,000,000 ESAL.
Location Predicted FN MEPDG Predicted AC Rutting (in.)
Grad 1 Grad 2 Grad 3 Grad 4 Grad 1 Grad 2 Grad 3 Grad 4
Phoenix, AZ 1,840 2,597 3,276 5,565 0.33 0.36 0.35 0.35
Flagstaff, AZ 2,574 3,632 4,583 7,784 0.21 0.23 0.23 0.22
Internat. Falls, MN 1,780 2,511 3,168 5,382 0.21 0.23 0.23 0.22
Indianapolis, IN 2,721 3,840 4,844 8,229 0.19 0.21 0.21 0.21
Madison, WI 1,784 2,518 3,177 5,396 0.23 0.25 0.25 0.24
Miami, FL 2,798 3,949 4,982 8,462 0.24 0.26 0.26 0.26
Seattle, WA 13,229 18,669 23,552 40,005 0.10 0.11 0.11 0.11
Grand Rapids, MI 2,373 3,349 4,225 7,177 0.18 0.19 0.19 0.19
Great Falls, MT 1,441 2,033 2,565 4,357 0.19 0.21 0.21 0.20
Redmond, OR 3,066 4,327 5,459 9,272 0.19 0.21 0.21 0.21
Corpus Christi, TX 2,485 3,507 4,424 7,515 0.23 0.25 0.25 0.24
Anchorage, AK 31,545 44,516 56,161 95,394 0.07 0.08 0.14 0.08
Los Angeles, CA 4,681 6,605 8,333 14,154 0.16 0.18 0.17 0.17
Houston, TX 5,368 7,575 9,556 16,233 0.21 0.23 0.23 0.22
Las Vegas, NV 1,790 2,526 3,186 5,412 0.30 0.33 0.32 0.32
Atlanta, GA 4,065 5,737 7,237 12,293 0.20 0.23 0.22 0.22
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Grand Forks, ND 1,382 1,950 2,460 4,179 0.21 0.23 0.22 0.21
Wichita, KS 2,084 2,941 3,711 6,303 0.24 0.26 0.25 0.25
Danbury, CT 3,544 5,001 6,309 10,716 0.21 0.23 0.23 0.23
Tulsa, OK 2,199 3,103 3,914 6,649 0.24 0.27 0.26 0.25
566

1 .6
1 .4
AC Layer =7.6 cm (3 in)
1 .2
MEPDG AC Rutting (in)
1 y = 1.785x-0.2763 y = 5.750x-0.3197 y = 9.565x-0.3168

R2 = 0.56 R2 = 0 . 6 7 R2 = 0 . 7 3
0 .8
0 .6
0 .4
0 .2
0
0 10000 20000 30000 40000
Predicted Flow Number
2E+06 ESALs 10E+06 ESALs 30E+06 ESALs
(a) AC Layer-7.6 cm (3 in.)
0 .8
0 .7
0 .6
0 .5
y = 2.599x-0.4286 y = 5.204x-0.4191 y = 8.991x-0.4243
0 .4
R2 = 0 . 7 5 R2 = 0.79 R2 = 0 . 7 9
0 .3
0 .2
0 .1
0
0 10000 20000 30000 40000
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

(b) AC Layer-15.2 cm (6 in.)
567

RODEZNO, KALOUSH
0 .8
0 .7
0 .6
0 .5
y = 2.078x-0.4182 y = 3.402x-0.3894 y = 6.194x-0.398
0 .4 R2 = 0 . 7 8
R2 = 0 . 7 0 R2 = 0 . 7 9
0 .3
0 .2
0 .1
0
0 10000 20000 30000 40000
(c) AC Layer-22.9 cm (9 in.)
Figure 2. Relationship between Flow Number and MEPDG AC rutting for 7.6, 15.2,
22.9 cm (3, 6, 9 in.) AC layer, 3 traffic levels and gradation 1.
Based on generally accepted goodness of fit statistics classifications found in

NCHRP Report 465, this approach appeared to provide a very good relationship
between rutting, Flow Number, ESAL and pavement thickness (Witczak et al, 2002).
5. Final Assessment of the Rutting Prediction Model
This section examines the relationship established in Equation 4 for all four gradations.
The wide range of gradations was selected based on the gradations of all mixtures that
were available in the Flow Number database. This range of gradations was presented in
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Table 1 and reproduced in Figure 3 using the FHWA 0.45 power gradation chart.
568

100
90
80
Percenta g e Pa ssing
70
60
50
40
Grad 1
30
Grad 2
20
Grad 3
10 Grad 4
0
N o 200
3/ 4 i n
3/ 8 i n
N o. 4
Sie ve Size
Figure 3. FHWA 0.45 Chart for Different Gradations.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Similar to the approach presented before, the relationship between AC rutting,
Flow Number, ESAL, pavement thickness, and the four gradations was established.
Figure 4 shows this relationship for the 7.6, 15.2, 22.9 cm (3, 6 and 9 in.) AC layers.
From these graphs it can be observed a similar power, non-linear, and rational
relationship exists for all thicknesses and traffic levels.
569

RODEZNO, KALOUSH
1 .6
1 .4 AC LayerA=C 7L.6aycem
r (3 in)
1 .2
y = 1.564x-0.2382 y = 3.629x-0.2445 y = 6.354x-0.2477
1 R2 = 0 . 5 2 R2 = 0 . 5 1 R2 = 0 . 5 5
0 .8
0 .6
0 .4
0 .2
0
0 20000 40000 60000 80000 100000 120000
(a) AC Layer-7.6 cm (3 in)
1
0 .9 AC Layer =15.2 cm (6 in)
0 .8
0 .7
0 .6
y = 1.628x-0.3421 y = 3.195x-0.3284 y = 5.163x-0.3232
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0 .5 R2 = 0 . 6 1 R2 = 0 . 5 6 R2 = 0 . 5 5
0 .4
0 .3
0 .2
0 .1
0
0 20000 40000 60000 80000 100000 120000
(b) AC Layer-15.2 cm (6 in)
570

0 .8
0 .7 AC Layer =22.9 cm (9 in)
0 .6
0 .5
y = 0.854x-0.281 y = 1.876x-0.2862 y = 3.231x-0.288
0 .4 R2 = 0 . 4 6 R2 = 0 . 5 2 R2 = 0 . 5 2
0 .3
0 .2
0 .1
0
0 20000 40000 60000 80000 100000 120000

(c) AC Layer-22.9 cm (9 in)
Figure 4. Relationship between Flow Number and MEPDG rutting for 7.6, 15.2, 22.9
cm (3, 6, 9 in.) AC layer, three traffic levels and four gradations.
A similar non-linear regression analysis was then performed for the complete data
set of all variables. A total of 1440 combinations were used in this analysis. The same
form of Equation 4 was used. Again, the AC rutting was the dependent variable, and the
independent variables being the Flow Number, traffic level and thickness of the AC
layer.
The final model form is presented in Equation 6:
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
R = 0.0038 * FN 0.242 * ESALs 0.485 * h 1.021 [6]
The final model has an associated R2adj = 0.86 and a Se/Sy= 0.37. These values are a
little lower than the ones obtained using only one gradation analysis (R2adj = 0.91; Se/Sy =
0.28), but still considered as very good measures of model accuracy.
571

RODEZNO, KALOUSH
Figure 5 shows the comparison of MEPDG AC rutting versus AC rutting based on

the Flow Number. It is noticed from this figure that the model tends to under predict AC
rutting values above 1 in. Most of these data points correspond to locations with the
highest effective pavement temperatures (Phoenix and Las Vegas). Despite this fact, the
model still represents a very powerful tool in the prediction of the AC rutting based on
laboratory or predicted Flow Number values for a specific mixture.
AC Rutting based on Predicted FN (in) .
1.6
1.4 N=1440
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Se/Sy=0.36
1.2 2
R adj=0.86
1
0.8
0.6
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Figure 5. Comparison of MEPDG AC rutting versus AC rutting based on Flow Number

for AC Layer. (1 in=25.4 mm)
6. Model Validation
Field rutting data from the NCAT pavement test track were used to validate the final
rutting model form presented in Equation 5. The rutting data used corresponds to test
sections that were part of the NCAT 2006 cycle (Willis et al., 2009). A total of ten test
sections were selected based on the availability of laboratory flow number test results.
Table 6 shows the gradation of the ten mixtures. Table 7 summarizes information that
includes binder grade, mix type, asphalt content, thickness of the layer, Flow Number
test results and field rutting data at five and ten million ESALs. For some of the sections,
no rutting information was available because the mixes of these sections were replaced
572

after five million ESALs due to excessive cracking and rutting. For the actual Flow
Number tests, all specimens were prepared in the laboratory using plant-produced mixes
sampled during construction (Willis et al., 2009). The tests were conducted using an
Asphalt Mixture Performance Tester (AMPT) with a deviator stress of 483 kPa (70 psi)
at 58 C. Three replicates were tested for each mix.
Table 6. Gradation of the ten NCAT sections used for validation.
Sieve Percentage Passing

Size N1 N2 N8 N1 0 S7A S7B S8A S8B S11 S12
11" 100 100 100 100 100 100 100 100 100 100
3/4" 100 100 100 100 100 100 100 100 100 100
1/2" 97 97 93 96 98 100 97 98 100 100
3/8" 82 85 71 83 90 91 88 88 100 97
No . 4 59 61 31 52 71 76 66 63 86 65
No . 8 49 50 22 33 58 53 53 49 67 38
No . 1 6 39 39 17 21 45 42 41 38 52 36
No . 3 0 30 31 15 14 32 30 30 28 37 18
No . 5 0 22 23 13 9 18 17 18 18 21 11
No . 1 0 0 14 15 12 7 12 11 11 12 13 10
No . 2 0 0 8 .8 9 .6 1 0 .5 5 .4 8 7 .4 7 .5 7 .8 8 .6 7 .4
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,
573

RODEZNO, KALOUSH
Table 7. Mix Information, Flow Number Test Results and Field Rutting
Data at two Levels of Traffic.
Rutting Rutting
% Thickness
Section Binder NM AS FN (mm) at 5 (mm) at 10
AC (cm)
million million
ESALS ESALS
N1 67-22 1 2 .5 4 .9 1 7 .8 493 0 .0 6 N/ A
N2 76-22 1 2 .5 4 .8 1 7 .8 1121 0 .0 6 0 .1 5
N8 76-28 1 2 .5 6 .9 2 5 .4 1840 0 .0 8 0 .0 9
N1 0 70-22 1 2 .5 5 .6 2 0 .3 618 0 .3 3 0 .4 8
S7A 64-22 9 .5 6 .5 5 .1 348 1 .3 1 N/ A
S7B 64-22 1 2 .5 6 .1 5 .1 407 0 .8 4 N/ A
S8A 64-22 1 2 .5 6 .2 5 .1 301 0 .9 1 N/ A
S8B 64-22 1 2 .5 6 .1 5 .1 280 0 .9 7 N/ A
S11 76-22 9 .5 6 .9 1 7 .8 752 0 .2 7 0 .4 3
S12 76-22 9 .5 7 .4 1 0 .2 731 0 .5 4 0 .7 7
Using the rutting model that was developed in Equation 5, the rutting values at the
two traffic levels were predicted for the ten NCAT sections for a total of 15 data points.
Figure 6 shows the comparison of field rutting versus predicted rutting. The final
prediction has an associated R2adj = 0.83 and a Se/Sy = 0.47 that indicates a very good
accuracy in the prediction. The availability of the NCAT laboratory Flow Number test results
along with measured field rutting values and traffic loadings for the same mixtures provided
an excellent venue for validating the rutting model developed in this study. Overall, the model
is rational and offers a good alternative of correlating the laboratory derived Flow Numbers for
asphalt mixture to their expected field rutting performance.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
574

40
35
Data points =15
Predicted Rutting (mm)
30 Se/Sy=0.47
R2adj= 0.83
25
20
15
10
0
0 5 10 15 20 25 30 35 40
Field Rutting (mm)
Figure 6. Comparison of field rutting versus predicted rutting.
7. Conclusions
Since the conclusion of the NCHRP 9-19 study, the challenge has been to indentify an
approach for using the Flow Number test for rutting prediction or evaluation of asphalt
mixtures. It is realized that an extensive, well planned, laboratory and field testing
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
program is needed to accomplish this; asphalt mixtures with a wide range of properties
need to be designed, tested for their Flow Number potential, and placed in the field for
long term monitoring of their rutting performance. In the absence of such an organized
and rather expensive endeavor, this study provided a modest approach to address this
issue, at least for the short term. A rutting prediction model was developed using the
Flow Number test input. A rational Flow Number predictive equation and the MEPDG
were important tools used in this process. Both allowed the generation of hundreds of
rutting and Flow Number values for different pavement structures, climatic locations,
and traffic levels. The Flow Number predictive equation has been published, used
extensively since its development, and proved to be a powerful assessment tool for
predicting flow numbers of various asphalt mixtures. The MEPDG rutting model is,
perhaps, one of the better models available to predict rutting of various asphalt mixtures.
575

RODEZNO, KALOUSH
The final rutting flow number model recommended had very good statistical
goodness of fit, R2adj = 0.86 and Se/Sy= 0.36. The model can provide guidance on the
anticipated rutting of an asphalt mixture with known Flow Number value, either from
laboratory testing or predicted using the asphalt mixture volumetric properties.
The model was validated using field rutting data obtained from the NCAT Test Track
2006 cycle. The final prediction also had very good statistical goodness of fit. The
availability of the NCAT laboratory Flow Number test results along with measured field
rutting values provided the best current venue for validating the model.
As additional future construction projects and tests sections become available, the
model can be recalibrated for a wide range of asphalt mixtures and field conditions.
8. References
El-Basyouny, M., Jeong, M., Effective Temperature for Permanent Deformation and
Fatigue Distress Analysis of Asphalt Mixtures, Transportation Research Record,
No. 2127, 2009, pp 155163.
El-Basyouny, M., Jeong, M., Development of Probabilistic Performance Related
Specifications Methodology Based upon AASHTO MEPDG, Transportation
Research Record, No. 2151, 2010, pp 93102.
Kaloush, K. E., and Witczak, M.W, Tertiary Flow Characteristics of Asphalt
Mixtures., Journal of the Association of Asphalt Pavement Technologists, Vol.71,
2002, pp. 248280.
Rodezno M.C, Kaloush, K., Corrigan, M., Development of a Flow Number Predictive
Model, Transportation Research Record, No. 2181, 2010, pp 7987.
Rodezno, M.C., Rutting Criteria for Asphalt Mixtures Based on Flow Number
Analysis, Ph.D. Dissertation, Department of Civil, Environmental and Sustainable
Engineering, Arizona State University, Tempe, Arizona, 2010.
Willis, R., Timm, D., West, R., Powell, B., Robbins, M., Taylor, A., Smit, A., Tran, N.,
Heitzman, M., Bianchini, A., Phase II NCAT Test Track Findings, NCAT Report
09-08, December 2009.
Witczak, M.W., Kaloush, K. E., Pellinen, T.K., El-Basyouny, M., and Von Quintus, H.
L., .NCHRP Report 465: Simple Performance Test for Superpave Mix Design,
Transportation Research Board, National Research Council, Washington, DC, 2002.
Witczak, M.W., NCHRP Report 547: Simple Performance Test: Summary of
Recommended Methods and Database, Transportation Research Board, National
Research Council, Washington, DC, 2005.
576
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Witczak, M. W., NCHRP Report 580: Specification Criteria for Simple Performance
Tests for Rutting, Transportation Research Board, National Research Council,
Washington, DC, 2007.
9. Discussion
DR. RAMON BONAQUIST: One of the reasons that people like the flow number is
that they feel that it identifies something in the mixture that is not stiffness based. If Im
following the model of rutting that youve developed, you essentially take flow numbers,
which people like, and develop a model on rutting resistance based on the mechanistic
empirical pavement design model, which is stiffness based, which is what engineers
dont want to base the rutting on. The model, to me, becomes circular; that you're
essentially taking what people consider a better test and developing a model using a
poorer methodology. Id like to hear your comments on that.
DR. MARIA RODEZNO: The only reason we use MEPDG to obtain rutting values was
that there was not really a lot of field data available. So this was kind of a shortcut to do
that because we knew we would have to wait, I dont know, ten more years to get all
these field data to really go and correlate flow number with this information. We have
seen that the MEPDG regarding rutting is probably fair. I mean the predictions of
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
MEPDG regarding rutting are probably based part of the MEPDG because fatigue is not
that good. So we thought it was like a fair approach to use that in the meantime until we
have more information to really go back and correlate flow number with field data.
DR. BONAQUIST: But wouldnt it just be simpler to use the modulus values and put
them into the design guide? Essentially what youve done is youve correlated the flow
number to a modulus and then created a model based on that. It just seems that were
negating the reasons that we like the flow number test.
DR. RODEZNO: We were just trying to find a way to really have some use for the flo w
number test. The initial study actually was also motivated by the fact that we never kno w
exactly what stress level for the flow number we should be doing or temperature we
should be doing, so its really a big debate up to this day on what are the real conditions
we should use. So the whole study was also trying to look into that. And that was the
reason that motivated the development of the flow number model. Probably that didnt
answer your question.
DR. BONAQUIST. Okay. Thank you.
DR. ALAEDDIN MOHSENI: I have a concern about flow number because in a three-
stage test where you have the primary stage of flow and then you have secondary and
then the tertiary. The primary state of loading data does not correlate with the field. Its a
combination of different things including viscoelastic effects and also consolidation,
577

RODEZNO, KALOUSH
which you can usually see in the later stage of compaction. So I dont think that portion
of the test would really correlate with rutting. The tertiary seems to be heavily dominated
by the edge effects, and there are fatigue cracks on the sample when it bulges before it
fails. Many people think that the flow number is also highly affected by the fatigue
properties even at the high temperature. I want to see your opinion about this and why
you think the flow number would be a good indication of the rutting in the field.
DR. RODEZNO: As I mentioned before, when we are doing all this, we are using the
flow number, and in order to have a flow number, you need to be in the tertiary zone. I
mean if you dont achieve that tertiary stage, actually you dont have a flow number.
DR. MOHSENI: Your flow number is right at the tertiary stage. So you have the primary
stage, which really does not relate it to field. Then you have the secondary and it fails
because of the boundary condition, because its a small sample. It does not necessarily
fail because of the flow properties.
DR. RODEZNO: Yes, but whatever we are doing here is that we are correlating rutting
to that flow number at the tertiary zone, so we are not really dealing with the primary
zone here. I mean I dont know if Im getting your
DR. MOHSENI: But the flow number would include the hundreds of cycles before you
reach the secondary, right?
DR. RODEZNO: Yes.
DR. MOHSENI: That would include consolidation that you would not normally see in
the field. Im talking about the first stage of the loading. The flow number would include
that.
DR. RODEZNO: I guess I dont follow what you're trying to ask.
DR. MOHSENI: All Im saying is that the test seems to be heavily dominated by the
edge effects and even the flow number I dont think it be a direct I think there are
some fatigue limits and probably sometimes lots of fatigue limits included in flow
number.
DR. RODEZNO: Thank you.
578
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Using Small Scale Specimens for AMPT
Dynamic Modulus and Fatigue Tests
Xinjun Lia and Nelson Gibson*b

a
ESC Inc.,Turner-Fairbank Highway Research Center, McLean, VA 22101
b
Federal Highway Administration, Turner-Fairbank Highway Research Center,
McLean, VA 22101
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
ABSTRACT: This study investigated the feasibility of performing dynamic modulus and fatigue
performance tests using reduced scale specimens in an Asphalt Mixture Performance Tester. Ten
different mixtures were characterized which had different nominal maximum aggregate sizes and
were laboratory prepared or field cored from accelerated pavement test sections. Indirect tension
dynamic modulus was included in the characterization of the field cored materials.
An aspect ratio for 38 mm diameter specimens was recommended based on an exploratory portion
of the study. The experimental results show that the modulus of the small scale specimens can be
quite similar to the full size specimens and tends to be slightly softer at high temperature and low
reduced frequencies. Phase angle is more comparable and, when different, tends to be slightly
higher for the small scale specimens. Data analysis comparing the full size and small scale
specimens revealed two out of three data quality indicators from small size dynamic modulus are
as good as that of full size dynamic modulus. The third quality indicator was worse for about one
third of the data points, but the majority of tests satisfied recommended values.
The fatigue test results showed the modulus reduction at failure and endurance limit are
comparable between full size and small scale specimens. There was no consistent trend where
small scale was larger or smaller than full scale specimens in fatigue resistance and the ranking
was mostly preserved between the two sized specimens. Overall, the small scale approach is very
promising which can allow the field-compacted fatigue and stiffness characteristics of pavements
to be assessed. The applications of this research are field validation of cracking tests, performance
based quality assurance and forensic investigations.
KEYWORDS: performance test, AMPT, dynamic modulus, phase angle, fatigue, size effects
The oral presentation was made by Dr. Gibson.
579

LI, GIBSON
1. Introduction
1.1 Background
NCHRP Project 9-19, Superpave Support and Performance Models Management,

recommended dynamic modulus as one of the simple performance tests to be
incorporated in the Superpave volumetric mixture design process (Witczak et al., 2002).
Mechanisticempirical pavement design uses dynamic modulus as the key parameter in
flexible pavement design that controls permanent deformation and fatigue cracking. A
standard dynamic modulus test protocol has been developed and provisionally adopted
by state highway agencies and industry (AASHTO, 2009). This test protocol calls for
the use of axial compression on 100 mm diameter by 150 mm tall specimens that are
cored from gyratory compacted mixtures.
The simple performance tester (SPT), now called the Asphalt Mixture Performance
Tester (AMPT) was developed as a result of the NCHRP Project 9-29, Simple
Performance Tester for Superpave Mix Design (Bonaquist and Christensen, 2008). The
AMPT is being implemented by highway agencies to characterize asphalt mixtures
through a national pooled fund project (FHWA, 2012). In addition to the three
performance tests recommended by NCHRP 9-19, the newest AMPT has the capability
to conduct cyclic direct tensile fatigue testing that was recently proposed by Underwood
et al. (2010). Similar to the dynamic modulus test, this new fatigue test requires usage of
100 mm by 150 mm cylindrical specimens.
One of the issues related to performance tests in the AMPT is the specimen size
when it comes to the characterization of as-built pavements. Generally speaking, most
construction lifts are only few inches thick and this limitation makes it very difficult, if
not impossible, to prepare the standard AMPT performance test specimens. To address
this problem, Kim et al. (2004) developed an analytical solution for the dynamic
modulus of asphalt mixtures tested in the indirect tension (IDT) mode using the theory
of linear viscoelasticity. Twelve asphalt mixtures commonly used in North Carolina
were tested for dynamic modulus with both axial compression and IDT test methods and
the test results showed that the two methods were in good agreement. Since the thickness
of the IDT test specimens could be as thin as 38 mm, this makes this test very promising
as an alternative test for measuring the dynamic modulus of field compacted mixtures. It
should be noted, however, that the IDT test employs four displacement transducers, two
(one vertical and one horizontal) on each surface of the specimen to measure the vertical
and horizontal displacements. Test specimens with smooth and parallel surfaces are
required to ensure that a quality measurement can be achieved in this test setup.
Otherwise, variation could result. In addition, an extra test fixture is required for the IDT
test to properly apply the load. Most importantly, a viable fatigue characterization
580
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Using Small Scale Specimens for AMPT Dynamic Modulus and Fatigue Tests
methodology in the IDT mode has yet to be realized. Therefore, it is of interest to

investigate the feasibility of performing performance tests in the AMPT using small
scale specimens with a cylindrical geometry. This will be ultimately helpful in
developing a protocol for field validation of cracking tests and implementing
performance tests toward performance-based construction quality acceptance.
In the present study, performance tests were carried out on small scale specimens in
the AMPT for ten asphalt mixtures with various nominal maximum aggregate sizes and
including both laboratory-prepared and field-cored specimens. The objective of this
study was to evaluate AMPT specimen size and geometry effects on dynamic modulus
and fatigue in order to implement the option of small sample testing to evaluate
engineering properties of constructed pavements.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
1.2 Performance Test Specimen Size
There is a general aggregate-based pavement material testing requirement that the

specimen size and boundary conditions ensure a representative volume. The minimum
specimen dimension of the asphalt mixture is usually required to be four to six times the
nominal maximum aggregate size for laboratory prepared specimens and this general
size requirement has been reduced for the specimens cut from oversized laboratory
samples or from field compacted materials. For example, in a Strategic Highway
Research Program (SHRP) study, 3 in. diameter specimens were cored from original
field cores 6 in. diameter resulting in a specimen diameter to maximum aggregate size
ratio as low as two (Lytton et al., 1993). In addition to the specimen diameter, the height
to diameter or aspect ratio of the test specimen is usually required to be greater than two
to minimize boundary (end) effects. However, aspect ratios smaller than two or near
unity are sometimes used by researchers. In a more recent study, test specimens with a
102 mm diameter and 102 mm height were tested in a repeated direct tension test to
characterize the fatigue cracking of asphalt mixtures (Luo et al., 2013). In addition to
uniaxial loading configurations, small asphalt mixture specimens are also used for
performance testing by researchers in other loading configurations. The newly developed
mixture bending beam rheometer (BBR) methodology utilizes 6.35 mm thick by 12.5
mm wide by 127 mm long specimens (Zofka et al., 2005). A Semi-Circular Bending
(SCB) test protocol characterizes specimens 25 mm thick by 150 mm wide and 75 mm
tall to determine the fracture energy at low temperatures (Li and Marasteanu, 2010).
Although various specimen aspect ratios and aggregate sizes have been investigated,
a literature review regarding the effect of specimen size on the properties measured in
contemporary performance tests, especially the dynamic modulus and axial fatigue test
using the AMPT, produced limited results. One of the most key studies on the effects of
aggregate size and specimen geometry on asphalt mixture properties was finished as a
581

LI, GIBSON
part of NCHRP Project 9-19 (Witczak et al., 2002). Asphalt mixtures with three nominal
maximum aggregate sizes (12.5, 19 and 37.5 mm) and four aspect ratios (1.0, 1.5, 2.0
and 3.0) were selected for test specimens with three diameters (70, 102 and 150). The
experimental design included 450 mm tall specimens that were fabricated by stacking
specimens together. These specimens were tested for dynamic modulus at only two
conditions, 16 Hz at 4C and 0.1 Hz at 40C. A significant diameter effect was found for
the uniaxial compression dynamic modulus and a minimum specimen diameter to
maximum aggregate size ratio of 3.0 and a minimum aspect ratio of 1.5 was also
recommended from this study. It is very important to note that this study was finished
more than a decade ago when the test and sample preparation equipment were different
from today. In other words, this was before the AMPT equipment specification and
AASHTO protocols PP 60, Preparation of Cylindrical Performance Test Specimens
Using the Superpave Gyratory Compactor (SGC), and TP 79, Determining the Dynamic
Modulus and Flow Number for Hot Mix Asphalt (HMA) Using the Asphalt Mixture
Performance Tester (AMPT). A more recent study that used smaller size specimens to
characterize the asphalt mixtures was in a uniaxial push-pull fatigue test (Kutay et al.,
2009), but in a universal test machine rather than an AMPT with standardized fixtures
and extensometers.
2. Materials and Specimen Preparation
A total of ten mixtures with different nominal maximum aggregate sizes (NMAS) were
investigated in this study. Most were gyratory-compacted and some were field cored
materials. The details of the materials are summarized in Table 1. All mixes are dense-
graded except for the Thin Hot Mix Asphalt Concrete Overlay (THMACO), which is a
high binder content, high film thickness, gap-graded mix used in thin surface layers; also
known as Ultra-Thin Bonded Wearing Course. The THMACO mix was selected to
expand on the variety of volumetrics used in the study. For each mix, test specimens
with two aspect ratios were prepared. The gyratory-compacted specimens were cored
from gyratory plugs that were compacted to 150 mm in diameter and 185 mm tall. Full
size AMPT specimens that are 100 mm in diameter by 150 mm high and small scale
specimens that are 38 mm in diameter were then vertically cored from the gyratory
sample. One gyratory sample can yield one full size AMPT specimen or up to six small
scale specimens with one from the center and five from around the center as shown in
Figure 1. The small scale specimen from the center is always of higher density and
usually does not meet the air void criteria of the other five. It was decided to core five
instead of six small scale specimens out of the gyratory sample without the central
specimen and move the coring direction closer to the center. This way, the small scale
specimens had more consistent air void contents since the coring is a little further from
the edge where the material is compacted less.
582
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 1. Summary of asphalt mixtures.

