Computers % Strrcrures Vol. 40, No. 2, pp. 313-327, 1991 s3.00+ 0.

00
0045-7949/91
Printedin GreatBritain. Pergamon Pressplc

NUMERICAL ANALYSIS OF HIGH STRAIN RATE

CONCRETE DIRECT TENSION TESTS

J. W. TEDESCO,~ C. A. Ross,$ P. B. MCGILL? and B. P. O’NEILt

tCivi1 Engineering Department, Auburn University, Auburn, AL 36849, U.S.A.
jU.S. Air Force Engineering and Services Center, Tyndall Air Force Base, FL 32403, U.S.A.

Abstract-Direct tension tests of plain concrete specimens were conducted on a split-Hopkinson pressure
bar (SHPB) to investigate the effects of increasing strain rate on the tensile strength of concrete.
A comprehensive finite element method (FEM) study was performed on the SHPB experiments. Both
linear and nonlinear analyses were conducted. The results of the numerical analyses disclose the dynamic
states of stress in the specimen prior to failure as well as the mode of failure.

1. INTRODUCTION range. To emphasize this statement the American

Society of Testing Materials (ASTM) does not recog-
The effect of increasing strain rate on the strength nize a standard method for testing mortar or concrete
properties of many materials has been recognized in direct tension. The splitting cylinder method, an
for a number of years. Efforts to experimentally test indirect tensile test, is recognized by ASTM as a
materials at high strain rates using bar impact began concrete tensile strength test in the low-strain-rate
approximately 75 years ago. Most materials, with the range.
exception of some work-hardened aluminum alloys, The direct tension test has seldom been used to
experience some increase in strength with increases in evaluate the tensile strength of concrete. This is
load or strain rate. These increases in strength begin because of the difficulties of holding the specimens
to appear in the strain rate range of lo-103/sec. to achieve axial tension and the uncertainties of
Mortar and concrete compressive tests are easier to secondary stresses induced by the holding devices.
conduct than tensile tests and some experimental Recently, however, direct tension tests on plain con-
compressive strength properties in the strain rate crete specimens, using a split-Hopkinson pressure
range of lo-‘-lO/sec have been determined. How- bar (SHPB), have been successfully conducted at
ever, in the intermediate strain rate range of strain rates between 10 and 103/sec [4]. However, a
IO-i-lO/sec, testing is very difficult and compressive critical shortcoming with experiments performed
data is scarce in this region. on the SHPB is the inability to ascertain the stress
In the study of conventional weapons’ effects upon condition in the material specimen at failure. There-
structures, loading pulses having rise times in the fore, in an effort to analyze the dyamic stress
vicinity of 1.Omsec may impose strain rates in the distribution in high rate SHPB direct tension tests,
range of 10 up to 103/sec on materials and structures. a comprehensive finite element method (FEM) study
Therefore, the understanding of material response was conducted [S] through implementation of the
to high amplitude, short duration, impulse loads is ADINA [6] computer programs. Both linear and
essential in the analysis and design of protective nonlinear analyses were performed.
shelters [l]. Moreover, very little data is available on
the tensile strength of concrete and mortar at strain
rates associated with close-proximity conventional 2. DIRECT TENSION TESTS ON THE SHPB
weapons’ explosions, thus prompting several recent
research efforts in that area [2,3]. An illustration of the SHPB device used in the
Tensile strength testing of concrete and mortar is experimental study [4] is shown in Fig. 1. The device
difficult, even in the low or quasi-static strain rate is operated by the Engineering and Services Center,

101.6 mm to 152.4mm. space?

7.15 mm. ID Pillow Block [Split]
Dodge Journal Erg. 110 Req’d] 1.52 m. Throw Off Bar

3.66m. lncldent

Fig. 1. Illustration of AFESC 51 mm diameter split-Hopkinson pressure bar.

313
314 J. W. TEDEXOer al.

I----- zS,::~~e-72LII----:p,“,I
-
+ f 1 Ear No.2 +

Both Barsr50.8 mm Diam. f

Striker Bar
~1
TanSiDfl

Strain Gage Strain Guage

Oscilloscope
Conditioner Conditioner
L r 1
Fig. 2. Schematic of split-Hopkinson pressure bar.

