Sunteți pe pagina 1din 201

Evaluation of Floor Vibrations in Concrete Buildings

Justin Clark

A thesis
submitted in partial fulfillment of the
requirements for the degree of

Masters of Science in Civil Engineering

University of Washington

2009

Program Authorized to Offer Degree:

Civil and Environmental Engineering Department


University of Washington
Graduate School

This is to certify that I have examined this copy of a master’s thesis by

Justin Clark

and have found that it is complete and satisfactory in all respects,


and that any and all revisions required by the final
examining committee have been made.

Committee Members:

John Stanton

Marc Eberhard

Laura Lowes

Date
In presenting this thesis in partial fulfillment of the requirements for a Master’s degree at the
University of Washington, I agree that the Library shall make its copies freely available for
inspection. I further agree that extensive copying of this thesis is allowable only for scholarly
purposes consistent with “fair use” as prescribed by the U.S. Copyright Law. Any other
reproduction for any purposes or by any means shall not be allowed without my written
permission.

Signature
Date
Table of Contents
List of Figures ................................................................................................................................ iv
List of Tables ............................................................................................................................... viii
Chapter 1: Intorduction and Previous Work ................................................................................... 1
1.1 Background ........................................................................................................................... 1
1.2 Previous Work ...................................................................................................................... 3
1.2.1 Acceptable Levels of Vibrations .................................................................................... 3
1.2.2 Procedures for Predicting Building Vibrations .............................................................. 7
1.3 Problems with Current Practice .......................................................................................... 10
1.4 Objectives of Research ....................................................................................................... 13
Chapter 2: Data Measurement ...................................................................................................... 14
2.1 Vibration Instrumentation and Testing Procedure .............................................................. 14
2.1.1 Fast Fourier Transform Process ................................................................................... 16
2.1.2 Human-Induced Loading Tests .................................................................................... 19
Ambient Test ..................................................................................................................... 19
Heel Drop Test .................................................................................................................. 20
Walking Test ..................................................................................................................... 23
2.2. Buildings Chosen for Testing ............................................................................................ 24
2.2.1 West Campus Parking Deck ........................................................................................ 25
2.2.2 University of Washington Faculty Club ...................................................................... 28
2.2.3 More Hall ..................................................................................................................... 31
2.2.4 Commercial Building #1(K1) ...................................................................................... 33
2.2.5 Commercial Building #2(K2) ...................................................................................... 35
2.3 Vibration Data ..................................................................................................................... 36
2.3.1 Test Results and Initial Analysis.................................................................................. 36
West Campus Parking Deck ............................................................................................. 37
UW Faculty Club .............................................................................................................. 42
Building K1 ....................................................................................................................... 47
Building K2 ....................................................................................................................... 60
2.3.2 Data Measurement Conclusions .................................................................................. 66
Chapter 3: Principles for Structural Modeling .............................................................................. 68
3.1 General Modeling Procedures............................................................................................. 68
3.1.1 Material Properties ....................................................................................................... 68
3.1.2 Geometry...................................................................................................................... 70

i
3.2 Representation of Loading .................................................................................................. 75
3.2.1 Heel Drop Load Simulation ......................................................................................... 75
Pulse Load ......................................................................................................................... 77
3.2.2 Walking Load Functions .............................................................................................. 79
3.3 Models of Tested Structures ............................................................................................... 81
3.3.1 West Campus Parking Deck ........................................................................................ 81
3.3.1.1 Initial Model.......................................................................................................... 81
3.3.1.2 Model Modifications ............................................................................................. 82
3.3.1.3 Analysis of Mode Shapes...................................................................................... 86
3.3.2 UW Faculty Club ......................................................................................................... 89
3.3.2.1 Initial Model.......................................................................................................... 89
3.3.2.2 Modal Identification.............................................................................................. 91
3.3.2.3 Model Modifications ............................................................................................. 93
3.3.2.4 Analysis of Mode Shapes...................................................................................... 95
3.3.3 Building K1 .................................................................................................................. 96
3.3.3.1 Initial Model.......................................................................................................... 96
3.3.3.2 Model Modifications ........................................................................................... 102
3.3.4 Building K2 ................................................................................................................ 112
3.3.4.1 Initial Model........................................................................................................ 112
3.3.4.2 Model Modifications ........................................................................................... 113
3.3.4.3 Analysis of Mode Shapes.................................................................................... 118
3.4. Modeling Conclusions ..................................................................................................... 118
Chapter 4: Modeling and Field Measurement Data Correlation ................................................. 120
4.1 West Campus Parking Deck ............................................................................................. 120
4.1.1 Heel Drop Test/Triangular Pulse ............................................................................... 121
4.1.2 Walking Test/Camel Hump ....................................................................................... 123
4.2 UW Faculty Club .............................................................................................................. 125
4.2.1 Heel Drop Test/Triangular Pulse ............................................................................... 125
4.2.2 Walking Test/Camel Hump ....................................................................................... 128
4.3 Building K1 ....................................................................................................................... 130
4.3.1 Heel Drop Test/Triangular Pulse ............................................................................... 130
4.3.2 Walking Test/Camel Hump ....................................................................................... 133
4.4 K2 ...................................................................................................................................... 134
4.4.1 Heel Drop Test/Triangular Pulse ............................................................................... 135

ii
4.4.2 Walking Test/Camel Hump ....................................................................................... 136
4.5 Building Comparison ........................................................................................................ 137
4.6 Comparison Conclusions and Sources of Error ................................................................ 140
Chapter 5: Summary and Conclusions........................................................................................ 142
5.1 Summary ........................................................................................................................... 142
5.2 Conclusions ....................................................................................................................... 142
5.3 Recommendations ............................................................................................................. 148
5.3.1 Recommendations for Practice .................................................................................. 148
5.3.1 Recommendations for Further Research .................................................................... 149
Appendices .................................................................................................................................. 151
Appendix A: Methods of Determining Damping of Structures ............................................. 151
Appendix B: Exterior Beam Effective Width Factor.............................................................. 158
Appendix C: Column Stiffness Determination ....................................................................... 161
Appendix D: Cardinal Point Analysis..................................................................................... 171
Appendix E: Heel Drop Load Simulation.............................................................................. 178
Bibliography ............................................................................................................................... 188

iii
List of Figures
Figure 1.2.1.1 .................................................................................................................................. 4
Figure 1.2.1.2 .................................................................................................................................. 5
Figure 1.2.1.3 .................................................................................................................................. 6
Figure 1.2.1.4 .................................................................................................................................. 6
Figure 1.2.2.1 .................................................................................................................................. 9
Figure 1.3.1 ................................................................................................................................... 12
Figure 2.1 ...................................................................................................................................... 15
Figure 2.1.1 ................................................................................................................................... 17
Figure 2.1.2 ................................................................................................................................... 18
Figure 2.1.3 ................................................................................................................................... 20
Figure 2.1.4 ................................................................................................................................... 21
Figure 2.1.5 ................................................................................................................................... 22
Figure 2.1.6 ................................................................................................................................... 23
Figure 2.1.7 ................................................................................................................................... 24
Figure 2.2.1 ................................................................................................................................... 26
Figure 2.2.2 ................................................................................................................................... 27
Figure 2.2.3 ................................................................................................................................... 29
Figure 2.2.4 ................................................................................................................................... 30
Figure 2.2.5 ................................................................................................................................... 32
Figure 2.2.6 ................................................................................................................................... 32
Figure 2.2.7 ................................................................................................................................... 34
Figure 2.2.8 ................................................................................................................................... 36
Figure 2.3.1 ................................................................................................................................... 37
Figure 2.3.2 ................................................................................................................................... 38
Figure 2.3.3 ................................................................................................................................... 38
Figure 2.3.4 ................................................................................................................................... 40
Figure 2.3.5 ................................................................................................................................... 41
Figure 2.3.6 ................................................................................................................................... 42
Figure 2.3.7 ................................................................................................................................... 43
Figure 2.3.8 ................................................................................................................................... 44
Figure 2.3.9 ................................................................................................................................... 45
Figure 2.3.10 ................................................................................................................................. 46
Figure 2.3.11 ................................................................................................................................. 48

iv
Figure 2.3.12 ................................................................................................................................. 49
Figure 2.3.13 ................................................................................................................................. 50
Figure 2.3.14 ................................................................................................................................. 51
Figure 2.3.15 ................................................................................................................................. 52
Figure 2.3.16 ................................................................................................................................. 53
Figure 2.3.17 ................................................................................................................................. 54
Figure 2.3.18 ................................................................................................................................. 55
Figure 2.3.19 ................................................................................................................................. 55
Figure 2.3.20 ................................................................................................................................. 57
Figure 2.3.21 ................................................................................................................................. 58
Figure 2.3.22 ................................................................................................................................. 58
Figure 2.3.23 ................................................................................................................................. 60
Figure 2.3.24 ................................................................................................................................. 61
Figure 2.3.25 ................................................................................................................................. 61
Figure 2.3.26 ................................................................................................................................. 62
Figure 2.3.27 ................................................................................................................................. 63
Figure 2.3.28 ................................................................................................................................. 64
Figure 2.3.29 ................................................................................................................................. 65
Figure 3.1.2.1 ................................................................................................................................ 72
Figure 3.1.2.2 ................................................................................................................................ 74
Figure 3.1.2.3 ................................................................................................................................ 75
Figure 3.2.1.1 ................................................................................................................................ 77
Figure 3.2.1.2 ................................................................................................................................ 79
Figure 3.2.2.1 ................................................................................................................................ 80
Figure 3.3.1.1.1 ............................................................................................................................. 82
Figure 3.3.1.2.2.1 .......................................................................................................................... 85
Figure 3.3.1.2.2.2 .......................................................................................................................... 86
Figure 3.3.1.3.1 ............................................................................................................................. 87
Figure 3.3.1.3.2 ............................................................................................................................. 87
Figure 3.3.1.3.3 ............................................................................................................................. 89
Figure 3.3.2.1.1 ............................................................................................................................. 89
Figure 3.3.2.1.2 ............................................................................................................................. 91
Figure 3.3.2.3.1 ............................................................................................................................. 94
Figure 3.3.2.3.2 ............................................................................................................................. 95

v
Figure 3.3.2.3.3 ............................................................................................................................. 96
Figure 3.3.3.1.1 ............................................................................................................................. 97
Figure 3.3.3.1.2 ............................................................................................................................. 98
Figure 3.3.3.1.3 ............................................................................................................................. 98
Figure 3.3.3.1.4 ............................................................................................................................. 99
Figure 3.3.3.1.5 ........................................................................................................................... 100
Figure 3.3.3.2.1 ........................................................................................................................... 103
Figure 3.3.3.2.2 ........................................................................................................................... 104
Figure 3.3.3.2.3 ........................................................................................................................... 105
Figure 3.3.3.2.2.1 ........................................................................................................................ 109
Figure 3.3.3.2.2.2 ........................................................................................................................ 109
Figure 3.3.3.2.2.3 ........................................................................................................................ 110
Figure 3.3.3.2.2.4 ........................................................................................................................ 111
Figure 3.3.4.1.1 ........................................................................................................................... 113
Figure 3.3.4.2.1 ........................................................................................................................... 115
Figure 3.3.4.2.2 ........................................................................................................................... 116
Figure 3.3.4.2.3 ........................................................................................................................... 116
Figure 3.3.4.2.4 ........................................................................................................................... 118
Figure 4.1.1.1 .............................................................................................................................. 122
Figure 4.1.2.1 .............................................................................................................................. 124
Figure 4.2.1.1 .............................................................................................................................. 126
Figure 4.2.1.2 .............................................................................................................................. 127
Figure 4.2.1.3 .............................................................................................................................. 128
Figure 4.2.2.1 .............................................................................................................................. 129
Figure 4.3.1.1 .............................................................................................................................. 132
Figure 4.3.1.2 .............................................................................................................................. 133
Figure 4.3.2.1 .............................................................................................................................. 134
Figure 4.4.1.1 .............................................................................................................................. 136
Figure 4.4.2.1 .............................................................................................................................. 137
Figure A1 .................................................................................................................................... 151
Figure A2 .................................................................................................................................... 152
Figure A3 .................................................................................................................................... 153
Figure A4 .................................................................................................................................... 155
Figure A5 .................................................................................................................................... 157

vi
Figure C1.1 ................................................................................................................................. 162
Figure C1.2 ................................................................................................................................. 163
Figure C1.3 ................................................................................................................................. 164
Figure C1.4 ................................................................................................................................. 165
Figure C2.1 ................................................................................................................................. 168
Figure C1.2 ................................................................................................................................. 163
Figure C1.3 ................................................................................................................................. 164
Figure C1.4 ................................................................................................................................. 165
Figure D1 ................................................................................................................................... 171
Figure D1.1 ................................................................................................................................. 172
Figure D1.2 ................................................................................................................................. 173
Figure D2.1 ................................................................................................................................. 175
Figure E1 .................................................................................................................................... 179
Figure E2 .................................................................................................................................... 184
Figure E3 .................................................................................................................................... 185
Figure E4 .................................................................................................................................... 185
Figure E5 .................................................................................................................................... 187
Figure E6 .................................................................................................................................... 187

vii
List of Tables
Table 2.2.1 .................................................................................................................................... 27
Table 2.3.1 .................................................................................................................................... 37
Table 2.3.2 .................................................................................................................................... 41
Table 2.3.3 .................................................................................................................................... 42
Table 2.3.4 .................................................................................................................................... 47
Table 2.3.5 .................................................................................................................................... 48
Table 2.3.6 .................................................................................................................................... 59
Table 2.3.7 .................................................................................................................................... 60
Table 2.3.8 .................................................................................................................................... 66
Table 3.1.2.1 ................................................................................................................................. 74
Table 3.3.1.1.1 .............................................................................................................................. 82
Table 3.3.1.2.1 .............................................................................................................................. 83
Table 3.3.1.2.1.1 ........................................................................................................................... 84
Table 3.3.1.3.1 .............................................................................................................................. 88
Table 3.3.2.3.1 .............................................................................................................................. 95
Table 3.3.3.2.1 ............................................................................................................................ 104
Table 3.3.3.2.2.1 ......................................................................................................................... 108
Table 3.3.3.2.2.2 ......................................................................................................................... 111
Table 3.3.4.2.1 ............................................................................................................................ 115
Table 3.3.4.2.2 ............................................................................................................................ 117
Table 4.1.1.1 ............................................................................................................................... 121
Table 4.1.1.2 ............................................................................................................................... 122
Table 4.1.2.1 ............................................................................................................................... 123
Table 4.2.1.1 ............................................................................................................................... 125
Table 4.2.1.2 ............................................................................................................................... 126
Table 4.2.2.1 ............................................................................................................................... 128
Table 4.2.2.2 ............................................................................................................................... 128
Table 4.3.1.1 ............................................................................................................................... 130
Table 4.3.1.2 ............................................................................................................................... 130
Table 4.3.2.1 ............................................................................................................................... 134
Table 4.4.1.1 ............................................................................................................................... 135
Table 4.4.1.2 ............................................................................................................................... 135
Table 4.4.2.1 ............................................................................................................................... 137

viii
Table 4.4.2.2 ............................................................................................................................... 137
Table 4.5.1 .................................................................................................................................. 138
Table 4.5.2 .................................................................................................................................. 139
Table A1...................................................................................................................................... 154
Table B1 ...................................................................................................................................... 158
Table C1.1 ................................................................................................................................... 165
Table C1.2 ................................................................................................................................... 167
Table C2.1 ................................................................................................................................... 168
Table C3.1 ................................................................................................................................... 170
Table D1.1................................................................................................................................... 174
Table D2.1................................................................................................................................... 176
Table E1 ...................................................................................................................................... 181
Table E2 ...................................................................................................................................... 181
Table E3 ...................................................................................................................................... 182
Table E4 ...................................................................................................................................... 182
Table E5 ...................................................................................................................................... 183
Table E6 ...................................................................................................................................... 186

ix
ACKNOWLEDGEMENTS

I would like to thank Dr. John Stanton for his continued guidance throughout this
research. I would also like to thank Dr. Andy Taylor and David Webster for
lending their professional expertise and knowledge. I finally would like to
express my thanks to my friends and family for their understanding and support
for me throughout this process.

x
1

CHAPTER 1: INTRODUCTION AND PREVIOUS WORK


1.1 Background
For most of the twentieth century, vibrations in buildings were seldom a problem. Spans were

short, floors were thick, and the resulting motions were innocuously small. However, over the

last three or four decades several changes have taken place that have caused vibrations to play a

much larger role in determining the fitness of a building for use.

The first is the advent of high strength materials. In materials such as steel and concrete, higher

strengths can now be obtained without a proportionate increase in price, so the stronger materials

are cost-effective and have become the designer’s choice. However, the higher strength is not

accompanied by a comparable increase in modulus, so the service strains are higher. In steel, for

example, Young’s modulus is essentially independent of steel strength, so a 50% increase in

strength leads to a 50% increase in strain and deflection under a given load. Such a reduction in

static stiffness also carries over into the dynamic field, where it translates into larger vibration

amplitudes.

Seismic design requirements have also had an influence. They have grown more stringent over

time. One way to mitigate the effects of earthquake motions is to reduce the inertial mass, and

the easiest way to do that is to make the floors thinner, because they embody almost half of the

structural mass in a typical commercial building. Thinner floors are usually more flexible, again

leading to larger vibration amplitudes. This trend is easily seen in the cold-formed metal-deck

floor systems that are almost universally used for steel framed buildings today. Unfortunately

the lower inertial mass that is desirable for seismic purposes is counter-productive for floor

vibrations, where mass is an effective means of limiting vibrations. As an example, mass has
2

deliberately been added to the foundations of some recent concert halls to prevent ground borne

vibrations from entering the structure (Glover 1998).

The pressure to maximize rentable space and minimize initial capital costs has led to designs in

which the story height is minimized. By doing so, more stories can be fitted into a given

building height, and the costs of external cladding and internal services are reduced. The

reduction in story height is achieved by keeping the floor system as thin as possible, and, for

concrete buildings, this has led to a proliferation of floor systems without beams, commonly

referred to as flat plates. These are necessarily less stiff than floors with beams.

While the increase in vibration amplitude due to greater floor flexibility is easy to envision, the

flexibility also has another, more subtle, effect. The natural frequency of a modern floor is

typically lower than its counterparts of yesteryear, and that change brings it closer to the

frequency of excitation, which is often induced by human walking or other activity. Thus

resonance effects are more likely. The influence of reduced stiffness in altering the natural

frequency is slightly modified, but not completely overcome, by counteracting effects such as the

reduction in mass and the widespread use of post-tensioning in concrete floors. The latter

reduces cracking and therefore increases stiffness over the value to be expected from a

comparable reinforced concrete floor.

The foregoing structural changes combine to render buildings more vibration-prone, and this

characteristic can be thought of as a reduction in buildings’ capacity to resist vibrations.

However, changes have also taken place on the demand side. The explosion in the field of real
3

estate since the late 1980s has expanded the high-priced luxury end of the residential market,

especially in condominiums. With high prices come high expectations, including the expectation

that the building not suffer from excessive vibrations. The simultaneous growth of activities

such as health clubs and gyms has unfortunately made this expectation more difficult to achieve

if the exercise facilities are located in the condominium building. This is particularly true

because human perception of vibration is greatest at relatively low frequencies, in the 4-8 Hz

range, and it is in just that range that the fundamental frequency of many modern buildings lies.

In the commercial sector, pressures have also been building for a reduction in floor vibrations,

often driven by the need to conduct high-resolution work, such as in electronics or bio-

technology. Using a microscope or manufacturing computer chips is difficult on a vibrating

floor, so many potential tenants now demand that the building they rent has floor vibrations that

are bounded by stringent criteria.

This increase in demand for reduced vibrations combined with a reduction in capacity has put

pressure on designers from both sides, and has led to the need for a serious evaluation of

methods of predicting vibrations in buildings.

1.2 Previous Work


1.2.1 Acceptable Levels of Vibrations
The need to limit and control human-induced vibrations has stemmed from an increased human

sensitivity to floor vibrations due to the structural modifications described above. Figure 1.2.1.1

shows the baseline acceleration for human perception of vibration as a function of frequency, as

prescribed by the International Organization of Standardization (ISO). This plot indicates that

human sensitivity to vibration is greatest at frequencies from 4-8Hz. Therefore, the drop in the
4

fundamental frequency of newer buildings, moving some structures into this range, is increasing

the likelihood that vibration will be an issue for human comfort. The region of the plot below

4Hz has limited relevance, because few structures have natural frequencies that low.

The different curves represent the level of vibration that is likely to be acceptable to a person

engaged in the specified activity. An acceleration of 0.25%g may be completely unacceptable to

someone in the reading room of a library, whereas it would not even be noticed by a dancer in a

night club.

Figure 1.2.1.1: Peak vibration acceleration levels


acceptable for human comfort (Murray et al. (1997))

Figure 1.2.1.2 shows the Generic Vibration Criteria (VC) Curves for vibration-sensitive

equipment. The curves are based on building use and the types of equipment that are present in

the area where vibration is a concern. The curves range from the Workshop criterion curve,

which is appropriate for areas in which vibrations are unimportant, to VC-E, which applies to
5

areas with extremely sensitive equipment which require extraordinary dynamic stability (Gordon

1991).