Mixture Material NMAS Binder Target Air
Source Test
ID Type (mm) PG Voids (%)
Plant Modulus +
THMACO Virginia 4 .7 5 76-22 9 .0
mix Fatigue
Plant Modulus +
N7 1 NCAT 9 .5 82-22 7 .0
mix Fatigue
Plant Modulus +
N7 2 NCAT 1 9 .0 82-22 7 .0
mix Fatigue
Plant
NY9 5 Ne w Yo r k 9 .5 64-28 3 .5 Modulus
mix
Plant
NY1 9 0 Ne w Yo r k 1 9 .0 64-22 6 .5 Modulus
mix
Lab Modulus +
AZ0 Arizona 1 9 .0 58-34 6 .5
prepared Fatigue
Lab Modulus +
CO0 Colorado 1 2 .5 58-28 6 .5
prepared Fatigue
Modulus (IDT +
ALF Field Aged
L8 1 2 .5 ---- Small & Full
Lane 8 co r es 70-22
Scale)
Modulus (IDT +
ALF Field Aged
L9 1 2 .5 ---- Small & Full
Scale)
Modulus (IDT +
ALF Field Aged
L10 1 2 .5 ---- Small & Full
Scale)
( a) (b)
Figure 1. Test specimen preparation from (a) Gyratory sample; (b) Field core.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
583

LI, GIBSON
The three field mixes investigated in this study are from the FHWA Accelerated
Load Facility (ALF). The details of the project can be found elsewhere (Gibson et al.,
2012). The pavements were built in 2002 and the cores for this study were taken in 2012;
thus at an age of approximately ten years. Lane 8, Lane 9 and Lane 10 were cored and
tested in the present study. These three lanes were designed to have the same structure,
aggregate and gradation where the only difference is the asphalt binder. Lane 8 is a
control unmodified asphalt mix, Lane 9 is an SBS modified asphalt mix and Lane 10 is
an air blown unmodified asphalt mix. The total asphalt pavement thickness is nominally
150 mm and was constructed in two layers each about 70 to 75 mm thick. Therefore,
each specimen from the ALF can be vertically cored to make one full size test specimen
or horizontally cored to form two small scale specimens as shown in Figure 1. All cores
were taken from a location away from the disturbance of the previous ALF wheel loads
so as not to include induced damage as a confounding factor.
At the beginning of the study, the small scale specimen geometry investigated was 38
mm in diameter and 140 mm in height. The 38 mm diameter was chosen to make
samples from 1.5 to 2 in. thick pavement layers. The 140 mm height would be the result
of slight end trimming from horizontally coring across a 150 mm wide field core. The
result is an aspect ratio of 3.7 (140/38) whereas the full size AMPT specimen is the
standard 1.5 aspect ratio (150/100). The 140 mm height was abandoned after two
mixtures were prepared and tested with this size. The minimum specimen height of an
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
AMPT was determined to be 110 mm; 70 mm extensometer gauge length plus about 20
mm space away from the extensometer mounting studs to the specimen edge. The result
is an aspect ratio of 2.9. As a matter of practicality, the acceptable range in air void was
set at 0.7% for the small scale specimens while full size AMPT specimens are
customarily taken at 0.5%.
For comparison purposes, IDT specimens were also prepared from ALF cores and
tested for dynamic modulus. Similar to the small scale specimens, the IDT specimens
were horizontally cut to make two specimens; 45 mm thick from the top and bottom
layer. Therefore, the small scale axial and IDT specimens are divided into two groups,
the top layer specimens and the bottom layer specimens.
3. Testing Setup and Procedure
The uniaxial dynamic modulus and direct tension cyclic fatigue tests for both full size
and small scale specimens were performed in an AMPT, while the IDT dynamic
modulus tests were conducted in a universal loading machine. The AMPT small scale
fatigue test setup is shown in Figure 2; platen adaptors were required to be machined so
as to ensure centering of the specimen and adequate load transfer as well as necessary
modifications for the extensometer stud gluing jig. No end lubrication between the small
scale specimens and the metal platens was used, such as Teflon sheets in the full size
584

AMPT dynamic modulus protocol, given the slender aspect ratio that decreases end
effects.
Figure 2. AMPT fatigue test setup for small scale specimen.
The uniaxial dynamic modulus was measured at 4.4, 21.1, 37.8 and 54.4C, and
frequencies of 25, 10, 5, 1, 0.5 and 0.1 Hz on three replicates following the AASHTO
TP 79. IDT dynamic modulus was not performed at 54.4C, but one additional
frequency of 0.01 Hz was added at 37.8C.
Cyclic, axial fatigue testing was developed as a practical alternative to flexural beam
fatigue since this axial test can be compatible with the advent of the AMPT. This study
followed the Simplified Viscoelastic Continuum Damage (S-VECD) direct tension
cyclic fatigue test protocol developed at the North Carolina State University (Hou et al.,
2010) which is actuator-controlled in a cyclic pull-pull manner. This study used one
test temperature (21C) and two actuator displacements that ideally target failure in
approximately 1,000 cycles and 10,000 cycles, which does not occur precisely. The
frequency of the cyclic test was 10 Hz.
4. Experimental Results and Data Analysis
4.1 Dynamic Modulus
Three replicates were tested on both full size and small size specimens for dynamic
modulus and phase angle. The average dynamic modulus data were then used to build
master curves by utilizing a sigmoidal model (Pellinen and Witczak, 2002). A
585
--`,`,,,``,,```,,`,,,

LI, GIBSON
comparison of the average dynamic modulus master curves and average phase angle
values will be discussed for each mix. Test quality statistics will be discussed in a
following section.
4.1.1 Thin Hot Mix Asphalt Concrete Overlay (THMACO)
The dynamic modulus and phase angle data for the full size and two heights of the 38
mm diameter specimens are plotted in Figure 3 and Figure 4. The upper and lower lines
in the plot above and below the line of equality are 80% and 120% of unity for the full
size dynamic modulus. The experimental results show that the dynamic modulus from
both small sizes is smaller than that of the full size. A more careful observation finds that
the difference between the 110 mm tall and full size is less than 20% whereas that
between the 140 mm tall and the full size is more than 20%. This could be due to the
difference of the aspect ratio and suggests stiffness decreases when the test specimen
aspect ratio increases. The difference in the phase angle values between the full size and
small size specimens is found to be a few degrees and the biggest difference occurs for
the highest test temperature.
Small Scale Dynamic Modulus,MPa
3.E+03
3.E+02
38X110
38X140
3.E+01
3.E+01 3.E+02 3.E+03
Full Size Dynamic Modulus, MPa
Figure 3. Dynamic modulus data for THMACO.
586
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

45
40
Small Scale Phase Angle, degree
35
30
25
20
38X110
15
38X140
10
10 15 20 25 30 35 40 45
Full Size Phase Angle, degree
Figure 4. Phase angle data for THMACO.
The dynamic modulus master curves are shown in Figure 5. It shows that the master
curve from the 38 mm x 110 mm small size is closer to that of the full size and is softer
in the low frequency/high temperature region. Again, it was found the 140 mm
specimens result in a lower stiffness master curve than the 110 mm specimens.
1.0.E+05
1.0.E+04
|E*|, MPa
1.0.E+03
1.0.E+02
Full Size
38X140
38X110
1.0.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 5. Dynamic modulus master curves for THMACO.
587

LI, GIBSON
4.1.2 N71 Mix
The dynamic modulus and phase angle data for this 9.5 mm NMAS mix are shown in
Figure 6 and Figure 7, respectively. It can be found that the small scale dynamic
modulus from the 110 mm tall specimens is closer to the full size than the 140 mm
specimens, which produced lower dynamic modulus as with THMACO. For both size
specimens, the phase angle values are very close to that from the full size specimens.
The dynamic modulus master curves are plotted in Figure 8. It shows that the three
master curves are similar. However, a more careful observation indicates that the 110
mm tall specimen master curve is closer to the full size and therefore the 38 mm by 140
mm size was not continued.
Small Scale Dynamic
3.E+03
Modulus,MPa
3.E+02
38X110
38X140
3.E+01
3.E+01 3.E+02 3.E+03
Figure 6. Dynamic modulus for N71.
45
40
Small Scale Phase Angle,
35
30
degree
25
20
15
10
38X110
5
38X140
0
0 5 10 15 20 25 30 35 40 45
Figure 7. Phase angle for N71.
588
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

|E*|, MPa 1.E+04
1.E+03
Full Size
38X140
38X110
1.E+02
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 8. Dynamic modulus master curves for N71.
4.3.3 N72 Mix

The dynamic modulus and phase angle data for the 19 mm NMAS N72 mix are shown
in Figure 9 and Figure 10 while the modulus master curve is plotted in Figure 11. It was
found that the moduli and phase angles from the small scale specimens are very close to
those from the full size and therefore the corresponding modulus master curves are very
close as well.
3.E+03
3.E+02
3.E+01
3.E+01 3.E+02 3.E+03
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 9. Dynamic modulus for N72.
589

LI, GIBSON
45
40
35
30
25
20
15
10
5
0
0 5 10 15 20 25 30 35 40 45
Figure 10. Phase angle for N72.
1.E+04
|E*|, MPa
1.E+03
Full Size
38X110
1.E+02
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 11. Dynamic modulus master curves for N72.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
4.1.4 AZ Mix
The dynamic modulus and phase angle data for this 19 mm NMAS mix are shown in
Figure 12 and Figure 13 and the dynamic modulus master curves are presented in Figure
14. It can be found that the data from both the small scale and full size specimens are
590

fairly close; within 20%. The dynamic modulus master curve plot indicates that there is
obvious difference between these two dimensions for the highest test temperature/lowest
frequency. Small Scale Dynamic Modulus,MPa
3.E+03
3.E+02
3.E+01
3.E+01 3.E+02 3.E+03
Figure 12. Dynamic modulus for AZ mix.
45
40
35
30
25
20
15
10
10 15 20 25 30 35 40 45
Figure 13. Phase angle for AZ mix.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
591

LI, GIBSON
1.E+05
1.E+04
|E*|, MPa
1.E+03
1.E+02
Full Size
38X110
1.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 14. Dynamic modulus master curves for AZ mix.
4.1.5 CO Mix
The dynamic modulus and phase angle data for the 12.5 mm NMAS CO mix are
presented in Figure 15 and Figure 16 and the dynamic modulus master curve is shown in
Figure 17. It can be found that the dynamic modulus data from both dimension
specimens are similar. However, the phase angle values at the highest temperature fro m
the small scale specimens are obviously higher than those from the full size specimens.
The dynamic modulus master curves match each other well except for the low frequency
(high temperature) area.
592


3.E+03
3.E+02
3.E+01
3.E+01 3.E+02 3.E+03
Figure 15. Dynamic modulus data for the CO mix.
45
40
35
30
25
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
20
15
10
10 15 20 25 30 35 40 45
Figure 16. Phase angle data for the CO mix.
593

LI, GIBSON
1.E+04
1.E+03
|E*|, MPa
1.E+02
Full Size
38X110
1.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 17. Dynamic modulus master curves for the CO mix.
4.1.6 NY95 Mix

The dynamic modulus and phase angle data for the 9.5 mm NMAS NY95 mix are
plotted in Figure 18 and Figure 19 and the dynamic modulus master curve is presented in
Figure 20. The experimental results show that both the dynamic modulus and phase
angle are very close for the specimens with two different dimensions. This can also be
seen from the modulus master curve.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
3.E+03
3.E+02
3.E+01
3.E+01 3.E+02 3.E+03
Figure 18. Dynamic modulus data for NY95 mix.
594

45

40
35
30
25
20
15
10
5
0
0 5 10 15 20 25 30 35 40 45
Figure 19. Phase angle data for NY95 mix.
7.E+03
|E*|, MPa
7.E+02
Full
Size
7.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 20. Dynamic modulus master curves for NY95 mix.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
4.1.7 NY190 Mix
The dynamic modulus and phase angle data for the 19 mm NMAS NY190 mix are
shown in Figure 21 and Figure 22 and the modulus master curve is presented in Figure
23. It was found that only one dynamic modulus data point from the small scale
specimens is more than 20% below and several points above 20% of full size specimens.
The dynamic modulus master curves match each other very well.
595

LI, GIBSON
3.E+03
3.E+02
3.E+01
3.E+01 3.E+02 3.E+03
Figure 21. Dynamic modulus data for NY190 mix.
45
40
35
30
25
20
15
10
5
0
0 5 10 15 20 25 30 35 40 45
Figure 22. Phase angle data for NY190 mix.

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
596

|E*|, MPa 6.E+03
6.E+02
Full Size
38X110
6.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 23. Dynamic modulus master curves for NY190 mix.
4.1.8 ALF Lanes
Three different geometries were characterized for dynamic modulus from the three ALF
lanes. Those labeled full size have the standard 100 mm diameter, but they are not the
standard 150 mm height because the nominal thickness of the pavement was 150 mm.
However, the ends of the core had to be trimmed smooth and thus the specimen
geometry was 100 mm diameter and approximately 140 mm tall. Small scale 38 mm by
110 mm specimens were cored horizontally from each lift. IDT specimens were 150
mm in diameter and nominally 45 mm thick. The full size and small size specimens were
tested on the AMPT while the IDT tests were conducted on a universal test system. Two
groups of the small size and IDT specimens were obtained and tested, one from the top 3
in. lift and the other one from bottom 3 in. lift. Three replicates were tested for each
sample size. Figure 24 presents the air void contents for all test specimens along with
one standard deviation plotted as the error bars. The air void data show that the top layer
in each lane had lower air voids than the bottom layer. It was also found that Lane 8 had
the lowest density (highest air voids) and Lane 9 had the highest density (lowest air
voids).
597
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

LI, GIBSON
6. 0 Lane 8
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Lane 9
5. 0 Lane 10
4. 0
Air Void (%)
3. 0
2. 0
1. 0
0. 0
Full Small - T Small - B IDT - T IDT - B
Figure 24. Air void content for the ALF dynamic modulus test specimens.
It is difficult to compare among the geometries since the small size and IDT
specimens are cored from either the top or bottom lift while the semi-full size specimen
contains material which spans the lift boundary and the two different densities. A direct
comparison of modulus master curves between the semi-full size and small size
specimens is plotted in Figure 25 while Figure 26 compares the data from the AMPT full
size specimens and the IDT specimens for Lane 8. Data from the top layer (T) and
bottom layer (B) are plotted as well as averaged data from top and bottom. In both plots,
it can be found that the top layer has higher stiffness than the bottom layer; this could be
attributed to the aging effect which has overcome the lower density effect. It can be seen
that the full size data has lower stiffness than that from the top layer and is relatively
close to that from the bottom layer data; this is true for all frequencies from the small
scale axial compression test and most of the frequency domain from the IDT test
method. The two-layer-averaged data are always between the two layers for both test
methods.
598

|E*|, MPa 7.E+03
7.E+02
AMPT Full Semi
38X110 T
38X110 B
38X110 (T+B)
7.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 25. AMPT dynamic modulus master curves for Lane 8.
7.E+03
|E*|, MPa
7.E+02
IDT T
IDT B
IDT (B+T)
AMPT Full
Semi
7.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 26. Modulus master curves for full size and IDT specimens for Lane 8.
The same types of comparison between the full and small size and full size and IDT
specimens for Lane 9 and Lane 10 are plotted in Figure 27 to Figure 30, respectively. It
was found that for these two lanes the modulus from semi-full size is always between the
top and bottom layers with the top layer stiffer than the bottom layer for the whole
frequency domain from the small scale axial compression test method and the high
frequency area from the IDT test method
599
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

LI, GIBSON
2.E+04
2.E+03
|E*|, MPa
2.E+02 Semi
AMPT Full
38X110 T
38X110 B
38X110 (T+B)
2.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
2.E+04
2.E+03
|E*|, MPa
2.E+02
IDT T
IDT B
IDT (B+T)
AMPT Full
2.E+01 Semi
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,
600

6.0.E+03
|E*|, MPa
6.0.E+02
Semi
AMPT Full
38X110 T
38X110 B
38X110 (T+B)
6.0.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
6.E+03
|E*|, MPa
6.E+02
IDT T
IDT B
IDT (B+T)
AMPT Full
Semi
6.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
A comparison between the small size and the IDT specimens is plotted in Figure 31
to Figure 33 for the three lanes, allowing a direct comparison between different
geometries based solely on the top lift or the bottom lift. A significant difference was
found between these two geometries for Lane 8 and Lane 9, whereas very consistent
results were obtained by both specimen configurations for Lane 10. A careful check with
601
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

LI, GIBSON
the coefficient of variation (COV) for the dynamic modulus and phase angle does not
support a consistent conclusion regarding one lane being more variable than another, but
it verifies that the variation from the small scale can be equal to or smaller than that from
the IDT configuration. It is noteworthy that a few IDT specimens were tested twice at
the highest temperature to achieve the target peak to peak strains and this may impact the
final resultant dynamic modulus and comparisons.
7.E+03
|E*|, MPa
7.E+02
IDT T
IDT B
38X110 T
38X110 B
7.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
Figure 31. Dynamic modulus comparison for Lane 8.
4.E+03
|E*|, MPa
4.E+02
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
IDT T
IDT B
38X110 T
38X110 B
4.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
602

7.E+03
|E*|, MPa
7.E+02
IDT T
IDT B
38X110 T
38X110 B
7.E+01
1.E-05 1.E-03 1.E-01 1.E+01 1.E+03
Frequency, Hz
4.1.9 |E*| Statistical Analysis
A statistical analysis was performed to evaluate the significance of specimen

dimension on dynamic modulus measurements. Dynamic modulus data from both full
size and small scale specimens were analyzed using the statistical hypothesis t-test. The
null hypothesis is that there is no difference in dynamic modulus from both specimen
dimensions. The t-test employed is a small-sample test considering both equal and
unequal variance. The results showed no difference for the equal and unequal variance
assumptions and therefore only the results with equal variance are presented. The level
of confidence used for all analyses was 95%. If a significant difference was found for a
data pair, it was further analyzed to test for an alternative hypothesis, whether the small
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
scale modulus was statistically larger or smaller.
The dynamic modulus data were also compared to the modulus obtained from the
smallest load cell force calibration point, which is 0.1 kN. A 100 strain level was used
to calculate the minimum dynamic modulus under the smallest load level of 0.1 kN. The
minimum dynamic modulus values are 882 MPa and 127 MPa for the two specimen
dimensions of 38 mm and 100 mm in diameter, respectively. Therefore, different types
of scenarios could occur; no significant difference between geometry moduli, full size
significantly larger, small scale significantly larger, full size modulus below smallest
calibration point and small scale modulus below smallest calibration point, are analyzed
in this work. The analysis results are shown in Table 2. The legend below the table
explains the mixtures laid out in each sub cluster of frequency-temperature cell as well
as the symbolic meaning describing each scenario. It should be noted that due to the
603

LI, GIBSON
layer thickness and aging gradient the three ALF field lanes cannot be statistically
analyzed.
Table 2. Statistical analysis results.
The statistical analysis shows that significantly different dynamic moduli do exist
between the two geometries, but this depends on the mixture, the test temperature and
the frequency. The heavily polymer modified mix N71was found to have significantly
different moduli for all frequencies under the two lowest test temperatures. A more
careful check on the experimental data found the dynamic modulus data for this mix
from the three test replicate specimens are very consistent and the COV for both
geometries under the two lower temperatures are less than 4%. This indicates that
relatively small differences in average moduli could produce a statistically different
comparison. It could be the difference of the aspect ratio of the two specimen geometries
604
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

(1.5 and 2.89 for the full size and small scale specimen, respectively) that leads to the
difference of modulus. For the other six mixtures, only a few data sets were found to be
statistically different and most of the statistically different data sets were from the two
higher temperatures.
The analysis also illustrated that very few full size data points were below the
smallest calibration point and these data points are all at low frequencies under the
highest temperature. The small scale specimens, however, were found to have a dynamic
modulus lower than the smallest calibration point for all frequencies at 54C and most
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
frequencies at 38C.
Small scale modulus data from peak dynamic forces below the smallest load cell
calibration point does not necessarily mean the data are unreliable given that statistically
equal moduli were found for scenarios where the small scale modulus was below the
smallest force calibration point while the large scale modulus was above the smallest
force calibration point. Additional justification for this was found during a routine load
cell calibration when, under static conditions, a calibration load cell and the AMPT load
cell were in agreement as low as 0.05 kN (11 lb).
The strain magnitudes controlled by the AMPT were analyzed as another evaluation
of consistency between the two geometries. The average and standard deviation of the
strains are shown in Table 3. The general trend for both full size and small size
specimens was for the average strain to decrease and the variation (standard deviation) to
increase with an increase in temperature. The small scale specimens had lower
controlled strains and higher variation than full scale specimens, but the variation itself
diminishes the relatively small differences.
Table 3. AMPT dynamic micro strain controlled by AMPT.
Full Size Small Scale

Temperature (C)
Mean Std. Dev. Mean Std. Dev.
4 98 3 96 7
21 97 3 96 6
37 93 10 87 21
54 83 16 72 29
4.1.10 |E*| Data Variation and Quality
The variation of dynamic modulus and phase angle was also analyzed to investigate the
repeatability and controllability of the small scale specimen configuration. The
coefficient of variation was calculated for both dynamic modulus and phase angle data
605

LI, GIBSON
for all tested mixtures except the three ALF field mixes. These three mixtures were
excluded because of the variation between the layers. Figure 34 and Figure 35 plot the
Coefficient of Variation (COV) data from the small scale specimens against the values
from the full size specimens. Two different sets of data are shown for the mixes that
have nominal maximum aggregate sizes less than 19 mm and three mixtures that are 19
mm NMAS. The results are fairly scattered and show that the mixes with NMAS less
than 19 mm have a smaller COV for both sizes, but those that are 19 mm NMAS have
higher COV for the small scale.
Less than 19mm NMAS 19mm NMAS

25
Small Scale Dynamic Modulus Data COV (%)
20
15
10
0
0 5 10 15 20 25
Full Size Dynamic Modulus Data COV (%)
Figure 34. COV for the dynamic modulus data.

15
Small Scale Phnase Angle Data COV (%)
10
0
0 5 10 15
Full Size Phase Angle Data COV (%)
Figure 35. COV for the phase angle data.
606
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

The dynamic modulus test data quality was also investigated in this study. Three data
quality indicators, standard error of the stress, average standard error for the strain
measurements and uniformity coefficient for the strain measurements were calculated
and plotted in Figure 36 to Figure 38. Also plotted in these figures are the data quality
requirements in AASHTO TP 79, which are less than 10%, 10% and 30% for these three
indicators, respectively. The standard error of stress data show that some small scale
specimens are above the recommended limit and nearly none of the full size, but most of
the small scale data points are within the range. Those that were outside of the range
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
were a combination of 19 mm NMAS and smaller than 19 mm NMAS. The average

standard error of the strain data show that both the small scale and the full size
specimens have points outside the recommended criteria but no obvious difference was
found between these two dimensions. As for the uniformity coefficient for strain
measurements, there are a number of data points from the small scale specimens that are
out of the required range compared to the full size specimen with a slightly greater
concentration of 19 mm NMAS mixes. This should be further investigated. Further,
inside the recommended range of COV it is clear that 19 mm mixes have higher
variation. Overall, the qualities of the data seem reasonable and do not point to an
invalid approach.

20
Small Scale Average Standard Error of
18
16
14
12
Stress
10
8
6
4
2
0
0 5 10 15 20
Full Size Average Standard Error of Stress
Figure 36. Standard error of stress.
607

LI, GIBSON

20
Small Scale Average Standard Error of
18
16
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
14
12
Strain
10
8
6
4
2
0
0 5 10 15 20
Full Size Average Standard Error of Strain
Figure 37. Average standard error of strain.
less than 19mm NMAS 19mm NMAS

40
Small Scale Uniformity Coefficient of Strain
35
30
25
20
15
10
0
0 10 20 30 40
Full Size Uniformity Coefficient of Strain
Figure 38. Uniformity coefficient for the strain measurements.
608

4.2 Cyclic Direct Tension Fatigue Test
Four of the ten mixtures characterized for modulus were further characterized for
fatigue. As stated in some previous publications (Li et al., 2011; Li et al., 2012), the
actual strains in this cyclic direct tensile fatigue test grow and the measured stress
decreases until complete failure. This is neither a strain-control nor a stress-control
fatigue test, but the test protocol is used because it is a stable test for AMPT instruments
to keep control as the specimen loses structural integrity to failure during the test. A
typical reduction in dynamic modulus due to fatigue damage is observed in Figure 39. It
is necessary to account for the test control characteristics using the viscoelastic
continuum damage (VECD) method. A damage characteristic curve (C vs. S) is
developed, which is independent of loading mode and frequency and temperature.
Figure 40 shows an example damage characteristic curve that represents the loss in
modulus or pseudostiffness (C) as a function of the damage parameter (S) that represents
the amount of micro cracking damage.
4.5
4.0
3.5
Modulus, GPa
3.0
2.5
2.0
1.5
1.0
0.5
0.0
0 2000 4000 6000
Cycles
Figure 39. Modulus reduction example during the fatigue test.
609
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

LI, GIBSON
1. 0
Full Size
38x110
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0. 8
Pseudo Stiffness C(S)
0. 6
0. 4
0. 2
0. 0
0 50,000 100,000 150,000 200,000 250,000 300,000
Damage Parameter, S
Figure 40. Example damage characteristic curve.
Once the damage characteristic curve is obtained, the VECD-based fatigue life of the
mixture can be determined by simulating a pure strain control fatigue test with inputs
such as test temperature, frequency and strain levels. A strain range from 50 to 700
was selected for the simulations. The test specimens exhibited a localized macro crack
and thus failure is interpreted from the peak in the phase angle at that point. Each
mixture exhibited a different pseudo stiffness at failure and the values are listed in Table
4. These different values for each mix were taken as the definition of failure for the
simulations.
Figure 41 shows the predicted fatigue life at 200 and 21C. The rank of the five
mixtures for both the full and small size specimens is as follows: full size (THMACO >
N71 > AZ > N72) and small size (THMACO > N71 > N72 > AZ). Three out of the four
mixes are ranked the same with the two specimen sizes. The only exception is AZ which
was predicted to have relatively higher fatigue resistance by the full size specimens but a
relatively very low fatigue life from the small specimen prediction. This could be
attributable to the chosen temperature of the test, 21oC. Recent recommendations to
refine the S-VECD methodology are to set the test temperature based on the PG grade of
the binder. Thus this particular mixture should have been run at 9oC because of the soft
PG 58-34 binder. The fatigue life of the full size specimens was longer than the small
scale specimens for two of the four mixes and vice versa. Although differences exist
between the full size and small scaled specimen configuration, the values for fatigue life
are quite reasonable and comparable.
610

Table 4. Pseudo stiffness and endurance limit at failure.

THMACO N7 1 N7 2 AZ
Full Small Full Small Full Small Full Small
Failure
0 .0 8 7 0 .0 9 3 0 .1 4 9 0 .1 4 3 0 .1 3 6 0 .1 1 6 0 .0 8 4 0 .1 2 5
Pseudo Stiffness
Endurance
459 494 302 384 322 433 371 278
Limit ()
Failure Cycles
(x1M) 24,000 22,800 5,730 9 ,1 7 0 3 ,2 1 0 6 ,2 5 0 4 ,6 2 0 512
21oC 200
1.E+11
Full Size
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Small Size
1.E+10
Fatigue Life, Cycles
1.E+09
1.E+08
THMACO N71 N72 AZ
Mixtures
Figure 41. Predicted fatigue life for 200 at 21C.
This study investigates the possibility of conducting performance tests using reduced
scale specimens in AMPT equipment. This was motivated by the need to be able to
assess the plant production, field-compacted fatigue and stiffness characteristics of
pavements beyond what can be produced in the laboratory. A preliminary study
evaluated two aspect ratios of small scale specimens and indicated that a 110 mm tall x
38 mm diameter specimen was more optimal than a more slender 140 mm tall x 38 mm
diameter specimen. Ten mixtures with NMAS aggregate sizes from 4.75 mm to 19 mm
were characterized. Seven mixes were laboratory prepared and three were field cored
611

LI, GIBSON
specimens. The dynamic modulus for three field mixes was also evaluated using the IDT
method. The dynamic moduli and phase angles from the full size, small scale and IDT
specimens were analyzed by comparing their master curves and crossplotting their
values within a +/-20% interval. The experimental results show that the modulus using
small scale specimens is consistently softer in the high temperature/low frequency
regions and occasionally stiffer than the full scale specimens. Phase angle of the small
scale specimens tends to be slightly higher than the full scale, but more comparable than
modulus.
Full scale cores taken from ALF pavements could not avoid sampling two
construction lifts with different air void contents and aging conditions and the effective
dynamic modulus measured provides an incomplete picture. IDT and small scale AMPT
specimens could accommodate the different construction lifts. Two out of three sections
showed the IDT and small scale AMPT dynamic moduli were not comparable where the
small scale specimens exhibited softer modulus especially in the high temperature / low
frequency region. The IDT and small scale geometries from the third section produced
very comparable dynamic moduli.
Two out of three data quality indicators, stress and strain standard error, from the
small scale dynamic moduli appear as good as those of the full size where the data from
the 19 mm NMAS exhibited higher errors than smaller NMAS mixtures. The strain
coefficient of uniformity of small scale specimens tended to be worse than that from the
full size specimens for about one third of the data with the majority satisfying
recommended quality. The small scale methodology could benefit from more research
and development such as providing a second load cell calibration over a smaller range of
force with gain amplifiers.
The fatigue test results show that it is very promising to characterize small scale
specimens in the same manner as full size specimens in the AMPT. Only four mixtures
were characterized thus far and there was no consistent trend where the smaller scale
specimens produced consistently higher or lower fatigue life than full scale. Both
endurance limit and pseudo stiffness at failure were very comparable. The ranking was
mostly preserved and one mixture had a very different fatigue life by an order of
magnitude.
6. Acknowledgements
The authors want to thank Scott Parobeck and Frank Davis in the Bituminous Mixtures
Laboratory; and Jason Metcalf and Mario Tinio in the Accelerated Load Facility at the
FHWA Turner-Fairbank Highway Research Center are recognized for providing their
expert attention in the performance data collection, specimen preparation and running
the laboratory tests.
612
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

7. References
AASHTO, Determining Dynamic Modulus of Hot Mix Asphalt (HMA). AASHTO TP

62-07, 2009 edition.
Bonaquist, R. and D. Christensen. Refining the Simple Performance Tester for use in
Routine Practice, NCHRP Report 614, Transportation Research Board, National
Research Council, Washington, D.C., 2008.
FHWA and Transportation Pooled Fund Study TPF5(178), Implementation of the
AMPT for Superpave Validation, 2012.
Gibson, N., X. Qi,, A. Shenoy, G. Al-Khateeb, M. E. Kutay, A. Andriescu, et al.,
Performance Testing for Superpave and Structural Validation, Final Report FHWA-
HRT-11-045, Federal Highway Administration, Washington, D.C., 2012.
Hou, T., S. Underwood and R. Kim, Fatigue Performance Prediction of North Carolina
Mixtures Using the Simplified Viscoelastic Continuum Damage Model, Asphalt
Paving Technology, Journal of the Association of Asphalt Paving Technologists,
Vol. 79, 2010, pp. 3580.
Kutay, M.E., N.H. Gibson, R. Dongre and J. Youtcheff, Use of Small Samples to
Predict Fatigue Lives of Field Cores-Newly Developed Formulation Based on
Viscoelastic Continuum Damage Theory, Transportation Research Record,
Transportation Research Board, National Research Council, No. 2127, 2009, pp.90
97.
Li, X. and M. Marasteanu, Using Semi Circular Bending Test to Evaluate Lo w
Temperature Fracture Resistance for Asphalt Concrete, Journal of Experimental
Mechanics, Vol. 50, No.7, 2010, pp. 867876.
Li, X., N. Gibson, X. Qi, T. Clark, and K. McGhee, Laboratory and Full-Scale
Evaluation of 4.75 mm NMAS Superpave Overlay, Transportation Research
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Record, Journal of Transportation Research Board, No. 2293, 2012, pp. 2938.
Li, X., T. Clyne, G. Reinke, E. Johnson, N. Gibson and E. Kutay, Laboratory
Evaluation of Asphalt Binders and Mixtures Containing Polyphosphoric Acid,
Transportation Research Record, Journal of Transportation Research Board, No.
2210, 2011, pp. 4756.
Luo, X., R. Luo, and R.L. Lytton, "Characterization of Asphalt Mixtures Using
Controlled-Strain Repeated Direct Tension Test," Journal of Materials in Civil
Engineering, Vol 25, No. 2, 2013, pp. 194207.
Pellinen, T. K., and M. W. Witczak, Stress Dependent Master Curve Construction for
Dynamic Modulus, Asphalt Paving Technology, Journal of the Association of
Asphalt Paving Technologists, Vol. 71, 2002, pp. 281304.
613

LI, GIBSON
Kim, Y. R. Y. Seo, M. King and M. Momen Dynamic Modulus Testing of Asphalt

Concreter in Indirect Tension Mode, Transportation Research Record, Journal of
Lytton, R., J. Uzan, E. Fernando, R. Roque, D. Hiltunen and S. Stoffels, Development
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
and Validation of Performance Prediction Models and Specifications for Asphalt
Binders and Paving Mixes, SHRP Report A357, National Research Council,
Wshington, DC. 1993.
Underwood, S., R. Kim and M. Guddati, Improved Calculation Method of Damage
Parameter in Viscoelastic Continuum Damage Model, International Journal of
Pavement Engineering, Volume 11, Issue 6, pp. 459476, 2010.
Witczak, M. W., K. Kaloush, T. Pellinen, M. El-Basyouny, and H. Von Quintus, Simple
Performance Test for Superpave Mix Design, NCHRP Report 465, Transportation
Research Board, National Research Council, Washington, D.C., 2002.
Zofka, A., M. Marasteanu, X. Li, T. Clyne and J. McGraw, Simple Method to Obtain
Asphalt Binders Low Temperature Properties from Asphalt Mixture Properties,
Asphalt Paving Technology, Journal of Association of Asphalt Paving
8. Discussion
PROFESSOR HAIFANG WEN: Nelson, very nice work again. Just a question on the
comparison between the IDT modulus and the small scale. Did you have the full-scale
modulus also?
DR. NELSON GIBSON: We did, and thats in the report. But you have to put an asterisk
on that because we did core vertically down through the pavement. You had differences
aging from the top of that core to the bottom, and then right in the middle of that core
there was a lift boundary. So we did run it. The data is in the report, and its kind of a
mishmash. It doesnt line up; its not consistent with one geometry versus another, as to
be expected. Personally, I would rather core horizontally because you're getting away
from any kind of aging gradients with depth. So we do have it, but I wouldnt do it if Im
going to test a field pavement.
PROF. WEN: Okay, thank you.
MR. NATHAN MORIAN: I might have one, if we have a minute. When you're testing
the small-scale in the AMPT, when you added those other plates onto the top and the
bottom, did you have to adjust your gains or control parameters, calibration factors so to
speak?
614