Tyndall AFB, Florida. The pressure bars are con-
structed of PH 13-8 MO stainless steel. Each pressure
bar is 51 mm in diameter. The principles of operation
of the compressive SHPB, illustrated in Fig. 2, are
detailed in refs [4,5], and these same principles apply
to the direct tension SHPB. The configuration of the
SHPB arrangement employed in the direct tension
study reported in ref. [4] is also illustrated in Fig. 2.
The tensile loading m~hanism consists of a hollow
cylindrical striker bar sliding on the compressive
transmitter bar (bar 2) of the SHPB. The striker bar
impacts a tup threaded into the end of what becomes
the tensile incident bar (bar 2). A tensile stress wave
then propagates toward the specimen, cemented
between the two bars.
The direct tension specimen, termed a square-
notch specimen, is illustrated in Fig. 3. The specimen
quasi-static material properties for the concrete spec-
imens are presented in Table 1. The notch, located
at mid-length of the specimen, is 3.175 mm square.
All specimens were cemented to the ends of the SHPB
with a non-epoxy concrete cement. The bar surfaces
and specimen surfaces were cleaned in a manner
similar to that used for surface cleaning before the
placement of foil-resistant strain gages.
SHPB tests were conducted for three different
loading conditions. The stress versus time histories
for these cases are illustrated in Figs 4-6. It is
assumed that the dynamic tensile strength at the
notch, f,, is proportional to the transmitted stress,
err through the expression

//-\\\
is 51 mm in diameter and 51 mm in length. The

/fn
I+
1’
\
\\
i
i!
where A, is the area ratio,
the Hopkinson bar, and D,
specimen at the notch.
Additionally, the loading
Db is the diameter of
is the diameter of the

rate, ci, and the strain

\ / rate, i, in the specimen can be estimated from the
\

‘. -” I
expressions [7]
~
1 .L
P=-i (a)
&=-
z
(3)

and
if
g =-.-, (4)
3.1?5%0508
I+- Etl

Table I. Concreate material properties

~ 3J75%0508 Concrete Property Value

E, Young‘s Wodulu8 (GPa) 37.93

f:, Compressive Stress (WPa) 57.7

tb) f;, Tensile Stress (WPs) 4.53

Fig. 3. Square notch direct tension specimen: (a) end view, 7, Unit Weight (kg/m’) 2405
(b) plan view.
 High strain rate concrete tension tests 315

JO.0

20.0
INCIUENT- ; ‘~-TRANSMITTED X4

10.0
(0
$

ui 0.0

iz
:
VI -10.0
INVERSE OF THE TRANSMITTED->
-REFLECTED
-20.0

-30.0

0.0s 25o.ous 5oo.ous 75o.ous

TIME
Fig. 4. Square notch test data trace for load case 1.

where r is the time lag between the start of the
reflected stress wave and the maximum transmitted
stress, and E, is the modulus of elasticity of the
specimen. A summary of the results obtained from
the SHPB tests is presented in Table 2.
3. STATIC ANALYSES
A partial illustration of the axisymmetric FEM
model employed in the study is depicted in Fig. 7.
The incident and transmitter bar are each comprised
of 1633 eight-node, two-dimensional finite elements.
The specimen is comprised of 476 elements. The
loading is applied uniformly at the free end of the
incident bar.
In order to ascertain the stress distribution along
various cross-sections through the specimen, and to
identify and quantify areas of stress concentration,
a linear static analysis was conducted. The concrete
material properties used in the static analysis are
presented in Table 1. A load of 29 MPa was applied
at the free end of the incident bar.
The distribution of the longitudinal stress, uz, in
the specimen along a cross-section adjacent to the

<-REFLECTED

-50.0 -

,,,,,, ~~ ,,,,,,;
I I I I I I
-250 .ous 0.0s 250 .Ous 500.0us 75o.ous
TIME
Fig. 5. Square notch test data trace for load case 2.
316 J. W. TEDESCOet af.