Figure 1.2.1.2: Vibration Criteria Curves for vibration-sensitive equipment (Gordon 1991)

While the criterion controlling the level of vibrations acceptable for human comfort is given in

acceleration, the equipment criterion is given in terms of velocity for frequencies greater than

8Hz. The overall shapes of the curves for human and equipment vibration tolerance are in fact

the same; namely constant acceleration sensitivity for frequencies less than 8 Hz and constant

velocity sensitivity above 8Hz. This is shown in the plots of the two criteria in terms of velocity

in Figure 1.2.1.3 and in terms of acceleration in Figure 1.2.1.4. However, a comparison of the

“Office” curves for both sets of criteria in the two plots also shows that they differ in magnitude,

which illustrates the fact that no single set of universally accepted criteria exists.
6

1000000

100000 Gordon VC-E

Gordon VC-D

Gordon VC-C
Velocity (mu.in/sec)

10000
Gordon VC-B

Gordon VC-A

1000 Gordon Op th

100
4 8 16 32
frequency (Hz)

Figure 1.2.1.3: Vibration Velocity Acceptability for Human and Equipment Sensitivity

10

Gordon VC-E

Gordon VC-D
1
Gordon VC-C
Acceleration (%g)

Gordon VC-B
0.1
Gordon VC-A

Gordon Op th

0.01 Gordon Res


Day

0.001
4 8 16 32
frequency (Hz)

Figure 1.2.1.4: Vibration Acceleration Acceptability for Human and Equipment Sensitivity
7

1.2.2 Procedures for Predicting Building Vibrations


Various design guides have been developed to ensure that floor vibrations do not exceed the

levels set for human and equipment sensitivity. Predicting the level of vibration in buildings,

however, can be difficult. The American Institute of Steel Construction (AISC) Design Guide 11

(Murray et al. 1997) and Applied Technology Council (ATC) Design Guide 1 (Allen et al. 1999)

are two sets of guidelines that evaluate floor responses to human induced vibrations. The intent

of these design guides is to predict the level of response that a given floor will provide and to

evaluate its acceptability.

Both guides indicate that an adjustment factor for the concrete modulus of elasticity must be

applied in calculations and in Finite Element Models of the buildings, if used, to account for the

increase in concrete stiffness due to the dynamic excitation. AISC sets this value at 1.35, while

ATC states that this value should be 1.2. Neville (1995) also gives a factor of 1.2, but indicates

that the dynamic effect on Ec increases with the concrete strength. Logic suggests that the

increase should depend on strain rate, which will vary with vibration amplitude and frequency.

It is therefore possible that use of a single value, while convenient and simple, is in fact an

approximation.

Both design guides also provide methods for estimating the structural damping, the peak

accelerations of floor, and the fundamental frequency of the structure. However, AISC Design

Guide 11 was written for steel framed structures with concrete decks. Also, the equations to

estimate the fundamental frequency and peak response values of the floor (acceleration, velocity,

and displacement) include factors that are not fully described. This, combined with the fact that

its frequency calculations are based on the behavior of one-dimensional members, such as steel
8

beams, makes the guide difficult to apply rationally to floors in concrete structures in which the

primary behavior is plate action.

ATC Design Guide 1 does not claim that it is meant specifically for steel framed buildings.

However, there are several problems with this design guide, rendering its application to concrete

structures difficult. First, it uses the same 1-D assumptions as does SDG-11 to compute the floor

frequency. Additionally, it does not include any applications to concrete structures because it

assumes that, in them, vibrations are less important due to their greater stiffnesses and

consequently higher frequencies.

Willford and Young (2006) of Arup developed A Design Guide for Footfall Induced Vibrations

of Structures which is applicable to steel and concrete structures alike. This guide implements

modal analysis and simplified calculations to predict the structural response to floor vibrations.

The initial step of the analysis process is to calculate the fundamental frequency of the structure,

which is checked using a Finite Element Model of the structure. The model is also used to

determine the modal shapes of the floor. Willford and Young’s recommendation to use Finite

Element analysis, rather than approximate methods, to compute the frequencies differs markedly

from the approach of Murray and Allen. However, they point out that, today, almost every

major structure will be modeled using FEA software as a matter of course, Therefore using it for

the extra step of determining the mode shapes and frequencies does not constitute a burden.

Once the fundamental frequency of the structure is determined, the Arup design guide places the

structure into one of two categories: low-frequency structures, with fundamental frequencies

less than 10 Hz, and high-frequency structures with fundamental frequencies grater than 10 Hz .
9

Both Willford and Young (2006) and Murray et al. (1997) agree that a structure’s fundamental

frequency will determine the type of loading that will most excite the floor. Based on the fact

that the frequency of normal walking is between 1.5 and 2.5 Hz, the low-frequency structures are

more susceptible to excitation due to the first four harmonics of the walking forcing function.

These structures are susceptible to resonance if the right walking speed is applied, as shown in

Figure 1.2.2.1 (a). There the intensity of the response can be seen to build up with every

additional step. High-frequency structures, however, are more susceptible to impulsive, or

transient, loading. Because these structures have such high frequencies, the response of the floor

to each individual step dies out before the next step occurs, as shown in Figure 1.2.2.1 (b).

Figure 1.2.2.1: Example acceleration responses to walking excitation for floors with (a) fn < 10 Hz,
and (b)fn > 10 Hz

Based on the ISO baseline acceleration standard shown in Figures 1.2.1 and 1.2.2, the frequency

range less than 10 Hz corresponds to the constant acceleration portion of the graph, while

frequencies above 10 Hz correspond to the constant velocity portion. Therefore, in the Arup

design guide, the peak acceleration for the floor is estimated for low-frequency structures and the

peak velocity for high-frequency structures.


10

For a steady-state test, such as the walking test, it is possible that an unexpected pulse during the

recording of the test will result in a peak in the response (for acceleration or velocity) that is

much larger than the rest of the data. Therefore, inspecting only the peak response values can

lead to a misinterpretation of results. Thus, Arup uses the peak root-mean-squared (RMS) value

from both measured and generated responses to characterize the walking test data and compare

with the vibration limits. The RMS value is the quadratic mean of a data record. For the

acceleration record, the RMS, aRMS, is defined

1 ((𝑎𝑎 0 )2 +(𝑎𝑎 1 )2 +⋯+(𝑎𝑎 𝑛𝑛 )2 )


𝑎𝑎𝑅𝑅𝑅𝑅𝑅𝑅 = � ∑𝑛𝑛𝑖𝑖=0 𝑎𝑎𝑖𝑖 2 = � (Eq. 1.2.2.1)
𝑛𝑛 𝑛𝑛

where ao is the initial acceleration and an is the acceleration at the end of the acceleration record.

This same calculation can be performed for the velocity. These RMS values are then compared

to the criteria set forth for the acceptable level of vibration for each specific building.

1.3 Problems with Current Practice


Unlike AISC Design Guide 11 and other similar design guides for controlling vibrations, the

procedure recommended by Arup pertains to both steel and concrete structures. Their guide also

provides examples of several simple structures in which the method was applied and produced

very favorable results, providing very good predictions of the structures’ fundamental

frequencies and responses to loading. However, those examples are very simple structures in

which the floor configurations are very regular. In buildings with more complex geometry, it is

likely that the modeling will be more challenging and that additional guidelines will be needed.

Current vibration design procedures suffer from other drawbacks as well. The first is that there

is no standard loading function to represent individuals walking. Under static loading, standard
11

nominal loads for many occupancy types have been established over time, and many are now

embodied in ASCE-7. They are in general satisfactory because, although the real loads vary

significantly, the nominal loads used for design are high enough to have a very low probability

of exceedence in practice.

The dynamic loading imposed by walking or running presents more difficulties. The chosen

forcing function cannot be too conservative or else it will cause floors to become thicker and

costs to increase, yet it cannot be allowed to underestimate the true loading in more than a small

percentage of cases. It therefore needs to represent the true loading as faithfully as possible but

that true loading is a moving target, depending as it does upon many variables. Aguinaldo et al.

(2002) indicate that the shoe type has an effect on the forcing function, as an individual wearing

running shoes will apply a different force than the force applied by the same individual wearing

high-heeled shoes. Different walking gaits, such as limping or shuffling compared to a normal

gait, also alter the forcing function (Alwan et al. 2003). Other variables affecting the walking

loading are the weight of the individual, the floor surface, and the individual’s walking speed

(Allen et al. 1999). Furthermore, its effect on the floor is not determined solely by the total

impulse delivered by each footfall, but also by the variation of contact force with time (Hanagan

2003).

This can be understood by noting that the contact force contains two humps, as shown in Figure

1.3.1, which correspond to the strike of heel and toe respectively. A typical step rate is about 2

Hz. A typical floor frequency is about 6 Hz. Therefore it is not the fundamental frequency of the

contact force, but rather its second harmonic, that is likely to coincide with the fundamental floor
12

frequency and induce the largest response. This is unfortunate, because quite small changes in

the details of the contact force profile may have a significant effect on the magnitude of its

second (or other) harmonic, which in turn will affect the floor’s response.

Figure 1.3.1: Camel-hump function corresponding to the force applied


by one step of an individual walking (Ebrahimpour et al. 1996)

The result is that lack of a standard walking function causes difficulties in both engineering

design and in the eyes of the law. Designers who are attempting to design a floor to suit a

particular purpose need a forcing function, or a series of them, that they can use to satisfy the

clients’ needs. Legally, a standard loading function is desirable in a dispute to determine

whether a building does or does not meet a criterion such as VC-B. Absent such a loading

function, a malevolent party to a lawsuit could cause any floor to fail the criterion simply by

choosing a large enough loading.

Additionally there are no universally accepted standards for the response. Acceleration limits are

checked for low-frequency structures, while the velocities are compared for high-frequency

structures. The acceleration response is defined as


13

𝑎𝑎 = 𝜔𝜔𝑖𝑖 × 𝑣𝑣 (Eq. 1.3.1)

where a is the acceleration, v is the velocity, and ωι is the frequency corresponding to the ith

mode. This relationship indicates that higher modes contribute more to an acceleration record

than to a velocity record. But correct prediction of the mode shapes and frequencies becomes

increasingly difficult for higher mode numbers, so it is likely that the acceleration response will

be predicted with less accuracy than will the velocity.

1.4 Objectives of Research


The finite element modeling aspect of the vibration analysis process is a major concern because

predicting the first few frequencies of the floor is very important in determining how it will

behave under human loading. The fundamental frequency identifies whether the building is a

low- or high-frequency building, thereby identifying the type of excitation that may cause

excessive vibrations in the structure. Discussions with engineers have suggested that the

fundamental frequencies of models can often differ from those of the prototype by a factor of up

to two. Predicting response accurately requires both a good structural model and a reliable

loading function. If the model cannot represent correctly the modal characteristics of a building,

predicting the response will be impossible.

Studies are needed to improve both the structural modeling and the development of

representative loading functions. The goal of this thesis is to concentrate on the first area and to

develop a set of modeling guidelines with which to create models that consistently predict the

modal properties of structures. The guidelines are to be developed by recording data from

buildings of various types, comparing it with predicted values from a range of Finite Element

models of the structures, and adjusting the models until a good match is achieved.
14

CHAPTER 2: DATA MEASUREMENT


Before computer models can be created to predict the behavior of future buildings, measured

vibration data must be gathered in existing buildings to correlate with the response simulated

within the models. Five buildings were chosen for this field analysis. These five buildings

incorporate different gravity and lateral structural systems and were constructed to perform

different services. Using vibration analysis software and equipment provided by the Seattle

KPFF office, the acceleration responses to various human excitations of these five buildings

were measured.

2.1 Vibration Instrumentation and Testing Procedure


The program RT Pro Version 4.3, which is a dynamic signal analysis program used for

measuring sound and vibration signals, was made available by KPFF in Seattle to conduct these

various vibration tests. The important information that must be determined about each building’s

floor is the floor’s acceleration response to loading, the amount of damping in the system, the

frequencies that are excited by the loading, and the properties of the loading. Though this

program does not generate information about the loads creating the vibrations, RT Pro makes it

possible to measure and estimate the other desired traits of the floor by measuring accelerations

and processing the results.

There are three major software and equipment components included in the analysis that enable

these floor characteristics to be measured. A flow chart of how these components work together

is displayed in Figure 2.1. The first of these components is the accelerometer, which is the

instrument that measures the accelerations at various locations on the floor. The accelerometers

are connected to the next component, the signal analyzer called the Dactron Photon. Up to four

accelerometers can be attached to this device at once, making it possible to register acceleration
15

readings from four different locations simultaneously. The Dactron Photon uses the Fast Fourier

Transform (FFT) technique to convert the real-time acceleration readings, which are in a time

domain format, into the frequency domain. This process will be outlined in more detail in

Section 2.1.1. The final component of the process is the computer itself. The FFT instrument is

attached to the computer, transferring the acceleration data and the converted frequency domain

data to the computer to be analyzed.

Figure 2.1: Vibration instrumentation and equipment components


16

2.1.1 Fast Fourier Transform Process


The Fourier transform is a mathematical process used to decompose a given signal of infinite

length into its harmonic components. The output of this process is a separate function which

displays the magnitude of each of these components. The Fast Fourier Transform methods

utilized in RT Pro perform this same task, but at a faster rate and in real time. FFT also allows

for the transform of finite length records to take place. The speed gain is achieved by highly

efficient numerical algorithms. The input function for this process within RT Pro is the

acceleration record measured by the accelerometer, which is then decomposed into a plot

displaying the magnitudes of the contributing frequencies.

The acceleration response recorded by the accelerometer is a continuous time-domain analog

signal which is sent to the Dactron Photon. An example is shown in Figure 2.1.1(a). Since an

analog signal is continuous, its signal is very susceptible to noise and other disturbances. Also,

an infinite number of data points are received. This creates the need for the analog signal to be

converted into a digital signal, which assigns specific values at distinct intervals over the entire

duration of the continuous signal. Before the analog signal can be converted to digital, a low-

pass filter must be applied to it to remove signals above a certain threshold, which may corrupt

or distort the data. The cut-off frequency chosen here was 10MHz. Once this filter has been

applied, the signal is digitized by an Analog to Digital Converter (ADC), creating the digital

signal that can be processed by the FFT. Figure 2.1.1 (a) shows a sample analog signal, and

Figure 2.1.1 (b) shows it after conversion to digital.


17

Figure 2.1.1: (a) Function from an analog signal; (b) Digital signal converted
from analog signal Power Spectrum for ambient test in West Campus Parking

With the signal now it digital form, the Dactron Photon can perform the FFT process to

reproduce the signal in frequency range. The plot produced by the FFT process is called a power

spectrum (PS), which provides a magnitude for each frequency that is excited according to the

acceleration record. For an acceleration signal, it is defined by

𝐹𝐹(𝜔𝜔 )×𝐹𝐹 ∗ (𝜔𝜔 )


𝑃𝑃𝑃𝑃 = (Eq. 2.1.1)
2𝜋𝜋

where

𝐹𝐹(ω) = ∫ 𝑎𝑎(𝑡𝑡)𝑒𝑒 −𝑖𝑖𝜔𝜔𝑡𝑡 𝑑𝑑𝑑𝑑 (Eq. 2.1.2)

F(ω) is the Fourier transform of the signal a(t) and F*( ω) is its complex conjugate.

The units of the PSD values are (gn)2/Hz, where the acceleration, gn, is a fraction of acceleration

due to gravity (386 in/s2). The bandwidth for the PSD, which is the interval between each

assigned spectral value, is 0.09 Hz in this case. From the power spectra, the predominant

frequencies and modes of vibration that contribute to the system’s response can be determined

because they show up as local maxima.

In the FFT process, it is possible that energy can leak out of its original signal spectrum into

other frequencies, thereby attributing energy to frequencies that are not excited. To reduce the
18

amount of spectral leakage, the original acceleration data, ao(t), is multiplied by a window

function, w(t), to produce the acceleration data, a(t), which is used for the FFT analysis:

𝑎𝑎(𝑡𝑡) = 𝑤𝑤(𝑡𝑡) × 𝑎𝑎𝑜𝑜 (𝑡𝑡) (Eq. 2.1.3)

The window function used for this analysis was the Hanning window, a bell-shaped window

described by

2𝜋𝜋𝑡𝑡
𝑤𝑤(𝑡𝑡) = 0.5 × �1 − cos � �� (Eq. 2.1.4)
𝑇𝑇

where T is the total time length of the window. Figure 2.1.2 displays what the Hanning window

looks like. The amplitudes reported by the FFT process depend upon where the pulse is located

within the Hanning window. For transient excitation, the magnitudes of the accelerations

generated will be different from test to test unless the pulse is centered in the window. If the

pulse is near the end or beginning of the record,

the magnitudes of the accelerations will be

minimized because the pulse is then factored by

one of the tapered sections of the window. The

existence of the window influences the

amplitude of the accelerations that are input for

the FFT process, thereby affecting the

magnitudes of the reported frequencies in the

power spectrum. It does not, however, affect Figure 2.1.2: Hanning window

which frequencies are excited, but simply the overall magnitudes of responses at those

frequencies. If the loading is a pulse, the magnitudes of the reported frequencies will vary,

depending on where within the window the pulse occurs. For continuous loading, such as a

random walking test, the response amplitude might be expected to be relatively repeatable.
19

2.1.2 Human-Induced Loading Tests


There is not a single loading test that can provide all the vibration characteristics of a floor.

Therefore, various tests that generate information about how the floor responds to different

loading situations were performed on each floor. These tests are the ambient test, the heel-drop

test, and the walking test.

Ambient Test
The ambient test is simply an “as is” analysis of the floor. The accelerometer measures the

natural response of the floor when there are no human induced loads applied. Since there is no

human-induced loading for this test, the measured response is due to ambient excitation, such as

equipment within the building and white noise, which includes the other effects that are constant

within the system, such as ground-based vibration caused by trucks and cars outside, airborne

noise, etc. Ambient tests identify possible frequencies of the system, providing an initial

indication of the modes which may be excited by the human loads.

The ambient test must last at least a minute in order to generate a reasonably repeatable power

spectrum. If the signal was too short, two successive tests would lead to very different spectra,

and little information could be obtained about the floor itself. Figure 2.1.3 shows the

acceleration response and the power spectrum created from the ambient test on one of the

buildings in this study. Because the floor is not excited intentionally, the measured accelerations

are very small.

The ambient loading is also not really true white noise. Often the excitation will be dominated

by motors of fans and AC units, which cause response at frequencies of 30, 60, and 120 Hz.

These must be recognized as forced response at those frequencies, and should not be assumed to
20

represent natural frequencies of the floor. This finding can be verified during tests with larger

loads, such as the walking test. Then the motor excitation becomes relatively unimportant and

the peaks largely disappear. That would not happen if the 30 Hz were a structural frequency.

Also, frequencies below 4 Hz are considered to be noise within the system because most floors

have a fundamental frequency greater than about 4Hz. Consequently, the equipment frequencies

and lower frequencies are not considered to be vibration modes of the system. In the example in

Figure 2.1.3, there is a peak around 30 Hz, as well as a large spike close to 0 Hz. These modes

are not considered modes of vibration for the floor, and are therefore ignored in the analysis

process.

Figure 2.1.3: Acceleration response and power spectrum for ambient test in West Campus Parking
Deck

Heel Drop Test


Once the ambient test has concluded, the response of the floor to deliberate excitation can be

determined. The first loading test performed is a heel-drop test. The heel drop test is a simple

way of creating a pulse loading on the floor. The energy of the excitation is much larger than

that of the ambient test, allowing the dominant frequencies of the floor to be identified from the

power spectrum, provided that the excitation is applied at a suitable location. Because it is a
21

pulse load, the damping, at least in the fundamental mode, can also be estimated from the decay

of the response.

Figure 2.1.4: Real-time acceleration response to heel-drop loading and power


spectrum from RT Pro Analysis software

In the heel-drop test, the individual performing the test stands at various locations on the floor,

rises on his or her toes, and drops his or her heel to the floor. Figure 2.1.5 is a photograph of a

heel drop test in progress. The height of the heel rise can range anywhere from two to four

inches depending upon the individual performing the test and the consistency of the individual’s

heel dropping technique. The location of the heel drop and accelerometers is important in

determining the modes and modal frequencies of the floor.


22

Figure 2.1.5: Photograph of heel drop test

For a single-span, simply supported slab, applying the load in the center of the slab will induce a

response of the first mode of the slab, as displayed in Figure 2.1.6(a). If beams are present

within the gravity system, the first mode that includes the beams can be excited similarly by

applying the heel drop at the mid-span of the beam. The location of the accelerometer is also

important, as it should be placed in locations of anticipated maximum response, which are

known as the antinodes. For first mode excitation, the antinode location is at the slab center or

beam mid-span. Figure 2.1.6(b) displays second mode excitation, which can be achieved by

applying the heel drop at the quarter points of the beam or floor bay. For second mode

excitation, the nodes, which are the points at which there is ideally no displacement, are located
23

at the beam mid-span for a loaded beam and the slab center for a loaded slab. Therefore, the

accelerometers should not be placed at this location, but rather at the quarter points of the slab or

beam to measure the response at the antinodes. Though this is true in principle, the

accelerometer will pick up more modes than simply the first or second modes in these locations;

however, the magnitude of the lower modes is anticipated to be the largest at these locations.

Figure 2.1.6: (a) First mode response; (b) Second mode response

Walking Test
The final test is the random walking test. In this test, the accelerometer is usually placed in the

center of the slab because this is the point of maximum response for the first mode, which is

expected to have the largest contribution to the overall response of the floor. One or two

individuals then proceed to randomly walk around the bay, usually in a criss-cross pattern,

pacing back and forth past the sensor and toward the edges of the bay. Figure 2.1.7 is a

photograph of a walking test in progress. While the damping cannot be determined from the

results of this test, other pertinent information is obtained from this test. Since this test presents

results on the behavior of a floor under constant walking forces, it provides the most realistic

results of how the floor responds to actual human-induced forces in actual buildings such as

office spaces, laboratories, hospitals, and residential buildings.


24

Figure 2.1.7: Photograph of walking test in progress

Walking tests, if done at a constant pace, introduce some excitation at the walking frequency and

its harmonics. Similarly to the frequencies in the ambient test around 30, 60, and 120 Hz which

are related to machinery, the excitation at these walking harmonics could be confused as

structural frequencies. In the case of the walking tests conducted in this study, however, two

individuals were walking at the same time, without intent of pacing in unison or even at the same

step rate. Therefore, the forcing function associated with this loading would not have a clear

forcing frequency, and the corresponding harmonics cannot be determined.

2.2. Buildings Chosen for Testing


Five buildings were chosen to provide field testing measurements for the analysis. There are

four concrete buildings, two of which are concrete beam and slab structures, and the other two

are flat-plate structures. Access to these two flat plate buildings was granted by the buildings’
25

design engineering firm, KPFF. These projects will remain unidentified for security reasons.

The remaining building is a steel beam, concrete slab building.

2.2.1 West Campus Parking Deck


The first building selected for field testing was West Campus Parking Deck, which is a six story

concrete parking structure located at the University of Washington. Because vibrations are

seldom a design criterion for parking decks, this choice may at first appear strange. However,

the structure proved suitable for a number of reasons.