DR. GIBSON: Thats a very good question. Okay, Im going to get very technical when
I answer this. I expected our technician to come to us and say, Hey, the waveform did not
look well. The particular manufacturer of AMPT we are using is from IPC, and from
what I understand, we had an earlier version of that AMPT. It had a servo valve that had
certain characteristics. We then, for various reasons, one of the reasons being we fried
the electronics, we had to change that servo valve, and they had said that this servo valve
is better. We noticed that in terms of the tuning, you had to be less meticulous in tuning
the various parameters. Shawn, if Im not mistaken, I believe that our technician
basically applied the same he didnt change anything on the No, wait. I think he had
to change Theres proportional and then there are two others that you really dont
mess with. I believe he had to change proportional, but it was no different from having
to switch from a soft mix to a stiff mix. So nothing out of the ordinary. Good question.
Thank you very much.
615
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

A Mechanistic Permanent Deformation Model
for Asphalt Concrete in Compression
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Yeong-Tae Choia* and Y. Richard Kimb

a
Graduate Research Assistant, Department of Civil, Construction, & Environmental
Engineering, North Carolina State University, Raleigh, NC 27695-7908
b
Distinguished University Professor, Department of Civil, Construction, &
Environmental Engineering, North Carolina State University, Raleigh, NC 27695-7908
ABSTRACT: Permanent deformation modeling research at North Carolina State University has
produced the so-called incremental model that fits the primary and secondary regions in
permanent strain growth. Triaxial repeated load permanent deformation (TRLPD) tests are
conducted on FHWA-ALF and NY9.5B mixtures to evaluate the effects of temperature, stress, and
load time on permanent deformation and, therefore, to determine the form of the incremental
model to account for these effects. The test results suggest that the slope in the log(vp)-log(N) plot
is constant regardless of these three major factors. This observation provides the basis for two
modeling approaches: the functionalized model and the shift model. The functionalized model is
formulated by expressing the coefficients of the incremental model in terms of reduced load time
and deviatoric stress. The shift model, based on the time-temperature-stress superposition
principle, uses the strain mastercurve and reduced load time and deviatoric stress shift functions.
A composite loading test that is composed of varying load times and deviatoric stresses is
proposed as the model calibration test. It is found that the permanent strain growth under the
complex loading histories predicted by the calibrated models is in good agreement with the
measured permanent strain growth.
KEYWORDS: permanent deformation, rutting, incremental model, functionalized model, shift model,
time-stress superposition
http://dx.doi.org/ 10.1080/14680629.2013.812847
617

CHOI, KIM
1. Introduction
Rutting and fatigue cracking are two representative distresses found in asphalt
pavements. Permanent deformation (or rutting) can lead to traffic accidents, especially in
bad weather such as heavy rain or snow. Heavy rain can cause dangerous driving
conditions, such as hydroplaning, and large amounts of water spray that can hinder
visibility. Snow can cause similar problems because snow and ice can collect in the
rutting path.
Various models are available that can predict permanent deformation in asphalt
concrete pavements in order to reduce rutting-related problems. They can be classified
into two categories: mechanistic (visco-) plastic models and power law-type models.
Mechanistic (visco-) plastic models are based on classical plasticity. Desai et al.
(1986) and Gibson (2006) applied hierarchical single surface (HiSS) yield criteria and a
Perzyna-type flow rule in order to predict three-dimensional behavior of asphalt concrete
in compression. Their models, however, are limited in their ability to simulate rate-
dependent behavior, such as rate-dependent softening, of asphalt mixtures (Yun and Kim
2011). Darabi et al. (2012) introduced the concept of a memory surface to overcome the
limitations associated with rate-dependent behavior. However, this model cannot
describe the effects of temperature despite the temperature sensitivity of asphalt concrete.
Yun (2008) and Subramanian (2011) adopted viscoelastic concepts, such as the
convolution integral, to explain the time-dependent viscoplastic behavior of asphalt
concrete. Such concepts are complex for practicing engineers to understand thoroughly
and apply easily, even though these models can predict permanent strain in terms of
various loading conditions, such as stress, load time, and/or temperature, and may be
able to shed light on the three-dimensional performance of a pavement structure through
numerical implementation. Furthermore, viscoelastic concepts demand high calibration
and computing costs. Therefore, despite providing good predictions, mechanistic (visco-)
plastic models have not been widely accepted for use by agencies.
In contrast, power law-type models are relatively simple and easily implemented.
The representative model is a strain ratio model implemented in the Mechanistic-
Empirical Pavement Design Guide (MEPDG), as shown in Equation 1.
p
= 1 T 2 N 3 [1 ]
r
where p is permanent strain, r is resilient strain, T is temperature, N is the number of

cycles, and 1 , 2 , 3 are the regression parameters. This model can describe the effects
of temperature and stress on permanent strain. The effect of stress is represented via the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`-
618

Development of Characterization Models for Incremental Permanent Deformation Model
resilient strain that is calculated from the dynamic modulus and stress calculated by the
theory of elasticity. The model assumes that the difference in dynamic modulus values
represents the rutting properties for the different mixtures. However, Von Quintus et al.
(2012) report that this assumption increases error; that is, the dynamic modulus itself is
important but cannot support the plastic deformation characteristics for an entire range of
mixtures and climates. In addition, the model cannot describe the primary region
behavior properly because the functional form of Equation 1 applies only to the behavior
of the secondary region.
Recent permanent deformation modeling research at North Carolina State University
has resulted in a viscoplastic model that is based on viscoelastic convolution integrals for
explaining the behavior of asphalt concrete in compression under repeated loading
(Subramanian, 2011). This model has been reduced further to a simplified form, the so-
called incremental model, which is an advanced power-law type model and represents
both the primary and secondary regions in the permanent strain growth behavior of
asphalt concrete (Choi et al., 2012). This paper presents a research effort to express the
incremental model as a function of load time, deviatoric stress, and temperature. The
underlying reason to select these three factors is that these three parameters change
within the depth of the asphalt layer in a pavement structure, and accounting for their
effects on permanent strain is important in the rutting prediction of asphalt pavements.
2. Objectives
The objectives of this paper are to develop simple mechanistic permanent deformation
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
models that can represent the effects of load time, stress, and temperature on the
permanent deformation of asphalt concrete and to verify the model using complex
loading histories.
3. Testing Program
Several types of triaxial compression cyclic tests are used in this study. These test
methods are described in the next section.
619

CHOI, KIM
3.1 Triaxial Repeated Load Permanent Deformation (TRLPD) Test
In this paper, the triaxial repeated load permanent deformation (TRLPD) test is defined
as a cyclic compression test under a constant confining pressure that employs a
haversine load pulse followed by a rest period. The TRLPD test is used in this study
because constant cyclic loading can provide the permanent deformation characteristics of
a specific testing condition. In order to obtain true permanent strain, 100 seconds is
applied as the rest period, and the permanent strains are measured at the end of the rest
period. Two mixtures, FHWA-ALF control mix (hereinafter called FHWA) and NY9.5B,
are used with three different load times and three different deviatoric stress levels at two
temperatures. A constant confining pressure is applied to avoid complexity in the initial
development of the models. Table 1 presents the mixture information and TRLPD test
conditions used in this study, and detailed information can be found elsewhere (Choi et
al., 2012).
Table 1. Mixture information and test conditions for TRLPD tests.
FHWA-ALF NY9 . 5 B
Mixture Information
Aggregate type Diabase Crushed gravel
NM AS 12.5 mm 9.5 mm
Asphalt Grade PG 70-22 PG 64-22 (modified)
% Air voids 4 .0 % 5 .4 %
Testing Conditions
0.1, 0.4, 1.0 at 37C

Load time (second) 0.1, 0.4a, 1.6
0.1, 0.4, 1.6 at 47C
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Rest period (second) 100 100

620, 827, 1034 482, 689, 896
Deviatoric stress (kPa)
(90, 120, 150 psi) (70, 100, 130 psi)
140 70
Confining pressure (kPa)
(20 psi) (10 psi)
Temperature (C) 40, 54 37, 47
Load shape Haversine Haversine
a
Data are not available for 0.4-second load time tests with 90 and 150 psi deviatoric stress
levels at 40C.
620

3.2 Composite Loading Test

Composite loading involves several loading blocks with changing deviatoric stresses and
load times (see Figure 1 for an example). Each loading block is composed of hundreds
of cycles with the same deviatoric stress and the same load time. Then the deviatoric
stress and load time are changed in the next loading block. For three deviatoric stresses
and three load times evaluated in this study, a total of nine loading blocks are applied to
each specimen.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
The hypothesis behind composite loading tests is that one loading block represents an
entire TRLPD test. In other words, the characteristics of permanent deformation
behavior that are obtained from one TRLPD test can be gained from one loading block
of a composite test. Therefore, the composite loading test can serve as a substitute for
several TRLPD tests with different deviatoric stresses and load times.
Figure 1. Schematic diagram of a composite loading history test.
3.3 Random Loading History Test
A random load history is composed of a random distribution of load times and deviatoric
stress values. The load history for each random load test is selected randomly based on
combinations of three load times (0.1, 0.4, 1.6 sec) and three deviatoric stress levels (90,
120, 150 psi). That is, the loading history changes from pulse to pulse; one loading
condition among nine different ones is assigned randomly to each pulse. The random
load test is performed at 54C for the FHWA mix. Two replicates are tested for
comparison. This test is intended for the verification of the proposed models.
Based on the results of testing of the effects of rest period on permanent deformation,
it is found that about 10 sec of rest for the FHWA mix produces the same permanent
strain as a 100 sec rest period (Kim and Guddati, 2011). In order to reduce the testing
time, a 20-second rest period is selected as a conservative amount of time for a rest
period for both the composite loading test and random loading test.
621

CHOI, KIM
4. Incremental Model
The incremental model, expressed as Equation 2, is introduced by Choi et al. (2012) to

describe the permanent deformation behavior of asphalt concrete.
A+ BN
=
vp , [2 ]
(C + N )
where 1- is a slope in log-log space and is related to hardening evolution; A and C
govern the permanent strain behavior in the primary region; B and control the
secondary region development; A is related to the initial permanent strain; and C
represents the number of cycles where the transition from the primary region to the
secondary region occurs. One of the strengths of this incremental model is that it fits
both the primary and secondary regions (see Figure 2).
Characteristics of the incremental model are documented in Choi et al. (2012), and a
summary of these characteristics is given below:
The time-temperature superposition (t-TS) principle is verified for the permanent

strain behavior of asphalt concrete that has significant viscoplastic strains.
Reduced load time is used to evaluate the behavior of asphalt materials by applying
the t-TS principle, instead of using temperature and load time separately.
The slope of the secondary region in log-log space is constant regardless of the
reduced load time and deviatoric stress; thus, is a material constant.
The regression coefficients, A, B, and C, change according to reduced load time and
deviatoric stress.
A notable observation is that the slope of the secondary region is constant regardless
of the test conditions. At high and intermediate temperatures, the FHWA and NY9.5B
mixtures show a constant slope. At a low temperature, the slope is different; however,
the constant slope assumption does not produce a significant difference in the permanent
strain prediction of asphalt pavement because most of the deformation occurs at the
higher temperatures. Von Quintus et al. (2012) conducted an exploratory program using
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
TRLPD tests and concluded that the average slope is approximately the same for the
different test temperatures. Therefore, the assumption of a constant slope seems to be
reasonable considering the simplicity it can provide in developing the permanent strain
models.
Experimental observations from the TRLPD testing and of the characteristics of the
incremental model result in two different permanent strain models in this study. The
characteristic that allows coefficients of the incremental model to change according to
reduced load time and deviatoric stress provides a straightforward approach, called the
622

functionalized model, which creates mathematical functions for the incremental model
coefficients (i.e., A, B, and C) in terms of reduced load time and deviatoric stress.
The other modeling approach, based on a single constant slope, is to shift the
permanent strain curves of the different reduced load times and deviatoric stresses
horizontally along the number of loading cycles in the x-axis. If the focus is on the
secondary region, all the permanent strain values can be translated to construct one strain
evolution curve in log-log space because they have the same slope. The strain curve after
the horizontal shift is called the permanent strain mastercurve, and the amount of
translation is defined as the shift factor. This approach is conceptually the same as that
taken to derive the dynamic modulus mastercurve and its prediction using the shift
function. This shifting strain approach is thereby referred to as the shift model.
5. Functionalized Model
The incremental model has four coefficients. Among these coefficients, is considered
as a material constant that is related to the hardening evolution. The other coefficients, A,
C, and B, are made into functional forms in terms of the reduced load time and/or
deviatoric stress.
5.1 Coefficients A and C
Figures 3(a)(d) illustrate the dependence of coefficients A and C on reduced load time
and deviatoric stress at two different temperatures. In Figure 3 (a) and (b), the A
coefficient exhibits clear changes as a function of both deviatoric stress and reduced load
time. The A coefficient increases as the reduced load time increases and deviatoric stress
increases. This tendency is expected given the characteristic of A that is associated with
initial permanent strain. With regard to the C coefficient, Figure 3(c) and (d) show that C
is dependent on reduced load time, but its dependence on deviatoric stress is much less
significant. Given that C is related to the transition point between the primary and
secondary regions, this trend is reasonable; that is, a longer pulse time requires fewer
cycles to reach the secondary region. It appears that, using the same load time, a higher
level of deviatoric stress produces more incremental strain, but this effect is not strong at
the transition point between the primary and secondary regions.
623
--`,`,,,``,,```,,`,

CHOI, KIM
1.E-01
(a) 90 psi
Permanent Strain
1.6s-54C
1.E-02
0.1s-54C
1.E-03 0.1s-40C
1.E-04
1 10 100 1000 10000
Cycles (N)
1.E-01
(b) 120 psi
Permanent Strain
1.6s-54C
1.E-02 0.1s-54C
0.1s-40C
1.E-03
1.E-04
1 10 100 1000 10000
Cycles (N)
1.E-01
(c) 150 psi
1.6s-54C
Permanent Strain
0.1s-54C
1.E-02
0.1s-40C
1.E-03
M easur ed
Fitted
1.E-04
1 10 100 1000 10000
Cycles (N)
Figure 2. Fitting results of incremental model for FHWA mixture: (a) 90 psi, (b) 120
psi, and (c) 150 psi (Choi et al., 2012).
624
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 3 demonstrates the distribution of A and C under the constant as well as the
fitting of the As and Cs of all the specimens using a power form. The reason the power
form is used is that the distribution of both A and C in a semi-log plot seems to exhibit
curvature rather than linear change. The A and C coefficients always should be greater
than or equal to 0 because A is related to the initial permanent strain and C is the
curvature location. Based on these reasons, the power form is applied to both reduced
load time and deviatoric stress.
1.E-02 1.E-02
9.E-03 90 psi @ 40 C (a) 9.E-03 70 psi @ 37 C (b)
120 psi @ 40 C 100 psi @ 37 C
8.E-03 150 psi @ 40 C 8.E-03 130 psi @ 37 C
7.E-03 7.E-03 70 psi @ 47 C
A: NY9.5B
90 psi @54 C
A: FHWA
6.E-03 120 psi @54 C 6.E-03 100 psi @ 47 C

150 psi @54 C 130 psi @ 47 C
5.E-03 5.E-03 70 psi-fit
90 psi-fit
4.E-03 120 psi-fit 4.E-03 100 psi-fit
3.E-03 150 psi-fit 3.E-03 130 psi-fit
2.E-03 2.E-03
1.E-03 1.E-03
0.E+00 0.E+00
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
Reduced Load Time (sec) Reduced Load Time (sec)
25.00 25.00
(c) (d)
20.00 20.00
C: NY9.5B
C: FHWA
15.00 15.00
10.00 10.00
5.00 5.00
0.00 0.00
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
8.E-03 90 psi @ 40 C 8.E-03 70 psi @ 37 C
120 psi @ 40 C ( e) 100 psi @ 37 C
(f)
7.E-03 150 psi @ 40 C 7.E-03 130 psi @ 37 C
70 psi @ 47 C
6.E-03 90 psi @ 54 C 6.E-03 100 psi @ 47 C
B: NY9.5B
120 psi @ 54 C
B: FHWA
130 psi @ 47 C
5.E-03 150 psi @ 54 C 5.E-03 70 psi-fit
90 psi-fit 100 psi-fit
4.E-03 120 psi-fit 4.E-03 130 psi-fit
3.E-03 150 psi-fit
3.E-03
2.E-03 2.E-03
1.E-03 1.E-03
0.E+00 0.E+00
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 3. Behavior and fitting of coefficients of incremental model: (a) A of FHWA, (b)
A of NY9.5B, (c) C of FHWA, (d) C of NY9.5B mixes, (e) B of FHWA, and (f) B of
NY9.5B.
625

CHOI, KIM
5.2 Coefficient B
The most important coefficient in the incremental model is B. As shown in Figure 3 (e)
and (f), the value of B clearly increases as the reduced load time increases. However, the
deviatoric stress effect is not as clear as the reduced load time effect due to the scatter in
the data. In general, the slope of the fitted curve between B and reduced load time
increases as the deviatoric stress increases. Numerous functions are investigated to
express the B coefficients dependence on reduced load time and deviatoric stress. Root
mean squared errors (RMSEs) in both arithmetic and logarithmic scales, as defined in
Equation 3, are used to select the best model for coefficient B. RMSE (Arith) indicates
the overall prediction results and puts more emphasis on the secondary region. RMSE
(Log) indicates the accuracy of the prediction by concentrating more on the primary
region. The final functional form of the B coefficient based on the minimum error is a
product of the power and exponential functions, as shown in Equation 4.
=RMSE (Arith)
( fitted strain measured strain) 2
,
N [3 ]
( log( fitted strain) log(measured strain) )

2
=RMSE (Log) .
N
To summarize the functional forms of the coefficients of the incremental model, all
the coefficients are expressed in power form and/or as an exponential function for both
reduced load time and deviatoric stress. The final model is expressed in Equation 4
where p is reduced load time and d is deviatoric stress.
( A + B)
=
vp
(C + N )

a
A a1 p p 2 p da1d
= [4 ]
c
= c1 p p2 p and
C
B= (b 1p
b
)
p 2 p exp b3d d ( (b1d p + b2 d )
)
When Equation 4 is applied to the composite loading test, coefficients A and C can
be assumed constant and the values of these coefficients from the first loading block can
be used for the remaining test. This assumption is valid because A and C represent the
behavior of the primary region, and in the composite loading test most of the primary
region is consumed by the first loading block. Therefore, after the first loading block in
the composite loading test, coefficients A and C are merely fitting parameters and have a
626
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

minimal effect on the permanent strain accumulation. Hence, the form of the
functionalized model for the composite loading test can be simplified to Equation 5.
( A + B)
=
vp
(C + N )

A, C
= constant and [5 ]
B = (b 1p
b
) (
p 2 p exp b3d d
( b1d p + b2 d )
)
6. Shift Model
The previous observations concerning the coefficients of the incremental model suggest
that a single, material-dependent, value exists. Because (1-) is the slope of the
permanent strain growth in the log-log scale, this finding also suggests that all of the
strain curves can be shifted on the log-log scale plot to construct a single strain
mastercurve. This so-called shifting approach may be useful in characterizing the model
with fewer experimental requirements. The following sections describe the time-stress
superposition (t-SS) principle, which forms the basis for the shift model, and the
experimental verification of the shift model.
6.1 Time-Stress Superposition Principle
The findings from the incremental model characterization provide a phenomenological

basis for shifting the permanent strain curves. Theoretical support for this approach is
investigated via a literature review.
In order to understand the t-SS principle, it is important first to understand the time-
temperature superposition (t-TS) principle. The t-TS principle is a well-known
characteristic of thermorheologically simple (TRS) materials. Asphalt concrete is known
as such a material, and the t-TS principle is often applied when assessing the behavior of
this material in the viscoelastic range. The t-TS principle also has been shown to be a
viable tool when significant viscoplasticity and damage are present (Chehab et al., 2002,
Zhao, 2002, Schwartz et al, 2002, Yun et al, 2010). To show the theoretical basis for
these findings, attention turns to the single integral nonlinear viscoelastic constitutive
equation proposed by Schapery (1969), Equation 6.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
627

CHOI, KIM
t dg 2
(t ) g 0 D0 (t ) + g1 D( )
= [6 ]
0 d
where: D = creep compliance,

t
dt '/ a ( (t ') )
0 ,

( ) dt / a ( (t ) )
0 , and
g0, g1, g2, a = material properties and functions of stress.
When g0, g1, g2, and a are equal to 1, Equation 6 reduces to a standard linear
viscoelastic equation. That is, g0, g1, g2, and a are factors that represent the nonlinear
behavior of viscoelastic material. In the case of a square-shaped creep loading input,
Equation 6 can be simplified to Equation 7. This simplified equation then suggests that
the measured incremental creep compliances at different stress levels can be made to
overlap by translation along the logarithmic scale plot in terms of time, t, and
incremental creep compliance, Dn. In order to superpose the curves, the vertical
distance is defined as the difference between the log(g1 g2) of two different stress
levels, and the horizontal distance equals to log(a) of the two creep compliance
curves. In this scenario, a is called the time-stress shift factor. If g1 g2 is equal to one
(i.e., log(g1 g2) = 0), a vertical shift is not necessary. One example where only
horizontal shifting is necessary is shown by Schapery (1969) using experimental data
from glass-reinforced phenolic resin. Furthermore, many researchers (Haj-Ali and
Mulianan, 2012, Darabi et al., 2012, Mahmoud et al., 2012, Zhao et al., 2012, Saadeh et
al., 2007, and Huang et al., 2007) have adopted the nonlinear viscoelastic equation to
explain the viscoelastic behavior of asphalt materials.
t
(t ) g 0 D0 + g1 g 2 D
=
a
t [7 ]
log(=Dn ) log( g1 g 2 ) + log D
a
(t )
=Dn g 0=
D0 D(t ) g 0 D0

Other researchers (Brostow, 2000, Akinary et al., 2001, Akinary and Brostow, 2001,
Wenbo et al., 2001, and Jazouli et al., 2005) prove the t-SS principle using a free volume
concept. Free volume changes with time, temperature, and stress in a viscoelastic
material under loading. That is to say, these three factors are related to one another such
628
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

that they are interchangeable. (It is noted that this phenomenon is the underlying
mechanism that also explains the t-TS principle.) Therefore, the t-SS principle is
applicable to viscoelastic materials. Only Wenbo et al. (2011) suggest horizontal and
vertical shifting where the vertical shifting indicates the effects of stress-induced
microstructural changes on material properties. The other researchers use only horizontal
shifting (i.e., time-scale shifting) to generate creep compliance mastercurves using creep
test results at different stress levels. Azari and Mohseni (2012) introduced
the TP variable, TP = temperature (C) deviatoric stress (MPa), to construct a
minimum strain rate (MSR) mastercurve for the flow number test results for asphalt
materials. They successfully created the MSR mastercurve using the TP variable on the
horizontal axis. Their work indicates that deviatoric stress is interchangeable with
temperature, and thus, the MSR mastercurve could be evidence of the t-SS
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
correspondence principle at work.

In summary, the literature suggests that the t-SS principle is applicable to viscoelastic
materials. In some cases, vertical shifting, i.e., translating the curve along the creep
compliance axis, is required to represent the nonlinearity or the stress-induced
microstructural changes of the materials. In other cases, only horizontal shifting is
needed. The determination of which shift process is necessary for a given material
depends on the material behavior observed through well-orchestrated experiments.
6.2 Validation of Shiftability Using TRLPD Tests
Observations and the two superposition principles (t-TS and t-SS) support the notion that
permanent strains at different temperatures and deviatoric stress levels can translate and
create a permanent strain mastercurve. However, this hypothesis has never been proven
for asphalt concrete by applying it to experimental data. In this study, a series of TRLPD
tests is used to verify the assumption. As heretofore mentioned, three factors are used for
model development: deviatoric stress, load time, and temperature. The temperature and
load time are condensed into reduced load time. Then, two factors, the reduced load time
and deviatoric stress, remain. Therefore, the hypothesis that underlies the shift approach
should be able to be proven using these two factors.
6.2.1 Reduced Load Time Shifting
It is obvious that asphalt material permanently deforms more as the load time increases
under the same stress conditions. For general plasticity, where the hardening function
does not change dramatically, the permanent strain grows in proportion to the length of
the load time or cumulative load time if the stress conditions are constant. However, Yun
629

CHOI, KIM
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
(2008) found that the apparent incremental strain rate, defined as the incremental strain
due to one cycle that is divided by the load time, decreases as the load time increases.
His observation suggests that the amount of permanent strain is not directly proportional
to the length of load time. This lack of proportionality is due to hardening during loading
and softening during unloading, and thus, the effect of load time must be determined by
experimentation.
The load time shift factor can quantify the effect of the reduced load time on
permanent strain. The number of cycles is related directly to the cumulative load time for
the TRLPD tests, and thus, horizontal shifting of permanent strain curves along the
number of loading cycles axis among different load times can represent the effect of load
time under the same deviatoric stress levels. This horizontal load time shift is expressed
by Equation 8, where Nred stands for the reduced number of cycles after load time
shifting. The load time shift factor is defined as the logarithmic horizontal distance
between the reference strain curve and the strain curve in cases where the deviatoric
stresses are the same, as depicted in Figure 4. The reduced load time shift factors ( a p )
are calculated using the error minimization algorithm in MATLAB.
a p
N red= N 10
[8 ]
( ( N ) vp ( N red ) )
refer 2
=Error vp
Figure 4 shows the reduced load time shift. The left-side graphs represent the strains
before translation, and the right-side graphs show the same data after horizontal load
time shifting. In order to make one representative curve, the average strain of the 0.4 sec
load time test at each deviatoric stress level is used as a reference curve. Through
horizontal load time shifting, all the strain evolution curves overlap relatively well. The
same degree of success in horizontal shifting is also found for the NY9.5B mix. The shift
shown in Figure 4 is not perfect; however, as is seen later in the verification study, this
amount of error from horizontal shifting is acceptable from an engineering point of view.
630

2.0% 2.0%
(a) 90 psi (d) 90 psi
Permanent Strain
Permanent Strain
1.6% 1.6%
1.2% 0.4s-54C 0.1s-54C 1.2%

1.6s-54C
0.8% 0.8%
1.6s-40C
0.4% 0.4%
0.1s-40C
0.0% 0.0%
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E-02 1.E+00 1.E+02 1.E+04
Cycles (N) Reduced Cycles (Nred)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
2.0% 2.0%
(b) 120 psi (e) 120 psi
Permanent Strain
Permanent Strain
1.6% 1.6s-54C 1.6%
1.6s-40C
1.2% 1.2%
0.4s-40C
} 0.4s-54C
0.8% 0.8%
0.1s-54C
0.4% 0.4%
0.1s-40C
0.0% 0.0%
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E-02 1.E+00 1.E+02 1.E+04
2.0% 2.0%
(c) 150 psi 1.6s-54C
0.4s-54C
(f) 150 psi
Permanent Strain
Permanent Strain
1.6% 1.6%
1.2% 1.6s-40C 1.2%

0.1s-54C
0.8% 0.8%
0.4% 0.4%
0.1s-40C
0.0% 0.0%
1.E+00 1.E+01 1.E+02 1.E+03 1.E+04 1.E-02 1.E+00 1.E+02 1.E+04
Figure 4. Load time shifting of FHWA mix: (a) before shifting of 90 psi, (b) before
shifting of 120 psi, (c) before shifting of 150 psi, (d) after horizontal shifting of 90 psi,
(e) after horizontal shifting of 120 psi, and (f) after horizontal shifting of 150 psi.
631

CHOI, KIM
6.2.1 Stress Shifting
The t-SS principle is suggested for modeling asphalt concrete behavior at different stress
levels. As the evidence in the literature suggests, the possible shift combinations are:
only horizontal shifting (H), and both horizontal and vertical shifting (H&V). These two
approaches are investigated to verify the t-SS principle and to find the best shifting
approach for the study mixtures. The same shifting approach as is used for the reduced
load time shifting approach is applied here. The cluster of strains that develops after the
load time shift is translated visually to build the strain mastercurve. The stress shift
factor ( a d ) is determined based on the logarithmic distance between the cluster of
strains of the individual deviatoric stress levels and the reference curve.
Figure 5 indicates that, regardless of the stress shift direction, the strains of the
individual deviatoric stress curves overlap one another fairly well. Some discrepancies
are observed at smaller strain levels; however, these discrepancies are small and take
place mostly within the primary region, which is expected because of the one constant
assumption. When considering simplicity and efficiency in the model, horizontal-only
shifting is selected for applying the t-SS principle to the viscoplastic strain of asphalt
concrete.
6.3 Characteristics of Shift Factors
6.3.1 Reduced Load Time Shift Factors
The horizontal distance between a certain strain curve and a reference curve in the
logarithmic scale under the same deviatoric stress is defined as the load time shift factor.
In other words, the load time shift factor can be described as the ratio of the number of
cycles between two load time conditions that produce the same amount of permanent
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
strain.
Figure 6 depicts the shift factors obtained through horizontal-only shifting. No clear
trend is seen in Figure 6 (a) and (b) with regard to deviatoric stress. The shift factors at
120 psi for the FHWA mix are slightly higher than for the other stress conditions.
However, considering the specimen-to-specimen variability and the sequence of
deviatoric stress, it is reasonable to conclude that the deviatoric stress does not affect the
load time shift factors. The NY9.5B mix shows this random behavior more clearly in
Figure 6 (b). Thus, it is concluded that the load time shift factor is a function of reduced
load time only. This conclusion implies that only one stress condition, instead of three
different deviatoric stress levels, can be used to determine the load time shift factors if
various load times are applied to the sample.
632

2.0%
(a) FHWA 150 psi
Permanent Strain
1.6%
1.2% 120 psi
0.8%
0.4% 90 psi
0.0%
1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Reduced Cycles
2.0% 2.0%
(b) H (c) H&V
1.6% 1.6%
Strain: H-H&V
Strain: H-H
1.2% 1.2%
0.8% 0.8%
0.4% 0.4%
0.0% 0.0%
1.E-02 1.E+00 1.E+02 1.E+04 1.E+06 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Reduced Cycles Reduced Cycles
2.5%
(d) NY9.5B 130 psi
Permanent Strain
2.0%
1.5% 100 psi

1.0%
70 psi
0.5%
0.0%
1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Reduced Cycles
2.5% 2.5%
(e) H (f) H&V
2.0% 2.0%
Strain: H-H&V
Strain: H-H
1.5% 1.5%
1.0% 1.0%
0.5% 0.5%
0.0% 0.0%
1.E-02 1.E+00 1.E+02 1.E+04 1.E+06 1.E-02 1.E+00 1.E+02 1.E+04 1.E+06
Reduced Cycles Reduced Cycles
Figure 5. Stress shifting verification: (a) permanent strain data from the FHWA mix, (b)
horizontal shifting only (FHWA), (c) horizontal and vertical shifting (FHWA), (d)
permanent strain data from the NY9.5B mix, (d) horizontal shifting only (NY9.5B), and
(f) horizontal and vertical shifting (NY9.5B).
633
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

CHOI, KIM
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
2.0 2.0
1.5 (a) FHWA (load time) 1.5
(b) NY9.5B (load time)
Load Time SF
Load Time SF
1.0 1.0
0.5 0.5
0.0 0.0
90 psi 70 psi
-0.5 -0.5 100 psi
120 psi
-1.0 150 psi -1.0 130 psi
log log
-1.5 -1.5 power
power
-2.0 -2.0
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
2.5 2.5
(c) FHWA (stress) (d) NY9.5B (stress)
2.0 2.0
Stress SF
Stress SF
1.5 1.5
1.0 1.0 0.1s-70psi