Table 2. Summary of SHPB results

Load Cam Incident Transmitted Dynamic Loading Rata Strain Rate

Nu&er Stress strslss Tensile Stress C (MPa/aac) & (sac-‘)

or (MW oT (MW f- (MW

1 26.5 4.5 9.31 6.01*10s 4.9

2 67.0 4.0 7.59 1.48*10* 5.3

3 75.0 4.1 7.93 2.08*106 5.8

incident bar is presented in Fig. 8. A similar plot of are also indicated in the specimen along the cross
the a, stress distribution at the transmitter end of the section immediately adjacent to the incident and
specimen is depicted in Fig. 9. Plots of the 0; stress transmiter bars (Figs 8 and 9, respectively). This
distribution along cross-sections of the specimen phenomenon is attributed to the fact the specimens
taken through the roots of the notch are presented in are rigidly attached to the bars, thus creating a
Figs 10 and 11. Poisson effect.
Figures 8-11 indicate a nonlinear stress distri-
3.2. Nonlinear analysis
bution in the specimen. This nonlinearity is more
pronounced at the sections through the notch, where A static nonlinear material analysis was conducted
a stress concentration factor of 2.93 was calculated at in order to identify possible failure m~hanisms in the
the root of the notch. Regions of stress concentration specimen. The concrete material model employed in

-50.0 - <-REFLECTElI

TIME

Fig. 6. Square notch test data trace for load case 3.

2
II--+
Y

IncidentBar Specimen TransmitterBar

Axis of
--- ---
Symmevy
1.3208m 0.0508 m 1.3208m
A h b
4 Y Y

Fig. 7. FEM model of square notch specimen and portion of the SHPB.
 32.0

31.0

P
z
30.0

b”

29.0

28.0

Distanca From Axis of Symmetry (mm)

Fig. 8. Horizontal stress (a,) distribution in cross-section at the incident end of specimen, linear static
analysis.

Incidentl3~~xwwniner Bar

32.0

31.0

3
3 30.0
d

29.0

Distance From Axis of Symmetry (mm)

Fig. 9. Horizontal stress (a,) dist~bution in cross-section at the ~smitter end of specimen, linear static
analysis.

317
31s J. W. TEDEXO et al.

100.0 I /

80.0 -

60.0 -

0.0 10.0 20.0 30.0

Distance From Axis of Symmetry (mm)

Fig. 10. Horizontal stress (a,) distribution in cross-section at root of notch, linear static analysis.

loo+o
+

0.0 1 1 I I
0.0 10.0 20.0 30.0

Distance From Axis of Symmetry (mm)

Fig. 11. Horizontal stress (uI) distribution in cross-section at root of notch, linear static analysis.
 High strain rate concrete tension tests 319

Tension

Compression

I
t-
%

_ t-
%
Fig. 12. Uniaxial stressstrain relation used in concrete model.

the nonlinear analysis is a hypoelastic model based
on a uniaxial stress-strain relation (Fig. 12) that is
generalized to take biaxial and triaxial conditions
into account. The model employs three basic features
to describe the material behavior: (i) a nonlinear
stress-strain relation including strain softening to
allow for weakening of the material under increasing
compressive stresses; (ii) a failure envelope that
defines cracking in tension and crushing in com-
pression; and (iii) a strategy to model post-cracking
and crushing behavior of the material.
In order to establish the uniaxial stress-strain
law accounting for multiaxial stress conditions,
an appropriate failure envelope must be employed.
Since failure of the specimen is tension dominated,
the tension failure envelope depicted in Fig. 13 was
used in the concrete model. To identify whether
the material has failed, the principal stresses are

Fig. 13. Three-dimensional tension failure envelope of concrete model.