• First was access. The structure belongs to the University of Washington and, therefore,

access at any time, even for repeat visits to confirm data, was simple.

• Second was the simplicity and regularity of the layout. The primary objective of the study

was to establish modeling principles and that is much more simply done if the geometry

of the structure contains few irregularities. The parking structure contains a central

region with a series of identical bays that consist of one-way slabs on beams.

• Third is the vibration frequency. The slender slabs (5” thick spanning 21ft) were deemed

likely to give relatively low frequency response, and therefore would avoid some of the

problems of multiple, closely spaced modes. (This assumption proved to be invalid, but

it was one of the reasons for selecting the deck).

• Fourth, the slabs and beams are post-tensioned. This was considered an advantage

because the concrete is less likely to be cracked. Cracking reduces the stiffness of the

slab, and adds a variable to the model, thereby making correlation between measured and

predicted results more difficult.


26

The parking garage incorporates moment frames and shear walls as its lateral load resisting

system. The gravity system consists of a post-tensioned T-beam and slab configuration.

Figure 2.2.1 shows the third floor plan, which is typical. In the east-west direction there are

three bays of 55 ft, 58 ft, and 57 ft, which are spanned by the beams. The slab spans north-south,

with a typical bay size of 21 ft. In that direction, the building contains five central bays with

level floors, and three additional bays at each end that slope to accommodate the ramps. The two

additional exterior bays on each of the two western transverse bays, which are also 17’-6’’ in

width, provide additional parking along the edges of the ramp/drive-way.

1
Level 3 Level 3 /2

Figure 2.2.1: West Campus Parking Garage, Level 3

Table 2.2.1 shows some of the nominal properties of the beams and slabs in the parking deck.

Whereas the compressive strength of the concrete used for all of the parking deck’s shear walls,
27

retaining walls, footings, and ramp slabs was 4000 psi, the slabs and beams have a specified

compressive strength of 5000 psi.

Table 2.2.1: West Campus Parking Deck Structural Member Properties

Depth/Thickness
Member Compressive Stress, f'c (psi) Top Width (in) Bottom Width (in)
(in)
Flat Slabs 5000 5 --- ---
Ramp Slabs 4000 5 --- ---
Beams 5000 28 18 16

Figure 2.2.2 shows the typical T-beam section.

The beam is 33 inches deep overall, 16 inches

wide at the bottom, and has a 28:1 slope on the

sides to ensure easy form removal after casting.

For the longest span of 58 ft, this creates an L/d

ratio of 21. The slab is five inches thick. The

thin slab and long beam spans are possible due

to post-tensioning.

The structure contains few non-structural


Figure 2.2.2: Typical Beam Section
elements, so the damping was expected to come

largely from the structure itself. The primary non-structural elements are non-bearing concrete

masonry walls that run N-S and separate each of the three main bays. Though these 6’’ CMU

walls do not contribute to the strength of the building, they likely provide damping to reduce the

deck’s dynamic response. They also represent line masses, which inhibit vibration where they

are located.
28

In order to get a preliminary idea of the fundamental frequencies to anticipate from FEM models

and vibration testing, an initial frequency calculation was performed. This closed-form

calculation determined the natural frequency of the beams of this system, assuming a simply-

supported system and including the slab mass associated with the beam’s tributary area. This

calculation yielded a fundamental frequency of 3.36 Hz. The beams in the parking deck are part

of a continuous system and therefore are not truly simply-supported. However, this calculation

does provide an initial estimate of the frequency expected for the first beam harmonic of the

parking garage. Because the model ignores continuity, the frequency estimate is likely to be low.

Furthermore, since the estimated frequency is below 10 Hz, it supports the initial belief that the

parking deck is susceptible to resonant behavior under footfall loading.

The presence of ramps, which use concrete with different physical properties than the slabs in

adjacent bays, will cause the behavior of this structure to vary from that of a building with a

more uniform floor plan. Because of the geometric issues this could cause in the modeling

process, the ramps were left out of the study of this building. The focus of the analysis will be

on the five identical interior bays.

2.2.2 University of Washington Faculty Club


The Faculty Club at the University of Washington is a two story, steel frame building. The focus

of this study was on concrete buildings, but performing testing and creating a model for a steel

frame building enable the concrete structures to be compared to a steel building. Secondly, the

thin steel members should create low frequencies, increasing the floor’s susceptibility to

vibration. In addition to this, this building met some of the same criterion that led to West
29

Campus Parking Deck being chosen for testing, including ease of access and regularity of the

layout.

As shown in Figure 2.2.3, a set of two openings separate the building into two sections. The

section located east of these openings is a 6 bay by 1 bay dining area. All of these bays are 29’-

3’’ by 18’-0”, except the south middle bay, which is 18’-8’’ in width. Two 13’-7’’ long hallways

connect the eastern portion of the building to the larger, western section. This western section

consists of a 6 bay by 6 bay area that includes an open patio, various rooms, corridors, a kitchen,

and the building’s entryway and lobby. This section of the building will not be included in the

testing or modeling of this building because of all the rooms and partitions which may affect the

purity of the floor’s response.

Figure 2.2.3: UW Faculty Club


30

Cross bracing at the ground level of the building, consisting of 1” diameter bars, provides some

lateral bracing. The ground slopes down towards the east, so the column height between the

ground and floor varies and is greatest on the east side. The variable column height, combined

with the small cross-section of the bracing bars, renders the value of the seismic bracing dubious.

The gravity system consists of steel beams and a concrete deck.

Many modern steel buildings are framed so

that the top flanges of both the large girders

and the small beams that span between them

are at the same level. The slab rests directly

on top of both beams, and is made composite

to increase the efficiency of the floor system.


Figure 2.2.4: Section of beam to beam connection in
The UW Faculty Club, however, was UW Faculty Club

constructed using a different approach. The main girders are W21x62, and span E-W between

columns. The W12x27 beams span N-S, but rest on top of the main W21x62 girders as shown in

Figure 2.2.4. The slab is cast on top of the W12x27 beams and is not equipped with shear studs.

The floor construction leads to only very limited composite action and continuity, which

therefore affects its stiffness. The concrete floor consists of 4 inches of concrete on a metal deck.

The metal deck spans 9’-9” between the beams, but no shear studs are shown on the plans.

Therefore composite action will be developed only through adhesion of the concrete to the top of

the W12x27 beams. Those beams are detailed as simply supported, because they are fabricated

in 18’-0” lengths and attached to the main girders only by small bolts through their bottom
31

flanges. The connection between the two members is therefore almost entirely by bearing. No

reinforcement is shown in the deck, so there is no known source of continuity in the W12x27

beams. Because the W21x62 girders are not even in contact with the concrete, they can develop

no composite action.

2.2.3 More Hall


More Hall is the Civil and Environmental Engineering Building at the University of Washington.

Similarly to West Campus Parking Deck and the UW Faculty Club, this building was chosen for

testing because of its location and the ease of access. Figure 2.2.5 shows the layout of the

classrooms in this building. The 11’-6’’ wide bay in the middle of the figure is the central

hallway of the building, and classrooms are located in the 3 bay by 1 bay areas on either side of

the hallway. Figure 2.2.6 is a cross-sectional view of the beams spanning N-S. In the

classrooms, the 22’’ by 17’’ rectangular beams span 28’, while the 8” thick slab spans 16’-0” in

the E-W direction.


32

Figure 2.2.5: More Hall typical classroom

Figure 2.2.6: Cross-sectional view of beam and slab in More Hall

Though the floor plan is also fairly regular for this building, the spans are relatively short with

deep beams and a heavy concrete floor. Also, paneling, represented in the figure by a dotted
33

line, was used for the partitions. This not only increases the amount of damping, but may also

stiffen the floor at the location of the partitions. The floor frequency was estimated by

approximate means to be 40 Hz, which is too high to be susceptible to significant vibrations.

This very heavy, very stiff building saw limited floor accelerations under loading, proving to be

useless for this analysis. Therefore, the measured data from this building is not included in this

analysis and the building was not modeled.

2.2.4 Commercial Building #1 (K1)


The first commercial Building, which will be referred to as Building K1, is a high-rise mixed use

building. This building was chosen for testing because it represents a current trend in high-rise

condominium projects, utilizing a RC concrete flat plate to create a minimal floor to floor height.

Therefore, the vibration response of such buildings is important. Additionally, the layout of this

building contains few irregularities, which minimizes the geometric complexities for modeling.

The typical floor plan of Building K1 is shown in Figure 2.2.7. Four of the five bays in the N-S

direction are 23’ while the northernmost bay spans 27’-4’’. In the E-W direction, four of the

bays are again 23’, while the bay furthest to the east is 16’-10”. An 8 inch slab is typical for all

floors and an f’c of 6000psi is specified. A shear wall core provides the lateral load resistance for

this building. The core is 23’ by 35’-4’’ and is located in the center of the building.
34

Figure 2.2.7: Floor plan for Building K1

At of the time of testing in this building, the project was under construction. Partition walls were

present in the first 20 floors, which were partially outfitted, but no countertops, sinks, or baths

were present within the housing units. Above this floor, no such partitions were present, creating

a nearly fully undressed state. Measuring the response in both a partially outfitted floor and a

bare floor provided an indication of the effect that non-structural components have on the

performance of the building.


35

2.2.5 Commercial Building #2 (K2)


The second commercial Building, referred to as Building K2, is also a high-rise mixed use

building, utilizing a post-tensioned concrete flat plate to create a minimal floor to floor height.

The post-tensioning of the slab reduces the ability of the concrete to crack, improving the

stiffness of the slab, and minimizing the uncertain influence of cracking on the stiffness. As

another concrete flat plate building, Building K2 was also chosen for its applicability to current

trends within the design industry, its relatively regular column grid, and the PT slab’s low

propensity to crack.

The building is separated into two distinct sections. The western portion has irregularly shaped

bays. Due to the geometric issues this would present, this section was not included in the

analysis. The five by two bay configuration of the eastern section of the building, shown in

Figure 2.2.8, is much more regular. For the two bays in the N-S direction, the slab spans 31’-

11’’ in one span and 28’-8’’ in the other. The lengths of the western four bays in the E-W

direction are 29’, while the final bay spans 19’-9’’. An 8 inch slab is typical for all floors. The

specified f’c is 6000psi, but cylinders tested at 56 days had strengths of f’c of 7500 to 8000psi. A

27’’ thick shear wall core provides the lateral load resistance for this building. Because of the

irregular shape of this building, the core is located in the western section of the building, and is

not shown in the figure.


36

Figure 2.2.8: Floor plan for Building K2

Similarly to Building K1, this building was under construction at of the time of testing. In the

lower floors, granite countertops, sinks, and baths were present along with partition walls,

increasing the mass and the amount of damping present in these floors. The upper floors were

totally bare except for small amounts of light construction materials. As is the case for Building

K1, the response for the dressed and undressed state of this building could be measured and

analyzed.

2.3 Vibration Data

2.3.1 Test Results and Initial Analysis


For each of the measured buildings, the test results are presented below, and the initial analysis

of these results, such as damping obtained from the heel drop acceleration record, is discussed.

The damping values are given to the closest 0.5% because this was the most accuracy that could

be achieved using the methods discussed in Appendix A.


37

West Campus Parking Deck

Heel Drop Test


In the West Campus Parking Deck, the heel drop test was conducted at two locations: the center

of the middle slab and at the mid-span of the beam adjacent to this slab. Figure 2.3.1 shows the

floor plan with the two testing locations, and Table 2.3.1 summarizes the output of these two

tests.

Figure 2.3.1: Testing Locations for West Campus Parking Deck

Table 2.3.1: Results of Heel Drop Test in West Campus Parking Deck

Fundamental Frequency, Predominant


Test Location Damping, ζ (%)
fn (Hz) Frequency (Hz)
1 Center of Slab 3 7.23 19.22
2 Beam Midspan 3 7.23 7.23

The acceleration record and power spectrum for Test 1 in the middle of the slab is shown in

Figure 2.3.2. The amount of damping in the system was determined to be 3%. Refer to

Appendix A to see the process used to determine the appropriate value for the damping. From

the power spectrum, the excited frequencies in the parking deck were found to be 7.23 Hz, 8.51
38

Hz, 13 Hz, 19.22 Hz, 25.27 Hz, 32.32 Hz, and 41.02 Hz. The mode with the largest magnitude

on the power spectrum is the 19.22 Hz mode, indicating that this mode makes up the primary

contribution to excitation at this point. The magnitude is approximately twice that measured at

the fundamental frequency.

Figure 2.3.2: Output for heel drop Test 1 for West Campus Parking Deck; test performed
at the center of the slab span

Figure 2.3.3: Output for heel drop Test 2 for West Campus Parking Deck; test performed at
the beam mid-span
39

Figure 2.3.3 displays the acceleration record and PSD for Test 2. The damping for this

acceleration record was 2%. The predominant frequencies from the power spectrum density are

7.23 Hz, 11.35 Hz, 14.83 Hz, 16.5 Hz, 19.68 Hz, 21.33 Hz, and 29.57 Hz. The mode with the

largest magnitude on the PSD is the fundamental frequency.

Power Spectrum Comparison


Now that the fundamental frequency and the other contributing modes are known from the heel

drop test, they can be compared to the modes produced in the ambient and walking test. For both

of these tests, the accelerometer was placed in the center of the slab; therefore, the results are

compared to heel drop Test 1 in Figure 2.3.4.


40

Figure 2.3.4: Power spectra for Heel Drop Test 1, Walking Test, and Ambient Test in West
Campus Parking Deck. Data recorded at mid-span of the slab.

The power spectra from the walking test and ambient test corroborate the results displayed in the

PSD from the heel drop test at least with respect to the fundamental frequency. When comparing

the walking and heel drop cases specifically, the peak magnitudes occur at the nearly the same
41

frequencies (7.23 Hz, 19.22 Hz, 25.27 Hz, and 32.32 Hz) as is further displayed in Figure 2.3.5;

the magnitudes are simply larger for the heel drop test due to the larger load impulse. One thing

to note is that while the fundamental frequency does not have the largest magnitude for the heel

drop test, it does for both the ambient and walking tests.

Figure 2.3.5: Power spectrum comparison between Heel Drop Test 1 and the Walking Test in
West Campus Parking Deck. Data recorded at mid-span of the slab.

Summary of Results
Table 2.3.2 is a summary table of the pertinent information determined from the three tests.

Included in this table are the minimum and maximum peak accelerations and velocities measured

in each test.

Table 2.3.2: Test Summary for West Campus Parking Deck


Peak Accelerations
Peak Velocities (in/s)
Test Location (in/s2)
min max min max
Heel Drop-1 Center of Slab -0.278 0.210 -19.3 19.3
Heel Drop-2 Beam Midspan -0.0724 .0685 -4.11 5.03
Walking Center of Slab -0.0679 0.0814 -5.18 5.80
Ambient Center of Slab -0.0144 0.0303 -0.348 0.348
42

UW Faculty Club

Heel Drop Test


The heel drop test was performed in three places in the UW Faculty Club. Due to the presence

of beams in both the longitudinal and transverse direction, tests were conducted at the mid-span

of both the W12x27 (Test 2) and W21x62 (Test 3) in addition to the test in the middle of the slab

(Test 1). The accelerometer was placed at each of the locations shown in Figure 2.3.6, and the

heel drop location was approximately two feet from the accelerometer in each case. The output

of these tests is summarized in Table 2.3.3.

Figure 2.3.6: Testing Locations for UW Faculty Club

Table 2.3.3: Results of Heel Drop Test in UW Faculty Club


Fundamental Frequency, Predominant
Test Location Damping, ζ (%)
fn (Hz) Frequency (Hz)
1 Center of Slab 1.5 5.67 5.67
Beam Midspan
2 (W12x27) 1.5 5.67 5.67
Beam Midspan
3 (W21x62) 1.5 5.67 5.67

Because this is a steel building, the damping would be expected to be low (1~2%). The log

decrement techniques used to determine the damping from the acceleration response indicate that

the damping is 1.5%. A component of this damping value is believed to be due to the presence

of carpet covering loose wood paneling above the concrete flooring which may increase the

damping of the response.


43

The acceleration record and power spectrum for Test 1 at the middle of the slab are shown in

Figure 2.3.7. Again applying the process outlined in Appendix A to the acceleration record, the

amount of damping in the system was determined to be 1.5%. From the power spectrum, the

predominant modes excited in the test were found to be 5.67 Hz, 7.14 Hz, and 8.69 Hz, with

small peak at10.625 Hz. The mode with the largest magnitude on the power spectrum is the

fundamental frequency, indicating that this mode makes up the primary contribution to response

at this point. The magnitude of the contribution of this mode is approximately two and half times

the magnitude of the next frequency, indicating that this frequency is the primary contributor to

the floor’s response to the heel drop excitation at this location.

Figure 2.3.7: Output for Heel Drop Test 1 for UW Faculty Club; test performed at the center of the slab
44

Figure 2.3.8 shows the acceleration record and power spectrum for Test 2, which was performed

at the mid-span of the W12x27 beams. The amount of damping at the location above the

W12x27 was determined to be 1.5%. The ultimate peak acceleration value is 342.13 in/s2, which

is nearly the acceleration due to gravity. It is highly improbable that human-induced vibrations

could cause this level of vibrations, indicating that there must be an error in this test. The

settings in the RT Pro Software were modified for this particular test, as acceleration values were

recorded at shorter time intervals than for the other tests. It is believed that this was done

incorrectly, thereby explaining these excessively large acceleration values. The excited

frequencies from the power spectrum were found to be 5.67 Hz, 7.14 Hz, 9.06 Hz, 10.625 Hz,

12.50 Hz, and 22.25 Hz. The frequency with the largest magnitude is the fundamental

frequency. Furthermore, there are four peaks within 0.025 seconds in the acceleration response,

which corresponds to a frequency of approximately 160 Hz. This again indicates that the

incorrect settings were used for this test because the power spectrum indicates that the

fundamental frequency is the predominant frequency

Figure 2.3.8: Output for Heel Drop Test 2 for UW Faculty Club;
test performed at the mid-span of the W12x27 beam
45

The acceleration record and power spectrum for the final heel drop test in the UW Faculty Club

are shown in Figure 2.3.9. The amount of damping at the location above the W21x62 was

determined to be 1.5%. The only frequencies excited by the heel drop were found to be 5.67 Hz,

and 7.14 Hz, with the largest contribution again coming from the fundamental frequency.

Figure 2.3.9: Output for Test 3 for UW Faculty Club; test performed at the mid-span of the W21x62 beam

Power Spectrum Comparison


The excited frequencies from the heel-drop, ambient, and walking tests are displayed in Figure

2.3.10. The results of the ambient and walking tests will be compared to the results of heel drop

Test 1 since the accelerometer was placed in the center of the slab all three tests.
46

Figure 2.3.10: Power spectra for Heel Drop Test 1, Walking Test, and Ambient Test in UW Faculty Club

The power spectra from all three tests detect the same fundamental frequency of 5.67 Hz. The

heel drop and walking tests identify the same four frequencies (5.67 Hz, 7.14 Hz, 9.06 Hz,
47

10.625 Hz). The second frequency, however, becomes the predominant frequency for the

walking test rather than the fundamental frequency, as is the case for the heel drop test.

Compared to the heel drop and walking tests, the ambient test is only in agreement with the

fundamental frequency. The very pronounced frequency around 30 Hz in the ambient test is

believed to be equipment-related, and, therefore, not a structural frequency.

Summary of Results
Table 2.3.4 summarizes the pertinent information determined from the three tests.

Table 2.3.4: Test Summary for UW Faculty Club


Peak Velocities (in/s) Peak Accelerations (in/s2)
Test Location
min max min max
Heel Drop-1 Center of Slab -0.113 0.113 -23.82 10.27
Heel Drop-2 W12x27 Midspan -0.454 0.261 -342.13 212.71
Heel Drop-3 W21x62 Midspan -0.112 0.111 -20.78 23.33
Walking Center of Slab -0.077 0.075 -3.17 3.54
Ambient Center of Slab -0.0522 0.0619 -0.5410 0.5796

Building K1

Heel Drop Test


Tests were performed on two floors in Building K1 because this building had some floors that

were partially outfitted and some floors that were totally bare. The outfitted floor that was tested

was the 18th Floor and the bare floor was the 34th. On each of these floors the heel drop was

performed in two places. Because the gravity system is a PT flat plate and no beams are present,

Test 1 was performed in the center of the bay and Test 2 was performed at the centerpoint along

the column line adjacent to this bay as shown in Figure 2.3.11. The output of these tests is

summarized in Table 2.3.5.


48

Figure 2.3.11: Testing Locations for Building K1

Table 2.3.5: Results of Heel Drop Test in Building K1


Fundamental
Predominant
Floor Test Location Damping, ζ (%) Frequency, fn
Frequency (Hz)
(Hz)
18 1 Center of Slab 2 12.63 16.85
2 Column Line 2 7.96 35.8
34 1 Center of Slab 5 10.07 13.64
2 Column Line 5 7.6 11.71
49

The acceleration record and PSD for Test 1 on Floor 18 is shown in Figure 2.3.12. The amount

of damping at this location was determined to be 2%. From the power spectrum density, the

excited frequencies were found to be 12.63 Hz, 14.37 Hz, 16.85 Hz, 20.14 Hz, and 20.56 Hz.

The 16.85 Hz mode contributed the most to the response at this location.

Figure 2.3.12: Output for Heel Drop Test 1 for Building K1 on Floor 18; test
performed at the center of the slab
50

Figure 2.3.13 shows the acceleration record and PSD for Test 1 on Floor 34. The amount of

damping at this location was determined to be 5%. The frequencies excited in the power

spectrum density were 10.07 Hz, 11.07 Hz, 13.64 Hz, and 37.26 Hz. The 13.64 Hz mode had the

largest contribution to the response at this location.

Figure 2.3.13: Output for Heel Drop Test 1 for Building K1 on Floor 34; test
performed at the center of the slab
51

Figure 2.3.14 shows the acceleration record and power spectrum for the second test on Floor 18.

The amount of damping from this test was determined to be 2%. The excited frequencies were

7.96 Hz, 10.43 Hz, 14.9 Hz, 17.94 Hz, 20.69 Hz, and 35.8 Hz. The frequency with the largest

magnitude on the power spectrum is 35.8 Hz.