0.1s-90psi
0.4s-90psi 0.4s-70psi
0.5 1.6s-90psi 0.5 1.6s-70psi
power-0.1s power-0.1s
0.0 power-0.4s 0.0 power-0.4s
power-1.6s power-1.6s
-0.5 -0.5
50 100 150 200 50 100 150 200
Deviatoric Stress (psi) Deviatoric Stress (psi)
Figure 6. Shift factors and fitting: (a) load time shift factors of FHWA mix, (b) load time
shift factors of NY9.5B mix, (c) deviatoric stress shift factors of FHWA mix, and (d)
deviatoric stress shift factors of NY9.5B mix.
The load time shift factor can explain the effect of load time on permanent strain,
especially the load time-dependent hardening effect that Yun (2008) found. If hardening
is independent of the length of load time, the reduced load time shift factors increase
linearly in arithmetic scale because permanent strain develops proportionally to the
length of load time. However, Figure 6 (a) and (b) demonstrate that the shift factor
increases quite linearly in semi-logarithmic scale, not in arithmetic scale, due to the
hardening evolution caused by the load time.
A functional form for this shift factor is needed for modeling purposes. The shift
factor increases linearly in the semi-logarithmic scale as the reduced load time becomes
longer. Both the logarithmic and power function expressions are examined. Figure 6 (a)
and (b) illustrate the results of fitting using these two models. These graphs indicate that
the power form shows better overall fitting than the logarithmic function. In addition,
considering the hardening behavior typically observed in a creep test of asphalt concrete,
that is, the creep strain increases but the strain rate decreases as the load time increases
in a creep test, a power form is more reasonable to represent the load time shift factor
function.
634

6.3.2 Stress Shift Factors
Stress shift factors are acquired simply by translating the strain clusters after load time
shifting to construct a mastercurve. Because this process involves moving the strain
clusters for the individual deviatoric stress conditions, only three stress shift factors are
calculated, as depicted in Figure 6 (c) and (d), but different loading times are used as the
reference. For example, legend 0.4s-90psi in the graphs indicates that the 0.4 sec load
time strain is used as the reference load time for the load time shift and that the strain of
90 psi is the reference stress for stress shifting.
Due to the large specimen-to-specimen variability, the shift factor for the NY9.5B
mix using the 0.4-s-70 psi reference appears to be an outlier at 130 psi. Thus, the 0.4
sec load time for the NY9.5B mix is excluded in the detailed analysis. Otherwise, the
stress shift factors behave similarly to one another.
The stress shift factors illustrate the effect of deviatoric stress on permanent
deformation. Von Quintus et al. (2012) report that permanent strain increases
exponentially rather than linearly as the deviatoric stress increases. The deviatoric stress
shift factors demonstrate this phenomenon. The FHWA mix shows such exponential
increments clearly; however, the stress shift factors of the NY9.5B mix increase linearly.
If 150 psi or higher deviatoric stress is applied, the stress shift factor may increase
dramatically. Therefore, the power form is selected to explain the behavior of the stress
shift factors.
In summary, the model that employs shift factors is comprised of the reference curve,
load time shift function, and stress shift function. The reference curve is defined as the
permanent strain growth curve obtained from a single TRLPD test under the reference
loading condition. In contrast to the manner by which a dynamic modulus mastercurve is
constructed, the reference curve is obtained by separate testing to determine the datum
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
line. The permanent strain is fitted using the incremental model. The incremental model
becomes the final reference curve form, which is considered to be a permanent strain
mastercurve. Strains obtained from the TRLPD tests are translated horizontally for load
time shifting and again stress shifting to construct the reference curve; accordingly, the
total shift factor is a summation of these two shift factors (i.e., atotal = ap + ad). Equation
9 expresses the final shift model form. The number of cycles, N, at a certain condition is
converted to a reduced number of cycles (Nred) by the total shift factor (i.e.,
atotal
N red= N 10 ); a1 , a2 , a3 , b1 , b2 , and b3 are the model parameters, and Pa is the
atmospheric pressure and is used to normalize the stress.
635

CHOI, KIM
A + B N red
v=
p
( C + N red )

a p a1 p a2 + a3 ,
= [9 ]
=a d b1 ( d / Pa )b2 + b3
7. Model Verification Using State Variables
The functionalized model and the shift model are proposed here on the basis of the
characteristics observed from the TRLPD tests where the load time and the deviatoric
stress remain constant within each test. In order to apply these models to realistic loading
conditions with varying load times and deviatoric stress levels, the state variable
approach is introduced in this section.
7.1 State Variable Selection
By definition, the state variable is a quantity that marks the state of the material, and
thus, whatever its history, its current value is that which affects the development of the
permanent strain at any given instant. In short, the state variable defines the condition of
the materials in the permanent deformation model. If different specimens have the same
value of the state variable, then the same amount of incremental strain occurs due to the
same load, even though their past loading histories may be different.
Permanent strain is a good indicator for characterizing the state of viscoplastic
behavior because it is simple and provides intuition about the viscoplastic state. In
general plasticity, the yield criterion or hardening function based on the plastic strain is
used to illustrate the state of the material, which is complex and may require additional
calibration and testing. In this sense, the yield function and/or hardening function are not
acceptable for this simple model development. In addition, it should be noted that
DARwin-ME (previously known as the MEPDG) uses the permanent strain as a state
variable to evaluate rutting. In addition, Gibson (2007) uses volumetric permanent strain
as a state variable by combining yield functions. Based on these observations, permanent
strain is applied as a state variable in this paper.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
636

7.2 Composite Loading Test
The proposed models are built on TRLPD tests and demand at least five (generally nine)
different conditions for calibration. An alternative method for model calibration, called
the composite loading test, is investigated to reduce the number of tests required. This
test is composed of several loading blocks that each represents one of the conditions
needed to characterize the model.
Figure 1 provides a schematic representation of the composite loading test. Here, it
is assumed that all results from any individual TRLPD test can be captured reliably
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
using only a small amount of data at that condition by applying the state variable
concept. Thus, for example, it is not necessary to perform 0.1 sec pulse tests at 90 psi for
thousands of cycles in order to characterize the model; only a few hundred is sufficient.
A composite test for the FHWA mix is designed to simulate nine different loading
conditions using only one test, as shown in Table 2. The test consists of three separate
loading groups, each consisting of the nine small load sequences shown in Table 2. Each
sequence consists of loading that is the same as the loading that is used in the TRLPD
test (i.e., pulse-rest-pulse-rest with one load time under one deviatoric stress). For each
loading group the load sequence starts at one and goes to nine. After the ninth sequence,
the next loading group begins at sequence one and goes to nine. The permanent strain
level at a certain combination of load time and deviatoric stress for different loading
groups provides information on the permanent strain evolution under the same load time
and deviatoric stress but at different hardening states. This test is performed at 54C.
Table 2. Composite loading history for the FHWA mix.
Load Dev. Number of Cycles

Load Time
Sequence Stress 1 Group 2nd Group 3rd Group
st
1 0.1 sec 200 300 100

2 90 psi 0.4 sec 200 300 100
3 1.6 sec 200 300 100
4 0.1 sec 200 300 100
5 120 psi 0.4 sec 200 300 100
6 1.6 sec 200 300 100
7 0.1 sec 200 300 100
8 150 psi 0.4 sec 200 300 100
9 1.6 sec 200 300 100
637

CHOI, KIM
The composite loading test is designed to perform only a part of the TRLPD test, and
so, a fitting process is required to extract the entire TRLPD test behavior from the small
loading block that is available. The most important part of the fitting process is to
determine which part of the entire TRLPD test is equivalent to the given loading block
within the composite test. The permanent strain (i.e., a state variable) is applied to link
together these two tests whose loading conditions are the same. Because the permanent
strain is considered to be a state variable, N cBomp can convert to N rBefer , which is the
equivalent number of cycles in a TRLPD test, as shown in Figure 7. The measured strain
of the loading block in the composite test with equivalent cycles provides all of the
information about the TRLPD test by fitting with the incremental model. In short, the
strains of the composite test are translated to the TRLPD test whose loading conditions
are the same, and then fitted by the incremental model to capture the full TRLPD test
results.
Figure 7. Illustration of linking composite loading test to TRLPD test.
7.3 Calibration of the Proposed Models
7.3.1 Functionalized Model
The calibration of the functionalized model requires permanent strain growth in nine
combinations of three load times and three deviatoric stresses. This information can be
derived from actual TRLPD tests under these nine conditions or from permanent strains
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
predicted from the composite loading test using the state variable approach, or both. The
first approach is described in Section 5. The second approach starts from the permanent
strain curve for a certain loading condition (i.e., one load time and one deviatoric stress)
for the composite loading test. Then, a portion of the permanent strain curve for the
638

TRLPD test under the same condition that yields the same range of permanent strains as
in the loading condition in question in the composite loading test is identified. Figure 7
illustrates this concept. This process is repeated for the nine loading conditions, which
results in portions of nine TRLPD tests for nine conditions. These portions of the
TRLPD tests are used to determine the incremental model coefficients as a function of
load time and deviatoric stress.
The third approach (i.e., the composite loading test only) employs an iterative error
minimization method. First, the initial values of the Nref for nine loading conditions are
assumed. Using these Nref values and the permanent strain growth for the nine conditions
of the composite loading test, the functionalized model coefficients in Equation 5 are
determined. The determined functionalized model is then used to develop the permanent
strain growth for the TRLPD tests under nine different loading conditions. These
virtually generated TRLPD permanent strain curves and the permanent strain curves
generated from the composite loading test are used to determine new Nref values. These
new Nref values are then used to go through the same steps presented above. This
iteration is repeated until there is convergence.
Among the three calibration methods discussed, the only method that can be
implemented by state highway agencies as a routine method is the third method
(composite loading test only), because the first two methods require TRLPD test results
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
under nine different loading conditions. The process of performing nine TRLPD tests
with replicates is too lengthy for state highway agencies to adopt these methods as
routine tests. The problem with the third calibration method, albeit a very simple and
practical method, is related to its convergence during the iteration. It is found that the
convergence is highly sensitive to initial values and thus renders the method unstable.
7.3.2 Shift Model
A similar approach to the calibration of the functionalized model is used to characterize

the shift model. The permanent strain again is applied as a state variable; however, only
one reference strain curve (instead of all the TRLPD test results) is needed in order to
derive a relationship between the composite test and one TRLPD test. The load time and
stress shift factors are computed through horizontal-only shifting, so the composite test
yields the total shift factor first. Figure 7 also illustrates the process of calculating the
shift factors; here, the TRLPD test provides the reference strain curve. By matching a
loading block from the composite test with a reference strain curve, the total shift factor
can be calculated by Equation 10. The total shift factor is then separated into two shift
factors by applying the assumption that the load time shift factor is independent of
deviatoric stress. The total shift factor, for which the deviatoric stress is the same as the
reference condition, becomes the load time shift factor. Then, the stress shift factor is
639

CHOI, KIM
calculated by subtracting the load time shift factor from the total shift factor. The model
parameters are determined using the reduced load time and deviatoric stress shift factors.
N ref
ato= log
tal N
comp
[1 0 ]
where: atotal = total shift factor,
E B
N=
ref ( N ref N ref ) = incremental number of cycles at reference curve, and
E B
N=
comp ( N comp N comp ) = incremental number of cycles for composite
loading test.
7.4 Verification of State Variable Approach
In this section, the functionalized model is used to verify the composite loading test
using the state variable approach. The first two calibration methods used for the
functionalized model, i.e., the method that use nine TRLPD tests and the methods that
uses both nine TRLPD tests and the composite loading test, are used for this verification.
These calibration methods are described in Section 7.3.1.
The first method that uses nine TRLPD tests does not require the state variable
approach, whereas the second method that uses nine TRLPD tests and the composite
loading test does require the state variable approach. Therefore, a comparison of the
prediction errors caused by these two methods can shed light on the validity of using
permanent strain as the state variable.
Data from nine TRLPD tests and the composite loading test for the FHWA mix are
used to calibrate the functionalized model using the two calibration methods. The
resulting model is used to predict the permanent strain growth in the TRLPD tests under
nine loading conditions. Table 3 presents a summary of the prediction errors and
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
calculations. As shown in this table, the difference in the magnitude of error is not
significant. This finding indirectly verifies the composite loading test and the state
variable approach. The ultimate verification of this method, however, is found in the
way the models that use the composite loading test and the state variable approach
predict the loading histories that are different from the loading histories used in the
model calibration. This additional verification is presented in the next section.
640

Table 1. Comparison of RMSEs according to calibration method.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Nine TRLPD Tests and
Nine TRLPD Tests
Error Type Composite Loading Test
(No State Variable)
(Using State Variable)
RMSE (Arith) 0 .0 2 4 0 .0 2 5
RMSE (Log) 1 .5 7 2 1 .7 5 7
7.5 Verification of Proposed Models
In this section, the calibrated functionalized and shift models are used to predict the
permanent strain for the composite loading history, random loading test, and TRLPD
tests. The functionalized model is calibrated using the composite loading test only
because the other calibration methods that require running nine TRLPD tests are
impractical. The shift model is calibrated using the composite loading test and one
reference TRLPD test.
Prediction for the composite tests is the first step in the verification process because
the first loading group in the composite tests is used for the model calibration and the
additional test results in the composite tests are used for the verification. Figure 8 (a)
shows that the models predict the composite loading history well even though the
differences amongst them increase in the second and third groups. The functionalized
model characterized by the composite test only exhibits more deviation from the
measured strains than the shift model.
The random loading history mimics actual field conditions with changing load times
and deviatoric stress levels. Thus, a good prediction for the random loading history is
essential for a model to be applied to pavement performance prediction. Figure 8 (b)
shows that the permanent strains predicted by the functionalized and shift models are
within the upper and lower boundaries created by the measurements taken from two
specimens.
In addition to showing the verification results for the composite and random loading
tests, Figure 9 displays the prediction accuracy of the two models in terms of the TRLPD
results under various conditions. Other than the over-predictions shown in Figure 9 (a)
and (b), the calibrated models seem to capture the permanent strain behavior under the
various conditions well. It is noted that the models are calibrated using the composite
loading test at 54C, and then the models are used to predict the permanent strain
behavior at 40C. The load time and temperature used to develop the data shown in
641

CHOI, KIM
Figure 9 (a) and (b) (i.e., 0.1 second load time and 40C) result in the reduced load time
that is at the highest limit of the reduced load time range that is used in calibrating the
models. The functions determined from the calibration might have greater errors at this
limiting condition.
1st Group 2nd Group 3rd Group

2.5%
(a)
Shift (Comp.+One TRLPD)
2.0%
Permanent Strain (%)
1.5%
Functionalized
1.0%
(Comp.Only)
Meas ured
0.5%
Func tion (Com p. Only)
Shift (Com p.+One TRLPD)
0.0%
0 900 1800 2700 3600 4500 5400 6300
Cycles (N)
2.0%
(b) Shift (Comp.+One TRLPD)
1.6%
1.2%
Functionalized
(Comp. Only)
0.8%
Meas ured (1)

0.4% Meas ured (2)
Func tion (Com p. Only)
Shift (Com p.+One TRLPD)
0.0%
0 500 1000 1500
Cycles (N)
Figure 8. Predictions by the models calibrated using the composite test: (a) prediction
for composite loading test, and (b) prediction for random loading test.
642
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

2.0% 2.0%
(a)

90 psi-0.1 s-40C (1) (b) 120 psi-0.1 s-40C (1)
90 psi-0.1 s-40C (2) Function (Comp. Only)
1.6% Function (Comp. Only) 1.6% Shift (Comp.+One TRLPD)
1.2% 1.2% Shift
Functionalized
Shift
0.8% 0.8%
Functionalized
0.4% 0.4%
0.0% 0.0%
0 500 1000 1500 2000 0 500 1000 1500 2000
Physical Cycles (N) Physical Cycles (N)
2.0% 2.0%

(c) 120 psi-1.6 s-40C (1) (d) Shift
Function (Comp. Only)
1.6% Shift (Comp.+One TRLPD) 1.6%
1.2% Shift 1.2%
0.8% 0.8%
Functionalized
Functionalized 150 psi-1.6 s-40C (1)

0.4% 0.4%
Function (Comp. Only)
0.0% 0.0%
0 500 1000 1500 2000 0 500 1000 1500 2000
2.0% 2.0%
(e) 90 psi-0.1 s-54C (1) (f) 120 psi-0.1 s-54C (1)

90 psi-0.1 s-54C (2) 120 psi-0.1 s-54C (2)
1.6% 90 psi-0.1 s-54C (3) 1.6% 120 psi-0.1 s-54C (3)
Function (Comp. Only) Function (Comp. Only)
Permanent
Strain (%)
Shift (Comp.+One TRLPD) Shift (Comp.+One TRLPD)

1.2% 1.2%
Shift Shift
0.8% 0.8%
0.4% 0.4% Functionalized

Functionalized
0.0% 0.0%
0 500 1000 1500 2000 0 500 1000 1500 2000
2.0% 2.0%
(g) (h) Shift

1.6% Shift 1.6%
1.2% 1.2%
Functionalized Functionalized
0.8% 0.8%
120 psi-1.6 s-54C (1) 150 psi-1.6 s-54C (1)
120 psi-1.6 s-54C (2) 150 psi-1.6 s-54C (2)
0.4% 120 psi-1.6 s-54C (3) 0.4%
Function (Comp. Only) Function (Comp. Only)
Shift (Comp.+One TRLPD) Shift (Comp.+One TRLPD)
0.0% 0.0%
0 500 1000 1500 2000 0 500 1000 1500 2000
Figure 9. Predictions by the models calibrated using the composite test for TRLPD tests
of FHWA mix: (a) 90 psi-0.1 sec at 40C, (b) 120 psi-0.1 sec at 40C, (c) 120 psi-1.6 sec
at 40C, (d) 150 psi-1.6 sec at 40C, (e) 90 psi-0.1 sec at 54C, (f) 120 psi-0.1 sec at
54C, (g) 120 psi-1.6 sec at 54C, and (h) 150 psi-1.6 sec at 54C.
643
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

CHOI, KIM
In summary, the calibrated functionalized model and shift model can predict
permanent strain growth under complex loading histories well, including a rando m
loading history. These accurate prediction results provide the evidence that the models,
the composite loading test method, and the state variable approach can capture the
permanent deformation behavior of asphalt mixtures.
8. Comparison between the Functionalized Model and the Shift Model
Two models, the functionalized model and the shift model, are proposed and evaluated
in this paper as permanent deformation models that capture the effects of load time,
deviatoric stress, and temperature on the permanent strain evolution in asphalt concrete.
The two models are verified using TRLPD tests, the composite loading test, and rando m
load tests. Predictions of the models are accurate for all the loading conditions.
However, the two models employ different calibration processes and overall test
requirements. These features of the models are summarized in Table 4.
First, the origin of both models can be traced to the incremental model form. In the
functionalized model, the coefficients, A, B, and C, are expressed as functions of reduced
load time and/or deviatoric stress. In theory, the functionalized model can be calibrated
using only one composite test. In practice, however, convergence issues make such
calibration impossible, and so, the model requires a full suite of TRLPD tests for
characterization. In contrast, the shift model demands only two types of tests: the
composite test and a single condition TRLPD test. The TRLPD test is used to obtain the
reference curve for fitting the incremental model. Thus, the shift model is simple and
easy to understand and implement. By separating the load time effect and the stress
effect on permanent strain growth, the shift model provides information about the
behavior of hot mix asphalt as a function of load time and deviatoric stress. The shift
model also provides a more accurate estimation of strains than the functionalized model
that is calibrated using the composite test only. In summary, the shift model is preferred
over the functionalized model for describing the permanent deformation behavior of
asphalt concrete and for its stability and accuracy.
644
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Table 4. Comparison of two suggested models.
Functionalized Model Shift Model
(A + B N)
Incremental model: =
vp
(C + N )
Basic
Model Each coefficient is a function of One reference curve: Incremental
load time and/or deviatoric stress model fitting using a TRLPD test
Slope, is constant (material Load time and stress shift function
property).
10 parameters (including ) 10 parameters (including A, B, C and
=
( A + B) , ):
vp
(C + N )
( A + B N red )
=
vp ,
a (C + N red )
A a1 p p 2 p da1d ,
=
M o d el c N red= N 10atotal ,
Eq . = c1 p p2 p dc1d ,
C
ato= a p + a d ,
B
= (b 1p
b
)
p 2 p exp b3d d ( ( b1d p +b2 d )
). tal
a p a1 p a2 + a3 ,
=
=a d b1 ( d / Pa )b2 + b3 .
Possible to calibrate with one Simple, easy to understand and

composite load test composed of 3 implement
P ro s x 3 loading conditions Inexpensive in terms of computing
cost
Stable and fast
Complex model form Two testing types required
Iteration is required to calibrate the Constant stress shift factor (or
model function)
Cons Possible convergence issues
High sensitivity
High computing cost for
characterization
645
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

CHOI, KIM
9. Conclusions
Temperature, load time, and stress are the three major factors that affect the permanent
deformation of asphalt concrete. To evaluate the permanent deformation behavior of two
asphalt mixtures (FHWA-ALF and NY9.5B) in this study, TRLPD tests are performed
by varying these conditions. The incremental model is used for both detailed analysis of
the experimental results and for developing models that account for the effects of these
variables. One constant slope (i.e., constant ) is observed between the permanent strain
and number of loading cycles and provides the basis for developing two models: the
functionalized model and the shift model. The shift model is based on the t-TS and t-SS
principles. Therefore, reduced load time is used instead of temperature and physical load
time, which reduces the effort needed for modeling and testing. The summary of the
findings from this study is as follows.
The composite loading test is proposed and verified to reduce the number of tests
needed for model calibration, thereby avoiding the need to conduct a series of
TRLPD tests.
Both the functionalized and shift models that are characterized using the composite
test can accurately predict the permanent strain evolution in TRLPD loading,
composite loading, and random loading histories. The results prove the applicability
of the models, the composite testing, and the permanent strain-based state variable
in predicting the rutting performance of the pavement structures.
The shift model is selected over the functionalized model as the preferable
permanent strain model because it is stable, accurate, and practically applicable.
10. Acknowledgements
This research is sponsored by the Federal Highway Administration under project No.
DTFH61-08-H-00005, Hot Mix Asphalt Performance-Related Specifications Based on
Viscoelastoplastic Continuum Damage Models. The authors gratefully acknowledge the
support of the FHWA.
11. References
Akinary A. E. and Brostow W., Long-Term Service Performance of Polymeric

Materials from Short-Term Tests: Prediction of the Stress Shift Factor from a
Minimum of Data, Polymer Engineering and Science, Vol. 42, No. 10, pp. 4527
4532, 2001.
646
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Akinary A. E., Brostow W., and Maksimov R., Prediction of Long-Term Service
Performance of Polymeric Materials from Short-Term Tests: Creep and Prediction
of the Stress Shift Factor of a Longitudinal Polymer Liquid Crystal, Polymer
Engineering and Science, Vol. 41, No.6, pp. 977981, 2001.
Azari H. and A. Mohseni., Incremental Repeated Load Permanent Deformation Testing
of Asphalt Mixtures, Paper #12-4381. Proceedings of the 2012 Transportation
Research Board Annual Meeting, Transportation Research Board of the National
Academies, Washington, D.C., 2012.
Brostow W., Time-stress Correspondence in Viscoelastic Materials: an Equation for the
Stress and Temperature Shift Factor, Materials Research Innovations, Vol. 3, No.
6, pp. 347351, 2000.
Chehab G., Kim Y. R., Schapery R. A., Witczack M., and Bonaquist R., Time-
Temperature Superposition Principle for Asphalt Concrete Mixtures with Growing
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Damage in Tension State, Journal of the Association of Asphalt Paving
Technologists, Vol. 71, pp. 559593, 2002.
Choi Y., Subramanian V., Guddati M., and Kim Y. R., An Incremental Model for the
Prediction of Permanent Deformation of Asphalt Concrete in Compression,
In press.
Darabi, M. K., Al-Rub R. K., Masad E. A., Huang C., and Little D. N., A Modified
Viscoplastic Model to Predict the Permanent Deformation of Asphaltic Materials
under Cyclic-Compression Loading at High Temperatures, International Journal of
Plasticity, Vol. 35, pp. 100134, 2012.
Desai, C. S., Somasundaram S., and Frantziskonis G., A Hierarchical Approach for
Constitutive Modeling of Geologic Materials, International Journal for Numerical
and Analytical Methods, Vol. 10, pp. 225257, 1986.
Haj-Ali R. M. and Mulianan A. H., Numerical Finite Element Formulation of the
Schapery Non-linear Viscoelastic Material Model," International Journal for
Numerical Methods in Engineering, Vol. 59, No. 1, pp. 2545, 2004.
Huang C., Masad E. A., Muliana A. H., and Bahia H., Nonlinear Viscoelastic Analysis
of Asphalt Mixes Subjected to Shear Loading, Mechanics of Time-Dependent
Materials, Vol. 11, No. 2, pp. 91110, 2007.
Gibson, N. H., A Viscoelastoplastic Continuum Damage Model for the Compressive
Behavior of Asphalt Concrete, Ph.D. Dissertation, University of Maryland, 2006.
647

CHOI, KIM
Jazouli, S., Wenbo L., Bremand F., and Vu-Khanh T., Application of Time-Stress
Equivalence to Nonlinear Creep of Polycarbonate, Polymer Testing, Vol. 24, No. 4,
pp. 463467, 2005.
Kim, Y. R. and Guddati M., Hot Mix Asphalt Performance-Related Specifications
Based on Viscoelastoplastic Continuum Damage (VEPCD) Models, Quarterly
Research Progress Report: January ~ April 2011, Project Number DTFH61-08-H-
00005.
Mahmoud F. F., El-Shafei A. G., Abdelrahman A. A., and Attia M. A., A Numerical
Solution for Contact Problem with Finite Deformation in Nonlinear Schapery
Viscoelastic Solids, Ain Shams Engineering Journal, Vol. 3, No. 2, pp. 141151,
2012.
Saadeh S., Masad E. A., and Little D. N., Characterization of Asphalt Mix Response
under Repeated Loading Using Anisotropic Nonlinear Viscoelastic-Viscoplastic
Model, Journal of Materials in Civil Engineering, Vol. 19, No. 10, pp. 912924,
2007.
Schapery, R. A., On the Characterization of Nonlinear Viscoelastic Materials, Polymer
Engineering and Science, Vol. 9, No. 4, pp. 295310, 1969.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Schwartz, C. W., Gibson, N., and Schapery, R. A., Time-temperature superposition for
asphalt concrete at large compressive strains, Transportation Research Record:
Journal of the Transportation Research Board, No. 1789, Transportation Research
Board of the National Academies, Washington, D.C., pp. 101112, 2002.
Subramanian, V., A Viscoplastic Model with Rate-Dependent Hardening for Asphalt
Concrete in Compression, Ph.D. Dissertation, North Carolina State University,
Raleigh, NC, 2011.
Von Quintus, H. L., Mallela J., Bonaquist R., Schwartz C. W., and Carvalho R. L.,
Calibration of Rutting Models for Structural and Mix Design, NCHRP Report
719, 2012, Washington, D.C.
Wenbo, L., Ting-Qing Y., and Aunli A., Time-Temperature-Stress Equivalence and Its
Application to Nonlinear Viscoelastic Materials, Acta Mechanica Solida Sinica,
Vol. 14 No. 3, pp. 195199, 2001.
Yun, T., Development of a Viscoplastic Constitutive Model Using a Rate-Dependent
Yield Criterion for HMA in Compression, Ph.D. Dissertation, North Carolina State
University, Raleigh, NC, 2008.
Yun, T., Underwood, B., and Kim, Y., Time-Temperature Superposition for HMA with
Growing Damage and Permanent Strain in Confined Tension and Compression,
Journal of Materials in Civil Engineering, Vol. 22, No.5, p. 415422, 2010.
648

Yun, T., and Kim R. Y., Experimental Investigation of Rate-Dependent Hardening-

Softening Behavior of Hot Mix Asphalt in Compression, Road Materials and
Pavement Design, Vol. 12 No.1, p. 99114, 2011.
Zhao Y., Liu H., and Liu W., Characterization of Linear Viscoelastic Properties of
Asphalt Concrete Subjected to Confining Pressure, Mechanics of Time-Dependent
Materials, 2012
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
649

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

_______________________________________________________________
AAPT Symposium:
An Industry Perspective on Sustainable
Asphalt Technologies
Howard Marks*
National Asphalt Pavement Association, Lanham, MD 20706-4407

________________________________________________________________
ABSTRACT: The presentation examines the factors that have led to the increased interest in the
sustainability of paving products and specifically the sustainability of asphalt pavements. Four
main sustainability factors have an impact on the pavement types that are selected:
Greenhouse Gases and Climate Change
Green Construction Mandates
Pavement Economics
Environmental Life Cycle Assessment (LCA)
The National Asphalt Pavement Association focuses on sustainability demands, specifically

recycled materials (RAP, RAS, and tire rubber), sustainable technologies (WMA, porous asphalt,
alternative binders), and green metrics (rating systems, LCA).
The presentation also examined the impact of pavements on fuel use. Although research
conducted by MIT advocates that flexible pavements might require more vehicle energy, most
other researchers recognize that MITs modeling effort is too simplistic. A number of independent
studies now confirm that the characteristic of smoothness accounts for the majority of pavement
impacts on vehicle fuel economy and that the smoother the pavement, the greater the fuel savings
for the driving public.
Finally, the presentation examined the issue of pavement reflectivity and the urban heat island
effect. Original research by Lawrence Berkeley National Lab was developed for buildings and
modeled for pavements. Subsequent research has contradicted some of the original conclusions.
______________________________________________________________________________
651
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
1. Discussion
PROFESSOR GABRIELE TEBALDI: You mentioned that warm-mix asphalt is a

sustainable technique, but you did not mention the cold technique, like foam asphalt or
recycling with emulsion, that allows you to save a lot of energy because you dont need
to dry the aggregate.
DR. HOWARD MARKS: I agree. I think thats a sustainable technology, no question,
but my membership focuses more on the hot side of things and thats where my focus
has been.
DR. MARKS: Thank you.
DR. JAN VAN DER ZWAN: Thank you for your fine presentation. Its good to see that
the United States and Europe are somewhere on the same track because you confirmed
in your speech what I saw yesterday. I would like to have two comments on your
presentation. One is that you refer to Japan for percentages to 50%. Yesterday, I
mentioned already that in the Netherlands but also in Germany, it is common practice to
have this kind of mixtures. In the Netherlands, we on the average have 50% RAP in a
base course material. And it has to do that at production, of course less than the United
States; it is about eight million total [tons] of hot mix and we produce four million total
[tons] of RAP, and we recycle that also. The average percentage in a base course
material is already for years 50%. We are at this time developing and have already
introduced special equipment, special plants, where you can go up to 100%.
The second remark I want to make is porous asphalts. Like you may know, we have used
porous asphalt in the Netherlands already for 30 years and a total main network is
covered with porous asphalt. I was a little bit surprised about the figures that you gave
because they are a little bit contradictory to what we have in the Netherlands. We found
out that porous asphalts have the same skid resistance approximately as our normal
surface course. [Porous asphalt] is being used especially for noise reduction. But the
most astonishing fact that we had is that porous asphalt was not safer than other
mixtures. So it surprising for me to find out that in the United States it is safer.
DR. MARKS: On the safety part, I think thats more in wet conditions. I think there have
been some studies that have shown that in wet conditions where there are potential skid
or whatever, that a porous friction course is superior or helps with safety.
DR. VAN DER ZWAN: Well, it was developed for safety and we found out later on that
its not safer because in wet conditions, people are compensating for the better visibility.
Theyre driving faster and are driving fairly close to each other. So what we expected
was that it would be safer, but in practice by the compensation of the driver, its not
safer. It may be something to keep in mind.
652
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Sustainable Asphalt Technologies
DR. MARKS. All right. Thank you.