320 J. W. TEDESCO
et al.

Table 3. Concrete model parameters

Paranmtar Value

EO, Initial tangent modulus (GPa) 37.93

at, uniaxial cut-off tensile strength (ma) 4.53

o,, Uniaxial maximum compressive strength (Mpa) 57.7

CT",Uniaxial ultimata compressive strength (Mpa) 28.9

e,, Uniaxial compressive strain at 0, 0.002

e,, Uniaxial compressive strain at 0, 0.003

Q. spZr o#,, Principal stresses in

directions 1,2,3 respectively Figure 13

CT;, U&axial cut-off tensile stress under

multiaxial conditions Figure 13

CT:, Uniaxiai compressive failure stress under

multiaxial conditions Figura 13

used to locate the current stress state in the failure
envelope. The tensile strength of the material in a
principal direction does not change with the introduc-
tion of tensile stresses in other principal directions.
The pertinent material parameters for the failure
envelope and the uniaxial stress-strain relation are
summarized in Table 3.
The loading was applied monotomically over 16
load steps at the free end of the incident bar as
illustrated in Fig. 14. First cracking appeared at the
root of the notch after load step 4, as illustrated on
Fig. 15a, at a load of approximately 1 MPa. Failure
of the specimen occurred by fracture through the
notched portion of the cross-section after load step 8
at an approximate loading of 2 MPa. The failure
sequence for the specimen is illustrated in Fig. 15. The
profile of the longitudinal stress, B,, along the cross
section through the notch is presented in Fig. 16. the
mode of failure observed in the nonlinear analysis
supports the results obtained in the linear static
analysis.
4. DYNAMIC ANALYSES
Dynamic nonlinear materiai analyses were con-
ducted on the SHPB direct tension specimens pre-
viously described. The nonlinear material model is
the same as described for the nonlinear static analyses.
The loading conditions for the dynamic analyses
were determined from the SHPB data curves of the
incident, reflected, and transmitted stress versus time
histories presented in Figs 4-6. The loading function

0.0 4.0 8.0 12.0 16.0

Load Step

Fig. 14. Loading sequence for static nonlinear analysis.

 High strain rate concrete tension tests 321

(4

63 (4
Fig. 15. Cracking sequence for square notch specimen, static nonlinear analysis.

3.0 -

2.0 -

Distance From Axis of Symmetry [mm)

Fig. 16. Horizontal stress (cr,) distribution in cross-section at the notch, static nonlinear analysis.
322 J. W. TEDESCO et al.

Table 4. Load function parameters

Load Case Rim Time Ramp Time Total Time Stress Level
Number t, (wee) tz (wee) 4 M-o) PO (ma)

1 45 145 190 26.5

2 45 145 190 67

3 35 135 170 75

used in the numerical analyses, representative of the within that time increment. The maximum time step
incident stress time history, is depicted in Fig. 17. is defined by
The rise time, ti, the duration of the ramp loading,
tr, the total time of duration, t,, and the stress level,
(At),,, = $
P,,, for the three load cases are summarized in
Table 4.
For problems in which an elastic body is subjected where L, is the length of an element in the direction
to a short duration impulse loading, propagation of of wave propagation, and c is the velocity of wave
stress-strain waves through the body must be con- propagation, given by
sidered in formulating the solution. Modal analyses
E
generally do not yield cost effective, accurate results
for wave propagation problems. Therefore, a direct
numerical integration procedure must be utilized.
c=
J(7> ’
In the present study the Newmark method of It has been determined from previous experience [8]
implicit time integration with a consistent mass for- that a time step of
mulation is employed. The time step selected for
temporal integration in a wave propagation problem At <&W,,,, (7)
is critical to the accuracy and stability of the solution.
Since the Newmark method is unconditionally stable, yields accurate results. In the present study a
selection of the time step can be based entirely upon time step of AL\~
= 0.1 psec was used for all dynamic
accuracy considerations. In a wave propagation analyses.
problem, the maximum time step is related to wave
speed in the material and element size. The maximum 4.1. Results of nonlinear analysis
time step is selected so that the stress wave propagates The results obtained for load case 2 are presented.
the distance between element integration points Time histories for the longitudinal stress, u,, at

80.0

70.0

60.0

0.0 50.0 100.0 150.0 200.0

p SBC

Fig. 17. Typical loading function used in dynamic analyses.