Figure 2.3.14: Output for Test 2 for Building K1 on Floor 18; test
performed at the mid-span of the column line
52

The acceleration record and power spectrum for the second test on Floor 34 is shown in Figure

2.3.15. The amount of damping at this location was determined to be 5%. The power spectrum

density indicated that the excited frequencies were 10.16 Hz, 11.71 Hz, 13.64 Hz, 14.55 Hz,

15.47 Hz, and 37.17 Hz. The frequency with the largest magnitude is 11.71 Hz.

Figure 2.3.15: Output for Test 2 for Building K1 on Floor 34; test performed
at the mid-span of the column line

Effect of Excitation Type


The excited frequencies from the heel-drop, ambient, and walking tests are shown in Figure

2.3.16 for Floor 18 and in Figure 2.3.17 for Floor 34. The results of the ambient and walking

tests will be compared to the results of the first heel drop test on each floor since the

accelerometer was placed in the center of the slab for both the ambient and walking tests.
53

Figure 2.3.16: Power spectra for Heel Drop Test 1, Walking Test, and Ambient Test in Building K1 on
Floor 18
The walking test and heel drop test both agree on the fundamental frequency and the other major

contributing mode. The ambient test, however, identifies 7 Hz as the fundamental frequency and

provides no pertinent information about any other contributing modes.


54

Figure 2.3.17: Power spectra for Heel Drop Test 1, Walking Test, and Ambient Test in Building K1 on
Floor 34

For the bare structure on Floor 34 (Figure 2.3.17), the walking test and heel drop test both agree

on the fundamental frequency and the other major contributing modes. The ambient test also
55

agrees with the range of predominant frequencies, but the predominant frequencies are not as

distinct as the two loading tests. The correlation between heel drop test and walking test for

Floors 18 and 34 are further illustrated in Figures 2.3.18 and 2.3.19, respectively. In Figure

2.3.18, the walking test spectral values have been factored in order to more clearly display the

comparison, because the amplitudes would otherwise be very different.

Figure 2.3.18: Modal comparison between heel drop Test 1, performed at the center of
the slab, and the Walking Test performed on Floor 18 of Building K1

Figure 2.3.19: Modal comparison between heel drop Test 1, performed at the center of
the slab, and the Walking Test performed on Floor 34 of Building K1
56

The most heavily excited frequencies lie in the range of 10-20 Hz. This range is much higher

than the typical walking frequency (~2Hz); therefore the response must represent a natural

frequency of the structure and not a forced vibration effect. The heel drop is essentially a pulse

load, thereby having no inherent frequency. This, combined with the fact that the response

frequency distribution for heel drop and walking are so similar, despite the different excitations,

presents further evidence that the frequency peaks in the power spectrum represent structural

frequencies and not excitation characteristics.

Effect of Non-Structural Elements


The effect of the nonstructural elements is shown in Figure 2.3.20, where the tests conducted on

Floor 18 (partially finished floor) and Floor 34 (bare floor) are plotted together. Figure 2.3.20 (a)

compares the power spectra from the heel drop tests on each floor, while (b) shows the walking

tests. The measured power spectra for both floors are quite similar in shape, each having two

major peaks at nearby frequencies. However, there is a noticeable frequency shift. On floor 18,

the frequencies at the two peaks in the heel drop tests were 1.52 and 1.48 times higher than those

on floor 34. This same shift is seen from the walking tests. This result suggests that the

partitions effectively multiplied the stiffness by at least a factor of 1.52, or 2.25, even if the

increase in mass was negligible. Further comparing the heel drop and walking power spectra

from the outfitted floor to the results from the bare floor, the bare floor had a much cleaner

signal, as the smoother plot enabling the predominant frequencies to be much more distinct.

The increase in stiffness for the fitted out floor is due to the presence of partitions. The partitions

are fixed to the fitted out floor, effectively creating an additional support between the columns

locations. These new supports reduce the lengths of some of the bays. This will stiffen these

spans, creating an increase in the frequencies of the modes which include the excitation of these
57

bays, and in some cases, may alter the shapes of the modes which are excited. Additionally, the

partition present in the fitted out floor may not only be fully connected to the fitted out floor, but

also to the floor directly above it. Therefore, the partitions transfers loads between the two

floors, generating an increase in stiffness for the fitted out floors. This causes the two floors to

act similarly to a giant plate girder, with the partition acting as the girder web connecting the two

floors, which are behaving as the flanges of the girder. The floor above braces the floor, causing

the floors to act together, thereby stiffening the fitted out floor.

One piece of information is anomalous. In the fitted out floor, the power spectrum of the

ambient test shows a spike at about 7 Hz that is not seen in the power spectrum of the bare floor.

This suggests the addition of a frequency that is lower than the fundamental one of the bare floor.

The existence of such a frequency is inconsistent with the fact that the partitions appear to stiffen

the structure, and the reasons for it remain unknown.

Figure 2.3.20: Modal comparison between Floor 18 and Floor 34 for (a) Heel Drop Test 1 performed at
the center for the slab, and (b) the Walking Tests for Building K1
58

Effect of Location of Excitation


Comparing the power spectra for Heel Drop Tests 1 and 2 indicates that different modes are

excited depending upon the location of the loading. Figures 2.3.21and 2.3.22 show this

comparison for Floors 18 and 34, respectively. The predominant frequencies are different for

Test 1 and Test 2 for both cases, which suggest that a floor must be loaded in various locations in

order to reveal all of the possible modes that can be excited.

Figure 2.3.21: Modal comparison between heel drop Test 1, performed at the center of
the slab, and Test 2, performed at the mid-span of the column line, for Floor 18 of K1

Figure 2.3.22: Modal comparison between heel drop Test 1, performed at the center of
the slab, and Test 2, performed at the mid-span of the column line, for Floor 34 of K1
59

In general, it may be expected that the center of a bay will displace in many modes. Therefore,

loading at the center will excite many modes, particularly the fundamental mode and other lower

frequencies. By contrast, the modes which are excited at the mid-point along the column line,

the location where Test 2 was performed, may be different than those at the center of the bay.

However, the column line might be a nodal line for many modes, thereby reducing the number of

modes that may be excited by loading at that point. On the bare floor (Figure 2.3.22), the excited

frequencies appear to be similar for both Tests 1 and 2, so it is unclear whether any significant

modes fail to be excited. The fitted out floor, however, indicates that the modes excited at the

center of the bay and at the mid-point of the column line are very different. This may be due to

the effect of the partitions modifying the modal characteristics (modal frequencies and mode

shapes) for the fitted out floors. As mentioned previously, the addition of partitions increases the

frequencies of some modes and may create additional modes or change the mode shapes from

those seen on bare floors. Therefore, the test at the mid-point between the column lines may be

picking up totally different modes in the fitted out floor than the bare floor.

Summary of Results
Table 2.3.6 is summarizes the pertinent information determined from the three tests.

Table 2.3.6: Test Summary for Building K1

Peak Velocities (in/s) Peak Accelerations (in/s2)


Floor Test Location
min max min max
18 Heel Drop-1 Center of Slab -0.102 0.0813 -15.07 15.77
Heel Drop-2 Column Line -0.1266 0.0586 -22.06 19.78
Walking Center of Slab -0.150 0.009 -1.39 2.67
Ambient Center of Slab -0.133 0.0319 -0.580 0.464
34 Heel Drop-1 Center of Slab -0.160 0.188 -24.77 15.80
Heel Drop-2 Column Line -0.223 0.105 -7.63 6.36
Walking Center of Slab -0.0836 0.0309 -1.93 1.85
Ambient Center of Slab -0.0893 0.0085 -0.464 0.464
60

Building K2

Heel Drop Test


Similarly to Building K1, tests were performed on two floors in Building K2 because it was

unfinished at the time of testing. The outfitted floor that was tested was the 10th Floor and the

bare floor was the 24th. On each of these floors the heel drop was performed in one location. As

shown in Figure 2.3.23, Test 1 was performed in the center of the bay. The output of these tests

is summarized in Table 2.3.7.

Figure 2.3.23: Testing Locations for Building K2

Table 2.3.7: Results of Heel Drop Test in Building K2

Fundamental
Damping, ζ Predominant
Floor Location Frequency, fn
(%) Frequency (Hz)
(Hz)
10 Center of Slab 4.5 6.56 9.68
24 Center of Slab 4 5.94 7.18

The acceleration record and PSD for the heel drop test on Floor 10 is shown in Figure 2.3.24.

The amount of damping at this location was determined to be 4.5%. Building K2 was post-

tensioned, compared to the RC flat plate in Building K1. The larger damping was therefore

surprising in view of the expected absence of cracking. The excited frequencies were 9.68 Hz,

14.06 Hz, 15.94 Hz, 19.06 Hz, 27.12 Hz, and 31.875 Hz. The fundamental mode contributed the

most to the response at this location.


61

Figure 2.3.24: Output for Heel Drop Test performed at center of slab for Building K2 on Floor 10

Figure 2.3.25 shows the acceleration record and PSD for the heel drop test on Floor 24. The

amount of damping from this test was determined to be 4%. The frequencies excited in the

power spectrum were 7.19 Hz, 13.75 Hz, 18.13 Hz, and 24.06 Hz. The 7.19 Hz mode had the

largest contribution to the response at this location.

Figure 2.3.25: Output for Heel Drop Test performed at center of slab for Building K2 on Floor 24
62

Effect of Excitation
The excited frequencies from the heel-drop, ambient, and walking tests are shown in Figure

2.3.26 for Floor 10 and in Figure 2.3.27 for Floor 24.

Figure 2.3.26: Power spectra for Heel Drop Test, Walking Test, and Ambient Test
in Building K2 on Floor 10
63

Though the power spectrum from the walking test is very noisy for Floor 10, the peak

frequencies are in unison with the three predominant frequencies in the heel drop test.

Figure 2.3.27: Power spectra for Heel Drop Test, Walking Test, and Ambient Test
in Building K2 on Floor 24
64

The three tests on Floor 24 are in agreement that the fundamental frequency is in the area of 7.19

Hz. As was found in the other buildings tested, the ambient data shows peaks near 30 Hz that

are not seen in the heel drop or walking tests. They are, thus, must likely to be the result of

ambient excitation at that frequency, for example from rotating machinery.

Effect of Non-Structural Elements


The correlation between heel drop test and walking test for Floors 18 and 34 are further

illustrated in Figures 2.3.28 (a) and (b), respectively. Though the heel drop data is much

smoother, the heel drop and walking tests show similar excited frequencies in both cases.

Figure 2.3.28: Modal comparison between the heel drop test and walking tests performed in Building K2
on (a) Floor 10 and (b) Floor 24

On the bare floor, all three tests have nearly the same fundamental frequency, which is also the

only mode making a significant contribution to the response. Comparing the results from the

outfitted floor to the results from the bare floor, the bare floor had a much cleaner signal,

especially in the case of the walking test. As was found in Building K1, the nonstructural

components have an effect on the response of the floors. The partially outfitted floor (Floor 10)

and the bare floor (Floor 24) are compared in Figures 2.3.29(a) and (b) for their response to heel

drop loading and walking, respectively. As was the case for Building K1, a frequency shift is

present, as the partially outfitted floor displays a higher fundamental frequency due to an
65

enhanced level of stiffness from the nonstructural elements. The fundamental frequency from

the heel drop test on Floor 10 is 1.35 times higher than fn on Floor 24, corresponding to a

stiffness factor of 1.352, or 1.82. This is similar to the factor of 1.52 (2.25) which was

determined from the comparison in building K1.

Figure 2.3.29: Modal comparisons between Floor 10 and Floor 24 for (a) the heel drop tests
performed at the center of the slab and (b) the walking tests in Building K2

The heel drop tests shown in Figure 2.3.19 (a) provide a much cleaner signal than do the walking

tests of Figure 2.3.29 (b). The period shift in the fundamental mode, from about 7 to 10 Hz

when the floor is fitted out, is clear. Interestingly, the peaks that correspond to the next modes

appear to be shifted much less, if at all.


66

Summary of Results
Table 2.3.8 is summarizes the pertinent information determined from the three tests in K2.

Table 2.3.8: Test Summary for K2


Peak Velocities (in/s) Peak Accelerations (in/s2)
Floor Test Location
min max min max
10 Heel Drop Center of Slab -0.0727 0.0737 -10.43 7.07
Walking Center of Slab -0.0815 0.0474 -1.24 1.70
Ambient Center of Slab -0.0551 0.0726 -0.502 0.580
24 Heel Drop Center of Slab -0.1957 0.205 -14.84 10.36
Walking Center of Slab -0.1301 0.0044 -2.74 2.47
Ambient Center of Slab -0.058 0.0384 0.464 0.618

2.3.2 Data Measurement Conclusions


From the data taken from the various buildings that were measured in this study, several

conclusions can be made about the results in general and concerning the effects that the type of

building (structural material, the floor type, material and geometric properties, etc.) have on the

response to human induced loading of the floors in each building.

• Heel drop and walking tests usually lead to similar excited frequencies, with the power

spectra from the heel drop test generating more distinct frequencies in most cases. The

ambient test, however, may not always agree with these frequencies. Also, ambient

frequency spikes at around 30 Hz and 60 Hz are attributed to fans and other machinery,

and are, therefore, dismissed from being actual structural frequencies.

• A frequency shift is present when floors are fitted out. This shift can be attributed to the

nonstructural elements, mainly partitions, which effectively increase the floors' stiffness,

thereby increasing the frequencies of the floor.

• There was a increase in the damping coefficient for the fitted out floor in Building K2,

which was expected due to the addition of the non-structural elements. However, this

damping coefficient increase does not occur in Building K1. Because the damping values
67

were based on fairly noisy heel drop response records, they are approximations and

should be used with caution.

• The concrete buildings all had similar peak acceleration levels for the heel drop tests at

the center of the slabs, ranging from 19.3 in/s2 in the West Campus Parking Deck to 24.77

in/s2 in Building K1. Building K2 saw slightly lower peak accelerations, however, with

10.43 in/s2 in the fitted out floors and 14.84 in/s2 in the bare floors. This decrease in

acceleration was coupled with an increase in frequency. The post-tensioning of the floors

in Building K2 increases the stiffness of these floors, in turn increasing the frequencies of

the floor and decreasing the response. However, the post-tensioning appears to increase

the damping. This was unexpected because the opening and closing of cracks constitute

a potential source of damping, but the level of cracking in Building K2 was reduced by

the post-tensioning.
68

CHAPTER 3: PRINCIPLES FOR STRUCTURAL MODELING


The four structures that were tested in the field were modeled using the Finite Element software

SAP2000 v. 12 with the goal of developing models that would reproduce the measured results.

The ability to do this is essential if the responses of other floors are to be successfully predicted

during design. Material properties and geometry are discussed separately in the following

sections.

3.1 General Modeling Procedures


3.1.1 Material Properties
The three material properties that are critical to estimating response are density, stiffness and

damping. The extent of cracking in the concrete is also important and, since it affects the

stiffness of the concrete, it is accounted for in the material properties.

The densities of the component materials are usually known with good accuracy. Here, normal

weight concrete was assumed to weigh 150 pcf. Most normal weight concrete weighs

approximately 145 pcf, but the additional 5 pcf represents an allowance for the steel

reinforcement embedded in it. In some cases, such as the University of Washington Faculty

Club, where the floor contained no reinforcement other than the cold formed metal deck, the unit

weight of the concrete was taken as 145 pcf. Structural steel was assumed to weigh 490 pcf in

all cases.

The stiffness of the materials is known with less accuracy. For steel, a value of 29,000 ksi was

used in all cases. For concrete, the value of Young’s modulus is generally not measured from

test cylinders, so there is little choice but to accept a nominal value based on the relationship

between f’c and Ec. f’c was specified for all of the buildings except the UW Faculty Club. For
69

these buildings, the modulus of elasticity was determined from the relationship provided by ACI,

Ec = 57,000√f’c. However, there are two reasons that the concrete modulus can vary. The first

deals with the compressive strength. As seen in the case of the cylinder tests for Building K2,

the actual concrete compressive strength taken from the cylinder test can be up to 30% greater

than the specified value, resulting in a variation of up to 15% in the modulus values.

Additionally, the dynamic modulus of elasticity of concrete may be larger than the static

modulus of elasticity as indicated in Section 1.2. Murray (1999) indicates the dynamic modulus

may be up to 1.35 times the static, and Neville (1997) states that this ratio may be around 1.2.

These two reasons provide a rationale to modify the actual Ec from the modulus values

determined from the specified f’c values for this dynamic analysis. Once the models of each

building were fully defined geometrically, the modulus of elasticity was modified so that the

fundamental frequency of the model matched the fundamental frequency from the FFT of that

building. For the UW Faculty Club, a f’c of 4 ksi and Ec of 4000 ksi was assumed.

Damping in structures arises from a variety of sources. This is evident, for example, from the

significant difference between the damping obtained on two nominally identical floors of the K1

structure, except that one floor was finished and the other was bare. However, in addition to the

damping from non-structural elements such as fittings and finishes, some damping arises from

internal hysteretic effects in the material itself. Values recommended by Murray are 1-1.5% for

steel and concrete floors with no nonstructural components, 1.5-2% for floors with few

nonstructural components, and up to 5% for offices and residences with nonstructural

components and full-height room partitions between floors. The first range of damping values

would apply to the bare upper floors of Buildings K1 and K2, while the lower floors of these
70

buildings and the floors of the other two structures would fall into the third category. The

damping values for each building determined from the heel drop data using the log decrement

methods described in Appendix A, do not correspond exactly with these ranges, but are close.

The damping values determined from the heel drop data were chosen as the input damping for

the computer models of each building. Damping generally exerts less influence over the peak

response than does stiffness, except in the case of true resonant response. However, resonance

occurs primarily under continuous rhythmic loading, such as in a dance hall or gym. In most

other cases, such as vibration in an office induced by walking, the lack of a precise damping

value affects the accuracy of the predicted response only slightly.

3.1.2 Geometry
The major geometric issue for the building models was the issue of transforming a three

dimensional structure into a two dimensional model. For simplicity and to save time in the

modeling and analysis processes, only 2-D horizontal models of the floor for each building were

created. Also, only a section of the floor was modeled for each building. The effects of the

columns and the sections adjacent to the modeled area that were neglected in the model had to be

accounted for.

The section of each building selected for modeling depended upon the location of the load test

performed on the building. These locations were selected because of the regularity of the bays in

this area. The conditions at the outer boundaries of the models had to be modified to account for

the portion of the floor that was not modeled. For the buildings with beam-slab configurations, a

modification factor had to be applied to the exterior beams to account for the mass of the

adjacent slab that was not included in the model. The factor includes the weight of the beam and

the weight of half of the effective width of the slab. The process of determining the appropriate
71

factor to apply to these beams is outlined in Appendix B. For the flat plate buildings, no beams

are present and the weight of the adjacent slab is carried by the columns. Therefore, no

modification was necessary.

In the absence of columns in the model, the boundary conditions at the locations where the

columns meet the floor needed to be determined. Because the forces generated in this analysis

are relatively small, the axial flexibility of the columns is expected to be minimal. Therefore, the

floor translation in the vertical direction is fixed at the column locations. Additionally, rotational

springs were placed at these locations to simulate the bending stiffness provided by these

columns. A vertical 2-D frame of each building was created in a separate analysis to determine

the rotational stiffness of the columns at these locations. The process of determining these

rotational stiffnesses is outlined in Appendix C.

Other geometric issues for the building models not associated with the boundary conditions were

the connectivity and integration of various structural components and materials, concrete

cracking, and the mesh selection. When area elements, which are used to model slabs, and beam

elements are connected, the center of the area element runs through the cardinal point of the

beam element. The beam cardinal point is the point at which the gravity center of the beam is

taken. To model the beam and slab correctly, the slab must be offset from the cardinal point so

that it rests on top of the beam. This can be done various ways which are outlined in more detail

in Appendix D. In the method that proved to produce the most effective model, the cardinal

point is placed at the centroid and the slab offset is half of the height of the beam plus half of the
72

slab thickness, as shown in Figure 3.1.2.1. Each of the building models with beam-slab

configurations incorporated the centroid cardinal point method.

Figure 3.1.2.1: Centroid cardinal point (CP) and slab offsets


for steel and concrete beam and slab buildings

Concrete cracking was also considered. For the concrete buildings that used post-tensioned

members, such as the West Campus Parking Deck and K2, the post-tensioning is usually

designed to balance the self weight. Under self-weight alone, the slab is therefore in uniform

compression and will not even experience tensile stress, let alone crack, under human-induced

vibrations. Therefore, the gross area of the concrete sections was used. However, for Building

K1, which has a reinforced concrete deck, cracking will likely occur, especially around the

columns where the moments are highest. The loss in stiffness due to this cracking must be taken

into account. For this building, the modulus of the concrete was reduced in all elements to

account for the stiffness drop due to the concrete cracking. Ideally, it would have been reduced

only in the elements that were cracked. However, only one criterion for comparison between the

measured and predicted data existed (frequency), so without measuring mode shapes, identifying

both the location and extent of cracking was impossible.

The final geometric issue was the mesh refinement used for the analyses. A mesh analysis was

performed on a steel W21x62 beam with a strip of concrete on top of it. This arrangement was
73

chosen because a closed form solution is available for comparison. The modulus of the steel

beam was 29,000 ksi, while the modulus for the concrete was 4000 ksi, with a specific weight of

145 pcf. The model is shown in Figure 3.1.2.2 and the results of the mesh analysis are shown in

Table 3.1.2.1. Because the strip of slab is so narrow, transverse bending effects are negligible

and the beam-slab model can be assumed to behave like a simply supported beam. In finite

element analysis, the mass of the slab and beam is lumped at each node instead of being

distributed uniformly over the entire length of the beam. Therefore, as the mesh is refined, the

FE solution may be expected to converge to the closed-form frequency and mode shape. The

discretization correction factor accounts for the fact that the continuous beam is being

represented by a finite number of nodes for each test in the mesh analysis. The behavior of the

beam nears that of a continuous beam as this factor goes to one.