DR. VAN DER ZWAN: Thank you.
PROFESSOR CARL MONISMITH: Howard, Id just like to call your attention to a
very definitive study on the heat reflectance at UC-Davis, and it really provides very
definitive data for some of the things that you have talked about. I think there has been a
paper published in the TRB, but the young man thats finished this just completed his
doctoral dissertation very recently. It was done in part by an actual study with
measurements of heat right in the pavement and about 6 ft above the pavement. So I
think you would find it very interesting and very beneficial.
DR. MARKS: Thank you. I have looked at that. I agree. Thank you.
PROFESSOR IMAD AL-QADI: Thanks, Howard. Very comprehensive presentation. I
have a couple of points. The first one is I noticed that your focus was on a reactive
approach. For example, the response to the recent study on concrete is not to perform a
study to respond to it. What is needed is thinking out of the box. For example, if you are
looking at the rolling resistance, why should it be better for flexible pavement? I think,
you said the MIT study was simplistic or funded by the industry. I dont think thats
maybe what the focus should be. The focus should be on the mechanism and real forces
and factors that are affecting the rolling resistance. We need to simulate actual
conditions including deformable tire over deformable pavement and how that affects
rolling resistance. As you mentioned, the user phase of LCA is really important. Some of
the work that were doing right now for the toll way in Illinois shows that although
construction emission is greater for asphalt concrete, materials emission for portland
cement production can be high. So, I think we need to look at it in a comprehensive way.
Your thoughts?
DR. MARKS: I would say that Im a big proponent of looking at it in a comprehensive
way, and I think that this Michigan Just by a show of hands, whos even [tried]
Michigan Techs lifecycle assessment? I mean, its pretty phenomenal. And I think
thats where agencies are going to start to look. If they can have one agreed upon system
to look at these things instead of just cherry picking certain results and certain things that
other types of institutions may be doing. I think a holistic approach is much better. I
agree with that, and I think we need to be a little bit more proactive. I would be more
than willing to chat with you after this to see if we can put our heads together on that.
DR. REBECCA MCDANIEL: I just wanted to comment on the porous friction courses.
The example you showed was from San Antonio and it was indicated that there could be
issues in more northerly climates. We did a study for the Indiana DOT and followed the
performance of a porous friction course on an interstate in Indiana for five years after
construction. We did see continued noise benefits over that time frame. At the end of
five years, it wasnt quite as quiet as it had been. I think we lost one or two decibels, but
it was still significantly quieter than the adjacent SMA. We did see higher friction
653
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
values on that porous friction course also. So I think the key is that it was an interstate,
with fast moving traffic. Also, it was an urban area, so we didnt use abrasives for snow
and ice control that could clog the pores. This project shows porous friction courses can
work in colder climates, as well.
DR. MARKS: Thank you.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
654

_______________________________________________________________
AAPT Symposium:
Alternative Binders for the Production of
Bituminous Materials
Elham Fini *
North Carolina A&T State University, Greensboro, NC 27411-0002

________________________________________________________________
ABSTRACT: The presentation began by discussing the need for alternative binders as a result of
increasing asphalt demand, prices, and coking technology. The motivation for alternative binders
is to :
enhance binder properties;
reduce dependence on crude oil;
improve the use of underutilized resources to supplement petroleum-based asphalt;
produce more environmentally friendly binders (reduced VOC); and
produce colored asphalt.
A review of the research that has been conducted within the past ten years on alternative binders
(bio-binders) indicates the following.
The physio-chemical properties of the bio-binders can vary significantly based on the
feedstock and processing methods.
Bio-binders derived from lignocellulosic bio-mass have higher viscosity range than bio-
binders derived from bio-mass with high lipid content.
Bio-binders show significantly lower temperature susceptibility than petroleum-based
binders, regardless of feedstock source.
The temperature range for viscous behavior is usually lower for bio-binders than
petroleum-based (bitumen) binders.
The high temperature performance grade for bio-binders is typically close to those of
bitumen binders.
The low temperature performance grade could vary significantly from bituminous
binders depending on the feedstock type and processing method.
Susceptibility of bio-binder to U.V. and oxidative aging varies based on their feedstock
source and processing methods. The storage stability of bio-binders could be improved
if lighter compounds (aldehydes and ketones) are removed.
The presentation concluded with questions yet to be addressed with bio-binders.

______________________________________________________________________________
The oral presentation was made by Dr. Geoffrey Rowe.
655
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
1. Discussion
DR. JOHN DANGELO: I applaud the work people are doing with biobinders, but lets
be realistic about what were talking about here. At this point, and I think for decades to
come, the ability to produce that material in any usable quantities will be very limited,
which is what you have to look at as a contractor. If hes going to replace 10% or 15 or
20% asphalt with biobinder, he has to do it on a regular basis. He cant do it, Well,
today Ive got some materials, so Ill replace it. Tomorrow, I dont have it, so I have to
go back. It has to be a consistent source, and right now they havent been able to
produce biobinders on a consistent basis. Are you going to tell me Im wrong? And then
the other major thing, and I think Ellie had it in her slides, is that right now, these
materials really arent that compatible. There is a significant phase separation problem
with them and so, the question is, with these very light oils, will they continue to
separate even at room temperature or in the roadway as opposed to just at elevated
temperatures?
DR. GEOFFREY ROWE: Thanks, John. Im going to let Ellie respond to that. I kno w
some folks have more consistent supplies. Maybe I should let Chris make a comment.
DR. ELHAM FINI: Reliability of the feedstock and the scalability of such business
model is a main drawback on many biobinder. As discussed during the presentation
there have been many bio-based binder which could not scale due to either technological
limitation for mass production or lack of a convincing energy balance in the process. In
addition, talking about alternatives we should not settle for lower quality or simply a
partial replacement to reduce the overall cost of tonnage of asphalt. Since we are
synthesizing the alternative bio-based binder, we could design materials with enhanced
properties.
PROFESSOR CHRISTOPHER WILLIAMS: I think Johns comments are fairly
reasonable. I think we have to step back a little bit even further and realize that if were
going to have a business model thats solely focused in on producing bioasphalt, that
model wont work. Weve seen this in many other industries, that this has just got to be
one component of a better, larger biorefinery complex, and this is one component of the
biorefinery complex that provides economic opportunity and value in balancing that risk.
I think that stability issues that John mentioned are solved. I think we can solve those
issues. Ellie made some very good points here. Were going to be looking at doing
demonstration projects that are, you know, large tonnages this construction season with
10 and 20% replacement of asphalt.
DR. ROWE: Thanks, Chris.
MR. CHRISTOPHER DEDENE: Ive had the pleasure of working with a couple of
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
these different biobinders, and one thing that keeps coming to mind as I continue to
encounter these things, is the biobinder retains an odor of whatever you started with, and
656

I think thats a social aspect that were going to have to address down the road. So the
wood one actually smells like barbecue once its been processed and the swine waste one
doesnt get any better.
DR. ROWE: I have not worked with any of these products, so I certainly havent got
close to their smells.
MR. DEDENE: Just things to think about as we go down the road.
PROF. WILLIAMS: I can answer that. Its not hard. You simply put in a reaction cell,
put a carbon filter in there, take it out, put in a GC mass spec, figure out whats coming
off, and you can screen those materials out. This would take the smell out. Thats not a
problem. I will say, if you're looking at this in a larger biorefinery complex. In fact, you
get carbon out, you can put it in the soils sequestering carbon, grow plants, take CO2 out
of the air, and we have a carbon negative situation here. So, in fact, it goes quite
positively on the green side of things.
DR. ROWE: Thanks for that comment, Chris.
PROFESSOR HAIFANG WEN: I think we have to think about a few things. Number
one, it would not be the same as traditional petroleum-based asphalt. We have to think
that way. Also, one single bioasphalt may not be sufficient. We may have to have
bioasphalt from different sources and collectively have enough for the future. So this
will be a long-term effort to develop the biobinder or bioasphalt. I think one important
thing, you may have many technologies as you have seen here that may be adding to
asphalt. But to make that work, we have to demonstrate what the cost is. Most likely, it
has to be some secondary or waste material, just like asphalt, which is residue from
crude oil, to make that work cost wise. Also the greenhouse gas emission and energy
have to be put on the table, not just look at performance. We have to look at cost,
energy, and greenhouse gas emission. We have to have those numbers on the table
before we can do something.
DR. FINI: That is definitely a great comment and that is mainly why not all of our
scientific solutions and great papers lead to business model and commercial products.
That is also a question which needs to be answered if a new development including
biobinder is going to be referred to as sustainable. Actually, without considering carbon
exchange and/ or states clean energy mandates, several existing product and business
models would not be able to survive on their own. However, we have to start
somewhere and I think starting with product performance and technical features would
be a good start.
DR. ROWE: I think those are good points, and I dont think most folks are looking at
these as 100% replacement binders. Theyre looking at using them in combination. Lets
have Mike and then Ludo, and then well perhaps stop the conversation there.
657
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
DR. MICHAEL SOUTHERN: I know Im not directing any of this at you, and Im not
even going to mention biobinders more than once or twice during what I say. But in this
universe, were governed by the laws of thermodynamics, and theyre not guidelines,
theyre laws, so you cant break them. Put simply, you cant win thats the first law.
One cant break even thats the second law. And you cant get out of the game thats
the third law. So were stuck. And anything we do will be less than 100% efficient. And
what we need to do is whats best for the planet. Now the title of this session is
Sustainability and Environment. And I saw very little in this presentation which actually
gave any hard information. Im going to throw out two little phrases which you need to
think about every time somebody says green: system boundaries and allocation
methodology where you start and where you finish counting your carbon and how you
chop it up and how you allocate when you do it. When were talking about swine
manure, you grow a pig, it produces manure. Now how you allocate the emissions
associated with growing that pig will determine where you put your CO2 and how much
you allocate. If you take that swine manure, you collect it, you store it, you move it, you
do pyrolysis, you distill it, you do all of these things. You cant win, you cant break
even, and you cant get out of the game. Theres energy involved at every step of the
way, and its really only how you chop that up. If you choose to call something a waste,
then you can make some mischief with these allocations. And there are people out there
who are making mischief. Now Jan showed yesterday that weve quantified for bitumen,
weve quantified it for asphalt 60 g/kg for asphalt, 1600 g/kg for orange juice, and 6 kg
for a burger. Just because it is bio doesnt mean that its green. And so just bear in mind
system boundaries and allocation methodology because if you have to spend $10 to get
$5, even in energy terms or emission terms, its not green.
DR. ROWE: Thanks, Mike. Im going to let Ellie address that.
DR. FINI: Our energy balance evaluation shows overall the energy balance is positive.
Michael has a good point on when we can count something waste and accordingly stop
counting the associated carbon for it. Some of the feedstock used to produce bio-binder
definitely could fall into such vague definition of waste, which in those cases we should
be cautious referring to them as sustainable or even green. However, when it comes to
swine manure, just by considering the level of methane generated by decay of manure in
lagoons you would see the green aspect of the bio-binder production. The estimated 60%
carbon sequestration through this process is quite significant in lieu of current manure
storage and application. Sustainability of such bio-binder would be further reflected in
its improved low temperature cracking resistance which would have significant
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
economic and social impact when considering the cost and traffic congestion associated
with pavement cracking and rehabilitation.
PROFESSOR LUDO ZANZOTTO: I see that Im the shortest one. I just want to say,
this whole thing has two aspects. One is not that serious, thats the technical. The other is
quite serious, and its philosophical. Technically, I remember my father telling me that
658

in the 30s there were always predictions that in a couple years the world will run out of
the crude oil and so on. Even ten years ago I had some presentation in Australia where
they were afraid we are peaking, etc. Right now, Alberta, with the third largest amount
of crude oil, is afraid that we wont have the customers for it because its so much.
Thats much about technical. Now all these things are nice. It will be 1.5% if everything
works and its fine and its commendable. But one thing which Im really afraid of and
where I see the trends going, and I had several proposals for European commissions and
request for funding. And we heard our European colleagues yesterday about the amount
of bureaucracy and I would say politically correct bureaucracy and so on. I am afraid this
is creeping here as well. When you want to get the money for the research and so on, it
doesnt matter that much what technical things you are going to do. What really matters
is that you have this politically correct garbage on And I am not Every engineer has
a duty to be as clean, I mean not personally, but in his work as possible. But I am afraid
that this will become for us the time as it became in Europe lets say 20 years ago that
first you will have to be terribly politically correct and whether it has the technical sense
is totally secondary. If we already want somehow to replace asphalt, I would suggest its
better to channel the research money into the development of zeppelins and go away
from ground transportation and try to put it up. That would be at least thinking out of the
box. Not this. So nice as it is, but does it not become a trend.
DR. ROWE: Thanks, Ludo. I wish Ellie was here to answer some of these questions
because Im certain she would do a lot better job than me. So Im just really the fill-in
and didnt want to exclude this from the session, but thank you all for your comments.
Its very much appreciated.
DR. FINI: The question here is not really whether we will have crude oil in years to
come. The reality is most refineries are further converting their supply of asphalt to
synthetic fuel as there is high demand for synthetic fuel. And this demand will only
continue to go up. As such the supply of asphalt has been shrinking which has been the
main driving force for the increase of asphalt price. Actually the correlation between the
price of asphalt and that of crude oil has been decreasing over the last five years. As
such there is an urgent need for alternative asphalt resources regardless of how long the
crude oil reserves will last. I believe thinking out of the box here would be to place the
advancement in science in the context of market need. And no one knows market need
as much those actually in the industry. We talked to 122 asphalt users, distributors and
contractors and they all commented the need is to have a reliable consistent source which
could go with a blanket recipe for various job site and yet comparable in price. Many of
them were concerned about the increase of the price, low quality and monopoly in
market. I believe bio-binder synthesis could have a significant merit in addressing such
industry pain while opening a new paradigm in science of asphalt pavement design and
application and even further enhancing its sustainability features.
659
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

AAPT Symposium:
Recycling with RAP, RAS and Secondary
Aggregates
Jo Sias Daniel*
*University of New Hampshire, Durham, NH, 03824
ABSTRACT: As the asphalt paving industry continues to move towards the use of more sustainable
asphalt technologies, it is important to understand the impacts that these materials have on the
construction and performance of asphalt pavements as compared with mixtures containing virgin
materials. This paper is intended to be an introductory overview on the use of RAP, RAS, and
secondary aggregates. Basic definitions are provided along with potential issues that may be
encountered when using these materials. The paper focuses on the use of these materials in the
asphalt concrete layers, but they may be used in other layers of the pavement structure as well.
The paper concludes by providing some resources for further information.
KEYWORDS: RAP, RAS, secondary aggregates
1. Introduction
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
This paper is a summary of a presentation given at the Technical Symposium on

Sustainable Asphalt Technologies during the Association of Asphalt Paving
Technologists Annual Meeting on April 9, 2013. This is intended to be an introductory
overview of the use of Reclaimed Asphalt Pavement (RAP), Recycled Asphalt Shingles
(RAS), and Secondary Aggregates. The paper provides some basic definitions, potential
issues to be aware of when using these materials, and some information on current
research and where to find additional information. The focus is on use of these materials
in the asphalt concrete layers, although many of these materials can be used in other
layers of the pavement structure as well.
The paper is organized similarly to the presentation where definitions are presented
followed by a more detailed discussion on each type of material, and finally suggested
resources for further information are presented.
661

SYMPOSIUM
2. Definitions
2.1 RAP: Reclaimed Asphalt Pavement
RAP is generated from pavement millings, full depth reclamation of pavements, and also
from plant waste. These materials typically contain approximately 5% asphalt binder
and 95% aggregate by weight, although the exact values will vary depending on the
proportions of the source materials. These recovered materials are processed at a plant
and stockpiled for use in future mixtures. Processing generally includes some kind of
crushing operation to reduce the agglomerated RAP particles to smaller sizes and sieving
to either scalp oversize material or separate the RAP into different size stockpiles.
High RAP is currently defined as a level of 25% RAP or more, by total weight, in a
mixture (Copeland, 2011).
2.2 RAS: Recycled Asphalt Shingles
RAS is generated either from manufacturer waste or from shingles that are removed
from roofs (end of service or storm damage). Shingles contain asphalt, mineral fillers,
fine aggregate (sand) particles and fibers. They are processed for inclusion in ne w
asphalt mixtures; this typically includes removing debris (wood, nails) from post-
consumer shingles and then grinding the shingles to a specific size. There are two
different types of shingles: those made with fiberglass mat and those with organic felt.
The relative proportions of the component materials vary with the different type of
shingle and typical values are shown in Table 1 below.
Table 1. Typical component percentages by weight for RAS.

Component Organic Felt Fiberglass Mat
Asphalt cement 30-36% 19-22%
Felt (fiber) 2-15% 2-15%
Mineral aggregate (#30) 20-38% 20-38%
Mineral filler/stabilizer 8-40% 8-40%
662
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

2.3 Secondary Aggregates
Secondary aggregates are produced as a by-product of other mining or industrial

processes. Examples of secondary aggregates include: blast furnace slag, steel slag, and
foundry sand. These materials are typically crushed and sized for inclusion in new
asphalt mixtures.
2.4 Recycled Aggregates
Recycled aggregates are produced by recycling construction and demolition waste.

Examples of recycled aggregates include: crushed concrete, bricks, and glass. These
materials are typically crushed and sized for inclusion in new asphalt mixtures.
3. Reclaimed Asphalt Pavement
3.1 Defining RAP Percentage
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Historically, the amount of RAP in a mixture has been defined as the percentage by
weight of aggregate or by weight of the total mix. This is a very simple and
straightforward measurement, but it does not account for the potential differences in the
asphalt contents of different RAP stockpiles and mixture designs. This is important
because the primary issue with higher RAP contents in asphalt mixes is the amount of
binder replacement available since the use of RAP can reduce the need for virgin binder
and impact the binder properties. Thus, it is becoming more common to define RAP
contents by the amount or percentage of recycled binder in the mixture. The percentage
of recycled binder in a mixture is calculated as:
% recycled binder = (AB)/C [1 ]
where: A = % binder content of the RAP
B = % RAP by total weight in mixture
C = total % binder content in mixture (from mix design)
For example: a mixture that has a total asphalt content of 5% and includes 20% RAP (by
total weight) of a RAP stockpile that has a 6% asphalt content would have a recycled
663

SYMPOSIUM
binder percentage of 24% (620/5 = 24). If that mixture were made with a RAP
stockpile that had an asphalt content of 4%, then the recycled binder percentage would
be 16% (420/5=16).
3.2 Current Usage
The amount of RAP that is used in an average mixture by states in the US has been
steadily increasing over the last several years, according to surveys conducted by the
National Asphalt Pavement Association (NAPA). Figure 1 shows the average
percentage (by total weight) in all mixtures for the years 2009-2011 (most recent
released NAPA survey numbers). Figure 2 shows the information from the 2011 NAPA
survey by state. Across the nation, states are comfortable using up to 20% RAP, but
there is potential to use up to 30% or more.
Figure 1. Average percent RAP in all mixtures in US (Source: NAPA Survey).
664
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 2. Average RAP content by state in 2011 (Source: NAPA Survey).
3.3 Considerations when Using RAP
There are several issues that must be considered when using RAP in mixtures. These
include:
Processing and Quality Control (QC)
Characterizing RAP
Mix Design
Selecting the Virgin Binder Grade
Blending/Co-mingling of Virgin and RAP Binder
Specific gravity of RAP Aggregate
Performance
3.3.1 Processing and Quality Control

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
RAP mixtures must meet the same mix design specifications required for virgin
mixtures. Producing quality asphalt mixtures containing RAP begins with the
acquisition, processing, and quality control of the RAP itself. Milling procedures should
be managed to complement or reduce the amount of processing and increase the
consistency of the RAP stockpiles. Sizing of the RAP materials is important to allow for
better control of the mixtures and controlling the amount of dust in a mixture. A
665

SYMPOSIUM
technique called fractionation is used to manage and maintain better control, especially
for mixtures with higher percentages of RAP. Fractionation is processing and separating
RAP into at least two sizes, typically a coarse fraction and a fine fraction.
3.3.2 Characterizing RAP
Just as virgin aggregate stockpiles and virgin asphalt are characterized for inclusion in
mixtures, RAP materials should be characterized as well. The characterization of RAP
stockpiles should include the following:
A. Asphalt Content: The RAP contains asphalt cement that contributes to the total
asphalt content of the mixture, reducing the amount of virgin binder required. There
is currently debate about how much of the asphalt binder in the RAP becomes active
in a mixture and various agencies are in the process of producing trial mixtures and
test sections to determine the appropriate changes, if any, to their specifications.
Asphalt content of the RAP stockpiles can be determined either by ignition oven or
solvent extraction.
B. Performance Grade of RAP binder: The RAP binder has been oxidized and aged in
the field and therefore is typically stiffer and more brittle than the virgin asphalt
binder. PG grading of the RAP binder is used to evaluate the stiffness of the RAP,
as well as to estimate properties of the final mixture that contains some blend of the
virgin and RAP binders. The amount of blending that occurs between the virgin and
RAP binder is the topic of current research. When RAP binder is extracted and
recovered, it is considered an unaged material because it will be processed through a
plant again as it is incorporated into a new mixture.
C. RAP Aggregate Properties: The gradation of the aggregate in the RAP is important
to know for proper control of the overall mixture gradation. The combined
aggregate properties for a mixture, including the RAP aggregate, must meet
Superpave consensus requirements to be acceptable for use. The bulk specific
gravity (Gsb) of the RAP aggregate must also be determined for use in calculations
of the VMA of the final mixture. Errors in determining the Gsb of the RAP
aggregate can have a significant effect on VMA calculation, and therefore may
significantly affect the final design asphalt content of the mixture. These errors
become more significant with higher amounts of RAP.
D. Moisture Content: The moisture content of the RAP stockpile is important for
efficient and effective plant operations and producing a consistent, quality mixture. --`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
666

3.3.3 Selecting the Virgin Binder Grade

Current guidance, research, and experience suggest that for RAP contents below 15%,
no adjustments to the virgin binder grade need to be made. As RAP content increases,
particularly as measured by amount of binder replacement, the inclusion of the aged
RAP binder in the mixture may significantly impact the properties of the overall
mixture. This will depend on the amount of blending that actually occurs between the
RAP binder and the virgin binder. The blending that occurs will be a function of the
properties of the virgin and RAP binders and will likely be impacted by plant production
parameters such as burner temperatures, mixing times, and storage times as well. The
amount of blending that occurs for a particular mixture is difficult to evaluate; testing the
extracted and recovered mixture produces a full blend of the RAP and virgin binders
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
through the extraction process itself. Mixture testing may be performed to back
calculate the effective binder properties, but this is an indirect measurement.
To compensate for the stiffer RAP binder in a mixture, the virgin binder grade may
be adjusted, or bumped down. This may be done on both the high and low ends. The
intent is to produce a mixture that has an effective binder grade that meets the virgin
binder grade requirement for that particular project. Current guidance on binder grade
adjustments in AASHTO M 323, Standard Specification for Superpave Volumetric Mix
Design, is shown in Table 2 below.
Table 2. AASHTO M 323 guidance on binder grade bumping.
Recommended Virgin Asphalt Binder Grade Percent (%) RAP

No change in binder selection < 15
Select virgin binder grade one grade softer than normal 15 25
Follow recommendations from blending charts > 25
This guidance is based on the assumption of significant blending between virgin and
RAP binder, and does not include the effects of plant production. There are practical
difficulties with this approach, as softer virgin binders may be more expensive or hard to
get in regions of the country. Blending chart analysis is time-consuming and also
requires expensive and hazardous chemicals. These impediments can effectively limit
the amount of RAP that is used in mixtures. Some states have developed their own
specifications and methods based on experience with local materials that allow for
higher amounts of RAP without binder grade adjustments. However, there is still not
clear guidance on how to design mixtures at higher RAP levels.
667

SYMPOSIUM
3.3.4 Determining Gsb of RAP Aggregate

The Bulk Specific Gravity (Gsb) of the RAP aggregate is used in the calculation of VMA
for the overall mixture. Calculation of the VMA requires the combined Gsb of all the
aggregates in the mixture, so the values for both the virgin aggregates and RAP
aggregates are required. There are three alternatives for determining the Gsb of the RAP
material:
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
A. Construction Records: If the source of the RAP material is known (i.e. millings
from a particular project), construction records can be used to determine the Gsb of
the original aggregate. This is the most accurate method, but typically is not
applicable because most RAP piles contain material from various sources.
B. Direct measure: This method uses the combined specific gravity that is determined
from an aggregate sample obtained following a solvent extraction or ignition oven
procedure. The recovered aggregate is split into coarse and fine fractions and the
individual specific gravity tests are performed. The combined RAP aggregate
specific gravity is then calculated. Studies have shown that the bulk specific gravity
of the RAP aggregate recovered from the ignition oven is typically significantly
lower than that of the original aggregate. The aggregate obtained from a solvent
extraction process may contain residual asphalt or solvent which may influence the
results.
C. Estimation: This method uses a fairly simple and quick three step process:
1. Determine the maximum theoretical specific gravity (Gmm) of the RAP mixture
using AASHTO T 209 or ASTM D 20141
2. Use the asphalt content determined from either extraction or ignition oven and
an assumed asphalt specific gravity (Gb) to calculate the effective specific
gravity (Gse) of the RAP aggregate
3. Based on experience with mix designs for the specific location, assume a
typical value for asphalt absorption (Pba), and use this value to estimate the bulk
specific gravity (Gsb) of the RAP aggregate from the calculated Gse.
Higher levels of RAP will magnify any errors in the assumptions of Gb and Pba in
the estimation of Gsb using this method.
3.3.5 Dust to Binder Ratio
RAP stockpiles typically have high levels of dust from the milling and crushing
processes. A typical value for a full gradation stockpile of RAP is at least 10% dust
(passing No. 200 sieve). Most virgin aggregates also carry dust, so using higher
amounts of RAP can create challenges in terms of controlling the amount of dust in a
668

mix. The dust fills void spaces and reduces the VMA, especially in the field. With the
high levels of dust in the RAP stockpiles and the fact that the dust cannot be removed
from the RAP, additional measures must be taken to control the amount of dust in the
final mixture. These include additional processing of virgin materials washing, or dust
removal at the plant during production but there are limits to these approaches and they
add cost.
3.3.6 Placement of RAP Mixtures
Construction issues with RAP mixtures are not different from issues encountered when
paving with asphalt mixtures produced with virgin materials. However, failure to
properly address processing, QC of RAP, and mixture design will significantly increase
the likelihood of problems in placement and compaction of the new pavement. No
special equipment or techniques are required when placing and compacting mixtures
containing RAP, but mixtures containing RAP, especially those with higher levels of
RAP, may require more attention than conventional mixtures due to increased stiffness
as a result of the RAP. Achieving density with RAP mixes is typically not a concern as
long as the proper techniques are used and attention is paid to the process. RAP
mixtures are typically stiffer and/or may be produced at slightly lower production
temperatures, therefore resulting in reduced compaction time. Like conventional mixes,
compaction should be monitored using a non-destructive device, calibrated to cores, to
ensure adequate density is achieved.
3.3.7 Performance Testing
The primary performance concerns with mixtures containing RAP are cracking (fatigue
and thermal) and durability (raveling). These are due to the addition of the more brittle,
aged binder that is introduced into the mixture from the RAP. If softer virgin binders or
higher asphalt contents are used to offset the impact of the RAP binder, permanent
deformation may become a concern as well. Table 3 below lists some of the more
commonly used performance tests that are available for evaluating different distress
mechanisms in the laboratory.
669
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
Table 3. Available performance tests.
Distress Test Reference

Asphalt Pavement Analyzer (APA) AASHTO TP 63
Permanent
Hamburg Wheel Tracking Device (HWTD) AASHTO T 324
Deformation
Flow Number (AMPT) AASHTO TP 79
Moisture Tensile Strength Ratio (TSR) AASHTO T 283
Sensitivity Hamburg Wheel Tracking Device (wet) AASHTO T 324
Beam Fatigue AASHTO T 321
Fatigue S-VECD Fatigue -
-
Overlay Tester
Creep Compliance & Strength Using Indirect
AASHTO T 322
Thermal Tensile Test (IDT)
Cracking Thermal Stress Restrained Specimen Test Old AASHTO
(TSRST) TP 10-93
3.4 Performance of High RAP Mixtures
There have been many research projects that have investigated performance of high RAP
mixtures and documented the field performance of various mixtures. This section
highlights a few of these research studies, but is not a complete list. Readers are
encouraged to review all the available literature relating to this topic.
West et al. (2011) examined the Specific Pavement Studies 5 (SPS-5) in the Long
Term Pavement Performance (LTPP) database. These SPS-5 pavement studies consisted
of 50 mm and 125 mm overlays constructed from both virgin and 30 % RAP mixtures.
Surfaces were both milled and non-milled, the oldest pavement was over 17 years old
and the pavements were in a variety of climates. Figure 3 shows a summary of the
results for this project. This study found that the overlays with recycled mixtures
performed as well as, or significantly better than, overlays with virgin mixes for the
following distresses:
Fatigue Cracking 71% of the sections
Longitudinal Cracking 85 % of the sections
Block Cracking 97 % of the sections
Raveling 93 % of the sections
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
670

Fatigue Cr acking Longitudinal Cr acking
Virgin performed 15% Virgin performed

significantly better significantly better
29%
than RAP than RAP
10%
RAP performed RAP performed
than Virgin than Virgin
61%
Difference Difference
10% between Virgin between Virgin
and RAP and RAP
insignificant 75% insignificant
Block Cr acking Raveling
3% Virgin performed 7% Virgin performed

1% significantly better
significantly better
than RAP than RAP
15%
RAP performed RAP performed
than Virgin than Virgin
Difference Difference
between Virgin between Virgin
and RAP and RAP
96% insignificant
78%
insignificant
Figure 3. LTPP 5 study results (West et.al, 2011).
Anderson and Daniel (2013) collected data from transportation agencies across the
United States regarding the existence and long term performance of high RAP sections.
These sections were then paired with nearby virgin pavements of similar mix design and
age so that suitable case studies could be conducted. The performance data from these
sections was then compared visually, and statistically when appropriate, in order to
evaluate if the RAP sections performed at a similar level to that of the virgin sections
over time. Table 4 gives the summary of findings from all of the case studies.
In many of the case studies presented, high RAP sections have performed to a level
equivalent to that of a virgin section, and have outperformed virgin sections in several
cases. The high RAP sections tend to exhibit slightly higher degrees of cracking and
lower ride quality as defined by present serviceability rating (PSR). The RAP sections
also tend to exhibit a higher deterioration rate with respect to their ride quality. In some
cases, these differences were statistically significant, and in other cases they were not.
However, none of these distress factors were to a great enough severity to be detrimental
to the long term performance or expected life span of a RAP section (typically less than
5% difference from virgin sections) after approximately 10 years in service.
671
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
Table 4. Summary results of high RAP case studies (Anderson and Daniel, 2013).
Case Study Summary of Results
Washington,
Renslow to Comparable performance rating after 9 Years of data
Ryegrass
Lower ride quality and more cracking in 45% RAP section, More rutting in
Wyoming I-25
30% RAP than Virgin and 45% RAP, otherwise comparable performance
Lower Ride quality and present serviceability rating (PSR) in RAP sections,
Wyoming I-80
otherwise comparable performance
Lower ride quality and PSR and less Rutting in 20% RAP sections than 30%
Wyoming I-90
RAP and Virgin
Wyoming US-85 Virgin section out performs 30% RAP section in every rating except rutting
Wyoming Indices RAP Sections not statistically different from one another, only 20% RAP
Statistics sections PSR was statistically worse than virgin sections
Wyoming Slopes RAP sections deteriorate in ride quality significantly faster than virgin
Statistics sections. 30% RAP PSR deteriorates significantly faster than virgin section.
Woodstock-
Presence of RAP does not affect long term performance
Lincoln, NH
Florida Recycled
Projects
Decreasing performance with increased RAP content in 30-50% range, but all
RAP levels show better average performance than virgin pavements.
(Nash et. al,
2012)
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
672