 High strain rate concrete tension tests 323

three locations in the specimen are illustrated in
Figs 18-20. Profiles for the longitudinal stress at two
cross sections in the specimen at selected times are
presented in Figs 21 and 22.
The cracking sequence for the specimen, from the
initiation of the first crack until failure, is illustrated
in Fig. 23. The first crack occurs along the axis of
symmetry at the interface of the specimen and inci-
dent bar (Fig. 23a), at time t = 258 psec. This result
is also evident in the stress time history presented in
Fig. 18. Cracks do not appear in the notch until time
t = 260 psec, as shown in Fig. 23b. At this same time,
however, cracking at the incident end of the specimen
has become more extensive. At time t = 265 psec,
cracking is extensive in both the incident end of the
cross-section as well as in the vicinity of the notch
(Fig. 23~). Finally, at t = 295 psec, the specimen fails
simultaneously at the notch and at the incident end
(Fig. 21d).
Table 5. Experimentally and numerically determined strain
rates
Load Case
Number
1 4.9 6.10
2 5.3 6.78
3 5.8 9.32
Similar nonlinear analyses were conducted for load
cases 1 and 3. Failure for load case 1 occurred at the
notch, while the failure predicted in load case 3
occurred at the incident end of the specimen. In all
three cases, the mode of failure predicted by the FEM
analyses is consistent with that observed in the SHPB
experiments [4]. These results indicate that the mode
of failure is highly sensitive to the rate of loading.
5. DISCUSSION OF RESULTS
In all three cases, the specimens failed or fractured
during the rise time of the loading (incident) pulse.
This early failure is evident in the strain gage traces
presented in Figs 4-6. For load case 1 (Fig. 4), the
transmitted pulse is inverted and illustrates that the
peak of the transmitted pulse occurs in the rise time
of the reelected pulse. The occurrence of failure early
in the incident pulse precludes the use of the reflected
pulse to calculate the strain rate in the specimen.
However, the peak reflected stress is indicative of
the concrete tensile strength and the slope of the
transmitted time curve is a measure of the strain rate,
assuming a linear elastic material. The strain rates for
the SHPB experiments were calculated by dividing
the slope of the straight line segment of the trans-
mitted stress by the quasi-static elastic modulus.
These experimentally determined strain rates are in
good agreement with those predicted by the FEM
analyses as shown in Table 5.

IncidentBar / n ) Transmitter Bar

6.0 . I /

Time (p set)
Fig. 18. Time history for horizontal stress (a,) at incident end of specimen, load case 2.
324 J. W. TEDESCO
et al.

6.0 I I

4.0

2.0

0.0

-2.0

-4.0

-6.0
0.0 too.0 200‘0 300.0

Time (r WC)

Fig. 19. Time history for horizontal stress (a,) at root of notch, load case 2.

Incident Bar 1 f 1 Transmitter Bat

6.0 I I

4.0 -

100.0 200.0 300.0

Time (Ir Set)

Fig. 20. Time history for horizontal stress (u2) at root of notch, load case 2.
 High strain rate concrete tension tests 325

Incident B~~nxnsmitter Bar

3.0 -

1.0 -

2 0.0 ,, I : : ; ~

B
-1.0 -
d - Time 257 wsec
-2.0 +-----t Time 260 lr set
- lime 266 in set

-3.0

-4.0 -

-5.0 , I
0.0 10.0 20.0 30.0

Distance From Axis of Symmetry (mm)

Fig. 21. Profiles for horizontal stress (u,) at incident end of specimen, load case 2.

0.0
2
3
-1.0 -
b” - Time 257 JAset
-2.0 - - Time 260 c set
- Time 266 IL set

-3.0 -

-4.0 -

-5.0
0.0 10.0 20.0 30.0

Distance From Axis of Symmetry (mm)

Fig. 22. Profiles for horizontal stress (cJ,) at notched cross section, load case 2.
326 J. W. TEomco et al.

04 fb)

(cl W
Fig. 23. Cracking sequence for square notch specimen, load case 2: (a) t = 258 psec, (b) i = 260 psec,
(c) t = 265 psec, (d) r = 295 psec.