The mesh of the slab, modeled as area elements, was continually refined along the length of the

beam until the fundamental frequency converged. Shear deformations were turned off so that the

model would replicate the flexural displacements that were the basis for the closed form

calculation. The FE frequency converged to a constant value of 20.409 Hz. Figure 3.1.2.3

shows a log-log plot of the error vs. mesh size is 3.0, implying that the error is proportional to

N3, where N is the number of elements along the length. With 16 elements, the frequency had

converged to 0.06% of its final value, which was taken to be that obtained with 512 elements.

The converged frequency of 20.409 Hz did not, however, agree exactly with the closed-form

value of 20.72 Hz. The difference is 2.972%. The reasons for that difference remain unknown.

Therefore, a 16 x 16 mesh size for square slab was selected as a compromise between run time
74

and precision. For rectangular slabs, the mesh included 16 elements in the longitudinal direction

and 8 elements in the transverse direction.

Figure 3.1.2.2: W21x62 beam with concrete slab model

Table 3.1.2.1: Mesh Refinement Analysis of W21x62 beam with concrete slab

Closed Form Frequency = 20.72 Hz

Mesh Refinement Discretization Correction


f1 (Hz) Error (%) Relative Error
(# of Elements, N) Factor = (1-0.8/N2) =
2 0.8 -100
4 0.95 20.1979 -4.97 1.03%
8 0.9875 20.3583 -3.45 0.25%
16 0.996875 20.3966 -3.09 0.06%
32 0.99921875 20.4059 -3.00 0.02%
64 0.999804688 20.4082 -2.98 0.00%
128 0.999951172 20.4087 -2.97 0.00%
256 0.999987793 20.4088 -2.974 0.00%
512 0.999996948 20.409 -2.972 0.00%
75

Figure 3.1.2.3: Plot of error for mesh refinement analysis

3.2 Representation of Loading


3.2.1 Heel Drop Load Simulation
The heel drop test was the first load test applied to each building. It is a somewhat artificial test

in that it represents a loading that is unlikely to occur in practice, but it has value in vibration

testing because it constitutes a pulse load that can be used to estimate the damping inherent in a

floor.

The floor system for each of the structures tested in this study is a Multi-Degree-of-Freedom

(MDOF) structure, with many modes and frequencies of vibration in each of which the damping

is likely to be different. Determining the damping in each mode from a pulse loading is

extremely difficult.

The system may be analyzed using modal analysis, with a spectrum constructed from an

assumed pulse shape. In that case each mode may be assigned an individual damping

coefficient, and identification techniques could theoretically be used to find the damping values

that lead to the closest match between the measured and predicted responses. The difficulty with

this approach is that spectral analysis combines peak modal values and loses the phase
76

information. For transient motion such as a pulse, this is critical and renders the procedure

valueless.

Alternatively, a time history analysis may be conducted, but then either a complete damping

matrix must be identified, which is numerically extremely challenging, or Rayleigh damping is

assumed and the two coefficients can be identified from the responses. However, that process

leads to the correct damping value in only two of the modes. A further difficulty with

identifying damping for higher modes is that the pulse shape is unknown, but is almost certainly

not a Dirac delta function. If the pulse duration, td, is longer than the period of any of the modes

that attract significant excitation, the response becomes more complex, and the identification

becomes more difficult.

A simpler, but approximate, approach is to assume that the first mode response dominates, and to

treat the response as a SDOF system. Then the damping in the first mode can be determined by

comparing the amplitudes of successive peaks in the decaying response signal. In a true SDOF

system, the amplitudes of successive peaks during the free vibration phase are related by

𝑑𝑑(𝑡𝑡) 𝑑𝑑 𝑖𝑖
𝑑𝑑(𝑡𝑡+𝑇𝑇𝐷𝐷 )
= 𝑑𝑑 = 𝑒𝑒 2𝜋𝜋𝜋𝜋 (1)
𝑖𝑖+1

where ζ is the damping ratio, di is the initial peak response, and di+1 is the peak following the

initial peak. In the responses measured in this study, the signals were typically not purely

sinusoidal within a decaying envelope, which indicated that modes other than the fundamental

one were contributing to the response. Nonetheless, various envelopes to the decay rate were

used to obtain an approximate damping value. That process is described in Appendix A. The
77

smallest damping value from this process was taken as the damping of the floor so that the

damping was not overestimated.

Pulse Load
Whereas human walking creates a loading function with multiple pulses, the loading function

associated with the heel drop tests is much different, consisting of one single pulse of unknown

shape. For any type of pulse loading, the time at which the maximum response of the SDOF

system occurs is dependent on the duration of the load, td, but is independent of the peak load

value, po. For short pulse durations (td < Tn/2), the peak response occurs during the free vibration

phase, after the pulse has finished. When td = Tn/2, the maximum response occurs when t = td.

Longer pulses create a peak response during the pulse, which corresponds to the forced vibration

phase. Figure 3.2.1.1 displays the shock spectra for three different pulse types (rectangular,

sinusoidal, and triangular) of equal area (impulse). As shown in the figure, for td /Tn < 0.25, the

response of the SDOF system to each of the various pulse shapes is almost the same. As the

td/Tn ratio increases to 0.5, the responses to the pulse types diverge, indicating that having

identical pulse areas is not enough to create identical responses for a given td/Tn value. This

indicates that, for small pulse durations, the pulse shape selection is not important, but that the

impulse is crucial.

Figure 3.2.1.1: Shock spectra for three force pulses of equal area from Chopra (2007)
78

It was hoped that the duration of a heel drop would prove to be short compared with the

fundamental period of the structure in order to avoid having to determine the shape of the pulse.

However, to investigate this hypothesis, response of one of the buildings was predicted using a

variety of pulse profiles and durations and those responses were compared with the measured

values. The process of selecting this pulse is outlined in Appendix E. The properties of the pulse

load are independent of the building on which it is applied, provided that the mass of the person

performing the heel drop is much less than the mass of the floor. Thus, the measured data from

any building could be used to identify the pulse properties, and the measured response and the

computer model of the West Campus Parking Deck were selected because of its regular

configuration and presumed lack of cracking due to the post-tensioning. The optimal pulse that

was selected is a two-peaked, triangular pulse with a time gap between peaks, which is shown in

Figure 3.2.1.2. The pulse duration is 0.125 seconds and the two peaks have peak forces of 0.368

and 0.184 kips. The impulse, given by the area under the curve, represents a weight of a 185 lb

individual dropping through a height of 3 inches.

The double-triangular pulse was found to match the measured data better than a single pulse. Its

shape suggests that the heel drop procedure includes a bounce, in which the individual’s heel

separates from the floor briefly, after the initial impact, then strikes the floor again. The smaller

second impact is then rational. The total pulse time of the 0.125 seconds is approximately equal

to Tn (1/fn = 1/7.23 Hz = 0.138 sec.). This finding means that, for a floor with fn = 8 Hz (Tn = 0.125

sec.), the pulse shape does influence the response. This was unfortunate, because the pulse shape

becomes more important, but this study proceeded using the double-triangular pulse shape as the

best available option.


79

0.4

0.35

0.3

Force (kip) 0.25

0.2

0.15

0.1

0.05

0
-2.78E-17 0.025 0.05 0.075 0.1 0.125 0.15
Time (sec)
Figure 3.2.1.2: Double triangular pulse function used for heel drop test simulation

In each of the building models, a unit force with a magnitude of 1 kip was applied downward in

the z-direction at the location where the heel drop test was conducted. The pulse forcing

function has a value at each time instant. Using a time history analysis, the forcing function then

applies a value for the force applied to that node at each time step, scaling the unit force by the

value of the forcing function at each particular instant. The analysis for the model can be

completed at this point, and the response to the forcing function can then be plotted for any node

in the model.

3.2.2 Walking Load Functions


The loading function that represents walking load function is best described as a camel hump

function, which is shown in Figure 3.2.2.1(a). This function was provided by David Webster of

KPFF and was obtained from experimental data using an instrumented segment of floor. Others

(Ebrahimpour et al. 1996, Pan et al. 2008) have found similar shapes. The function used in the

models, shown in Figure 3.2.2.1(b), includes five successive steps, and the step frequency (SF) is

1.74 Hz, slightly smaller than the 2 Hz approximation made by Murray. The period is 0.575
80

seconds. Because some of the walking tests were conducted with one person walking and others

with two people walking, a function was made for both cases. The Two Walkers function is

simply double the Single Walker function. It therefore represents a worst case, in which both

people are walking in step. The maximum force for the Single Walker function is 0.33 kips.

Figure 3.2.2.1: (a) Single Camel Hump forcing function for a walking period of 0.575
seconds; (b) Walking test forcing function for multiple steps by one or two people walking

Similarly to the triangular pulse tests, a unit force was applied in the z-direction at the center of

the slab in each model and the walking force function applies a value for the force applied to that

node at each time step. Because the walking function was placed at the center of the slab in each

of the models, it actually simulates a person (or persons) walking in place at this position as

opposed to walking around the entire bay. However, this simulation was assumed to be

sufficient because the accelerometer was placed in the center of the slab for the walking tests and

the maximum response measured by the accelerometer is expected to be from steps that occur

near the sensor. Therefore, the peak response of the walking tests is included in the simulation.
81

3.3 Models of Tested Structures


The modeling process for the four buildings in this study and the response of each of the models

to loading is outlined below.

3.3.1 West Campus Parking Deck


3.3.1.1 Initial Model
The five interior bays in the North-South direction in the West Campus Parking Deck are

identical in the eastern section of the deck. Thus, the vibration testing was performed in and

around the center bay of this identical-bay section because this would provide the most useful

indication of the behavior of a continuous system.

Initially, only the three interior bays of this section were modeled to determine whether the

measured response could be replicated adequately. These three bays are 21’ by 57’ to replicate

the dimensions of the bays in the building’s row of bays furthest to the east, which is where the

vibration testing was conducted. The beams in the model were simply-supported. The modulus

of elasticity was chosen as 5000 ksi. This initial model is shown in Figure 3.3.1.1.1. The static

and modal output for this three-bay model is shown in Table 3.3.1.1.1. The computed beam

mid-span displacements are under the self-weight of the structure alone. Since the interior beam

displacement is similar to the exterior beam displacement, the boundary conditions at each end

of the model were assumed to be sufficient to replicate the behavior of a continuous system. The

dynamic error is calculated the error between the fundamental frequency generated by the model

and the closed-form solution for a single simply supported beam with the additional mass from

the tributary width of slab. This comparison was selected because the closed-form calculation

generates an initial estimate of the fundamental frequency equal to 7.46Hz, which is within 3%

of the measured fundamental frequency for WCPD.


82

Figure 3.3.1.1.1: Three bay model of West Campus Parking Deck

Table 3.3.1.1.1: Output for 3-bay model of West Campus Parking Deck
Beam Static Mid-span Displacement (in) Model
Support Measured/Model
Cardinal point Fundamental
Type Exterior Interior fn
Frequency (Hz)
Centroid SS 0.8044 0.8087 3.672 2.03

3.3.1.2 Model Modifications


From the three-bay model, the exterior and interior beams have nearly identical displacements,

indicating that the model is performing like a continuous system. Though the desired static

response has been achieved, the modal characteristics of the model do not match the measured

response of the deck. The model’s fundamental frequency is 3.67 Hz, which is approximately

half of the measured fundamental frequency. The first modification to improve the modal

response of the model was to increase the number of bays. To maintain geometric simplicity, the

deck ramps were neglected in the model. Therefore, only the five identical interior bays were

included in the modified model. As seen in Table 3.3.1.2.1, this had a minimal effect on the

fundamental frequency and mode shape; however, it resulted in as much as a 10% drop in the

frequencies of higher modes. Also, the additional two bays affected the mode shapes. Because
83

the three-bay model does not adequately produce the correct modal frequencies and shapes, the

five-bay model was selected.

Table 3.3.1.2.1: WCPD 3-Bay and 5-Bay


Comparison
Frequency (Hz)
Mode
3 Bay 5 Bay f3/f5
1 3.67 3.66 0.996
2 3.84 3.75 0.978
3 4.53 4.04 0.891
4 5.76 4.51 0.784
5 6.96 5.18 0.744
6 7.92 6.04 0.763
7 7.96 6.70 0.842
8 8.06 7.77 0.964
9 8.48 7.92 0.934
10 8.96 7.93 0.885

In addition to increasing the number of bays from three to five, the boundary conditions and

material properties were modified in order to replicate the measured response more closely.

3.3.1.2.1 Boundary Conditions


The beam boundary conditions in this model remained simply-supported. However, rotational

springs were also placed at the beam-column locations to simulate the stiffness provided by these

columns. A vertical 2-D model of the structure was created to determine the effect that multiple

stories and columns have on the flexural stiffness at the beam-column points of connection. This

process is further discussed in Appendix C, the rotational stiffness values for these springs were

determined. The values used for this building, along with the response of the model with these

springs inserted, are shown in Table 3.3.1.2.1.1. The stiffness of the springs is given as a ratio of

the spring stiffness, kθ, over the flexural stiffness of the column, (EI/L)c. Because these ratios

are larger than 4EI/L, the springs create boundary conditions for the beam which are very close

to fully-fixed against rotation. Thus, minor changes in the spring stiffness will have almost no
84

effect on the beam frequency, so any differences between the model and measured frequencies

must be attributable to other modeling errors. The insertion of rotational springs increased the

fundamental frequency from 51% of the measured value to 86%.

Table 3.3.1.2.1.1: Output for West Campus Parking Deck model


with rotational springs
Fundamental Frequency, Error (%) from
kθ/(EI/L)c @ Supports
f1 (Hz) Measured
Interior Exterior Model Measured Frequency
5.98 11.06 6.227 7.23 13.90

3.3.1.2.2 Material Properties


After optimizing the geometry and boundary conditions of the model, the fundamental frequency

of the final model was 6.227 Hz, approximately 86% of the measured natural frequency. The

only other plausible properties that could affect the frequency are density and constitutive

properties. The density of normal weight concrete varies only in the range of ± about 2%, which

is not enough to explain the discrepancy. Thus, the assumed material properties of the model

must be in error. The stiffness of the members is directly proportional to the modulus of

elasticity, which therefore has a large effect on the fundamental frequency of the system. The

Young’s Modulus was therefore adjusted until the computed and measured frequencies matched.

Figure 3.3.1.2.2.1 shows how the computed frequencies of the modes which correspond to the

first two measured frequencies vary with Ec. Mode 5 in the model was chosen to compare to the

second measured frequency because the model generates modes which are very close in

frequency, and Mode 5 was the first mode which was in the range of the second measured

frequency. The first 8 modes of the model are shown in Figure 3.3.1.3.1. Mode 3 has a similar

shape as Mode 5, however, the frequency is very near that of Mode 1, therefore the level of

excitation from this mode is assumed to be low. Modes 2 and 4 are anti-symmetric from each
85

other and also have no excitation in the loaded bay; therefore, it is unlikely that these modes

were excited during the measurements.

9
8.5
8
Frequency (Hz)

7.5
Mode 1
7
6.5 Mode 5

6 Mode 1, Measured (7.23 Hz)


5.5 Mode 2, Measured (8.51 Hz)
5 Optimization (Ec =6859 ksi)
4.5
2 3 4 5 6 7 8 9
Ec (x10E3) ksi

Figure 3.3.1.2.2.1: Modulus of Elasticity Determination for WCPD

A value of Ec = 6859 ksi was found to give a computed fundamental frequency within 0.05% of

the measured value and a mode within 4% of the second measured frequency. This is 1.37 times

higher than the 5000 ksi obtained from the ACI equation. This ratio is comparable to the 1.35

correction factor for dynamic modulus as outlined by Murray. Because no measured f’c data was

available, some of the 1.37 factor may be accounted for by the concrete being stronger than

specified.

With the determination of the concrete modulus of elasticity, the model is fully defined with

regard to both geometric and material properties. The model is now ready to be loaded to

determine its response to the triangular pulse and camel hump loading function so that its

response can be compared to the measured response for the heel drop and walking tests. A wire

frame model and extruded view of the final model is shown in Figure 3.3.1.2.2.2. The center bay
86

is divided into four sections to create a node at the center of the bay so that the heel drop load

may be applied to the model.

Figure 3.3.1.2.2.2: Wire frame and extruded view of final model for WCPD

3.3.1.3 Analysis of Mode Shapes


The first eight mode shapes for the WCPD model are shown in Figure 3.3.1.3.1. The mode

shapes follow the typical sinusoidal mode shapes with a single peak at the mid-point of each bay

for the first five modes and with a peak at both ends of each bay for the subsequent modes. The

regularity of the floor for this building enables these very clean mode shapes to be generated.
87

Figure 3.3.1.3.1: First eight mode shapes for WCPD model

Figure 3.3.1.3.2 displays the power spectrum of Heel Drop Test 1 performed at the center of the

slab in the center bay of the West Campus Parking Deck. The fundamental frequency and other

predominant frequencies from this plot are shown in Table 3.3.1.3.1. The modal frequencies

from the WCPD model that correspond to the measured frequencies are also shown in this table,

along with a description of the type of behavior that mode represents.

Figure 3.3.1.3.2: Power spectrum for Heel Drop Test


at the center of the slab in WCPD
88

Table 3.3.1.3.1: Measured and Model Frequency


Comparison for WCPD

Measured Model
Mode
Frequency fmodel/fmeas Description
Frequency
(Hz) Mode
(Hz)
Beam 1st
7.23 1 7.237 1.00 mode
Slab 1st
8.51 5 8.15 1.04 mode
Beam/Slab
13 17 11.93 1.09 1st mode
Slab 4th
19.22 32 19.91 0.97
mode

The mode shapes for the fundamental frequency and the model mode that corresponds to the

second measured frequency are shown in Figure 3.3.1.3.1. Though there is deflection in the slab

in the mode shape of the fundamental mode, this mode describes the first mode of the beams

because the predominant motion is occurring in the beams. There is significantly more

deflection in the center slab in the second mode. Figure 3.3.1.3.3 (a) shows the modal shape of

the model mode that matches the third predominant frequency from the power spectrum. This

mode also has major deflection occurring at the center of the slab, where the loading took place.

The mode shape of the model mode that matches the final predominant frequency from the

power spectrum is shown in Figure 3.3.1.3.3 (b). This mode has almost no beam action and is

the fourth mode of the slab.


89

Figure 3.3.1.3.3: Mode shapes for the model modes that correspond to the (a) 3rd measured
predominant frequency and the (b) 4th measured predominant frequency.

3.3.2 UW Faculty Club


3.3.2.1 Initial Model
The UW Faculty Club has the most regular layout out of all of the buildings in this study. There

are no shear walls to be modeled and all of the bay shapes are rectangular and are the same size,

except for the middle bay. Figure 3.3.2.1.1 shows the initial model along with the actual floor

plan. The steel modulus was set at 30,000 ksi, and the concrete modulus at 3600 ksi because ithe

age of the building suggests that high strength concrete was not used.

Figure 3.3.2.1.1: UW Faculty Club (a) initial model and (b) actual floor plan
90

The four inch concrete slab on metal deck was modeled as a series of concrete plate elements.

The ribbing of the metal deck spans the E-W direction. Therefore, the flexural stiffness in the N-

S direction is less than in the E-W direction. To account for this, a stiffness modification factor

of 0.75 was applied to the flexural stiffness in the N-S direction. This value was chosen

arbitrarily but, because of the slab spans one way between the secondary steel beams, its

transverse stiffness was expected to make little difference.

The column locations were simply-supported, so the western columns were fixed against

translation in all directions and the eastern columns had rollers in the E-W direction.

The beam-girder connections created modeling difficulties. The W12x27 beams sit atop the

W21x62 girders, with the four inch concrete slab on metal deck sitting on the W12x27 beams.

To achieve this, the cardinal point was set at the bottom center for the beams and at the top

center for the girders. The slab was then offset vertically from the bottom of the W12x27 beam

so that it sat directly on top of the beam. This configuration generated a fundamental frequency

of 11.19 Hz, and second and third modal frequencies of 11.32 Hz and 11.45 Hz, respectively.

An additional configuration, shown in Figure 3.3.2.1.2, was tested to determine the effect of

using different cardinal points. The centroid was selected for the cardinal point for both the

W12x27 beams and the W21x62 girders. The W21x62 girders were then offset vertically so that

the W12x27 beams could rest on top of them. The slab was again vertically offset to place it

directly on top of the W12x27 beams. The first three frequencies from this configuration were

11.23 Hz, 11.34 Hz, and 11.42 Hz, which, when compared to the model with the centroid

cardinal points, generates differences of -0.34%, -0.21% and 0.30%, respectively. Therefore, the
91

choice of cardinal point has no large effect on the modal characteristics of the model. The

fundamental frequency, however, was approximately twice the measured fundamental frequency

of 5.67 Hz. Therefore, modifications to the model had to be made to match the model to the

measured values.

Figure 3.3.2.1.2: Cross-section of the beam-girder


connections in the UW Faculty Club model

3.3.2.2 Modal Identification


The measured fundamental frequency of 5.67 Hz was found in the power spectra from the heel

drop tests at each of the three testing locations. Therefore, it was unclear whether this frequency

corresponds to a mode in which the main excitation is occurring in the W12x27 beam, W21x62

girder, in the slab, or a combination of these. An analysis was performed to determine which

mode the fundamental frequency corresponds to. A closed-form frequency calculation was

performed for the first two “beam” frequencies for both the W12x27 beam and W21x62 girder.

The weight of the slab that each beam supports was included in the calculation. However, only

the moment of inertia of each beam and girder was used because the deck does not act

compositely with the beams or girders. The closed-form fundamental frequencies for the

W12x27 beam and the W21x62 girder, shown in Table 3.3.2.2.1, are 7.768 Hz and 5.47 Hz,

respectively.
92

Each beam was also modeled separately, with an additional mass factor to account for the

additional weight of the slab. The first two frequencies from these models are included along

with the closed-form first and second modal frequencies in Table 3.3.2.2.1. The differences

between the FE model and the measured values for the second modes of these beams are

similarly small. This small error indicates that the closed-form solution is an effective initial

estimate for the frequencies of the first mode for beams and girders.