Bennert and Maher (2013) conducted an in-depth study of the LTPP SPS 5 sections
in New Jersey. That study showed that cracks appeared at approximately the same time
for both virgin and 30% RAP sections, but the cracking progressed faster in RAP
sections. Summary data for alligator cracking is shown in Figure 4 for this study.
Figure 4. Alligator cracking for NJ LTPP SPS5 sections. (Bennert and Maher, 2013).
673
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
There is currently a pooled fund project underway to evaluate properties of plant

produced high RAP mixtures: TPF 5(230): Evaluation of Plant Produced High-
Percentage RAP Mixtures in the Northeast. University of New Hampshire, North
Carolina State University, Rutgers University, and University of Massachusetts at
Dartmouth are the research team. Seven states and FHWA have contributed to this
project that focuses on the fatigue and low temperature cracking of RAP mixtures. As
part of this project, the research team has been testing specimens that were compacted
immediately at the plant: plant mixed, plant compacted, (PMPC) and specimens
fabricated from loose mix that was reheated in the laboratory: plant mixed, laboratory
compacted (PMLC). Testing has included extraction and recovery of the binders for PG
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
grading, critical cracking temperature (CCT), G* master curve construction, and ABCD
testing. Mixture testing has included |E*|, S-VECD fatigue, overlay tester, beam fatigue,
low temperature IDT creep and strength, and TSRST. Preliminary data from this project
indicates that plant production variables such as mixing temperature and silo storage
time may have a significant impact on the mixture properties, as well as the differences
between the PMPC and PMLC specimens (Mogawer et al., 2012). Mixture variables
that have been investigated include RAP content and PG grade of the virgin binder.
Preliminary data also indicates that mixtures with higher total asphalt contents have
smaller differences between the RAP percentages.
4. Recycled Asphalt Shingles
4.1 Current Usage
The amount of RAS that is used by producers in the US has been steadily increasing
over the last several years, according to surveys conducted by NAPA. Figure 5 shows
the tonnage used for the years 2009-2011 (most recent released NAPA survey numbers).
Producers in 32 states report using RAS in mixtures. RAS has much higher asphalt
content than RAP, so the percentage of RAS typically used in asphalt mixtures by total
weight is on the order of 2-5%.
674

Figure 5. Tons of RAS used.
4.2 Considerations when Using RAS
Many of the issues that must be considered when using RAS are similar to those for
using RAP. These include:
Processing & Quality Control (QC)
Characterizing RAS
Mix Design
Changing Binder Grade
Blending/Co-mingling of Virgin and RAS Binder
Performance
The approach to addressing these issues is similar to the descriptions above for RAP
mixtures, with a few additional points related to the differences between RAP and RAS
materials. Only the additional issues are discussed below.
4.2.1 Processing and Quality Control

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Shingles should be ground so that 95-100% passes the 3/8 sieve for effective mixing
and inclusion in mixtures. The ground shingles can tend to clump together because of
the high asphalt content; blending the shingles with a sand stockpile can help mitigate
the clumping. Ground shingles will also retain higher moisture contents than RAP
stockpiles.
675

SYMPOSIUM
RAS material can be generated from both post-consumer (shingles that have been
on roofs and have been removed) and manufactured waste (during manufacture of new
shingles). These materials typically have different properties because of the aging that
occurs on a roof. In many cases the two sources are kept separate and may be used
separately or blended in specific proportions.
4.2.2 Characterizing RAS
RAS stockpiles must be characterized to determine the material properties, similar to

those measured on RAP stockpiles:
A. Asphalt Content and PG Grade: Shingles do not contain a typical paving grade
asphalt binder. It is very stiff, oxidized asphalt that typically exceeds the capacity of
equipment used to grade paving asphalts. The asphalt content in RAS is also much
higher than that found in RAP materials. Thus shingles are used in asphalt mixtures
in low percentages (less than 7%) due in part to challenges in grading the shingle
binder and determining the binder content.
B. Fine Aggregates: Shingles contain sand size particles that must be considered in
terms of the overall aggregate structure and consensus properties.
C. Fibers: Shingles contain fibers that will be incorporated into the asphalt mixture.
Typically these are not characterized or measured separately, but may impact the
overall performance of the mixture.
D. Moisture Content: Moisture content of the RAS stockpile must be known for
production considerations.
4.3 Performance of RAS Mixtures
Researchers, agencies, and contractors have investigated performance of RAS mixtures

and documented field performance of these various mixtures. This section summarizes a
pooled fund project that is nearing completion, but there are many other smaller projects
that have been completed or are under way. Readers are encouraged to review all the
available literature relating to this topic.
TPS-5(213): Performance of Recycled Shingles in HMA is being led by researchers
from Iowa State (Chris Williams, Andrew Cascione, Debra Haugen) and University of
Minnesota (Mihai Marastaneu) along with Jim McGraw from MnDOT. Eight states,
mostly in the upper Midwest region of the country, have participated in this project. The
project includes both laboratory testing and demonstration projects. Laboratory testing
that has been conducted on RAS mixtures includes PG grading of extracted and
676
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

recovered binders, and dynamic modulus, beam fatigue, flow number, and semi-circular
bend (SCB) testing of the mixtures. Several demonstration projects are also included in
the study. These are mostly overlays with a thickness of 1.5-3 and also include an
SMA project in Illinois. Based on the results obtained to date (Cascione et al., 2010 and
2011), the researchers have observed that the RAS mixtures typically have higher flow
numbers than the control mixtures (some virgin, some with RAP). The laboratory
measured cracking performance from the beam fatigue and SCB testing is inconclusive;
in some cases the RAS mixtures perform as well as or better than the control mixtures
and in some cases the RAS mixtures performed worse. Field performance evaluated to
date shows that there is more transverse cracking in the sections containing RAS for
some sections, and in some sections RAS mixtures are performing as well as the control
mixtures. The field sections do show that a coarse ground RAS has more cracking than
a fine ground RAS and that a RAS mixture produced using a Warm Mix Asphalt
foaming technology has more cracking than the comparison HMA mixture.
5. Secondary and Recycled Aggregates
Both secondary and recycled aggregates are used to partially replace virgin aggregates in
the asphalt mixture. The usage of these materials is not well tracked because some of
these materials are considered normal aggregates in different areas of the country. The
use of particular recycled or secondary aggregates is very regional, depending on
industry in that area and availability of specific materials.
There are several potential issues that must be considered when recycled or
secondary aggregates are used:
Absorption
Stripping potential
Frictional characteristics
Wear characteristics
The severity of each of these potential issues is dependent upon the specific material
being considered. Some secondary or recycled aggregates will have improved properties
over locally available virgin aggregates, so it is important to characterize the specific
materials to be used in a mixture and the overall mixture properties.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
677

SYMPOSIUM
6. Resources
There are many available resources and guidance documents for the use of RAP, RAS,
and recycled aggregates in asphalt mixtures. These include:
Designing HMA Mixtures with High RAP Content: A Practical Guide, Publication
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
QIP-124
How to Increase RAP Usage and Ensure Pavement Performance, NAPA Publication
P S 34
Uses of Waste Shingles in HMA: State-of-the-Practice, Special Report 179
Guidelines for the Use of Reclaimed Asphalt Shingles in Asphalt Pavements,
Information Series 136
Reclaimed Asphalt Pavement in Asphalt Mixtures: State of the Practice Publication
no. FHWA-HRT-11-021 April 2011 http://www.fhwa.dot.gov/publications/
research/ infrastructure/pavements/11021/11021.pdf
High Reclaimed Asphalt Pavement Use FHWA Publication No.: FHWA-HRT-11-
057
http://www.fhwa.dot.gov/publications/research/infrastructure/pavements/11057/
11057.pdf
McDaniel, et al. Investigation of Low and High Temperature Properties of Plant
Produced Mixtures. http://www.fhwa.dot.gov/publications/research/infrastructure/
pavements/11058/index.cfm
NCHRP 9-46 Final Report: Improved Mix Design, Evaluation, and Materials
Management Practices for Hot Mix Asphalt with High Reclaimed Asphalt Pavement
Content http://onlinepubs.trb.org/onlinepubs/nchrp/docs/NCHRP09-46_FR.pdf
www.moreRAP.us
http://www.shinglerecycling.org
http://rmrc.wisc.edu/ug-materials/
7. Acknowledgements
The author would like to thank Audrey Copeland, Jeff Melton, and Chris Williams for
providing information, materials, and resources for this symposium presentation and
paper. Thank you also to Geoff Rowe for his review and comments.
678

8. References
Anderson, E. and J.S. Daniel, Long-Term Performance of High-RAP Pavements: Case

Studies, Transportation Research Record: Journal of the Transportation Research
Board, Transportation Research Board of the National Academies, Washington,
D.C., 2013, in press.
Bennert, T. and A. Maher, Forensic Study on the Cracking Distress of New Jerseys
LTPP SPS-5 Sections 30% RAP vs Virgin Hot Mix Asphalt (HMA),
in press.
Cascione, A., R. Williams, S. Gillen, R. Bentson, D. Haugen, Utilization of Post-
Consumer Recycled Asphalt Shingles and Fractionated Recycled Asphalt Pavement
in Hot Mix Asphalt, Green Streets and Highways 2010. An Interactive Conference
on the State of the Art and How to Achieve Sustainable Outcome, American Society
of Civil Engineers, 2010, pp. 349359.
Cascione, A, R. Williams, W. Buttlar, S. Ahmed, B. Hill, D. Haugen, S. Gillen,
Laboratory Evaluation of Field Produced Hot Mix Asphalt Containing Post-
Consumer Recycled Asphalt Shingles and Fractionated Recycled Asphalt
Pavement, Journal of the Association of Asphalt Paving Technologists, Vol. 80,
2011.
Copeland, A., Reclaimed Asphalt Pavement in Asphalt Mixtures: State of the Practice,
Publication no. FHWA-HRT-11-021, April 2011.
Mogawer, W., T. Bennert, J.S. Daniel, R. Bonaquist, A. Austerman, and A.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Booshehrian, "Performance Characterisitcs of Plant Produced High RAP Mixtures,"

International Journal of Road Materials and Pavement Design, Vol. 13, 2012, pp.
183208.
Nash, T., G. Sholar, G. Page, and J. Musselman, Evaluation of High RAP Asphalt
Mixture Performance in Florida, Transportation Research Record: Journal of the
National Academies, Washington, D.C., 2012, pp. 1625.
West, R., J. Michael, R. Turochy, and S. Maghsoodloo, Use of Data From Specific
Pavement Studies Experiment 5 in the Long-Term Pavement Performance Program
to Compare Virgin and Recycled Asphalt Pavements, Transportation Research
Record: Journal of the Transportation Research Board, No. 2208, Transportation
Research Board of the National Academies, Washington, D.C., 2011, pp. 8289.
679

SYMPOSIUM
9. Discussion
MS. JILL THOMAS: Thank you, Jo, for a very good presentation. I just wondered if
upon looking into this presentation, you found any studies about using recycled asphalt
shingles in porous pavement rather than using fibers? I know Missouri does some of
that. I dont know if you looked at any of that.
PROFESSOR JO DANIEL: I didnt look at specific studies. I have heard that some of
that is ongoing, Im betting Chris would have a good idea of some of that information.
Chris, do you know some of the shingles?
PROFESSOR CHRISTOPHER WILLIAMS: Yes, when you step back and look at the
use of shingles in asphalt mix, any mix that calls upon using fibers, theres no better use
of shingles than to replace the use of the fibers in asphalt mixes. One of the reasons is
because of the substantial complete coating of the fibers contained in the shingles which
I dont think that we do as well in an asphalt mix production plant. Its a cost/economic
side of things and using fibers in an asphalt plant is fairly challenging. So, yes, whether
its porous asphalt or SMA, I think both mixes are excellent candidates for shingle use.
MS. THOMAS: Thank you both.
DR. GARY FITTS: Very nice summary there. One thing I was wondering about, though,
if youve looked into is, I know in some other parts of the world, like in Europe, I
believe the European Asphalt Pavement Association has come up with some guidelines
for evaluating the safety aspects with respect to handling RAP. I dont know about RAS.
But as we use more and more of it, some of these things are going to have to come into
play or come into consideration as were looking at utilizing these materials and some of
the things that could be out there. I believe if you go to perhaps Ontario, Canada, there
have been some issues with asbestos in shingles. Coal tar is something thats been used
widely to protect surfaces where there have been fuel spills and that sort of thing, ground
tar, rubber, sulfur, and so on. Do you know of any guidelines that have been developed
for providing someone with a set of instructions or a checklist for things to check on to
identify whether any of these substances exist in RAP and how those might be dealt
with?
PROF. DANIEL: Yes, good point. I didnt go into specifics on that. I know there is
some guidance out there on some of those issues. I know in some of the early work that
was done with shingles, the asbestos question was a big question, and there is some
guidance in terms of testing for asbestos. I know thats becoming less and less of a
concern, but I know there is some guidance out there on that. In terms of some of the
coal tar and other things, Id have to go back and look. That wasnt the focus of what I
was trying to put out in this presentation, but I know there is some guidance out there on
some of those. And I think the suggestion would be to look at some of the guidance
thats coming out of Europe and some other regions on those materials.
680
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

DR. FITTS: Thank you.

JIM SCHEROCMAN: I have two questions. One is on the fractionization of RAP. You
didnt mention that. But, if I have a large RAP particle and I break it into two or three
smaller particles, but I feed all of those pieces in the asphalt plant at the same time, does
fractionization make any difference?
PROF. DANIEL: It depends. This is the processing, yes. When you talking about
fractionization, you will have those bigger clumps that are made up of the smaller
clumps. But part of the fractionization you can break some of that out, but you cant take
care of all of it. Some of the fractionization also helps you with some of the asphalt
binder because you can do a good job of characterizing that. But no matter what, even if
you're fractionating your RAP stockpiles, you still have to characterize that stockpile in
terms of what the overall gradation is going to be. So your point is correct. Youre
fractionating the whole clump, but you still have to characterize that stockpile, and
depending on the materials and depending on whats happening, there may or may or
may not be differences. So some regions do fractionate, and some regions of the country
do not fractionate.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
MR. SCHEROCMAN: My second question is on RAS. We know that the old felt type
shingles have up to 38 percent asphalt binder in them. The fiberglass shingles currently
being manufactured typically have 18 to 20 percent binder in them. If I have a pile of
combined scrap shingles and I shred then to a particle size of 3/8 inches or less, and I
add shingles at a rate of 5 percent by weight of the mix, how do I know what the
percentage of recovered binder is going to be in the RAS mix?
PROF. DANIEL: Again, Im going to go back to youve got a pile that youre using and
you have to have a good characterization of what that material is in your stockpile thats
going into your mixture, and if that stockpile is changing, you have to be continuously
characterizing that stockpile. So it goes back to the quality control of whats happening
at the plant, and youve got to know what that material is and if thats changing over
time, youve got to know how thats changing over time.
MR. SCHEROCMAN: So you actually expect the contractor to do all of the testing
required?
PROF. DANIEL: Yes, I do.
MR. SCHEROCMAN: Remember that one definition of QC/QA is often taken to be
Quit Checking/Quit Asking.
PROFESSOR LUDO ZANZOTTO: I remember when we doing the roofing asphalts. So
of course in the flat roofs, there was used the glass or there was used the polyester but
there was also used paper. For the paper, there were some problems with
biodegradability. I believe that some shingles are still used with the paper base. I just
681

SYMPOSIUM
wonder if there ever came any issue with some kind of biodegradability that we might be
putting some amount of paper into the pavement, whether it even occurred.
PROF. DANIEL: Well, if you're talking about tear-offs, then youve got the issue of
wood, nails, other material that are in that waste stream, so absolutely there is the
possibility. Again, it gets back to the quality control and knowing what is in that
stockpile and what you're putting into the material.
PROF. ZANZOTTO: Thank you.
DR. GEOFFREY ROWE: Thanks very much, Ludo. And, Jo, for that presentation.
682
--`,`,,,``,,```,,`,,,

___________________________________________________________________
AAPT Symposium:
Sustainable Energy Use in Asphalt Production
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Adam Hand*
Granite Construction Inc., Sparks, NV 89431-550

___________________________________________________________________
ABSTRACT: The objective of the presentation was to answer the question As an asphalt
contractor what are the company objectives for using sustainable technologies? For Granite, the
sustainability goal optimizes economic (return on investment or ROI), environmental
(compliance), and social (citizenship) impacts. Granites sustainability plan includes:
Operation of energy-efficient asphalt production facilities on a per-unit consumption
basis;
Responsible use of sustainable materials technologies without compromising the quality
of the end-product;
Increased use of recycled materials in asphalt pavements; and
Increased use of WMA consistent with quality objectives.
As an example of the economic component of the sustainability goal (ROI), Dr. Hand noted that
Granite invested in modifications to ten mixing plants to reduce losses, improve energy
consumption, reduce energy costs, and account for peak demands. Examples of improvements
included insulating drums and hot oil piping and installing variable frequency drives on baghouse
fans.
The presentation also examined Granites use of RAP. Dr. Hand noted that RAP was initially
separated into three piles depending on the source. A and B piles are formed from known sources
(projects) and are expected to have lower binder stiffness and good aggregate qualities which
permits maximized use of RAP with minimized virgin binder grades. C piles include imported
RAP and is more typically used in EnviroBase. Dr. Hand also noted that fractionation of RAP was
common as the percentage of RAP used exceeds 25 %. Fractionating provides reduced variability
and allows the use of the finer RAP fraction (<9.5 mm) which has a higher binder content
(allowing a greater binder replacement).
Dr. Hand also reviewed the sustainable technologies in use today including RAP, RAS, WMA,
Crumb Rubber Modified mixes, and Ground Tire Rubber modified asphalt. Dr. Hand focused on
WMA as an example and noted how the selection of WMA technology impacted the ROI (cost of
equipment and technology), and Compliance/Citizenship (reduced energy use and emissions)
while maintaining quality (improvement of workability and attainment of density).
______________________________________________________________________________
683

SYMPOSIUM
1. Discussion
DR. HOWARD MARKS: Good presentation. Just curious, has Granite or other types of
producers looked at compressed natural gas to run, for example, maybe trucks and/or
facilities, plants?
DR. ADAM HAND: I can tell you that we have not, Howard. So I dont know what our
competition is doing, but we have not.
DR. MARKS: Have you heard about that as a potential?
DR. HAND: Yes, but we havent tried it. Were a very conservative company. Were
always Were a very conservative company.
DR. ERVIN DUKATZ: Adam, I think you did an excellent job explaining the
technologies and also pointing out the fact that they come at a cost, and you have to take
into account that cost, which kind of echoes a theme that a couple of others have made
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
today. But we need to follow the path and be technically correct. As to Howards
question, the answer is yes. Thank you again.
DR. HAND: That was an easy one, Erv. Thank you.
684

AAPT Symposium:
Binder Additives for Warm Mix Asphalt
Technology
Gaylon L. Baumgardner* and Gerald R. Reinkea
*Paragon Technical Services, Inc., Jackson, MS

a
Mathy Technology and Engineering, Inc., Onalaska, WI
ABSTRACT: The more common science used in determination of a materials coefficient of friction
is known as tribology. Tribiology is defined as the science and technology of interacting surfaces
in relative motion, including the study of friction, wear, and lubrication and is derived from the
Greek tribo (to rub) and the Latin logia (study or learning). The coefficient of friction of a
material is empirical it has to be measured experimentally and cannot be calculated. Rough
surfaces like hot-mix aggregates tend to have higher coefficients than smooth surfaces. Boundary
fluids (like asphalt binders) can serve to reduce the coefficient of friction of aggregates during
production of asphalt mixtures.
Reinke and Baumgardner first introduced the concept of testing asphalt binders in thin films and
suggested lubricity and internal friction reduction as a potential explanation of the mechanism
allowing production of asphalt mixtures at reduced temperature, warm-mix asphalt (WMA), at the
Warm-Mix Technical Working Group (WMA TWG) in Baltimore, MD, December 2007. Further
work in this area was later reported by Hanz, Faheem, Hahmoud and Bahia at the 89th Annual
Meeting of the Transportation Research Board, January, 2010.
This paper presents a continuation of the work of Reinke and Hanz moving from gap-dependent
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
rheology to tribology utilizing a standard tribology fixture and new methods for testing asphalt
binders with a dynamic shear rheometer. The work consists of evaluations of various asphalts
used to produce hot-mix asphalt (HMA) mixtures and the effect they might have on friction
characteristics, allowing production of warm-mix asphalt through non-chemical techniques such
as foaming; and evaluation of Lubricity Optimized Asphalts (LOA) and the effect additives have
on friction characteristics in asphalt mixtures produced with LOA asphalt binders.
KEYWORDS: warm asphalt mixture, rheology, tribology, lubricity, lubricity optimized asphalt
685

SYMPOSIUM
1. Introduction
Warm mix asphalt (WMA) is the generic title given to technologies developed for
production and placement of hot-mix asphalt (HMA) at temperatures substantially lower
than those conventionally accepted. WMA is typically produced at temperatures that are
20 to 55C (35 to 100F) lower than conventional HMA, and has been used successfully
in all types of asphalt concrete, including dense-graded, stone matrix asphalt, porous
asphalt and mastic asphalt over a range of layer thicknesses and traffic loadings
(USDOT, 2008). Original speculation was that WMA technologies reduced the
viscosity of the asphalt binder to provide complete coating of aggregates at lower
temperatures. However, it has been shown that viscosity reduction is not the primary
mechanism of WMA and that lubricity characteristics of asphalt binders may be the
mechanism allowing for reduced asphalt paving temperatures (Reinke and Baumgardner,
2007; Baumgardner and Reinke, 2008a; Baumgardner and Reinke, 2008b; Reinke, 2010;
Bahia et al., 2011). This has prompted research and tribological testing of asphalt
binders in thin films, less than 100 microns, as compared to typical bulk rheological
binder specification testing thicknesses of one millimeter and greater.
Tribology, is the science and technology of interacting surfaces in relative motion,
including the study of friction, wear, and lubrication, is derived from the Greek tribo (to
rub) and the Latin logia (study or learning) (Czichos, 2002). Friction is the force that
resists the relative motion of two solid surfaces in contact with one another. Surfaces of
solids, even those that appear smooth, are always microscopically irregular. When two
such surfaces are in contact under pressure, the peaks or asperities adhere to one another
at an atomic level. When those surfaces move with respect to one another, friction arises
from shearing of the adhered junctions, with the frictional force required to shear those
adhesions varying dramatically depending on material involved. The ratio of frictional
force to the load, known as the coefficient of friction, is a constant that depends on the
material that comprises the surfaces being contacted. The coefficient of friction of a
material is empirical it has to be measured experimentally and cannot be calculated.
Rough surfaces like hot-mix aggregates tend to have higher coefficients than smooth
surfaces. Boundary fluids (like asphalt binders) can serve to reduce the coefficient of
friction of aggregates during production asphalt mixtures. Relatively recent research has
focused on asphalt binders as friction reducing media in asphalt mixtures at typical
HMA mixing and compaction temperatures.
Reinke introduced the term lubricity as it relates to asphalt binders and a thin film
rheological test using a conventional dynamic shear rheometer (DSR) to evaluate the
effectiveness of warm mix additives at reduced temperature (Reinke and Baumgardner,
2007; Baumgardner and Reinke, 2008a; Baumgardner and Reinke, 2008b; Reinke,
2010). The test incorporates modification of an improved tribo-rheometry test geometry
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
686

for thin film testing developed by Chasen Kavehpour and McKinley (Chasenet al., 2004;
Kavehpour and McKinley, 2004). In the test a sample of binder sufficient to perform the
thin film rheology test ( 1 gram for a 25 mm diameter plate) is placed in a stainless
steel cup as shown in Figure 1. Depending on the binder grade being evaluated, the test
temperatures will vary. Typical test temperatures are on the order of 105C (221F),
115C (239F), or 125C (257F) for typical straight run binders but can range as high as
160C (320F) for highly polymer modified binders. The cup assembly has been found
to be the most utilitarian for this test because the high rotational speeds can spray small
amounts of binder onto inner surfaces of the oven or Peltier enclosure. The binder
sample is loaded into the cup and brought to the initial gap thickness. For screening
purposes, generally four test gaps are used on the same sample for a single temperature.
The initial gap is set at 500 m and the steady shear test is performed at rotational speeds
of from 1 to 150 radians/sec (0 to 1433 rpm) and repeated at 100 m, 50 m and 25 m
gap settings.
Figure 1. Tribo-Rheometry testing fixture and sample loading.

Data collected during a test consists of rotational speed (in radians/sec), viscosity (in
Pas) and normal force (in Newtons). The test outcome for the Control binder in this
study is shown in Figure 2. Observing Figure 2, it can be seen that as the rotational
speed increases, the normal force exerted by the binder increases along with resultant
torque. The viscosity decreases exponentially with increasing rotational speed, and the
normal force reaches a peak value. When the temperature is decreased to 110C, there is
a sharp breakpoint in the viscosity plot and sharp drop in the normal force and torque
after achieving a maximum value. This occurs with most binders at some temperature
and rotational speed and appears to be related to the upper plate breaking loose of the
binder sample at some rotational speed. Two factors affecting binder performance
appear to be the magnitude of rotational speed or shear rate that can be sustained before
687
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
this slippage occurs and the magnitude of normal force produced by the binder at that
shear rate.
1.000 1.250
1.000
0.7500
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
0.5000
normal force(N)
viscosity(Pa.s)
0.2500
-0.2500
-0.5000
0.1000 -0.7500
1.000 10.00 100.0 1000
velocity(rad/s)
10.00 4.000
3.500
3.000
2.500
1.000
2.000
al force(N)
viscosity(Pa.s)
1.500
norm
1.000
0.1000
0.5000
-0.5000
0.01000 -1.000
1.000 10.00 100.0 1000
velocity(rad/s)
Figure 2. Thin film (25 m) tribo-rheology plot for the Control binder @ 130C
(266F) and @ 110C (230F) respectively, viscosity and normal force.
688

Hanz (2010) reported a four-ball asphalt lubricity test based on a modified version of
an American Society for Testing and Materials (ASTM) test method commonly used to
test lubricants (ASTM D5183-2005). As with Reinke and Baumgardner (2007, 2008,
2010), a thin film rheological test was employed using a conventional dynamic DSR
with a specially designed testing fixture. The apparatus consists of three balls clamped
in a cup, with a forth ball held in a chuck which is placed in contact with the three balls,
with a sufficient amount of asphalt binder added to produce a film between the chuck
and clamped assembly. The four-ball testing fixture is shown in Figure 3.
Figure 3. Asphalt lubricity testing fixture (from Hanz 2010).

The chuck is rotated at a constant shear rate of 50 rpm with the resistance provided
by the fixed balls in the cup. The testing temperature ranged from 80 to 100C (176 to
212F) and two levels of normal force, 20N and 30N, were selected to ensure that
sufficient force was present to maintain contact between the chuck and the clamped
balls.
These two methods rely on gap-based methods. With Reinke and Baumgardner
(2007, 2008, 2010) gap dependence is the main variable due to the large contacting
surface area of the rotating 25 mm (1 in.) plate and the stationary cup. The method
employed by Hanz (2010) is also gap dependent due to test temperature limitations of
from 80 to 100C (176 to 212F), which is well below the operating temperature range
689
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
for production of hot-mix asphalt and most warm-mix asphalt technologies. This paper
presents a continuation of the early groundbreaking work of Reinke and Baumgardner
(2007, 2008, 2010) and the following related work of Hanz (2010) by moving from gap
dependent rheology to tribology utilizing a standard tribology fixture and new methods
for testing asphalt binders with a DSR.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
2. Materials
The primary (control) asphalt binder used in this study was a PG 64-22 supplied by Lion
Oil Company, El Dorado, Arkansas. Additional binders were also selected from sources
that would be considered to have considerably different chemistry and characteristics
than the control binder. These binders consisted of a PG 64-22 supplied by NuStar
Asphalt, Savannah, Georgia; a PG 64-16 from Shell Oil Company, Martinez, California;
and an AR4000 (PG 64-16) from San Joaquin Refining, Oildale, California.
3. Experimental
3.1 Rheometer
As with the work of Reinke and Baumgardner (2007, 2008, 2010) and Hanz (2010), a
DSR was used in this study as described by Heyer (2009). An MCR 301 DSR from
Anton Paar was used for all tests. The instrument controls the rotational speed and
measures the resulting torque very accurately. In addition, torque-controlled
measurements are possible by applying a torque and measuring the resulting speed. The
normal force can be set and recorded during all tests. The instrument used features the
following measurement ranges: rotational speed, 10-6 to 3000 rpm; torque, 10-7 to 0.2
Nm; normal force, 0.01 to 50 N.
3.2 Temperature Control
For tribological testing, temperature was maintained by a Peltier controlled bottom plate
to which the cup fixture is affixed and an additional Peltier-controlled hood. Accurate
temperature is therefore ensured by a combination of conduction and convection heating.
The Peltier-controlled hood ensures a uniform temperature distribution within the
sample over the whole measurement range. This is crucial since a temperature gradient
within the sample will induce misleading results.
690

3.3 Tribology Fixture
The tribology set-up is based on the ball-on-three-plates principle (or ball-on-pyramid)

consisting of a geometry in which a steel ball is held, a cup where three small plates can
be placed and a bottom stage movable in all directions on which the cup is affixed. The
ball-on-three-plates set-up was previously used by Stehr (2004) in a dedicated device to
measure static friction coefficients. Figures 4 depicts the tribometer fixture
schematically; Figure 5 presents photos of the fixture.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 4. Schematic setup tribology accessory.
Figure 5. Cup and plate assembly for the tribology fixture.