An interesting aspect of the SHPB experiments is the FEM analyses do indicate the presence of these
the occurrence of oscillations at the trailing end of oscillations at the trailing end of the transmitted
the transmitted pulse. Upon fracture of the specimen, pulses, as illustrated in Figs 18-20.
an unloading pulse is transmitted toward the speci-
men/transmitter bar interface. Due to the character-
istic impedance mismatch at this interface, a portion 6. CONCLUSIONS
of the pulse is trapped between the segment of the
The results of the linear static analysis indicate that
specimen which is cemented to the transmitter bar
areas of stress concentration exist in the vicinity of
and the fracture plane through the notch. This pulse
the notch and at the sp~imen-incident bar interface.
oscillates back and forth within the broken specimen,
This prediction is confirmed by the results of the
but decays rather rapidly. The shape of the oscillation
static nonlinear analysis which reveals cracking in
is different in each load case. For the lower strain
both those areas of stress concentration. However,
rate test, load case 1, the specimen fractured at the
failure occurs by fracture of the cross section through
notch giving rise to a comparatively larger oscillation
the notch.
(Fig. 4) than was exhibited in the other two load
In the dynamic nonlinear analyses, the predicted
cases. For the intermediate strain rate test, load
mode of failure is dependent on the rate of loading.
case 2, the specimen fractured at both the notch
For a low strain rate, failure occurs at the notch; for
and the incident end of the specimen resulting in a
an intermediate strain rate, failure occurs at both the
different signature at the end of the transmitted pulse
notch and the incident end of the specimen; and for
(Fig. 5). The specimen in the high strain rate test, load
a high strain rate, failure occurs at the incident end
ease 3, fractured at the incident end, resulting in a
of the specimen. Regardless of the rate of loading,
longer length of fractured specimen which remained
however, failure always occurs during the rise time of
attached to the trans~tter bar. This in turn yielded
the incident pulse.
a different pattern at the end of the transmitted pulse
(Fig. 6).
It is doubtful that this experimentally observed Acknowledgements-The experimental work was conducted
phenomenon has any significant influence on the at the Air Force Engineering and Services Center, Tyndall
tensile strength of the concrete. However, it does Air Force Base, Florida. The numerical analysis was spon-
sored by a research contract from the Air Force Engineer-
serve as a check on the ability of the FEM simu- ing and Services Center, Contract No. FO8635-90-C-0120.
lations to accurately predict the true behavior of Computational resources for the numerical analyses were
the experiments. Indeed, the results obtained from provided by the Alabama Supercomputer Network.
 High strain rate concrete tension tests
REFERENCES
1. Fundamentals of protective design for conventional
weapons. Department of the Army, Waterways
Experiment Station, Corps of Engineers, Vicksburg,
MS (1984).
2. C. A. Ross, S. T. Kuennen and J. W. Tedesco,
Experimental and numerical analysis of high strain-rate
concrete tensile tests, pp. 353-364. In Micromechnnics of
Failure of Quasi-Brittle Materials, Elsevier, New York
(1990).
3. C. A. Ross, S. T. Kuennen and W. S. Strickland,
High strain rate effects on tensile strength of concrete.
Fourth International Symposium on the Interaction of
Non-nuclear Munitions with Structures, Panama City
Beach, FL, April 17-21 (1989).
327
4. C. A. Ross, Split-Hopkinson pressure bar tests. Final
Report, Air Force Engineering and Services Center,
Tyndall AFB, Fl, ESL-TR-88-82 (1989).
5. J. W. Tedesco, Numerical analysis of dynamic splitting-
tensile and direct tension test. Final Report, Air Force
Engineering and Services Center, Tyndall AFB, FL,
Contract FO8635-88-C-0195 (1990).
6. ADINA-A Finite Element Program for the Automatic
Dynamic Incremental Nonlinear Analysis. Report
ARD 87-1, ADINA R&D Inc., Watertown, MA (1987).
7. J. W. Tedesco, C. A. Ross and R. M. Brunair,
Numerical analysis of dynamic split cylinder tests.
Comput. Struct. 32, 609-624 (1989).
8. J. W. Tedesco, Stress wave propagation in layered
media. Final Report, AFOSR, Contract F49620-85-c
0013 (1988).