Table 3.3.2.2.1: Frequency Comparison for UW Faculty Club


Frequencies Error Calculations
1 2 3
Closed- Single Beam Measured Error1-2 Error2-3 Error1-3
Mode
Form Model (FFT)
1 7.768 7.7629 7.14 0.066% 8.024% 8.084%
W12x27
2 31.072 31.043 --- 0.093% --- ---
1 5.47 5.45 5.67 0.366% -4.037% -3.656%
W21x62
2 21.86 21.81 --- 0.229% --- ---

These closed-form and single beam model frequencies also match rather closely when compared

to the first two frequencies from the FFT. The differences between the closed-form W21x62

frequency and the measured fundamental frequency is 4.037 %, while the difference between the

closed-form W12x27 frequency and the second frequency from the FFT is 8.024 %. The

difference between the model and measured frequencies is 3.656 % for the W21x62 girder and

8.084 % for the W12x27 beam. The larger errors for these comparisons indicate that the beams

and girders in actual buildings do not behave exactly like a single simply-supported beam since

they are connected to other beams, girders, and slabs which, along with the many non-structural

elements present in the building, affect the response of the floor. This correlation is, however,

close enough to indicate that the fundamental frequency of the floor corresponds approximately
93

to the first mode of the W21x62 girder, while the second overall frequency of the floor

corresponds to the first mode of the W12x27 beam.

3.3.2.3 Model Modifications


Since the excited modes had been identified, the entire model of the UW Faculty Club had to be

modified to generate frequencies that matched the measure values. In the UW Faculty Club, the

slab simply sits above the W12x27 beams, which rest on top of the W21x62 girders. There are

no shear studs to connect these elements and create composite behavior between the concrete

slab and the W12x27 beam. Furthermore, the only means of transferring longitudinal shear

between the W21x62 and the elements above is by transverse shear through the web of the

W12x27 beam, and that is likely to be minimal because those webs are so thin.

At the beam-girder connection points, the initial model for this building had the slab, W12x27

beams, and the W21x62 girders all connected at one node. Though there were offsets which

enable the slab to rest on the W12x27 beams which rest upon the W21x62 girders, the offsets

were fully rigid, making the beams, girders, and slab in the model act compositely, even though

they do not in the field. This unintended composite action stiffens the floor and is the reason that

the fundamental frequency of the initial mode was twice that of the measured value.

To remove this composite action from the model, the slab was physically shifted above the

W12x27 beams and the W21x62 girders shifted below the W12x27 beams so that the these three

element types no longer had nodal connectivity, as each element received its own node at the

location where the elements come into contact. In this configuration, the beams, girders, and

slab are not composite. However, they still act independently of one another, which is still an

incorrect description of their behavior. To remedy this, a body constraint was created within the

model between the slab node, the beam node, and the girder node, at the locations at which the
94

elements are in contact with one another, as

shown in Figure 3.3.2.3.1. This body

constraint effectively forces the slab, beam,

and girder at the constrained nodes to have

identical vertical and transverse translations

and rotations. However, the elements remain

unconnected longitudinally, which enables


Figure 3.3.2.3.1: Cross-section of the nodal body
constraints in the UW Faculty Club Model
each element to behave independently in that

direction. End releases were also created for the ends of the W12x27 beams because these

beams are not connected and cannot transfer moments or loads.

The power spectra from the heel drop tests at the three different locations on the floor in the UW

Faculty Club had very similar predominant frequencies, but each test included additional

frequencies that were not excited in the other two tests. The predominant frequencies from all

three tests are compiled in Table 3.3.2.3.1, along with the frequencies from the model which

correspond to the measured values. The frequencies from the model are merely the frequencies

which most closely match the measured excited frequencies. However, it is unknown whether

these modes from the model match the actual modes because the actual mode shapes are

unknown. The largest error between the measured excited frequencies and those generated by

the model is 3.3%, while most of the errors are less than 1%.
95

Table 3.3.2.3.1: Measured and Model


Frequency Comparison for UW Faculty Club
Measured Model
Frequency Frequency fmodel/fmeas
Mode
(Hz) (Hz)
5.67 1 5.682 1.00
7.14 4 7.375 0.97
8.69 5 8.81 0.99
12.45 10 12.058 1.03
13.8 13 13.705 1.01
15.56 18 15.533 1.00

3.3.2.4 Analysis of Mode Shapes


The mode shapes of the model modes tabulated in Table 3.3.2.3.1 are displayed in Figure

3.3.2.3.2. From the modal identification analysis, the 5.67 Hz frequency was matched to a

W21x62 girder mode, while the 7.14 Hz frequency was linked to excitation of the W12x27

beams. Therefore, the model modes which correspond to these measured modes are expected to

exhibit this type of excitation. This is the case, as the first mode of the W21x62 girders is

excited in Mode 1 and the first mode of the W12x27 beams is excited in Mode 4. Due to the

modal frequency and mode shape correlation, the modified model of the UW Faculty Club,

which includes body constraints between the slab, beams, and girders, adequately generates the

modal characteristics of the actual floor.

Figure 3.3.2.3.2: Mode shapes for UW Faculty Club Model


96

Figure 3.3.2.3.3: Mode shapes for Mode 1 (W21x62 excitation) and


Mode 4 (W12x27 excitation) from UW Faculty Club Model

3.3.3 Building K1
3.3.3.1 Initial Model
Three different 2-D models of Building K1 were created in order to determine the most basic

model that could adequately replicate the response of the loaded areas of the floor. Model 1,

shown in Figure 3.3.3.1.1 alongside the actual floor plan, is a model of an entire floor. The bay

where measurements were taken is highlighted. The open space in the center of the model is the

location of the concrete core wall. The boundary conditions along the edge of the core wall are

fully fixed against translation and rotation. At the column locations, the columns were assumed

to be sufficiently stiff to resist translation. Therefore, supports that restrict vertical and

horizontal displacement were placed at the nodes located at the center of each column.
97

x
Figure 3.3.3.1.1: (a) Computer model 1 for Building K1; (b) Actual building floor plan for K1

Along with the fixed displacement conditions, rotational springs, which account for the rotational

stiffness of the columns, were inserted about the two horizontal axes. The value of the stiffness

was determined according to the process outlined in Appendix C. Because no cylinder strength or

stiffness test results were available for this building, an initial value of Ec = 5300 ksi was

assumed for the concrete modulus of elasticity. This modulus is based on a concrete strength of

5000 psi and includes a dynamic factor of 1.35. Since Model 1 is a model of the entire floor, the

modal response of this model was used as the baseline to judge the other two models described

below to determine the simplest model that could be used to provide an accurate response.

Displayed in Figure 3.3.3.1.2(a), Model 2 is a simplified version of Model 1, consisting of only

the half of the floor that includes the bay that was loaded for the heel drop and walking tests.

The boundary conditions for the core wall and the column locations remain the same as for

Model 1. Finally, Model 3 further simplifies the floor, only including only one bay on either side

of the loaded bay. This model is shown in Figure 3.3.3.1.2(b). Again, the bay where
98

measurements were taken is highlighted. Because of the complexity of the boundary conditions

along the edges of the exterior bays in Models 2 and 3 and the difficulty in representing them,

the slab edges were left free, apart from the column support, at these locations.

Figure 3.3.3.1.2: (a) Computer Model 2 and (b) Model 3 for Building K1

Modal analyses were run for each of these models and the mode shapes and frequencies

produced for Models 2 and 3 were compared to those of Model 1 to determine the adequacy of

these simplified models. The first two modes from Model 1 are shown in Figure 3.3.3.1.3. The

movement in these modes is on the opposite side of the building from where the field

measurements were taken. Because of the location of these bays with respect to the measured

area, it is unlikely that these modes could be excited by loading in the measured bay and it is

impossible for Models 2 and 3 to replicate this response because these sections of the floor were

not modeled. Therefore, these two modes and the higher modes from Model 1 that included

activity primarily on the opposite side of the floor were left out of the comparison.

Figure 3.3.3.1.3: Modes 1 and 2 from the modal analysis of Model 1 of Building K1

The remaining mode shapes and matching frequencies from Model 1 are compared to the first

six mode shapes and frequencies for Model 2 in Figure 3.3.3.1.4. The bays excited in each mode
99

for Model 2 are almost identical to the corresponding modes in Model 1. The frequencies for

these corresponding modes are within 0.4% of each other for the first five modes. The frequency

of the sixth mode in Model 2, though slightly less accurate, is still within 2.46% of the frequency

from Model 1. This comparison shows that Model 2 can be used to reproduce the response of

the measured section of the floor with great precision.

Figure 3.3.3.1.4: Mode shape comparison between Model 1 and Model 2 for Building K1

Model 3 was created to determine whether it is possible to further simplify the floor model from

the initial simplified version, Model 2. Figure 3.3.3.1.5 shows the mode shape comparison

between Model 2 and Model 3. Mode 1 from Model 2 has been omitted from the comparison

because the predominant motion occurring in this mode is in the bays that have been left out of

Model 3. The mode shapes generated for Model 3 are very similar for the first three modes, then
100

the similarity in mode shapes diverge. Though the frequency for Mode 3 from Model 3 is within

1% of the frequency from the similar mode in Model 2, the error for remaining modes ranges

from 6% in Mode 1 to 47% in Mode 5. After the third mode, the frequency error increases

along with the divergence of the mode shapes, indicating that the ability of Model 3 to replicate

the response of Model 2, and in turn Model 1 and the actual floor, decreases. Therefore, Model 3

cannot be used to model the measured section of the floor with any kind of precision, and Model

2 is chosen as the most basic model that can effectively validate the measured behavior of the

floor. This finding suggests that a successful model must include at least two bays either side of

the loaded bay. The same result was found for the WCPD.

Figure 3.3.3.1.5: Mode shape comparison between Model 2 and Model 3 for Building K1

Analysis of Mode Shapes


The column layout of this building is fairly regular, forming mostly rectangular bays. This

regularity is especially true around the measured bay and is the reason that this section of the
101

floor was selected for the load tests. Though the apparent regularity of the bay sizes and shapes

was a reason for the use of the building in this analysis, a few irregularities are present which

account for atypical modal shapes. Building K1 has a large floorplan with a concrete core wall

located directly in the center of the floor. The stiffness and location of the core wall creates a

barrier between the north and south sides and between the east and west sides of the floor,

effectively isolating motion in each section. The stiffness of the panels adjacent to the core wall

is increased by the proximity to this shear wall. The presence of the core wall creates essentially

fixed boundary conditions, creating additional stiffness against the overall movement and

rotation of these bays. Due to this higher stiffness, the amount of strain energy needed in order

to generate the mode shapes in these bays is larger than for the unstiffened bays. The core wall

acting as a barrier and the added stiffness to the adjacent bays hinders the ability of the floor to

deflect in a typical fundamental sinusoidal mode shape; therefore the fundamental frequency and

lower modes of the floor will not include significant displacements in the bays adjacent to the

core wall.

Two additional sections of the floor are less stiff than the rest of the bays, adding to the modal

irregularity. A ten foot overhang is located in the southwest corner of the floor and a 27’ long

bay, which is four feet longer than the typical bay for this building, is located in the northwest

corner of the floor. These more flexible sections require less strain energy to generate movement

than any of the other bays. Therefore, the lower modes of the entire floor will include the

movement of these bays. This, along with the stiffened core wall bays, creates the localized

movement in the most flexible sections for the lower modes of instead of a more regular

sinusoidal shape that includes multiple bays.


102

The modal comparison between Models 1 and 2 showed that the core wall effectively isolates

motion to the section of the floor that is being loaded. Additionally, the mode shapes and

frequencies of Model 2 have proven to be representative of the behavior of the loaded side of the

floor. Therefore, the entire floor does not need to be modeled in order to determine the response

of the measured section of the floor.

3.3.3.2 Model Modifications


3.3.3.2.1 Geometric Properties
In Models 1, 2, and 3, vertical support at each column location was provided at a single node at

the center of the column. The column was thus treated as a “stick” member, with zero lateral

dimensions. The consequence is that plate bending is permitted in the plate elements that lie

within the perimeter of the real column (24” x 24” in this case), and the model floor is more

flexible than the floor in the actual building.

To study the effect of this shortcoming, Model 2 was modified (to Model 2A). In it, the mesh in

the region of each column was changed by placing nodes at the perimeter of the column, as

shown in Figure 3.3.3.2.1. All plate elements inside the boundary were give Ec = 1x109 ksi

(more than five orders of magnitude higher than the floor plate elements). This change causes

plane sections in the column to remain plane and stiffens the floor. In frame models that consist

of “stick” beam and column elements, a similar result can be achieved by using rigid end offsets.

However, this feature is not available for plate elements. As in Model 2, the node at the center of

the column was fixed against translation and rotation in all three directions
103

Figure 3.3.3.2.1: Finite-sized column model for K1

The modal characteristics from this Model 2A compared to the stick column model with the

same modulus of elasticity of 5300 ksi (Model 2) are shown in Table 3.3.3.2.1. For the first four

modes, the inclusion of the rigid zones increases the frequency by about 30%. This is almost

twice as large as the change introduced by the dynamic factor on concrete modulus. The mode

shapes for the first four modes from both models are compared in Figure 3.3.3.2.2.
104

Table 3.3.3.2.1: Mode shape comparison


Frequency (Hz)
Finite-Sized Column
Stick Column Model
Mode Model f3a/f3 ratio
(Model 2)
(Model 2A)
1 11.776 9.025 1.30
2 11.883 9.116 1.30
3 12.003 9.331 1.29
4 12.097 9.482 1.28

Figure 3.3.3.2.2: Mode shape comparison

Though the mode shapes remain similar, the increase in frequency for each mode indicates that

the additional rigidity at the column locations must be included to generate accurate modes and

frequencies of floors with columns. The additional stiffness at the column locations due to the

finite-sized columns is further displayed in Figure 3.3.3.2.3, which includes the displaced shapes

for both the stick column and finite-sized column models.


105

Figure 3.3.3.2.3: Displaced shapes for (a) stick column model and (b) finite-sized column model.

3.3.3.2.2 Material Properties


The variation between the fundamental frequency of the measured data and the finite-sized

column model, which is fully constrained geometrically, indicated that concrete cracking of the

reinforced concrete slab and columns must be accounted for in order to decrease the stiffness of

the model to match the measured frequency. To account for the amount of cracking in the slab, a

slab cracking factor, αcr, was created. This factor, which represents the ratio of the cracked

stiffness with respect to the stiffness of the gross section, is given by

𝐸𝐸𝑐𝑐 ×𝐼𝐼𝑐𝑐𝑐𝑐 𝐼𝐼𝑐𝑐𝑐𝑐


𝛼𝛼𝑐𝑐𝑐𝑐 = = (𝐸𝐸𝐸𝐸. 3.3.3.2.2.1)
𝐸𝐸𝑐𝑐 ×𝐼𝐼𝑔𝑔 𝐼𝐼𝑔𝑔

where Ec is the static modulus of elasticity of the concrete, Ig is the moment of inertia of the

uncracked section, and Icr is the cracked moment of inertia. This factor was applied to the static

modulus of elasticity of the slab in the model to give the slab the cracked stiffness. This was
106

done in lieu of reducing the gross moment of inertia of the slab to the cracked moment of inertia;

however, the stiffness of the slab is reduced by the same factor.

The αcr factor was found necessary to match the frequencies from the model to floor frequencies

in Building K1, which has an RC (non-post-tensioned) floor. Ideally, each plate element in the

floor would have been given an individual αcr value. These values would be low (<<1.0) near

the columns where cracking is likely to be more prevalent, and higher (≈1.0) in the regions near

the inflection points, where cracking is unlikely. However, reliable identification of such

individual αcr values would require very detailed measurements of all the mode shapes (in

addition to the frequencies), and that was not possible with the resources available. Use of an

average αcr value, applied to all elements in the floor, was the best that could be done.

The effect of dynamic loading on the modulus must also be addressed. According to Neville

(1997) and Murray et al. (1999), this dynamic modification factor should range from 1.20 to

1.35. The West Campus Parking Deck model is in agreement with this range, as a modification

factor of 1.30 was used to match the fundamental frequency of the model to the measured value.

Therefore, this value was used to determine the adequacy of the dynamic modification factor,

αdyn, used for Building K1, as well. Applying both the cracking factor, αcr, and the dynamic

modification factor, αdyn, generates the slab’s effective dynamic modulus, Eslab:

𝐸𝐸𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = 𝛼𝛼𝑐𝑐𝑐𝑐 𝛼𝛼𝑑𝑑𝑑𝑑𝑑𝑑 𝐸𝐸𝑐𝑐 = 𝛼𝛼𝑡𝑡𝑡𝑡𝑡𝑡 𝐸𝐸𝑐𝑐 (𝐸𝐸𝐸𝐸. 3.3.3.2.2.2)

where αtot is the combination of αcr and αdyn. This is the modulus of elasticity input for the model

of Building K1 to account for the cracking and dynamic effects.


107

The final material constraint that must be addressed is the cracking of the columns. This will

affect the rotational stiffnesses of the columns, which are represented by the rotational springs at

the column locations in the model. The amount of cracking in the columns, along with the

variables that make up αtot, the amount of cracking in the slab and the actual dynamic factor that

should be used, is unknown. Therefore, a series of test models was created to optimize these

values. Each test varies the αtot applied to the floor plate elements to adjust for dynamic effect

and for the amount of concrete cracking together. The αtot values used were 1, 5/6, 2/3, and 1/2.

The amount of column cracking was also accounted for in each test. The tests vary with the

kθ/kθo ratio, where kθo is the rotational spring/column stiffness for an uncracked column and kθ is

the stiffness assuming column cracking. The ratios used here are 1, 0.7, and 0.5, which span the

range suggested by ACI 318-08 (Section 10.10.4.1).

Each combination of αtot values and kθ/kθo ratios was tested and the resulting mode shapes and

frequencies were recorded. The test variables and first mode frequencies generated from each

test are shown in Table 3.3.3.2.2.1.


108

Table 3.3.3.2.2.1: Building K1 Test Output


Variables
Measured Model fn
Eslab fmodel/fmeas
αcr αdyn αtot kθ/kθO fn fn √α
(ksi)

0.741 1.35 1.00 5000 100 10.07 11.996 1.19 12.00


0.617 1.35 0.83 4167 100 10.07 10.951 1.09 12.00
0.494 1.35 0.67 3333 100 10.07 9.795 0.97 11.99
0.370 1.35 0.50 2500 100 10.07 8.483 0.84 12.00
0.741 1.35 1.00 5000 70 10.07 11.823 1.17 11.82
0.617 1.35 0.83 4167 70 10.07 10.793 1.07 11.83
0.494 1.35 0.67 3333 70 10.07 9.653 0.96 11.82
0.370 1.35 0.50 2500 70 10.07 8.36 0.83 11.82
0.741 1.35 1.00 5000 50 10.07 11.614 1.15 11.61
0.617 1.35 0.83 4167 50 10.07 10.602 1.05 11.62
0.494 1.35 0.67 3333 50 10.07 9.482 0.94 11.61
0.370 1.35 0.50 2500 50 10.07 8.212 0.82 11.61
0.559 1.35 0.76 3775 70 10.07 10.067 1.00 11.62

Two plots of the results for Mode 1 of each test are shown in Figures 3.3.3.2.2.1(a) and (b). The

fundamental frequency is plotted as a function of the amount of column cracking in (a) and then

again in (b) as a function of the Eslab used in the model. In Figure 3.3.3.2.2.1 (b), each curve

represents a different kθ/kθo ratio. The fact that they lie very close to each other shows that

column cracking has only a minimal effect on the fundamental frequency. These shifts are

slightly larger than the FFT frequency bandwidth of 0.09 Hz, and therefore, the magnitude of this

effect is about the same as the precision of the measured data. Because changes in the column

stiffness had a minimal effect on the frequencies of the model and since the ACI

recommendation for the amount of column cracking is 70%, a kθ/kθo ratio of 0.7 was chosen for

the model.

The αtot value, however, has a much larger effect on the fundamental frequency. This can be

seen in Figure 3.3.3.2.2.1 (b), in which each curve represents a different αcr. The curves are
109

spaced quite widely, indicating that the slab stiffness has a much larger effect that the column

stiffness. Further evidence that the slab behavior is dominated by α rather than kθ is provided by

the fact that fn is almost exactly proportional to √α, as shown in Figure 3.3.3.2.2.2.

Figure 3.3.3.2.2.1: System identification plots for Building K1 models

Figure 3.3.3.2.2.2: System identification plots for Building K1 models

In order to achieve a fundamental frequency that matched the measured frequency using the

chosen kθ/kθo ratio of 0.7, the plots indicate that a modulus of approximately 3800-4000 ksi, or

an αtot value of about 0.77, must be used. If αcyn is taken as 1.3, αcr must be 0.77/1.30 = 0.59.
110

Figure 3.3.3.2.2.3: Mode shapes for Building K1 model

The mode shapes, which remain the same for each test, are shown in Figure 3.3.3.2.2.3. There is

no movement in the measured bay in the mode shape for Mode 1. Therefore, the first mode with

excitation in the measured bay (Mode 2 in this case) was assumed to be the fundamental

frequency measured by the FFT because this would be the lowest mode that could be excited by

loading in this bay. Therefore, the αtot value was modified until the frequency of this mode

matched the measured frequency. The αtot value to achieve this was 0.77 and the results of this

test are shown in the final row of Table 3.3.3.2.2.1. The frequencies for the first eight modes

using αtot = 0.77 are shown in Table 3.3.3.2.2.2.

Table 3.3.3.2.2.2: Building K1 Frequencies


Model Measured
Frequency Mode Frequency fmodel/fmeas
Mode
(Hz) Group (Hz)
1 9.98 1 10.07 0.991
2 10.07 1 10.07 1.000
3 10.17 1 10.07 1.010
4 10.25 1 10.07 1.018
5 10.81 2 11.07 0.977
6 11.54 2 11.07 1.043
7 12.72 3 13.64 0.933
8 13.08 3 13.64 0.959
111

The measured data are shown in Figure 3.3.3.2.2.4, and demonstrate that there are multiple

modes with closely spaced frequencies. The same was found to be true in the model. From the

FFT of Building K1, the modes are grouped in three frequency ranges which are very near the

three predominant frequencies. The modes produced by the model fall into these same three

ranges, validating the accuracy of the model. Therefore, each mode is compared to the nearest

predominant frequency that was determined from the FFT. This comparison is shown along with

the modal frequencies in Table 3.3.3.2.2.2.