Flexibility of the bottom plate is required to ensure that the normal force acts evenly
on all the three contact points of the upper ball, the rotating sphere is adjusted
automatically, and the forces are evenly distributed on the three friction contacts. An
691

SYMPOSIUM
overload of one contact point would result in erroneous friction values. The ball, as well
as the plates for the cup, can be exchanged so that the system can be adapted to desired
material combinations such as aggregate types used in asphalt mixtures. The rotational
speed applied to the shaft is producing a sliding speed of the ball with respect to the
plates at the contact points. The resulting torque can be correlated with the friction force
by employing simple geometric calculations. The axial force of the rheometer is
transferred into a normal force acting perpendicular to the bottom plates at the contact
points. The relations between the axial force (FN) and the torque (M) of the rheometer to
the normal force rectangular to the plates (FL) and the friction force (FF) experienced by
the sample are:
2 ,
With rball being the radius of sphere and the angle of the plates, respectively. The
following dimensions have been used: = 45; rball = 6.35 mm (0.25 inch). Based on the
geometrical dimensions, the tribological properties can be calculated:
Normal force (N): 2
Friction force (N): 2/
Sliding distance (m): 1/2 ( : deflection angle in rad)
Sliding speed (M/s): 1/2 (: angular velocity in rad/s)
3.4 Sample Testing -- Rotational Mode
A sample of binder sufficient to perform the tribology test ( 1 gram) is placed in a

stainless steel cup as shown in Figure 5. As with Reinke and Baumgardner (2007, 2008,
2010), depending on the binder grade being evaluated, test temperatures may vary. Test
temperatures used in the work being reported were 115C (239F), 130C (266F),
145C (293F) and 160C (320F). The cup assembly has a cover (not shown in figures)
which provides protection from spray of binder onto inner surfaces of the Peltier
enclosure at the higher rotational speeds. After the binder sample was loaded into the
cup, the upper ball was brought into contact with the cup and plate assembly and an
applied normal force of 10 N was maintained. The Peltier enclosure was lowered, and
temperature was allowed to equilibrate. A steady shear test was performed at rotational
speeds of from 0 to 150 radians/sec (0 to 1433 rpm).
3.5 Sample Testing -- Oscillatory Mode
An oscillatory test procedure was evaluated to optimize the accuracy and precision of the
tribology fixture. This method can be used to determine the kinetic coefficient of friction
692
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

of the stainless steel ball on the stainless steel plates with asphalt binder samples placed
between the stainless steel ball and stainless steel plates.
Amontons Law is defined as the ratio the shearing force to the normal force. The
oscillatory material function of the shearing force is the complex modulus G* (Pa), and
thrust in grams is converted to the normal force in Newtons (N). The plate pressure P
(N/m2) is calculated by dividing the force in Newtons by the contact area in square
meters such that:
COF=G*/P [1]
As with rotational mode testing, a sample of binder sufficient to perform the

tribology test (1 gram) is placed in a stainless steel cup as shown in Figure 5. Again, as
with Reinke and Baumgardner (2007, 2008, 2010), depending on the binder grade being
evaluated, test temperatures may vary. Test temperatures used in the oscillatory work
being reported were 115C (239F), 130C (266F), 145C (293F) and 160C (320F).
The cup assembly provides protection from spray of binder onto inner surfaces of the
Peltier enclosure; however, this may not be an issue in oscillatory mode testing. After
the binder sample was loaded into the cup the upper ball was brought into contact with
the cup and plate assembly. The Peltier enclosure was lowered, and temperature was
allowed to equilibrate. The normal force was adjusted to 10 0.05 N.
An amplitude sweep using stress control mode at 1 Hz was made in order to
determine the strain range where the plates provided a reasonable G*. Too high strains
result in out-of-control tests. The preferred strain is no greater than 1 strain unit. Next, a
5 minute (300 second) 1 Hz time sweep was performed at the selected temperature while
maintaining an applied normal force of 10 0.05 N. It is not possible to keep the load
exactly at 10 N but a variation of 5 N will not affect the coefficient of friction
calculated values.
4. Measurements and Discussion
Before discussing measurement results from asphalt binder tribology testing, a brief
discussion of lubrication and what the data from tribology testing represents will be
helpful. This discussion is provided by Anton-Paar via an application note for DSR
tribology testing (Heyer, undated). The Stribeck curve (Gelinck and Schipper, 2000)
from tribology testing of lubricants can be divided into four areas, Figure 6. At low
speeds, in the boundary lubrication regime (1) no hydrodynamic pressure has built up in
the lubricant due to the movement. Boundary lubrication is lubrication by a liquid under
conditions where the solid surfaces are so close together that appreciable contact
between opposing asperities is possible. The friction in boundary lubrication is
693
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
determined predominantly by interaction between the solids and between the solids and
the liquid. The bulk flow properties of the liquid play little or no part in the friction
behavior.
Figure 6. Friction coefficient vs. speed .
Figure 7. Film thickness dependency on speed.

At higher speed in the mixed regime (2), the lubricant (asphalt binder) builds up a
hydrodynamic pressure so that direct contact between the two bodies is reduced. In the
elasto-hydrodynamic lubrication regime (3) at intermediate speed, a thin lubrication film
has built up and at high speed, the hydrodynamic pressure is in equilibrium with the
normal load so that a thicker lubrication film has formed to separate the sliding bodies
from each other (Figure 7).
4.1 Rotational Mode
Figure 8 presents the friction characteristics of the steel ball in contact with the steel
plates in absence of lubricant in the tribology fixture.
694
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 8. Coefficient of friction, steel ball on steel plates.
Figure 9 presents graphical data from testing of the control binder. These graphs
reveal that the asphalt binder indeed behaves as described in the preceding discussion of
expected tribology test results and that asphalt binder is considerably effective at
reducing the coefficient of friction of steel surfaces in the testing apparatus. It is
apparent from Figure 9 that the asphalt binder behaves primarily in the elasto-
hydrodynamic lubrication regime. At the 160C (320F) test temperature, Figure 9 (a),
the asphalt binder appears to show lubrication behavior predominantly in the boundary
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
lubrication regime at the lower shear speeds from 0 to 600 rpm. As the test temperature
is decreased the lubricity behavior at these shear speeds trends to the elasto-
hydrodynamic regime, which may be attributed to both increase of thickness in asphalt
film due to shear speed and resultant normal force and an increase in film thickness due
to temperature. This is most apparent at the 115C (239F) test temperature, Figure 9
(d), as the lubricity behavior is almost entirely in the elasto-hydrodynamic regime and
may actually be a measurement of film viscosity or laminar flow in the binder as the
binder behavior transitions from a lubricant to an adhesive media.
695

SYMPOSIUM
696
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Figure 9. Tribology results for rotational testing of Lion Oil PG 64-22 (Neat) at (a)
160C (320F), (b) 145C (293F) , (c) 130C (266F), and (d) 115C (239F).
These data are explanatory of WMA technologies that employ foamed asphalt binder
as the method for production of WMA mixtures. It appears that foaming may not be the
primary mechanism for producing WMA in the plus 135C (275F) range as even
without foaming or additives the Lion Oil PG 64-22 control asphalt or asphalt with
similar properties could be used for successful production of WMA in that temperature
range. The primary function of foamed asphalt binder may be limited to coating of the
697
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
bulk aggregate structure. In US Patent 3,868,262, Karl Gunner Ohlson (Ohlson, 1975)
reports a sequential mixing method for batch plant mixed asphalt mixtures that divides
the bulk aggregate of the mixture into two portions which are coated in sequence of
coating of the coarse fraction with the entire binder content with subsequent addition of
the finer aggregate fraction for a mixture that shows improved aggregate coating. It
should be noted that the aggregate is dry and no added water or damp aggregates are
used in this process. Ohlson also reports improved coating of aggregates at bulk binder
viscosities in the normal HMA range of 200-500 cSt depending on composition of
mineral aggregate and at bulk binder viscosities of as high as 2000 cSt. While Ohlson
does not specifically refer to WMA (Ohlson, 1975), it has recently been suggested by the
Swedish Transport Administration (Wendel, 2011) that the process reported in Ohlson
(1975) does in fact enable reduction of mixture temperatures without foaming or the
presence of moisture in the aggregates. Successful results applying this method to
reduced temperature mixes were also reported by Howard at the 90th annual meeting of
the Transportation Research Board (Howard, 2011). Methods and results reported by
Ohlson (1975), Wendel (2011) and Howard (2011) all support that binder lubricity, not
foaming, is the primary factor supporting these reduced temperature processes that do
not employ additional WMA additives.
In practice, the sequential mixing and foaming approaches conducted at around 135
C (275F) enable some reduction in mixing temperatures likely because both processes
spread the asphalt more effectively than simply adding asphalt at 135C (275F) to
aggregate and then attempting to get it to coat all of the particles (large and small alike).
Since the asphalt is cool relative to normal hot mix temperature, coating coarse
aggregate as in the sequential mixing process is fairly easy. At this point the elasto-
hydrodynamic film is in place not due to shear rate but due to the thicker film of asphalt
spread over the surface area of the coarse aggregate. In foaming processes, water
injected into asphalt at 135-140C (275-285F) will tend to cool the asphalt as the water
is converted to steam and the coarse aggregate is more readily coated than the fine
aggregate; lack of agglomeration of the fine aggregate is probably indicative of this.
Essentially the large aggregate being coated spreads the asphalt to a thick film while
additional mixing in the drum enables the fines to coat more readily as they are sheared
between the asphalt films on the coarse aggregate. Figure 9 helps make this case in that
at 130C (320F) and 145C (293F) the untreated asphalt binder exhibits all of the
lubrication phases. Consequently, asphalt films have to be thinner which is probably
why untreated mixes cannot be easily compacted at reduced temperatures. Thicker film
mixes (foam, sequential mixing) are lubricated by the thicker layer of asphalt on the
coarse aggregate and therefore can be compacted at temperatures below those of
normally produced HMA.
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
698

Figures 10 and 11 show results from the control binder modified with various WMA
chemical additives, 0.5% Sasobit, (b), and 0.5% of two different Evotherm 3G additives,
(c) and (d). Comparison of results from testing of binder without additives at 115C
(239F) and 130C (266F) to the same binder modified with lubricating chemical
WMA additives reveal the impact these WMA additives have on the lubricity
characteristics of asphalt binders. It is readily seen that addition of WMA additives
alters the lubricity characteristics of the treated binder as compared to those of the
untreated binder at elevated HMA temperatures. Figure 10 shows the impact of
chemical additives on the ability to provide a complete lubricating effect at 115C
(239F), similar to the untreated asphalt at 160C (320F).
699
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 10. Tribology results for Lion Oil PG 64-22 at 115C (239F) Neat (a), Sasobit
(b), Evotherm J-1 (c) and Evotherm M-1 (d).
Figure 11 (a) shows the untreated asphalt at 130C (266F) with a narrow range of
rotational speed when it is in the elasto-hydrodynamic regime and then it quickly move
to the hydrodynamic regime, at which point it might be very difficult to further coat
and/or compact the aggregate particles. The additives evaluated, especially the Sasobit,
Figure 11 (b), and Evotherm M-1, Figure 11 (d), remain in the elasto-hydrodynamic
700

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
regime through the end of the test with the Evotherm J-1, Figure 11 (c), showing only a
slight increase at the very highest rotational speed.
701

SYMPOSIUM
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Figure 11. Tribology results for Lion Oil PG 64-22 at 130C (266F) Neat (a), Sasobit
(b), Evotherm J-1 (c) and Evotherm M-1 (d).
These data make it clear that while foaming and sequential mixing processes can
enable mixing and compaction at reduced temperatures, the reason (i.e. thick films on
coarse aggregate) is mechanically much different than what appears to be occurring with
the addition of chemical additives. Chemical additives, even at substantially reduced
temperatures, enable the binder to exhibit the complete range of lubricating
702

characteristics similar to those of the asphalt binder at elevated temperature. From this it
is apparent that lubricity optimization or the concept of Lubricity Optimized Asphalt
(LOA) provides additional improvement of WMA binders beyond that of simple
foaming or sequential mixing. An expanded study by Howard (2011) on the concept of
hot-mixed, warm compacted asphalt mixtures for disaster response is currently
underway to evaluate the limits of LOA WMA asphalt mixtures.
4.2 Oscillatory Mode
Oscillatory testing was performed at the steady frequency of 1 Hz, opposed to the
ramped rotational speed typically used in standard tribology testing. This was elected
with the assumption that typical mixing and compaction of asphalt paving mixtures
occurs within a narrow range of shear rates. Figure 12 presents graphical data from
oscillatory testing of the control binder. These graphs also reveal that the asphalt binder
indeed behaves as described in the preceding discussion of expected tribology test
results and that asphalt binder is considerably effective at reducing the coefficient of
friction of steel surfaces in the testing apparatus. It is apparent from Figure 12 that the
asphalt binder behaves primarily in the elasto-hydrodynamic lubrication regime. At the
160C (320F) and 145C (293F) test temperatures the asphalt binder appears to show
lubrication behavior predominantly in the boundary lubrication regime in the early
stages of testing which equilibrates in the elasto-hydrodynamic regime after about 3
minutes. As the test temperature is decreased, similar to the rotational testing, the
lubricity behavior trends more to the elasto-hydrodynamic regime, which again may be
attributed to both increase of thickness in asphalt film due to shear speed and resultant
normal force and an increase in film thickness due to temperature.
Comparison of steady rotation and oscillatory tests showed that the latter mode of
measurement may be preferred. The criteria for making this determination are based on a
comparison to the measured friction coefficients, as well as the precision of
measurement between the two modes of testing. Furthermore, steady rotation will
probably cause the plates to wear out faster than oscillatory testing, thereby making
subsequent results change with time.
703
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
Figure 12. Tribology results for oscillatory testing of Lion Oil PG 64-22 (Neat) at
160C (320F), 145C (293F) , 130C (266F), and 115C (239F).
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
4.3 Effect of Asphalt Source
This study focused on applicability of tribology testing to primarily the control

binder. It has been reported that binders from some sources perform well (with respect
to lower WMA mixing and compaction temperatures) in additive WMA technologies but
not as well in foamed WMA processes. This is considered to be due to binder source
and the contribution the source chemistry may have on lubricity properties, therefore,
additional binders from sources known to have different chemistry and performance
characteristics in asphalt mixtures were also tested for their specific lubricity character.
Figure 13 presents comparative graphs from tribology testing of four binders from this
study tested at 130C (239F) and 115C (266F). These binders are all known to
provide substantially lower temperature performance with chemical additive WMA
technologies and typical low temperatures ( 135C (275F)) for foaming technologies,
however, the NuStar binder does not generally perform as well in foaming technologies.
From the data presented the Lion Oil, Shell Martinez and AR4000 are expected to
provide lower compaction temperatures as compared to the NuStar binder. Considering
that this binder, from a Venezuelan source, is chemically different than the other three
binders, it is believed that binder source and composition play an appreciable role in
704

binder WMA performance and that binder chemistry with respect to lubricity is a topic
that merits further study. It should be noted that the NuStar binder tends to perform
better compared to most binders when chemical additive WMA technologies are
utilized.
Figure 13. Tribology results at 130C (266F) (a) and 115C (239F) (b) for asphalt
from different sources.
705
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
Asphalt binder testing in bulk, thick films, is a common practice for grading purposes
throughout the world but does not address binder performance in thin films common in
asphalt mixtures. The advent and proliferation of WMA technologies has made
understanding of asphalt binder performance in thin films necessary in order to
understand the mechanisms of WMA binder performance. Thin film testing, tribo-
rheology and tribology, have not been commonly used to evaluate asphalt binders.
Reinke first reported the concept lubricity and thin film tribo- rheology testing of asphalt
binders in 2007, followed by Hanz in 2010 with a method akin to tribological methods.
This paper has presented results of a conventional tribology method and fixture,
typically used in testing of lubricants, which can be applied to studying asphalt binders
and additives used in HMA and WMA technologies.
The tribology procedure and fixture reported here refine and extend the ground-
breaking parallel plate gap-based work of Reinke and the testing temperature/gap
dependent work of Hanz Further this work provides a mechanistic explanation for the
performance of reduced temperature technologies which rely on sequential mixing or
water created foaming procedures in comparison to chemically enabled WMA
production procedures. These results support the theory of binder lubricity as a factor in
asphalt mixing and compaction at all applicable temperatures further demonstrating
dependence on the technology being employed to produce the bituminous mixtures. The
fixture described herein is fully adaptable to asphalt binder testing and provides
tribological data useful in evaluating binder performance for binders used in asphalt
mixtures. This valuable new tool is not only applicable to WMA technology evaluation,
but may also have merit in determination of mixture and compaction temperatures for
asphalt binders used in all asphalt mixtures.
6. Acknowledgements
The authors would like to thank Dr. Judge Brown and the staff of Paragon Technical
Services, Inc. (PTSi), Richland, Mississippi, for sample preparation and assistance in
testing in this study.
7. References
American Society for Testing and Materials, Standard Test Method for Determination
of the Coefficient of Friction Using the Four-Ball Wear Test Machine, ASTM,
D5183, 2005.
706
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Bahia, H., Hanz, A., and Clopotel, C., Effect of Warm Mix Asphalt Additives on
Binders Workability and Performance Properties, 2nd International Warm Mix
Conference, St. Louis, MO, 2011.
Chasen, C., Kojic, H., Kavehpour, H. and Mckinley, G., Micro-rheometry: Gap-
Dependent Rheology and Tribology of Complex Fluids, Proceeding XIVth
International Congress on Rheology, Seoul Korea, August 2004.
Czichos, H., Tribology and its Many Facets: From Macroscopic to Microscopic and
Nano-Scale Phenomena, Mechanica 36, 2002, pp 605615.
Gelinck, E.R.M., and Schipper, D.J., Calculation of Stribeck Curves for Line Contacts,
Tribology International No.33, 2000, pp. 175181.
Hanz, A., Ahmed, F., Enad, M., and Bahia, H., Measuring Effects of Warm Mix
Additives Using Additives Using a Newly Developed Asphalt Binder Lubricity
Test for the DSR, 89th Annual Meeting of the Transportation Research Board,
Washington D.C., 2010.
Heyer, P. and Luger, J., Correlation Between Friction and Flow of Lubricating Greases
in a New Tribometer Device, Lubricating Science 21, 2009, pp. 253268.
Heyer, P., Application Note: Physica Rheometers, Available from Anton-Paar High
Precision Instruments.
Howard, I., Doyle, J., Hemsley, M., Baumgardner, G., and Cooley, A., Use of Hot-Mix
Warm Compacted Asphalt in Construction, 90th Annual Meeting of the
Transportation Research Board, Paper #11-3522, Compendium of Papers DvD,
2011.
Kavehpour, H.P., and McKinley, G.H., Tribo-rheometry: From Gap-Dependent
Rheology to Tribology, Tribology Letters, Bol. 17, No. 2, Aug 2004.
Ohlson, K.G., Methods in Production of Plant-Mixed Asphalt Concrete, United States
Patent and Trademark Office, Patent US3,868,262, February 25, 1975.
Reinke, G., and Baumgardner, G.L., REVIX Waterless Warm Mix, Presentation,
Warm Mix Technical Working Group (WMATWG), Baltimore, MD, December
2007.
Reinke, G., and Baumgardner, G., REVIX/Evotherm 3G Waterless Warm Mix
Technology, Presentation, Annual Meeting of the Transportation Research Board,
Washington D.C. January 2008a.
Reinke, G., and Baumgardner, G., REVIX Waterless Technology for Production of
Reduced Temperature Asphalt Mixtures, Presentation, Western Research
707
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
Institute/Federal Highway Administration Pavement Performance Prediction

Symposium, Laramie Wyoming, July 2008b.
Reinke, G., Warm Asphalt Binder Compositions Containing Lubrication Agents,
United States Patent 7,815,725, October 19, 2010.
Stehr, W., Friction Phenomena at Lowest Sliding Speeds Tribological Fingerprint of
Test Specimen, 14th International Colloquium Tribology, Stuttgart/Ostfildern,
Germany: 20-22, 13-15 January 2004.
U.S. Department of Transportation, Warm Mix Asphalt: European Practice, U.S.
Department of Transportation Federal Highway Administration Publication,
M ar ch 2 0 0 8 .
Wendel, M., Performance of the Flow Mixing Technology in Asphalt Concrete,
Presentation, Petersen Asphalt Research Conference, Laramie, WY, 2011.
8. Discussion
MR. JEAN PAUL FORT: I have a few remarks about warm mix efficiency of foam and
additive. First, I agree that water foaming is not decreasing the viscosity of the binder.
To the contrary, it is increasing its apparent viscosity. The more water is added into the
foam, the more viscous the foam becomes. We did measure this with a mobile
viscometer. However, behind the paver, you realize into foam properties. Foam, any
foam is shear thinning, which means that the more shearing you apply to the foam, the
lower is the viscosity. This opened my eyes about what was happening behind the
rollers and behind the paver. So is foam useful for warm mix? Yes, it is. Behind the
paver, it increases the workability of the mix because of the shear thinning properties of
the foam. In the silo or the trucks, in absence of shearing, foam viscosity prevents drain-
down. I dont know if this relates to your conclusion there.
MR. GAYLON BAUMGARDNER: It does. I agree with that. In fact, I know of some
patents where they actually foam the mix in the paver because it gives better workability
of the paver. Theyre actually foaming the asphalt in the mixture. That is in France,
actually, that they do that.
MR. FORT: Yes. If you compare the same mix side to side, same binder and you pave
400 tons with hot mix, 400 tons with warm mix with foam, no matter what, you are
going to see an increased workability with the foam. And the only explanation I found is
this shear thinning, but certainly not the loss of viscosity with the water in the binder
because thats exactly the contrary.
708
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

MR. BAUMGARDNER: I agree. Our main thing was to determine if it is necessary.

And, yes, it may help, but its not necessary to be able to produce low-temperature
mixes. Thats why I said, I understand why we use it at the hot-mix plants, but, you
know, our concern was we wanted to know whats going on. Is it the foam thats
actually making the warm mix or are there other characteristics? One thing I didnt sho w
when we were looking at the tribology is the different binders make a difference also. In
other words, just the chemistry of the binder. For example, we look at a California
Valley asphalt versus a Venezuelan asphalt; you see a very big difference between the
lubrication characteristics between those asphalts. In fact, the Venezuelan asphalt would
take less additive or less foam, if you will, than maybe the California Valley material
with the lower asphaltenes.
DR. FORT: The same with the foam, aptitude to foaming varies from binder to binder
MR. BAUMGARDNER: Exactly.
709
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

___________________________________________________________________
AAPT Symposium:
Warm Mix Asphalt Technology
Implementation from a State DOT
Perspective
Bill Schiebel*
Colorado DOT, Denver. CO 80216-6408

___________________________________________________________________
ABSTRACT: The presentation focused on the implementation of WMA in Colorado. The process
included:
the use of WMA as part of an experiment on I-70;
formation of a WMA Task Force
information gathering;
documentation of performance;
development of a specification; and
incorporation of WMA as an approved technology.
In 2007, the Colorado DOT initiated an experimental project on I-70 from the west end of the
Eisenhower Tunnel to the town of Silverthorne. Three WMA technologies were used: Advera,
Sasobit, and Evotherm. In 2009 a WMA Task Force was formed to: (1) develop the submittal
criteria and approval process for WMA technologies; and (2) develop/modify specifications to
allow WMA on Colorado DOT projects. The Task Force used current research efforts to guide the
WMA requirements. Included were considerations for volumetric property adjustments and
reduced aging. A maximum temperature drop of 100F (60C) is permitted for WMA to allow for
binder blending with RAP. The revised specification allows the use of any approved WMA
technology as desired by the Contractor (subject to any stated restrictions). Currently the
Colorado DOT has a list of nine approved WMA technologies.
The presentation concluded with a summary of the implementation process. Pilot projects were
used to introduce WMA locally. Criteria were developed for WMA approval and project
acceptance. Benefits and risks expected from WMA were considered in developing specifications
(reduced emissions, increased haul distances, extended paving season, and improved compaction).
Specifications will be re-evaluated regularly.
______________________________________________________________________________
711
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

SYMPOSIUM
1. Discussion
DR. ADAM HAND: Nice job, Bill. I wanted to clarify on the temperature item. What
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
we were seeing in specifications was minimum production temperatures, not maximums.
I think you made a distinction there that you were talking about maximums. I think
thats really important. For what its worth, we did some of the same work with crack
filler, where on one of our haul roads we poured rubberized crack filler the day before,
and we produced up to about 280F and didnt have any issues with swelling of the
sealer. So in terms of finding a number, theres some more evidence.
MR. WILLIAM SCHIEBEL: Excellent. Thank you.
712

___________________________________________________________________
AAPT International Forum:

Specifications and Use of Bituminous
Materials in Europe
European Bitumen Specifications
Mike Southern, Eurobitume
Hot Mix Asphalt: CEN Harmonized Standards

Jean-Luc Delorme, CEREMA-CETE Ile de France- Ecomaterials
Laboratory
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Insights into Bitumen Characterization in Europe
Jean-Pascal Planche, Western Research Institute
Sustainability in Asphalt Road Construction, European

Developments
Jan van der Zwan, Rijkswaterstaat
______________________________________________________________________________
713

AAPT 2013 INTERNATIONAL FORUM
1. Discussion
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
PROFESSOR MIHAI MARASTEANU: Very good and informative presentation. I
would like to ask you if you are aware of the following issue. We worked on a couple of
projects in which we investigated the fracture properties of binders at low temperature,
and we found out that the cooling liquid, the ethanol, has a significant effect, as
published by Dongre more than ten years ago. Although it has been two years since we
looked at this, we still dont know what happens. Some researchers believe ethanol
reacts with the asphalt binder at low temperature and initiates environmental stress
cracking. Other researchers believe the cooling rates in ethanol vs. air are very different,
although this does not explain the difference between ethanol and potassium acetate.
Would you please comment? There is a recent European specification on single edge
notched test; in the published report different laboratories use different media. I dont
know if you have any insight into this work
DR. JEAN-PASCAL PLANCHE: Are you asking what they are using in Europe?
PROF. MARASTEANU: Yes.
DR. PLANCHE: I believe that it is ethanol.
PROF. MARASTEANU: It is ethanol.
DR. PLANCHE: I think so.
PROF. MARASTEANU: Does it make sense to worry that ethanol chemically reacts
with the binder, even at these low temperatures?
DR. PLANCHE: I wouldnt think in terms of chemical reaction, but interaction is a
possibility, like swelling or something, and particularly if the binder is aged, for
example. But I dont have an answer. Its just a guess.
PROF. MARASTEANU: It would be something worth exploring; maybe a good topic
for cross-fertilization as you mentioned before.
DR. PLANCHE: Lets talk together on cross-fertilizing.
PROF. MARASTEANU: Sounds very good. Thank you.
PROFESSOR REYNALDO ROQUE: I have just a couple of quick questions that I dont
want to forget later on. They relate to the crystalline softening that you showed. Im not
sure what testing was done it was but it showed an increase in compliance with repeated
testing.
DR. PLANCHE: That was the repeated creep test.

PROF. ROQUE: It was a repeated creep test?
714

Specifications and Use of Bituminous Materials in Europe
DR. PLANCHE: Yes, for a hundred cycles.

PROF. ROQUE: Is that what we commonly refer to as reversal of steric hardening?
Would the crystallization then be the steric hardening?
DR. PLANCHE: Im sorry, that was in the case of a polymer modified asphalt with
EVA, so it was more attributed to the fact that EVA can soften at this temperature range.
In fact, it was done at different temperatures and the effect was more pronounced at
higher temperature than at lower temperature.
PROF. ROQUE: Okay. So I should not visualize the crystallization as steric hardening.
Im asking out of complete ignorance. It is an empirical effect that we see all the time
that I typically refer to as steric softening. However, Ive never had a way to visualize it.
DR. PLANCHE: It could be. It could be related because if it is a link to crystallinity,

then there is some crystallinity in asphalt due to waxes if it waxy. And then that type of
behavior could occur for some particular asphalt.
PROF. ROQUE: Okay.
DR. PLANCHE: Not for all of them.
PROF. ROQUE: Yes, I understand. And one other thing, do you know whether these
fairly small differences in KIC from the little binder notch test?
DR. PLANCHE: Yes.

PROF. ROQUE: Have those been shown to translate to a mix level at low temperatures?
DR. PLANCHE: No, that study was binder oriented.
PROF. ROQUE: Binder only?
DR. PLANCHE: Yes. I dont think they were in a mix testing with TSRST, for example.
PROF. ROQUE: My experience is that you dont see the difference in fracture energy
with the polymer addition, especially at low temperatures. You did see it at the binder
level but you dont know if these folks have, in fact, taken it to the mastic level?
DR. PLANCHE: I dont think they did.

PROF. ROQUE: Okay, thank you.
DR. PLANCHE: Youre welcome.
PROFESSOR CARL MONISMITH: My question is directed to Dr. Jean-Luc DeLorme.
My question is, what procedures do you use to prepare your specimens (other than the
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
one for the compression test) for your fatigue tests, thermal stress restrained specimen
tests and other performance related tests?
715

DR. JEAN-LUC DELORME: For the fatigue test and also for the stiffness, we use the
French rolling compactor.
PROF. MONISMITH: Thank you. I wanted the audience to hear this.
PROF. HERVE DI BENEDETTO: Is there another question, Mihai?
PROF. MARASTEANU: I have a business question. If we are going to build durable
roads that last forever, whats going to happen to the contractors? Im serious. I mean
you look at other areas. You know, you look at electronics and things like that. There is
always something that you buy after that. I mean, they have to stay in business. So how
do you see this? I mean is there something? Im not saying contractors are bad. I think
we have excellent contractors. But Im just curious.
DR. JAN VAN DER ZWAN: If you build roads that last forever, then the contractors
will be out of business. But you are not able to build roads that will last forever. I can tell
you a little bit about the philosophy that we have in the Netherlands. We try to build a
road with an eternal life for structural capacity. That means that we dont allow any
structural damage in our roads. We dont have any structural damage in the road. But we
have a thin layer on the top of it and you have to replace that on a frequent level, because
in the Netherlands it is porous asphalt. You have failure mechanisms so you cannot
have an eternal life. So there is not a problem. The contractors in the Netherlands are not
afraid for that. They dont look at quantity. They look at quality and they look at money,
and then they charge us more.
PROF. DI BENEDETTO: Anybody else for a question?
DR. MICHAEL SOUTHERN: If there are no other questions, I have a kind of
philosophical point, and Ive been enthused by Jans circular arguments. I was struck by
the presentation from the chap from Colorado DOT this morning, where he talked about
the fact that we dont have enough money. We hear that a lot, not just in the US but
everywhere I go, that theres not enough money. Now back in 2006, I saw a presentation
which on the face of it was really, really dull. It was an accounts-based approach to road
maintenance, and I thought this is going to be a mega yawn, but in fact the guy had a
very simple and clear point. When you look at the value of the road infrastructure that
you have in Denver, in Colorado, in the United States, in Europe, in the world, there is
simply not enough money to build that again. Nobody can afford to build it again. There
are trillions of dollars involved, and we dont have that sort of money. I come back to
something that Jan said, this is a vicious circle. That we have to look after the roads and
we have to make them last longer because we dont have the money to replace them
again. We have to find a way as an industry, as industry people, of getting the political
message back to the politicians who make the strategic decisions that you cant afford
not to look after the roads. You have to look after them because you cant afford to
rebuild them. Unless were going to go back to the days of horse and carts and rutted
roads, weve got to find a way to improve rolling resistance. You saw from Jans
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
716

presentation, again, the biggest emission in the whole process is from the cars. The
reason behind that is that there are millions, tens of millions, even hundreds of millions
of cars using the highways over the lifetime of the road. Anything you do that improves
something which you multiply by 109 or 1012 is a big win. Rolling resistance is a big
win. And well-maintained pavements. Ive worked with bitumen producers, so I have a
clear bias. But anything we do to improve road maintenance and rolling resistance has
got to be a win. I think we just need to think as an industry about how were going to get
that message back to the decision makers, and thats my rant.
MR. GERALD HUBER: Actually, you sort of set up the question that I was going to
ask. It deals exactly with these ideas and these concepts of rolling resistance and fuel
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
savings and whatnot. Of course smoothness plays a large role in rolling resistance.
Theres no doubt about that. If you talk to someone from the concrete industry, theyll
tell you that deflection also plays a very large role. What this underscores in my view is
gamesmanship and marketing amongst the industries. Of course, we as the asphalt
industry are probably viewed with suspicion by, say, an agency-type person equally as
much as perhaps some of the concrete information thats being spread around. So my
question is: In all this evaluation of components that contribute towards the lifecycle
costs, in Europe, how are you approaching the question of developing an unbiased
estimate of the various properties that feed into the total carbon footprint?
DR. DELORME: Im even not quite sure if I really understand your question. Can you
try to repeat in other words, please?
MR. HUBER: Let me just talk around it a little bit. So one group says, Our roads are
very smooth so therefore yada yada yada. The next group says, Your roads are very
flexible, our roads are very stiff, and so you're always driving uphill when you're driving
on an asphalt road. So, you know, both industries can make statements, but the question
is: How do you come up with some sort of unbiased information that is accepted by
agency people?
DR. DELORME: One word research. Thats the only thing you can do. I told you
about what were doing in Europe, this kind of model rutting resistance. It has nothing to
do with deflection because I think that we in Europe dont believe deflection [is a
problem], maybe because our heavy trafficked roads are really stiff. Also the asphalt
roads are really stiff. But in the project we are not looking into the deflection. Were just
looking at surface characteristics and maintenance and that kind of thing. This project is
trying to combine objective measurements. What is the objective? But with some kind
of a machine and try to relate that to fuel consumption. That is fact gathering. You try to
get the facts. After that, it is up to the politicians to decide what they want to do with
that. Maybe to add a little bit to what Mike said, for me, the largest problem is the
politicians dont care. They are sitting there for four years and they want to be reelected.
We make our roads so durable that normally it lasts three, four, or five times that before
we get problems. Maybe we let the total well not do that but let it collapse. Because
717

then it becomes an issue, and now its not an issue for them. We have the same in our
country, and I can imagine especially now in the crisis time, lack of money, so they have
to distribute it, to stay at that task, and looking at what the people want. And as long as
roads are good enough, we diminish the amount of money. I hope I answered your
question a little bit.
PROF. ROQUE: A very related question, and it has to do with how you in Europe feel
about your level of understanding of or the quality of the tools that you have to deal with
cracking and rutting. Are they of sufficient quality to give you the answers that Gerry
indicated that we need to have? Are you satisfied at this point? And how much effort or
money do you feel, from your side of the ocean, that we should be putting in to
furthering that understanding? Or are you satisfied with the fatigue tests and the tests
that you have right now and we have sufficient understanding and tools? Because I think
thats very important. If we convey that message, then theres not much, then were
going to be playing with the same tools. And as you saw yesterday and even Tony
Kriech this morning said, Weve been talking about rutting and cracking for 50 years,
ever since hes been coming here, meaning we dont have a full understanding of these
things. We had a session yesterday trying to get our hands around these things. But
anyway, Ill stop there. What level of confidence do you have in those tools?
PROF. DI BENEDETTO: There are more and more people in Europe, at least in France,
that think that the fatigue characteristics and crack approaches that are used in practice
for design are not really good tools. If it seems to works in practice, its maybe very
lucky. The physical phenomena describing fatigue and cracking are not well understood
and modeled. The tricky point is that during laboratory tests artifact effects are a great
bias for interpretation of results. These artifact effects do not exist on the road. More and
more people are aware of this fact. There is absolutely a need to continue research in that
field.
I dont know if you want to say something more. But in addition, its not only thermo-
mechanical properties that are important. We are in a process of lifecycle analysis and
we must think of future generations.
DR VAN DE ZWAN: Like I said in the beginning, Europe does not exist. So theres not
one European view. You talked a little bit for France. I think that two countries, France
and the Netherlands, were really dealing with functional design of materials and
constructions. As far as I am concerned from the Netherlands, I am more or less satisfied
with the fatigue test that we have and the stiffness test that we have. Like Herve said,
empirically validated, something like that so we dont have structural damage. And in
the Netherlands the research is really on surface characteristics like porous asphalt
raveling. Nevertheless we have not a blind faith in the four-point bending machine and
its tests; there are a lot of questions around that that we have to solve. But that is not the
main priority of our research yet. We can control cracking. We can control rutting
because by using porous asphalt, we dont have rutting. Porous asphalt itself doesnt rut
718
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

and it has some kind of an isolating value. So we dont have ruts in our material. In our
countries, they do it in a different way. So there is no European view and dont copy just
like that what we are doing in Europe because we still have a lot of problems with that.
DR. DI BENEDETTO: Okay. So I think its time to close the session. Thank you to the
speakers. We had very nice presentations. Thank you to all of you for attending this
session, and I wish you a good basketball game.
719
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Recipients of W.J. Emmons Annual Awards
1949 J.A. Lettier, D.F. Fink, N. B. Wilson and F.F. Farley
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
1950 Alvyn Day
1951 E.C. Hughes and H.F. Hardman
1952 N.W. McLeod
1953 James E. Rice
1954 F.M. Williams and F.W. Kimble
1955 R.L. Griffin, T.K. Miles, and C.J. Penther
1956 O.B. Andersland and W.H. Goetz
1957 Jean Lefebvre
1958 Luke Corbett and R.E. Swarbrick
1959 Robert F. Baker and H.S. Papazian
1960 L.E. Corbett and R.E. Swarbrick
1961 Carl L. Monismith, K.E. Secor and E.W. Blackman
1962 M. Livneh and E. Shklarsky
1963 F.N. Hveem, E. Zube and J. Skog
1964 James F. Shook
1965 C.L. Monismith, G.A. Secor and K.E. Secor
1966 R.J. Schmidt and L.E. Santucci
1967 John Skog
1968 A.W. Sisko and L.C. Brunstrum
1969 P.S. Pell and I.F. Taylor
1970 R.G. Hicks and F.N. Finn
1971 John A. Deacon
1972 James L. Duthie
1973 Willem Heukelom
1974 T.R.J. Fabb
1975 M.W. Witczak
1976 A.I.M. Claessen, R. Ditmarsch and C.P. Valkering
1977 Frank R. Meyer, Edward T. Hignell, G.J.A. Kennepohl and Ralph C.G. Haas
1978 J.A.N. Scott
1979 V.P. Puzinauskas
1980 Harry S. Pink, Richard E. Merz and David S. Bosniack
1981 Dallas N. Little, Jr., R.J. Holmgreen and J.A. Epps
1982 G.R. Morris, J.A. DiVito and N.J. Chen
1983 Teh-Chang Lee, Ronald L. Terrel and Joe P. Mahoney
1984 John H. Tenison, Jr.
1985 C.L. Monismith, J.A. Epps and F.N. Finn
1986 Harold W. Muncy, Gayle N. King and J.B. Prudhomme
1987 G.A. Huber and G.H. Heiman
1988 Joseph L. Goodrich
1989 P.S. Kandhal, C. Lubold, Jr. and F.L. Roberts
1990 C.B. Dawley, B.L. Hogweide and K.O. Anderson
1991 John D'Angelo and Ted Ferragut
1992 Donald W. Christensen, Jr., and David A. Anderson
721

1993 Ted Vinson and D. Jung
1994 Ronald Reese
1995 Timothy Aschenbrener, Robert McGennis and Ronald Terrel
1996 Manual G. Rodrquez, Geoffrey Morrison, Joshua vanLoon and Simon Hesp
1997 Y. Richard Kim, Hyun-Jong Lee and Dallas Little
1998 Leslie Myers, Reynaldo Roque and Byron Ruth
1999 Reynaldo Roque, Zhiwang Zhang and Bhavani Sankar
2000 Mary Stroup-Gardiner and M. Law
2001 L. Tashman, E. Masad, B. Peterson and H. Saleh
2002 K.J. Jenkins, A.A.A. Molenaar, J.L.A. de Groot, M.F.C. van de Ven
2003 D. Christensen, T. Pellinen and R. Bonaquist
2004 R. Roque, B. Birgisson, C. Drakos and B. Dietrich
2005 R. Dongre, L. Myers, J. DAngelo, C. Paugh
2006 Y.R. Kim, H. Lee, D. Little, Y. Kim
2007 B. Delaporte, H. DiBenedetto, P. Chavert, G. Gauthier
2008 E. Kutay, N. Gibson, J. Youtcheff
2009 A. LaCroix, Y.R. Kim, M. Sakhaeifar
2010 F. Zhou, S. Hu, T. Scullion
2011 A. Montepara, E. Romeo, M. Isola, G. Tebaldi
2012 Y. Chen, G. Tebaldi, R. Roque, G. Lopp, Y. Su
Runners-up for W.J. Emmons Annual Award
1949 J.M. Rice and W.H. Goetz

1950 N.W. McLeod
1953 E.G. Williams and L.E. Gregg
1954 Charles Mack
1955 Ward K. Parr, Paul J. Serafin and Tom Humphries
1956 J.O. Izatt, C.C. Evans and C.T. Metcalf
1957 J.A. Dillard and R.L. Alwood
1958 Carl L. Monismith
1960 Thomas L. Speer
1961 Austin Douglas and Egons Tons, R.N. Traxler
1963 J. Skog and E. Zube, W.J. Kari and L. E. Santucci
1964 Bonner S. Coffman, David C. Kraft and Jorge Tamayo
1965 Kamran Majidzadeh and Herbert E. Schweyer
1966 Willem Heukelom
1967 V.P. Puzinauskas
1968 R.J. Schmidt, L.J. Painter, J.B. Skog and V.P. Puzinauskas
1969 Jon A. Epps and Carl L. Monismith
1970 P.F. Dickson and J.S. Corlew
1971 H.J. Fromm and J.T. Corkill
1972 H.J. Fromm and W.A. Phang
1973 Matthew W. Witczak
1974 P.J. van de Loo
1975 H. van Dijk
722
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

1976 F.N. Finn, K. Nair and J. Hilliar
1977 W. van Dijk and W. Visser
1978 Richard D. Barksdale
1979 H. Plancher, A.J. Hoiberg, S.C. Suhaka and J.C. Petersen
1980 R.C. Deen, H.F. Southgate and J.G. Mayes
1981 William Whitcomb, R.G. Hicks and Steven J. Escobar
1982 J.W.H. Oliver
1983 Ervin L. Dukatz, David A. Anderson and James L. Rosenberger
1984 George A. Way, John Eisenberg, James Delton and John Lawson
1985 W.R. Meier, Jr., and R.A. Jimenez
1986 J.L. Goodrich, L.H. Dimpfl and P.E. Graf
1987 J.P.J. van der Heide
1988 Bernard Brule, Yvonnick Brion, and Anne Tanquy
1989 E.R. Brown and S. Cross
1990 R. Roque and B.E. Ruth
1991 R.M. Anderson, D. Walker, J.A. Scherocman and L.E. Epley
T.W. Kennedy and W.V. Ping
S.F. Brown, J.N. Preston and K.E. Cooper
1992 H.U. Bahia, D.W. Christensen and D.A. Anderson
1993 Gayle King, H. King, O. Harders, W. Arand, P. Chaverot and J. Planche
1994 Chris A. Bell, M.J. Felling and A. Wieder
1995 Basil Harris and Kevin Stuart
1996 William Buttlar and Reynaldo Roque
1997 E. Ray Brown, John Haddock, Rajib Mallick and Todd Lynn
1998 Hussain Bahia, Timothy Friemel, Pehr Peterson and Jeffrey Russell
1999 H. Bahia, H. Zhai, K. Bonnetti and S. Kose
2000 R.M. Anderson, G.A. Huber, D.E. Walker and X. Zhang
2001 H. Bahia, H. Zhai, M. Zeng, Y. Hu and P. Turner
2002 Jo Sias Daniel and Y. Richard Kim
2003 D. Christensen and R. Bonaquist
2004 K. Nam and H. Bahia
2005 W. Buttlar and E. Dave
2006 R.M. Anderson, M. Huner
2007 E. Masad, E. Arambula, R. Ketcham, A. Abbas and A. Epps Martin
2008 N. Kringos, A. Scarpas and A. De Bondt
P.J. Yoo and I. Al-Qadi
2009 L. Mohammad, A. Bae, M. Elseifi, J. Button, J. Scherocman
2010 S. Caro, E. Masad, A. Bhasin, D. Little and M.S. Sanchez-Silva
2012 G. Baumgardner, J. Hemsley, W. Jordan III, I. Howard
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
723

Honorary Members
Honorary Members in the Association are those people recognized by the membership as having
outstanding eminence and of long experience in the practice of some phase of asphalt paving
technology.
Clifford Richardson 1927

Prevost Hubbard 1948
W.J. Emmons 1948
W a r d K. P a r r 1984
Francis N. Hveem 1985
Norman W. McLeod 1985
Carl L. Monismith 1988
J. York Welborn 1989
Fred N. Finn 1990
Woodrow J. Halstead 1991
Bernard A. Vallerga 1993
William H. Goetz 1994
Vaughn Marker 1994
Jon A. Epps 1996
Bob M. Gallaway 1996
William Kari 1999
Thomas Kennedy 1999
Moreland Herrin 2000
Charles Hughes 2000
R.A. Jimenez 2001
Eugene L. Skok 2001
Peter Pell 2002
Byron Ruth 2003
David A. Anderson 2004
Stephen F. Brown 2004
J. Claine Petersen 2004
Charles F. Potts 2004
V.P. Puzinauskas 2005
Lawrence Santucci 2005
Matthew Witczak 2007
Elton R. Brown 2008
R. Gary Hicks 2010
Prithvi Kandhal 2012
724
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---

Award of Recognition
Presented to those members recognized by the Board of Directors as having contributed to the
field of Asphalt Technology and the Association for many years.
William H. Campen 1972

Charles F. Parker 1973
E. T. Hignell 1974
Robert J. Schmidt 1975
John M. Griffith 1976
Charles R. Foster 1977
Richard H. Lewis 1980
Herbert E. Schweyer 1981
Francis N. Hveem 1981
Bernard E. Gray 1982
Harold J. Fromm 1984
Richard L. Davis 1985
Robert M. Nady 1986
James F. Shook 1987
J. Claine Petersen 1988
Bernard A. Vallerga 1989
David G. Tunnicliff 1990
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Kenneth O. Anderson 1990

Ronald L. Terrel 1991
Jack N. Dybalski 1992
Jack E. Stephens 1992
Robert L.. Dunning 1993
William A. Phang 1993
Eugene L. Skok 1994
Leonard E. Wood 1994
Roy S. Hodgson 1995
James A. Burton 1996
Charles W. Smoot 1996
Elizabeth (Betty) Skok 1996
Gerhard J.A. Kennepohl 1997
Stephen Brown 1998
Bernard McCarthy 1998
Ilan Ishai 1999
James Moulthrop 1999
Frank Fee 2000
Roger Olson 2000
John Weigel 2001
Thomas White 2001
Imants Deme 2002
Andreas Kemp 2002
Frederick Hugo 2003
725

Louay Mohammad 2003
Harold Skip Paul 2004
David E. Newcomb 2005
John Bukowski 2005
Manfred Partl 2007
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Douglas Coleman 2008
Dallas Little 2008
Michael Heitzman 2010
Gerald Reinke 2010
Jo Sias Daniel 2012
Amy Epps Martin 2012
Hussain U. Bahia 2013
726

AAPT Scholarship Fund
In 1992 the AAPT scholarship fund was established. There are now three levels of sponsorship:
Permanent Sponsor: Cumulative Contributions of $20,000 or more. These are:
Chicago Testing Laboratory in honor of Ward K Parr

Southeast Asphalt User Producer Group
Douglas Coleman, Retired Michigan DOT
Annual Sponsor: Contribution of $500 or more in a calendar year. The 2013 Annual Sponsors
are:
Louay Mohammad, Louisiana State University
Audrey Copeland, National Asphalt Pavement Association
Annual Contributor: Contributions of $100 or more in a calendar year. The 2013 Annual
Contributors are:
John Baxter James Hedderich

Rod Birdsall R. Gary Hicks
Wayne Byard David Newcomb
John DAngelo Donald OConnor
John Deatherage Charles Reilly
Dale Decker Eugene L. Skok
Frank Fee Marshall Thompson
Richard Gribbin William Troxler
Kenneth Grzybowski J. Don Weathers
The scholarships that have been awarded are:
Randy Clark West, University of Florida-Gainesville 1992

Todd Scholz, University of Nottingham, England 1993
John Haddock, Auburn University, Alabama 1994
Jo Ellen Sias, North Carolina State University 1995
Cynthia Lynn, Auburn University, Alabama 1996
Robert L. Schmitt, University of Wisconsin-Madison 1996
Brian Birdsall, North Carolina State University 1997
Weng on Tam, University of Texas at Austin 1997
Leslie Myers, University of Florida-Gainesville 1998
Angela Waldhoff, University of Illinois-Urbana 1998
Glen A. Malpass, North Carolina State University 1999
Stacy G. Williams, University of Arkansas 1999
Fenella Long, University of California-Berkeley 2000
Zhiming Si, Texas A&M University 2000
David Timm, University of Minnesota 2000
Ghassan Chehab, North Carolina State University 2001
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
727

Xue Li, University of Minnesota 2001
Daniel Sherman, University of Illinois 2001
Michael Waggoner, University of Florida 2001
Aslam Al-Omari, Washington State University 2002
Il-Seok Oh, Iowa State University 2002
Laith Tashman, Washington State University 2002
Young Guk Seo, North Carolina State University 2002
Aaron LeChance, University of New Hampshire 2003
Nelson Gibson, University of Maryland 2003
Mark King, North Carolina State 2003
Xinjun Li, University of Minnesota 2003
Amit Bhasin, Texas A&M 2004
Angela Priest, Auburn University 2004
Christopher Robinette, Michigan Tech University 2004
Benjamin Underwood, North Carolina State 2004
Edith Arambula, Texas A&M University 2005
Audrey Copeland, Vanderbilt University 2005
Lucas-Jan Ebels, University of Stellenbosch (South Africa) 2005
Rodrigo Delgadillo, University of Wisconsin-Madison 2005
Andrew Braham, University of Illinois at Champaign/Urbana 2006
Eshan Dave, University of Illinois at Champaign/Urbana 2006
Niki Kringos, Delft University of Technology 2006
Adam Zofka, University of Minnesota 2006
Eli Fini, University of Illinois at Urbana-Champaign 2007
Pengcheng Fu, University of California- Davis 2007
Enad Mahmoud, Texas A&M University 2007
Elena Romeo, University of Florida 2007
Kamilla Vasconcelos, Texas A&M University 2007
Silvia Caro, Texas A&M University 2008
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Raul Velasquez, University of Minnesota 2008

James Willis, Auburn University 2008
Mary Robbins, Auburn University 2009
Adriana Vargas-Nordcbeck, Auburn University 2010
Maryam Sakhaeifar, North Carolina State University 2010
Behzad Behnia, University of Illinois at Urbana-Champaign 2011
Arash Motamed, University of Texas at Austin 2011
Salman Hakimzadeh, University of Illinois at Urbana-Champaign 2012
Brian Hill, University of Illinois at Urbana-Champaign 2012
Matthew Gersch, Auburn University 2013
Mirella Villani, Delft University of Technology 2013
728

Past Presidents
Francis P. Smith C. Merritt Hewett

Hugh W. Skidmore Woodrow J. Halstead
Leroy M. Law Ward K. Parr
Walter J. Emmons Vaughn Marker
Prevost Hubbard Charles W. Beagle
Charles A. Mullen Bob M. Gallaway
Henry L. Howe Paul J. Serafin
Roger L. Morrison Larry L. Kole
George W. Craig Moreland Herrin
Alan W. Dow R.A. Jimenez
Victor Nicholson Bernard F. Kallas
J. Strother Miller Jon A. Epps
Walt Hempelmann William J. Kari
Earl F. Kelley Charles S. Hughes
Gene Abson V.P. Puzinauskas
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Claude L. McKesson Richard E. Root

Thomas E. Stanton Byron E. Ruth
Lloyd F. Rader Richard W. Smith
Charles M. Baskin Charles F. Potts
Victor A. Endersby Thomas W. Kennedy
Roland Vokac Lawrence E. Santucci
Walter H. Flood James A. Scherocman
Jewell R. Benson Gale C. Page
Norman W. McLeod David A. Anderson
John Goshorn H. Fred Waller, Jr.
Frank R. Olmstead R. Gary Hicks
Arnold J. Hoiberg Dale S. Decker
George H. Dent Robert B. McGennis
Carl A. Carpenter Gayle N. King
Henry G. Nevitt Prithvi S. Kandhal
William H. Goetz E. Ray Brown
M.F. MacNaughton Gerald A. Huber
James E. Ward Larry L. Michael
James M. Rice Ervin L. Dukatz
Frank M. Williams John D Angelo
Roy E. Bollen Rita B. Leahy
John O. Izatt Frank Fee
J. York Welborn Stephen F. Brown
W.L. Hindermann R. Michael Anderson
Fred N. Finn James S. Moulthrop
Carl L. Monismith Reynaldo Roque
L.C. Krchma David Newcomb
Rebecca McDaniel
729

Life Members
Waddah Akili James E. Hampton J. Claine Petersen

David A. Anderson William J. Harper Richard W. Petrarca
Kenneth O. Anderson James H. Havens Alfredo Pinilla
Earl G. Angell M.J. Hensley Charles F. Potts
Guneri Artunkal Moreland Herrin V.P. Puzinauskas
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
John Baxter Stanley M. Herrin John T. Reinhard
Gary Behnke R. Gary Hicks Gerald Reinke
Chris A. Bell Roy S. Hodgson J.F.(Jack) Renner
Wade B. Betenson Richard J. Holmgreen, Jr. Z. Luis Rodriguez
Rod Birdsall Howard B. Hopps, Jr. Richard E. Root
Donald P. Blasland Eugene Y. Huang Reynaldo Roque
Christopher W. Bowker John E. Huffman Jerry G. Rose
Lennart S. Brenner Charles S. Hughes Byron E. Ruth
Pinckney J. Brewer Frederick Hugo Lawrence E. Santucci
J. Donald Brock Matti S. Huhtala James A. Scherocman
Stephen F. Brown Robert Humer Richard J. Schreck
Elton R. Brown Ilan Ishai Gary H. Schmidt
John Bukowski Rudolf A. Jimenez Russell Schnormeier
Joe W. Button Edwin H. Jones Kenneth E. Secor
William T. Camm Robert H. Joubert James F. Shook
Robert E. Campbell Nabil Kamel Eugene L. Skok, Jr.
Juan A. del Campo Prithvi S. Kandhal Charles W. Smoot
Paul F. Cecchini James E. Kelly Anthony Stock
Andrew C. Clayton Andreas Kemps Kai Kong Tam
Mervin F. Clark Thomas W. Kennedy Clifford A. Taylor
William J. Coffman Gerhard J.A. Kennepohl Tom K.G. Teo
Joseph Craus Irving Kett Mang Tia
Campbell Crawford N. Paul Khosla Walter Tommasini
Michael I. Darter Takashi Kobayashi Gerald S. Triplett
J. Keith Davidson Larry L. Kole Bernard A. Vallerga
Dale S. Decker Dennis D. Krivohlavek Cecil J. Van Til
John F. Deatherage Dah-Yinn Lee Per Ullidtz
Imants J. Deme Kang-Won Wayne Lee Harold L. Von Quintus
Ervin Dukatz Giorgio Legnani, Jr. H. Fred Waller, Jr.
Robert L. Dunning William M. Lehmkuhl George Way
Jack N. Dybalski Rafael Limon Steven Weidenhammer
Robert A. Eaton Robert A. Lindquist Alden L. West
Frank A. Enloe Dallas Little Thomas D. White
Larry E. Epley Robert P. Lottman Robert A. Wilkinson
Jon A. Epps Robert L. Lytton Matthew W. Witczak
Steven Escobar Donaldson MacLeod Ken Wonson
Jorge Fanlo M.R. Malya John Yeaman
Frank Fee J. Paul Martin Ludo Zanzott
Fred M. Fehsenfeld Jack R. Mathews Robert A. Wilkinson
Phillip D. Feliz Richard W. May Matthew W. Witczak
Fred N. Finn Bernie M. McCarthy Ken Wonson
Milton Fletcher Philip E. McIntyre John Yeaman
C. Royce Fugate Jacques de Medina Ludo Zanzotto
Arthur M. Furber Juan E. Ramos Medina
Thurmond J. Gates John Miner
Mike Geller Carl L. Monismith
Pedro Diaz Gonzalez Gene R. Morris
Geoffrey Griffiths David Newcomb
Barry L. Gundersen Donald OConnor
Ralph C.G. Haas John. W.H. Oliver
Douglas Hanson Peter S. Pell
730

Index
Contributors and Discussors
_________________________________________________________________________________________________________
Underlined page numbers refer to formal papers and presentations. Other page
numbers refer to discussion presentations.
Al-Qadi, Imad, 250, 323, 397, 398, 399, 653

Anderson, David, 55, 56, 497, 498
Andriescu, Adrian, 327
Arambula, Edith, 177
Asadi, Somayeh, 151
Austerman, Alexander J., 253, 403
Azari, Haleh, 127, 146, 147, 148, 149
Bahia, Hussain U., 23, 173, 465
Baumgardner, Gaylon L., 685, 708, 709
Behnia, Behzad, 91
Blankenship, Phillip, 201, 401, 402
Bonaquist, Ramon, 530, 531, 577
Booshehrian, Abbas, 403
Buttlar, William G., 91
Button, Joe, 367
Carlson, Doug, 437
Choi, Yeong-Tae, 617
Chowdhury, Arif, 177
Claros, German, 367
Copeland, Audrey, 327
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Cox, Ben C., 213

Crawford, Kenneth, 360, 361
Cucalon, Lorena Garcia, 177
DAngelo, John, 56, 207, 282, 322, 656
Daniel, Jo Sias, 327, 360, 361, 362, 363, 364, 661, 680, 681, 682
Daranga, Codrin, 437
Dave, Eshan V., 91, 124, 125, 126, 324, 325
Decker, Dale, 246, 247, 364
Dedene, Christopher, 175, 176, 656, 657
Delorme, Jean-Luc, 713, 716, 717
731

Di Benedetto, Herve, 499, 500, 716, 718, 719
Doyle, Jesse D., 213, 289, 322, 323, 324, 325
Dukatz, Ervin, 174, 363, 364, 684
Dylla, Heather, 151, 172, 173, 174, 175, 176
Epps, Jon, 367
Fee, Frank, 147, 148, 208, 246, 248, 361, 402
Fini, Elham, 655, 657, 659
Fitts, Gary, 680, 681
Fort, Jean Paul, 708, 709
Gibson, Nelson, 286, 327, 498, 499, 579, 614, 615
Glidden, Stacey, 400
Guddati, Murthy N., 1
Haddock, John E., 59, 89, 90
Hajj, Elie, 399
Hakimelahi, Hamed, 533
Hand, Adam, 683, 684, 712
Harvey, John, 533
Hassan, Marwa, 151, 172, 173
Heitzman, Michael, 89, 90
Hill, Brian, 283
Hintz, Cassie, 465, 496, 497, 498, 499, 500, 501
Howard, Isaac L., 213, 246, 247, 248, 249, 250, 251, 289, 324
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Hu, Sheng, 367

Huang, Baoshan, 250
Huber, Gerald, 717
Jackson, Alexander, 249
Johnston, Arthur, 209
Kaloush, Kamil, 503, 557
Kim, Y. Richard, 1, 617
Kutay, M. Emin, 126, 253, 286, 360, 435
Lee, Robert, 367
Leon, Sofie E., 91
Li, Hongshend, 367
Li, Xinjun, 579
Liu, Hua Qin, 462, 463
Mamlouk, Michael, 503
732

Marasteanu, Mihai, 146, 147, 435, 436, 714, 716
Marks, Howard, 651, 652, 653, 654, 684
Martin, Amy Epps, 177, 208, 210
McDaniel, Rebecca, 90, 365, 653
Mogawer, Walaa S., 253, 281, 282, 283, 284, 285, 286, 287, 403,
435, 436
Mohammad, Louay N., 151, 175, 253
Mohseni, Alaeddin, 127, 147, 149, 282, 496, 497, 577, 578
Molayem, Mohammad, 124, 125
Monismith, Carl, 57, 462, 653, 715, 716
Morian, Nathan, 614
OBrien, Joan P., 59
Page, Gale, 208, 401, 463
Park, Eun Sug, 177
Paulino, Glaucio H., 91
Planche, Jean-Pascal, 173, 713, 714, 715
Plemmons, Clayton, 437
Prowell, Brian, 399, 400, 529, 530
Puchalski, Sebastian, 462
Reinke, Gerald, R., 685
Rodezno, Carolina, 437
Rodezno, Maria C., 557, 577, 578
Roque, Reynaldo, 56, 125, 126, 361, 528, 529, 714, 715, 718
Rosenmayer, Tom, 437
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
Rowe, Geoffrey, 285, 286, 362, 363, 656, 657, 658, 659, 682
Saadeh, Shadi, 464, 533, 556
Sabouri, Mohammadreza, 1
Scherocman, Jim, 681
Schiebel, Bill, 711, 712
Scullion, Tom, 367
Sefidmazgi, Nima Roohi, 23, 55, 56, 57
Solaimanian, Mansour, 247, 248
Souliman, Mena I., 503, 528, 529, 530, 531
Southern, Michael, 658, 713, 716
Tarbox, Sean, 327
Tebaldi, Gabriele, 148, 149, 248, 249, 556, 652
733

Teymourpour, Pouya, 23
Thomas, Jill, 680
Turner, Pamela, 437
Van Der Zwan, Jan, 652, 653, 713, 716, 718
Wahidi, Siavash, 403
Wen, Haifang, 281, 500, 501, 614, 657
West, Randy, 175, 286
Williams, Christopher, 322, 323, 656, 657, 680
Willis, J. Richard, 437, 462, 463, 464
Xie, Hongbin, 172, 173, 209, 210, 283, 284, 285
Yin, Fan, 177, 207, 208, 209, 210
Zanzotto, Ludo, 658, 681, 682
Zeiada, Waleed, 503
Zhang, Jun, 1
Zhou, Fujie, 367, 397, 398, 399, 400, 401, 402
--`,`,,,``,,```,,`,,,,,`,``,````-`-`,,`,,`,`,,`---
734


ASphalt

Încărcat de

Informații document

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

ASphalt

Încărcat de

Drepturi de autor:

Formate disponibile

Volume 82

DEStech Publications, Inc.

Copyright DEStech Publications, Inc.

Copyright 2013 by the Association of Asphalt Paving Technologists

No part of this publication may be reproduced, stored in a

Printed in the United States of America

ISSN No. 0270-2932

Copyright DEStech Publications, Inc.

1974 Williamsburg, Vol. 43 15.00 1999 Chicago, Vol. 68 75.00

"The Association of Asphalt Paving Technologists is not responsible for the

Reprints of Articles included in Volumes which are out-of-print may be

Printed in the United States of America by DEStech, Lancaster, PA, US

Copyright DEStech Publications, Inc.

W.J. Emmons Annual Award

The Board of Directors announces the recipients of the W.J. Emmons

Effects of Interface Condition Characteristics on Open-Graded

Yu Chen, Gabriele Tebaldi, Reynaldo Roque, George Lopp and

Laboratory Evaluation of Asphalt Mixtures Containing Dry Added

Gaylon L. Baumgardner, J. Michael Hemsley, Walter Jordan III,

AAPT Scholarship Fund

The Board of Directors announces the 2013 winners of the AAPT

Matthew Gersch, Auburn University

Mirella Villani, Delft University of Technology

Copyright DEStech Publications, Inc.

2013-2014 Board of Directors

President: Ramon Bonaquist

Copyright DEStech Publications, Inc.

It is my pleasure to recognize Mr. Prithvi (Ken) Kandhal who has

Kens interest in asphalt pavements began in India where he was first

Copyright DEStech Publications, Inc.

Copyright DEStech Publications, Inc.

Ken is very worthy of this recognition and it is now my pleasure to

Prithvi S. Kandhal Remarks

I am humbled by this honor bestowed upon me especially when it

It is time to reflect back on circumstances and people who helped

As a young highway engineer in a desert region of India, which did

Having developed interest in asphalt, I came to Iowa State University

Copyright DEStech Publications, Inc.

Then in 1970 I joined the Pennsylvania Department of Transportation

Fortunately, in Penn DOT I was in charge of almost all asphalt-related

There was another motivation to write research papers. Every time a

Copyright DEStech Publications, Inc.

As Associate Director of NCAT and a member of the graduate faculty

While at NCAT, I was also fortunate to work with Freddy Roberts,

Copyright DEStech Publications, Inc.

I am also thankful to my wife, Uma, who had to put up with my

In closing, I again thank the AAPT Board of Directors and all

Copyright DEStech Publications, Inc.

Review and Implications of IARC 103 for the Asphalt Pavement

Anthony J. Kriech, Director of Research, Heritage Research

Bitumen (asphalt in the U.S.) was reviewed by the International

Copyright DEStech Publications, Inc.

Copyright DEStech Publications, Inc.

Copyright DEStech Publications, Inc.

Copyright DEStech Publications, Inc.

MENA I. SOULIMAN, WALEED ZEIADA, MICHAEL MAMLOUK

Copyright DEStech Publications, Inc.

Using Small Scale Specimens for AMPT Dynamic Modulus and

A Mechanistic Permanent Deformation Model for Asphalt

SYMPOSIUMSUSTAINABLE ASPHALT TECHNOLOGIES

Alternative Binders for the Production of Bituminous

Recycling with RAP, RAS and Secondary Aggregates . . . . . . . . 661

Sustainable Energy Use in Asphalt Production . . . . . . . . . . . . 683

Binder Additives For Warm Mix Asphalt Technology . . . . . . . . 685