Figure 3.3.3.2.2.4: Power spectrum for heel drop test at center of slab for K1 on Floor
34
Not only does the fundamental frequency from the model match that from the measured data, but

the model contains a mode that is within 1.5% of the second measured frequency and a separate

mode within 3.3% of the third measured frequency. This correlation suggests that the real

geometric constraints and the material properties for Building K1 have been replicated as closely

as possible, so the model can now be loaded to determine how well the response of the model

corresponds to the measured response to the heel drop and walking loadings.
112

3.3.4 Building K2
3.3.4.1 Initial Model
As mentioned in Chapter 3, the measurements in this building were taken in the section of the

building with a relatively regular column grid and rectangular bay shapes, while the irregular

section of the building was left out of the study. Because of this, only the regular portion of the

floor was included in the model. The initial model, along with the actual floor plan of this

section of the building, is shown in Figure 3.3.4.1.1 with the measured bay highlighted. Based

on the assumption that the 56 day concrete strength was similar to that of Building K1 (8000

psi), the concrete modulus of elasticity used for the initial model was 5000 ksi (Note that no

dynamic factor for Ec was used at this stage). The elements in the regions of the columns were

modeled in the same manner as in Model 2A for Building K1. Vertical support was provided at

a single node at the column center, but all the plate elements within the column boundary were

made very stiff. The square columns consisted of four stiffened plate elements (Ec=1E9 ksi),

with rotational springs at the center node to account for the flexural stiffness of the columns.

The short walls consisted of multiple square plate elements of the same modulus. However, the

rotational springs were spread out along the length of the walls.

The concrete shear wall is present in the north-east corner of the model. The boundary conditions

along the edge of the core wall are fully fixed against translation and rotation. The shear walls

within the model around the two north-western bays were modeled similarly to the finite-sized

columns, with the cross-section of the core-walls modeled with very stiff plate elements and

rotational springs along the length of the wall to account for the rotational stiffness at these

locations.
113

Figure 3.3.4.1.1: (a) computer Model 1 for K2 building; (b) Actual building floor plan for K2

A fundamental frequency of 6.941 Hz was generated from the modal analysis of this initial

model. This frequency was 3.5% lower than the measured fundamental frequency of 7.194 Hz,

so the model must be made stiffer.

3.3.4.2 Model Modifications


With the inclusion of the finite-sized columns and the boundary conditions at both the shear

walls within the model and along the edge of the model, the geometry for Building K2 was

believed to be represented as well as possible. Therefore, the increase in stiffness needed to

match the model’s fundamental frequency to that from the FFT had to come from modifications

to the material properties. An incorrect initial assumption also increased the amount of stiffness

that was needed to meet this requirement. The initial assumption of a 56 day concrete strength of

8000 psi proved incorrect, as cylinder test information later became available which indicated

otherwise. Cylinders tested at seven days had strengths of 5560, 4800, and 5450 psi, with a 56

day strength specified at 6000 psi. This decreases the static modulus of elasticity from 5000 ksi
114

to 4415 ksi. Using this modulus of elasticity further decreases the stiffness of the model,

generating a model fundamental frequency of 6.459Hz (10% too low).

To account for the deficiency in the stiffness of the model, a series of system identification tests

were run, as was the case with Building K1. The variables for these tests were again the amount

of cracking in the columns (kθ/kθΟ) and the actual modulus of elasticity of the slab when

accounting for slab cracking and dynamic effects (αtot = Eslab/Eco). The kθ/kθΟ ratios used for

these tests were 1, 0.7, 0.5, and 0.25. The αtot value consists of αcr , the slab cracking factor, and

αdyn, the dynamic modifier. Since the floors in Building K2 are post-tensioned, it was assumed

that little to no cracking exists in the slab. Therefore, the αcr value was set equal to 1 for all tests,

making the dynamic component the only modifier for the modulus (αtot = αdyn). Because the

dynamic modulus of elasticity is always larger than the static modulus, all of the αtot values used

in the tests were greater than 1.

The mode shapes of the model of Building K2, which remain the same for each test, are shown

in Figure 3.3.4.2.1. The lowest mode with excitation in the measured bay, Mode 2, was selected

as the fundamental mode to be compared to the measured fundamental mode. The test variables

for each test, the resulting frequency for Mode 2, and the error between the model frequencies

and the measured fundamental frequency are shown in Table 3.3.4.2.1.


115

Figure 3.3.4.2.1: Mode shapes for Building K2 model

Table 3.3.4.2.1: Building K2 Test Output


Variables Measured Model
fmodel/fmeas
αcr αdyn αtot Eslab (ksi) kθ/kθO fn (Hz) fn (Hz)
1.000 1.15 1.15 5098.235 100 7.194 7.298 1.01
1.000 1.27 1.27 5608.058 100 7.194 7.605 1.06
1.000 1.39 1.39 6117.882 100 7.194 8.129 1.13
1.000 1.50 1.50 6627.705 100 7.194 8.268 1.15
1.000 1.15 1.15 5098.235 70 7.194 7.186 1.00
1.000 1.27 1.27 5608.058 70 7.194 7.537 1.05
1.000 1.39 1.39 6117.882 70 7.194 7.8722 1.09
1.000 1.50 1.50 6627.705 70 7.194 8.194 1.14
1.000 1.15 1.15 5098.235 50 7.194 7.071 0.98
1.000 1.27 1.27 5608.058 50 7.194 7.416 1.03
1.000 1.39 1.39 6117.882 50 7.194 7.746 1.08
1.000 1.50 1.50 6627.705 50 7.194 8.062 1.12

Maintaining 70% cracking in the column, the αtot value was modified until the frequency of the

second mode matched the measured frequency. The plots in Figure 3.3.4.2.2 again show that the

floor stiffness, manifested here by αtot, has much more effect on the frequency than does the

column stiffness. They indicate that that the optimal αtot value is approximately 1.15,

corresponding to Ec = 5098 ksi. Inputting this αtot value into the model generated a Mode 2
116

frequency within 0.11% of the measured fundamental frequency. Figure 3.3.4.2.3 further

displays that the dynamic factor has a larger effect than the change in column stiffness.

Figure 3.3.4.2.2: System identification plots for mode 2 from the Building K2 model

Figure 3.3.4.2.3: System identification plots for mode 2 from the Building K2 model

Similarly to Building K1, the values in this plot remain at a constant value of 6.43 for all values

of αtot for a constant kθ/kθO =0.7. Therefore, a relationship can be created to determine the

approximate αtot value that is needed to match the measured fundamental frequency when kθ/kθO

= 0.7:

𝑓𝑓 𝑛𝑛 ,𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 2
αtot = � � (Eq. 3.3.4.2.1)
6.43
117

Table 3.3.4.2.2: Building K2 Frequencies


Measured Model
Mode of
Mode Frequency Frequency fmodel/fmeas
Comparison
(Hz) (Hz)
1 1 7.194 6.914 1.04
2 1 7.194 7.186 1.00
3 1 7.194 7.446 0.97
4 1 7.194 7.562 0.95
5 1 7.194 7.672 0.94
6 --- --- 9.062 ---
7 --- --- 9.222 ---
8 --- --- 9.368 ---
18 2 13.75 13.579 1.01
26 3 18.13 18.05 1.00

The frequencies for the first eight modes, the 18th mode, and the 26th mode are shown in Table

3.3.4.2.2 and are compared to the nearest predominant frequencies that were determined from the

FFT. Modes 6, 7, and 8 were not used in the comparison because these modes contained no

significant motion in the bay where measurements were taken. Similar to the FFT of Building

K1, the FFT for Building K2 in Figure 3.3.4.2.4 shows that multiple modes are grouped around

the predominant frequencies. The FFT indicates a high concentration of excited frequencies

centered about the fundamental frequency of 7.194 Hz and extending up to approximately 10 Hz.

This concentration of closely spaced modes makes it difficult to compare the modes from the

model to the nearby predominant frequencies. However, since the fundamental frequency

matches that from the FFT, because the following modes have frequencies which fall in this

concentrated range of frequencies, and finally, because there are modes generated by the model

that are near each of the excited modes from the FFT, the model is believed to accurately

simulate the response of the floor in Building K2.


118

Figure 3.3.4.2.4: Power spectrum for floor 24 for Building K2

3.3.4.3 Analysis of Mode Shapes


Though the column layout in Building K2 is typical and the bay shapes are typically rectangular,

the presence of shear walls creates irregularities within the mode shapes. Both the shear wall

along the north-eastern edge of the model and the two shear walls within the model in the north-

western portion of the model increase the stiffnesses of the adjacent bays. It is not until the sixth

mode that the bay adjacent to the edge shear wall is excited. Similarly to the core wall in the

center of Building K1, the shear walls around the two north-western bays create a barrier which

prevents motion in the adjacent bays from exciting these bays. The eighth mode is the first mode

in which there is any excitation in these bays. This barrier effect and the added stiffness prevent

the formation of typical modes which are sinusoidal in shape and excite all of the bays in the

floor.

3.4. Modeling Conclusions


The final models for each of the buildings in this study effectively replicate the modal

characteristics of the actual floors from the measured buildings. Each model, however,

underwent major geometric and material modifications from the initial models before becoming
119

completely defined. Each model added both insight to the necessary procedures and an

additional modeling principles that must be applied to 2D floor models in order to generate a

response that effectively mimics the actual response of the floor being modeled. The criteria

created from these models are described below:

• A 3-D model of the building is unnecessary because a 2-D model of the floor is adequate.

• At least two bays on each side of the loaded floor must be modeled.

• If only part of the floor is modeled, the mass and stiffness of any beam located at the cut

line must be modified to compensate for the material omitted.

• To simulate the behavior of non-composite floors, the beams, girders, and slab of these

systems must each have their own nodes at the locations where these elements come into

contact. Constraints must be used to allow these elements to be connected to or to move

with respect to the other elements to simulate the actual behavior of a non-composite

system.

• Finite-sized columns must be used at the column locations for 2-D floor models of

concrete flat plate buildings to represent the stiffness contribution of the column at these

locations.

• The elastic modulus of the concrete must be increased to account for dynamic behavior

and decreased to account for cracking (if any). Post-tensioned floors are likely to be

uncracked, but a significant reduction in E is likely to be needed for RC floors.

• The effects of column bending stiffness can be accounted for in the 2-D models by use of

rotational springs of equivalent stiffness. The column stiffness has a large effect in floors

with beams, but only a small effect in flat plate construction.


120

CHAPTER 4: MODELING AND FIELD MEASUREMENT DATA CORRELATION


To achieve a perfect match between the predicted and measured responses, both the structural

model and the loading function must replicate exactly those that exist in the field. In practice,

perfect matching is unlikely, so, to achieve the best match possible, the process was conducted in

two stages. In the first stage, reported in Chapter 3, the structural properties were identified

using a loading function (e.g. the heel drop test) to the characteristics of which the floor response

is relatively insensitive. By this means, structural models were created that were able to match

closely the frequencies of the first few modes of the floors. In this chapter, the models’ loading

responses are compared to the measured responses to further evaluate the structural models and

to try to identify the characteristics of the loading functions. For each of the buildings tested,

these comparisons are discussed below.

As mentioned in Section 1.3, the peak RMS value is a much more effective indicator of the

steady response than is the ultimate peak response value. Therefore, the peak RMS response

values from the models and measured data are compared for the walking tests along with the

ultimate peak values. Since the heel drop test is a transient test and the response is not a steady-

state, the peak RMS values do not provide much insight as comparisons. Therefore, the RMS

values are only compared for the walking tests. The values from the model and measured data

are compared using a Measured/Model acceleration amplitude ratio. These ratios indicate the

severity of the models’ underestimation or overestimation of each building’s response.

4.1 West Campus Parking Deck


To simulate the heel drop and walking tests performed in field testing conducted in the WCPD,

pulse loads and walking loads were applied to the model. The amount of damping determined

from the measurements for this building was 3%. This value is assumed to be a property of the
121

structure, and not affected by the loading type, and was applied to the model for both loading

cases. The results for these loading cases are described below and compared to the measured

response data.

4.1.1 Heel Drop Test/Triangular Pulse


The double triangular pulse that was described in Section 3.2 to represent the heel drop pulse

was applied to the model at the same two locations at which the heel drop tests were performed

in the parking deck (center of the slab and mid-span of the beam). The results of these two tests

are compared with the measured data from the heel drop tests performed at the same locations in

Table 4.1.1.1 and the averages of the magnitudes of the positive and negative peak response

values are shown in Table 4.1.1.2. Negative acceleration indicates downward motion, and the

first major peak in the response to a heel drop is therefore negative. Measured displacements

were obtained through integration of the acceleration record in an internal operation within the

RT Pro software. A ratio of the measured values over the model’s values is provided to show

how the model response values compare to the measured values. The model and measured

acceleration response plots are compared for the pulse at the center of the slab and for the pulse

at the midspan of the beam in Figures 4.1.1.1 (a) and (b).

Table 4.1.1.1: WCPD Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmax dmax dmin
Pulse Location
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 19.3 -19.3 0.21 -0.278 0.00285 -0.0032


Center of Slab Model 23.09 -16.21 0.1841 -0.239 0.00186 -0.00465
Meas/Model 0.84 1.19 1.14 1.16 1.53 0.71
Measured 5.03 -4.11 0.0685 -0.0724 0.0011 -0.0011
Midspan of
Model 5.285 -3.156 0.0543 -0.0532 0.00051 -0.0011
Beam
Meas/Model 0.95 1.30 1.26 1.36 2.21 0.97
122

Table 4.1.1.2: Average of Positive and Negative


Peak Response Values for WCPD
Pulse Average Average Average
Location peak a peak v peak d

Measured 19.30 0.2440 0.0031


Center of
Model 19.65 0.2119 0.0033
Slab
Meas/Model 0.98 1.15 0.94
Measured 4.57 0.0705 0.0011
Midspan
Model 4.2205 0.0537 0.0008
of Beam
Meas/Model 1.08 1.31 1.36

Figure 4.1.1.1: WCPD measured and model acceleration response comparison to heel
drop/triangular pulse at (a) center of slab, and (b) midspan of beam
For both tests, the magnitudes and the times at which the first two positive and negative peak

acceleration values occur match generally well for both acceleration histories. This is expected

for the test performed at the center of the slab for this building because the pulse was optimized

using the measured acceleration response, as discussed in Appendix E. However, the similarities

displayed in the comparison for the test at the beam mid-span further supports the choice of this

pulse, indicating that the pulse duration of the pulse function chosen to simulate the heel drop

pulse is very similar to the actual heel drop pulse. The difference for the average peak
123

accelerations in both tests is low, as the ameas/amodel ratio is near 1 for the peak average

accelerations, while the velocity and displacement differences are larger.

4.1.2 Walking Test/Camel Hump


The walking test for West Campus Parking Deck was performed with two individuals.

Therefore, the Two Walkers camel hump function described in Section 4.2 was applied to the

center of the slab in this model. The results of the measured data and the model response are

shown in Tables 4.1.2.1 and 4.1.2.2, and the acceleration responses are compared in Figure

4.1.2.1.

Table 4.1.2.1: WCPD Measured and Model Comparison to Walking Loading


Loading amax amin vmax vmax dmax dmin
Function (in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 5.8 -5.18 0.0814 -0.0679 0.0011 -0.00107


Two Walkers Model 6.324 -7.659 0.161 -0.198 0.00109 -0.012
Meas/Model 0.92 0.68 0.51 0.34 1.01 0.09

Table 4.1.2.2: Average of Positive and Negative Peak Response Values for WCPD
Loading Average aRMS Average vRMS Average dRMS
Function peak a (in/sec2) peak v (in/sec) peak d (in)

Measured 5.49 4.06 0.0747 0.0282 0.0011 0.0004


Two
Model 6.99 2.14 0.1792 0.0405 0.0065 0.0017
Walkers
Meas/Model 0.79 1.90 0.42 0.70 0.17 0.25
124

Figure 4.1.2.1: WCPD measured and model acceleration


response comparison to walking/camel hump loading

Though the model produced slightly larger maximum accelerations, the intermediate peak

acceleration values from the model tend to be smaller in magnitude than those from the

measured test along the entire record. This is displayed by the peak arms ratio for this test, which

indicates that the steady-state response is lower for the model than for the measured data, while

the highest peak is larger for the model. The plots for the measured and model are not in phase

with one another, and were not expected to be, because the walking in the measured test was a

random walking test, while the function applied to the model assumes that the two were walking

at the same step frequency and at the same location, and therefore, has a constant step frequency.

However, the period between highest peaks is between 0.5 seconds and 0.6 seconds for both

acceleration records. This is similar to the period of the model’s forcing function, which is 0.575

seconds, indicating that this is an appropriate approximation for the period of the actual walking

excitation. As with the heel drop tests, the variation between the ultimate peak acceleration

values for the measured data and the model’s output was the smallest out of the three
125

comparisons. The difference increased for the velocity comparison and increased again for the

displacement response.

4.2 UW Faculty Club


From the measured response, the amount of damping in the UW Faculty Club was determined to

be 2.5%. As mentioned previously in UW Faculty Club section in Chapter 3.3, a large

component of this damping is believed to be attributed to the loose wood paneling that is present

above the concrete floor. This damping value was applied to the model for all three heel drop

load cases and for the walking load case. The results for these loading cases are described below

and are compared with the measured response.

4.2.1 Heel Drop Test/Triangular Pulse


The double triangular pulse chosen to represent the heel drop pulse was applied in the model at

the center of the slab, at the midspan of the W12x27 beam, and at the W21x62 girder. Table

4.2.1.1 has the results of these pulse tests and provides a comparison with the measured peak

response values. The plots of the acceleration responses for these tests are compared with the

measured acceleration records of the tests performed at the same locations in Figures 4.2.1.1,

4.2.1.2, and 4.2.1.3. Because the measured acceleration values are much larger than those

generated by the model, the measured accelerations were factored to make comparisons easier to

make.

Table 4.2.1.1: UW Faculty Club Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmax dmax dmin
Pulse Location
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)
Measured 10.27 -23.82 0.113 -0.113 0.0043 -0.0033
Center of Slab Model 32.32 -20.48 0.193 -0.312 0.0019 -0.0060
Meas/Model 0.32 1.16 0.58 0.36 2.26 0.55
Measured 212.71 -342.13 0.262 -0.454 0.0016 -0.0018
W12x27 Mid-
Model 15.19 -16.19 0.155 -0.179 0.0021 -0.0039
span
Meas/Model 14.00 21.13 1.69 2.53 0.78 0.45
Measured 23.33 -20.78 0.11 -0.112 0.00029 -0.00027
W21x62 Mid-
Model 32.43 -19.47 0.21 -0.26 0.0026 -0.0045
span
Meas/Model 0.72 1.07 0.53 0.44 0.13 0.06
126

Table 4.2.1.2: Average of Positive and Negative


Peak Response Values for UW Faculty Club
Pulse Average Average Average
Location peak a peak v peak d
Measured 17.05 0.113 0.0038
Center of
Model 26.4 0.25 0.0039
Slab
Meas/Model 0.65 0.45 0.96
W12x27 Measured 277.42 0.35 0.0017
Mid- Model 15.69 0.17 0.0030
span Meas/Model 17.68 2.14 0.57
Measured 22.06 0.112 0.0003
W21x62
Mid- Model 25.95 0.23 0.0034
span Meas/Model 0.85 0.48 0.08

Figure 4.2.1.1: UW Faculty Club measured and model acceleration response comparison
to heel drop/double triangular pulse applied at the center of the slab.

As seen in Figure 4.2.1.1, the first positive and negative peaks for both acceleration records

occur at nearly the same time and the magnitudes of the initial negative peaks are comparable;

however, the initial measured positive peak is approximately three times larger than the peak

from the model.


127

As mentioned in Section 2.3, the measurement settings were erroneously modified during the

heel drop test above the W12x27 beam creating very large accelerations. The measured

acceleration level was much too high to be realistically caused by human excitation, making the

comparisons to the peak values from the model impossible. The first positive and negative peaks

also occurred at similar times, but the peak frequency is much lower than the measured record

and the damping is much lower.

Figure 4.2.1.2: UW Faculty Club measured and model acceleration response comparison to heel drop/triangular pulse
applied at the midspan of the W12x27 beam. (a) Un-scaled response; (b) Scaled down modified response

The plots for the tests at the W21x62 girder are very similar to the tests at the center of the slab,

in that the initial peaks occur at similar times for both the measured and model responses.

However, in this case, both the first positive and negative peak values are much more

comparable. The frequency of the peaks in the measured response is much higher than for the

model’s response. This indicates that the measured response has a contribution from some

higher modes which are not contributing in the model.


128

Figure 4.2.1.3: UW Faculty Club measured and model acceleration response comparison
to heel drop/triangular pulse applied at the midspan of the W22x62 girder.

4.2.2 Walking Test/Camel Hump


Two individuals were walking during the walking tests in the Faculty Club. Therefore, the

walking load used in the model to simulate this test was the Two Walkers camel hump function.

The results from the loading of this function are shown in Tables 4.2.2.1 and 4.2.2.2, and the

acceleration response is displayed in Figures 4.2.2.1 (a) and (b).

Table 4.2.2.1: UW Faculty Club Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmin dmax dmin
Loading Function
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 3.54 -3.17 0.075 -0.0771 0.0022 -0.0022


Two Walkers Model 11.53 -11.11 0.2832 -0.2948 0.001093 -0.02032
Meas/Model 0.31 0.29 0.26 0.26 2.01 0.11

Table 4.2.2.2: Average of Positive and Negative Peak Response Values for UW Faculty Club

Loading Average aRMS Average vRMS Average


dRMS (in)
Function peak a (in/sec2) peak v (in/sec) peak d

Measured 3.36 3.53 0.0761 0.0720 0.0022 0.0018


Two
Model 11.32 10.47 0.2890 0.1815 0.0107 0.0084
Walkers
Meas/Model 0.30 0.34 0.26 0.40 0.21 0.21
129

The RMS displacements are not similar for the model and the measured data, as the peak RMS

values are off by a factor of 5, while the RMS velocities are slightly more comparable. The

measured arms value is also five times larger than the value from the model. Also, the peak

values for the model are approximately three times those of the measured data, and since the

model loading is a worst case scenario, this outcome is not surprising. Therefore, the measured

response was tripled in Figure 4.2.2.1 (b) to ease the comparison of the two acceleration records.

The large pulses in the model response that occur every 0.6 seconds are believed to correspond

to each step of the walkers in the Two Walkers function. The response dampens out between

each large pulse in the model response, while it remains relatively constant along the measured

response. This disparity is believed to be due to the assumption of the Two Walkers forcing

function that the two individuals walk in stride with one another and at the same exact location.

Figure 4.2.2.1: Acceleration response of UW Faculty Club Model to Two Walkers camel hump loading function applied at
the center of the slab. (a) Un-scaled response; (b) Scaled up measured response
130

4.3 Building K1
The bare floor of Building K1 was chosen as the measured response comparison for the output

from the model. This is due to the difficulty in modeling partitions, cabinets, countertops, and

other non-structural elements that affect the response of the fitted-out floors. The amount of

damping determined from the measurements for the bare floor (Floor 34) in Building K1 was

5%, which was applied to the model for both loading cases. The results for these loading cases

are described below and compared with the measured response.

4.3.1 Heel Drop Test/Triangular Pulse


To simulate the two heel drop tests conducted within Building K1, the double triangular pulse

was applied at the center of the slab and at the midpoint along the column line. Table 4.3.1.1 has

the results of these pulse tests along with the measured response values, and Table 4.3.1.2

provides the average of the magnitudes of the positive and negative peak response values for

each test.

Table 4.3.1.1: Building K1 Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmax dmax dmin
Pulse Location
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 15.8 -24.77 0.188 -0.16 0.001760 -0.00171


Center of Slab Model 22.1 -30.5 0.3642 -0.3676 0.005085 -0.005686
Meas/Model 0.71 0.81 0.52 0.44 0.35 0.30
Measured 6.36 -7.63 0.105 -0.223 0.000993 -0.00103
Column Line Model 26.9 -18.86 0.255 -0.222 0.00155 -0.00397
Meas/Model 0.24 0.40 0.41 1.00 0.64 0.26

Table 4.3.1.2: Average of the Magnitudes of the


Positive and Negative Peak Response Values for
Building K1
Pulse Average Average Average
Location peak a peak v peak d
Measured 20.29 0.1740 0.0017
Center of
Model 26.3 0.3659 0.0054
Slab
Meas/Model 0.77 0.48 0.32
Measured 7.00 0.164 0.0010
Column
Line Model 22.88 0.239 0.0028
Meas/Model 0.31 0.69 0.37
131

The difference in the average peak values for the measured velocities and displacements between

the two test locations is not large, while the measured acceleration undergoes a significant

decrease in magnitude for the test conducted at the midpoint of the column line. This decrease in

the average of the peak acceleration values is not reproduced in the model, as the same two tests

in the model produce only a slight drop in average peak acceleration for the test over the column

line. The decrease in peak measured acceleration indicates that there is an increase in stiffness

along the column line in the building that is not represented within the model.

This effect is again displayed in the plots of the acceleration responses for these tests. The

responses for the model and measured data are provided in Figure 4.3.1.1 for the heel drop at the

center of the slab and Figure 4.3.1.2 for the heel drop at the midpoint of the column line. The

measured data for the heel drop acceleration response at the column line is modified (multiplied

by 5) to make comparisons easier to make because the model generated accelerations are much

larger for this test. At the center of the slab, the measured and model initial peak accelerations

are similar both in magnitude and in the time at which the first peak values occur. At the

midpoint of the column line, the peak acceleration from the model is almost the same as it was at

the center of the slab, but the measured value was much smaller.
132

Figure 4.3.1.1: Acceleration response of Building K1 to heel drop excitation applied at the center of the slab.

The reasons for the discrepancy between the measured and model response over the column line

are unknown. One possibility, which was not substantiated, is that a significant band of

reinforcement existed along the column line. Such reinforcement would be expected to reduce

the crack widths and stiffen the floor locally, thereby reducing the response. In the model, the

cracking was assumed to be uniformly distributed throughout the floor and was represented by a

single number, αcr.


133

Figure 4.3.1.2: Acceleration response of Building K1 to heel drop excitation applied at the
midpoint of the column line. (a) Un-scaled response; (b) Scaled up measured response

4.3.2 Walking Test/Camel Hump


The Two Walkers camel hump function was applied to the Building K1 model to simulate the

two individuals who performed the walking test in Building K1. The results from this test and

the measured walking tests are shown in Tables 4.3.2.1 and 4.3.2.2, while the acceleration

response is displayed in Figure 4.3.2.1. Similarly to the data for the UW Faculty Club, in

Building K1 the model and measured RMS displacements differ significantly, as the peak RMS

values are off by a factor of ; the RMS velocities are slightly more comparable. As for the

acceleration response, both the responses from the model and the measured data have a period of

approximately 0.5 seconds between successive maximum peak accelerations. However, the peak

values for the model are approximately twice those of the measured data. Therefore, the

measured response was doubled in Figure 4.3.2.1 (b) to ease the comparison of the two

acceleration records. Surprisingly, however, the peak RMS acceleration values are much closer,

indicating that the two acceleration records display similar steady-state responses, despite the

highest peak acceleration for the model being twice that of the measured data.
134

Table 4.3.2.1: Building K1 Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmin dmax dmin
Loading Function
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 1.85 -1.93 0.0309 -0.0836 0.00039 -0.00039


Two Walkers Model 4.132 -4.079 0.0639 -0.094 0.00055 -0.0041
Meas/Model 0.45 0.47 0.48 0.89 0.71 0.10

Table 4.3.2.2: Average of Positive and Negative Peak Response Values for Building K1
Loading Average aRMS Average vRMS Average
dRMS (in)
Function peak a (in/sec2) peak v (in/sec) peak d

Measured 1.89 1.21 0.0573 0.0110 0.0004 0.0002


Two
Model 4.11 1.40 0.0790 0.0224 0.0023 0.0008
Walkers
Meas/Model 0.46 0.87 0.73 0.49 0.17 0.21

Figure 4.3.2.1: Acceleration response of Building K1 to walking load case.


(a) Un-scaled response; (b) Scaled up measured response

4.4 K2
The bare floor of Building K2 was chosen as the measured response comparison for the output

from the model. The amount of damping determined from the heel drop tests performed on

Floor 24 of Building K2 and applied to the model for both the heel drop and walking load cases

was 7%. The results for these loading cases are compared with the measured data below.
135

4.4.1 Heel Drop Test/Triangular Pulse


The heel drop was only applied at the center of the slab in Building K2. Therefore, the double

triangular pulse which represents the heel drop pulse load was only applied at this location.

Tables 4.4.1.1 and 4.4.1.2 compare the results of this pulse test with the measured data. The

differences for the average displacements, velocities, and accelerations are all nearly the same,

with the measured values being three times as large as the values from the model in most cases.

Table 4.4.1.1: Building K2 Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmax dmax dmin
Pulse Location
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 10.36 -14.84 0.205 -0.196 0.00336 -0.00338


Center of Slab Model 4.148 -4.426 0.0668 -0.0674 0.000602 -0.00157
Meas/Model 2.50 3.35 3.07 2.90 5.43 2.26

Table 4.4.1.2: Average of Positive and Negative


Peak Response Values for Building K2

Pulse Average Average Average


Location peak a peak v peak d

Measured 12.60 0.2004 0.0033


Center of
Model 4.287 0.0671 0.0010
Slab
Meas/Model 2.94 2.99 3.17

The plot of the acceleration response for this test is shown in Figure 4.4.1.1. The measured data

for the heel drop acceleration response is divided in half to make comparisons easier because the

model generated accelerations are approximately half those of the measured values for this test.

Though the measured peaks are approximately three times larger than the peaks from the model,

the times at which these first few peaks occur are similar for the model and the measured tests.
136

Figure 4.4.1.1: Acceleration response of Building K2 to heel drop applied at the center of the slab

4.4.2 Walking Test/Camel Hump


The Two Walkers camel hump function was again used in the model to simulate the two

individuals walking in the walking test conducted in Building K2. The measured response for

the walking test is compared with the results from this test in the model in Tables 4.4.2.1 and

4.4.2.2, while the acceleration response is displayed in Figure 4.4.2.1. The measured response is

again factored, multiplied by three in this case, to ease the comparison between the

measurements and model. In contrast to the heel drop test for this building, the measured

accelerations and velocities are approximately 25% smaller than the values from the model,

while the peak RMS accelerations are three times larger than the model. Both the responses

from the model and the measured data have a period of slightly more than 0.5 seconds between

successive maximum peak accelerations.


137

Table 4.4.2.1: Building K2 Measured and Model Comparison to Heel Drop Loading
amax amin vmax vmin dmax dmin
Loading Function
(in/sec2) (in/sec2) (in/sec) (in/sec) (in) (in)

Measured 2.47 -2.74 0.0044 -0.1301 0.00056 -0.00051


Two Walkers Model 3.753 -3.174 0.078 -0.098 0.00084 -0.0057
Meas/Model 0.66 0.86 0.06 1.33 0.66 0.09

Table 4.4.2.2: Average of Positive and Negative Peak Response Values for Building K2

Loading Average aRMS Average vRMS Average


dRMS (in)
Function peak a (in/sec2) peak v (in/sec) peak d

Measured 2.61 2.69 0.0275 0.0124 0.0005 0.00


Two
Walkers Model 3.46 1.09 0.0879 0.0201 0.0033 0.0009
Meas/Model 0.75 2.47 0.31 0.62 0.16 0.22

Figure 4.4.2.1: Acceleration response of Building K2 to walking loading case

4.5 Building Comparison


Table 4.5.1 provides the ratio of the average peak response values of the measured data over that

of the model response for the heel drop/double triangular pulse tests performed at the center of

the slab. There is no visible trend from these ratios, as the ratio for the WCPD is near unity,

while the ratios for Building K1 is less than one and for Building K2 is well above 1. The

measured/model comparisons are closest for WCPD, which is expected since the properties of
138

the pulse applied in the model were optimized using the measured data from this building. The

measured/model comparisons are near one for all three categories for the WCPD. These same

comparisons for Building K1 indicate that the model overestimates the response of the model,

while the models for the Faculty Club and Building K2 underestimate the buildings’ responses.

Table 4.5.1: Peak value Measured/Model


ratio comparison for HD test at center of
slab by building
Average Average Average
Building
peak a peak v peak d

WCPD 0.98 1.15 0.94


UW Fac. Club 0.34 0.40 0.21
K1 0.77 0.48 0.32
K2 2.94 2.99 3.17

The measured/model ratios for the average peak and peak RMS response values for the walking

tests performed in each building are displayed in Table 4.5.2. The average peak values for

acceleration, velocity, and displacement are all less than 1 for the three concrete buildings

(WCPD, K1, and K2). Because the walking forcing function applied within the model is a

“worst-case scenario” with the pulse being applied at the center of the slab and the two walkers

walking in stride, the occurrence of the larger highest peak values in the model is consistent.

This is also the case for the peak RMS displacement, as all of the measured peak dRMS values

were in the range of 20-25% of the RMS values from the model. This ratio increased slightly for

most of the vRMS comparisons.

The ratio for aRMS, however, does not follow this trend, as the model underestimates the RMS

acceleration for WCPD, the UW Faculty Club, and Building K2. The larger highest peaks in the

model response, coupled with lower peak RMS values, indicate that the pulse created by each

footstep is dying out before the next step is applied. This opposes the measured responses,
139

which tend to have similar highest peak and peak RMS acceleration values. The similarity

between the highest peak and RMS peak accelerations indicate that the response remains more

constant between each step, as the pulses for each step do not die out before the next step is

applied. The variability from building to building makes it difficult to detect specific trends

within the responses.

Table 4.5.2: Peak value Measured/Model ratio comparison for walking test by building

Average aRMS Average vRMS Average


Building dRMS (in)
peak a (in/sec2) peak v (in/sec) peak d

WCPD 0.79 1.90 0.42 0.70 0.17 0.25

UW Fac. Club 0.31 0.29 0.26 0.26 2.01 0.11

K1 0.46 0.87 0.73 0.49 0.17 0.21

K2 0.75 2.47 0.31 0.62 0.16 0.22

Generally, for both the heel drop tests and walking tests, the modeled and measured accelerations

were the best comparison, as the differences increased for velocity and again for displacement.

This is opposite to what was expected from this analysis. As mentioned in Section 1.2, the

higher modes have the most affect on the acceleration response, and due to the difficulty in

replicating these higher modes in the modeling process, the difficulty to replicate the acceleration

response increases. The velocity and displacement responses, however, are much less affected

by these higher modes, thereby making the likelihood of achieving these responses more

probable. However, this was not the case, as the displacement and velocity differences were

larger than the difference in the acceleration.

It is possible that the higher modes are not represented well within the models, in turn generating

an incorrect response. The results are also strongly dependent upon the loading functions.

Because there are no standard loading functions and the accuracy of the pulse and walking load
140

functions that were used is unknown, it is difficult to determine how the variability of results can

be corrected.

4.6 Comparison Conclusions and Sources of Error


Several conclusions can be made from the comparisons between the measured data and the

responses of the models of these measured buildings:

• The matching of the measured and model frequencies indicates that the structural

properties of the models were adequately represented. However, the variability of the

loading functions made it difficult to match the measured response to loading.

• Peak displacement and velocity values cannot be simulated with much accuracy within

the models for the chosen heel drop and walking loading. Though there were a few cases

of differences of less than 5%, there are several cases where the values differed by an

order of magnitude.

• The magnitudes of the peak accelerations were also difficult to recreate for both the heel

drop tests and walking tests. However, for the heel drop tests, the initial positive and

negative acceleration peaks occurred at similar times as the measured data for most cases,

indicating that the pulse duration for the chosen pulse function was a good estimate of the

actual pulse generated by the heel drop.

• The period between the highest peak accelerations in the walking test records were

consistently in the range of 0.5 to 0.6 seconds for both the measured data and from the

models. This indicates that these overall peak values occur during the impulse of each

step taken by the individuals performing the test. The models’ response for the walking

tests agree with this since the step period for the camel hump function used in the model

was 0.575 seconds, which falls within this range of 0.5 to 0.6 seconds. Therefore, the
141

step frequency of the camel hump function is similar to that of the actual walking forcing

function.

There are also several sources of error that could have contributed to preventing a closer

correlation between the measured and modeled pulse and walking responses:

• Since both the heel drop and walking tests were performed by humans, the pulse

unquestionably varied from test to test. Therefore, matching the response of the models

to the measurements taken for tests with unknown loading functions becomes difficult.

Though the structural properties of the floor will not change from these varying pulses,

the acceleration, velocity, and displacement response values are highly dependent upon

the applied pulse.

• Because the two individuals performing the walking test walk around randomly, without

the intent of walking in step or at the same frequency, the Two Walkers loading function

can only produce a worst case scenario in which the two individuals are walking at the

same point with the same step frequency. This could have accounted for the larger

response values for all of the model walking tests.

• The dual-peaked triangular pulse that was chosen to represent the heel drop forcing

function may be a reasonable representation of the pulse function generated by a heel

drop if the initial impulse from the heel drop was accompanied by a second bounce, in

which the individual’s heel separates from the floor briefly and then hits the floor again.

However, it is unknown whether this is the actual profile of the pulse created by the heel

drop test.
142

CHAPTER 5: SUMMARY AND CONCLUSIONS


5.1 Summary
This goal of this research was to develop recommendations for numerical modeling of floors in

buildings, to enable accurate prediction of vibration response. This was achieved by conducting

vibration studies in three concrete buildings and one steel-framed building in which acceleration

response to human excitation (“heel drop” and walking tests) was recorded. The structural

frequencies that were excited by human loading were determined from the measured acceleration

histories through Fast Fourier Transform (FFT). In two cases, the buildings were under

construction, and it was possible to obtain results for a bare structural floor and another,

nominally identical, floor in which the tenant’s improvements had been partially completed.

Another building had a regular floor plan, simplifying the modeling process. The final structure

was the steel building, which provided a comparison between the responses of different

structural types.

Finite element models of the buildings were then created. The material and geometric properties

of the models were modified in order to match the modal characteristics gathered from the

vibration studies. Because the precise load histories of the tests conducted in the vibration

studies were unknown, estimated load functions were created and applied to the models. The

amplitudes of the responses to these loads were compared with those measured in the buildings.

These comparisons led to recommendations for numerical modeling for successful prediction of

dynamic response.

5.2 Conclusions
The conclusions of this study are separated into three sections according to the subject matter.

The Measured Response section covers the measured vibration response of the buildings and the
143

testing process itself. Modifying the geometric and material properties from the initial models of

the measured buildings was necessary in order to generate fundamental frequencies in the

models that matched the measured fundamental frequencies before the models could be loaded

with the heel drop and walking load functions. The conclusions from the process of matching

the measured and model frequencies are provided in the Modeling for Frequency section.

Modeling for Loading and Prediction of Response provides conclusions from the model loading

process and the comparison of the simulated and measured responses to human loading.

Conclusions on Measured Response

• Effect of Loading Types. Heel drop and walking tests usually led to similar excited

frequencies, with the power spectra from the heel drop test generating more distinct

frequencies in most cases. This agreement indicates that the frequencies in the power

spectrum represent structural frequencies, and not forcing frequencies, because the same

frequencies are excited by two different types of loading. The ambient test, however, did

not always agree with these frequencies. Also, ambient loading often showed frequency

spikes at around 30 Hz and 60 Hz that did not occur in the walking and heel drop records.

These spikes are attributed to fans and other machinery and are not considered to be

structural frequencies.

• Type of Structure. The level of acceleration response varied depending upon the floor

and structural type. The flat-plate buildings experienced the lowest accelerations, with

the stiffer, post-tensioned flat-plate building having slightly lower amplitudes of

vibrations than the reinforced concrete flat plate. The long-span concrete beam-and-slab
144

structure had higher amplitudes of vibration than the flat-plate. As expected, the level of

vibration of these concrete buildings was much less than that of the steel building.

• Non-Structural Elements. Fitted-out floors with full-story partitions could be expected to

produce higher frequencies and lower accelerations than bare floors due to the stiffness

increase caused by the partitions’ connecting the floors above and below. Here, the ratio

of frequencies for the bare to fitted-out stories was found to be as high as 1.5.

Modeling for Frequency

• Use of a 2-D floor model. A 2-D model can be used to predict satisfactorily the response

of a floor, provided that the support conditions are modeled appropriately. The columns

may be represented by nodes that are fixed vertically, but it is essential to provide floor

elements that are very stiff within the perimeter of the actual column. This is particularly

important in flat-plate buildings, leading to a frequency increase of approximately 30% in

most cases. In buildings with beam/slab configurations, beam rigid end offsets may be

used, thereby decreasing the effective length of the beam to simulate the presence of the

column.

• Minimum number of bays. It is possible to model only a portion of the floor and still

generate the correct mode shapes and frequencies. The number of bays required to do

this depends on the way in which the boundary conditions at the edge of the modeled

region are treated. In this research, the use of artificial springs at the boundary was

considered but rejected, because determining their stiffness would be complicated and

because, while they might improve the modeling in some modes, they would make it

worse in others. It was found that including two bays on either side of the loaded bay,
145

and leaving the boundaries rotationally free, led to good results. Use of fewer bays

caused the mode shapes to be significantly different from those found with a more

complete model.

• Contribution of column bending stiffness. In the model, the bending stiffness of each

column may be represented by rotational springs along with the rigid, finite-sized column

plate elements in flat-plate buildings and with beam rigid-end offsets in beam/slab

buildings. Basing the spring stiffness on 70% of the gross column stiffness produced

good results. In modeling the flat-plate buildings, the actual value for the stiffness of the

spring had only a small effect on the floor frequency. However, neglecting to include the

springs altogether results in an underestimate of the stiffness and in turn, an

underestimate for the frequencies of the structure. In the beam-and-slab building,

inclusion of a rotational spring to model the column bending was essential. The need to

include such a spring was increased by the fact that the beam span was much larger than

the column height in the one beam-and-slab building that was tested

• Dynamic material stiffness. The concrete material stiffness must be modified by a factor,

αtot, in order to match the measured frequencies. This factor accounts for both the

increase of the dynamic modulus from the static modulus and the decrease in stiffness

due to concrete cracking. The dynamic component of the factor was found to be

approximately 1.3. This value lies within the range recommended by others.

• Effect of cracking. Comparison between two flat-plate buildings, of which one was post-

tensioned and the other was not, showed that cracking was negligible in the post-

tensioned floor, because no correction factor, other than the dynamic increase in Ec, was

needed to match the measured frequency. In the non-post-tensioned floor, the dynamic
146

Ec had to be multiplied by a factor of approximately 0.60, which was attributed to the

effects of cracking. This is a crude approximation, because it assumes that the cracking

is uniform over the entire slab. Determining the true distribution of cracking would be

desirable, but was not possible within the scope of the project. The 40% drop in average

stiffness due to cracking corresponds to a frequency drop of 23%.

• Use of rigid offsets between beams and slabs to simulate composite action. In cast-in-

place concrete construction, the beams and slabs act compositely. Therefore, modeling

these elements as connected is appropriate. Care should be taken to ensure that the

various elements are modeled at the correct vertical locations to achieve the correct

degree of composite action, but these offsets must be rigid to maintain the composite

behavior.

• Non-Composite Action. In steel construction, the behavior may or may not be composite.

In cases where it is not, care should be taken to dissociate in the model those members

that are non-composite in the field. Failure to do so can result in fundamental frequencies

that differ from the true values by a factor of up to 2.0.

• Damping. Damping was estimated from the rates of decay in the responses to heel drop

impulsive loading. Damping values between 2% and 3% were found in most cases;

however, there were a few cases where higher values were found. Those higher values

should be treated with caution, because the decay was not a clean sinusoidal signal

bounded by an exponential envelope, as might be expected from a SDOF system

subjected to a pulse load. The range of 2% – 3% is in agreement with the findings of

other researchers.
147

• Non-Structural Elements. The fitted-out floors of the buildings were not modeled

because the non-structural elements are difficult to represent and because their

configuration is highly building-specific. Therefore, nothing explicit can be said about

modeling them. However, as described above, they may cause a significant increase in

the system frequencies, which typically could be expected to reduce the vibrations from

human excitation, thus making the models of the bare systems conservative.

Modeling for Loading and Prediction of Response

• Amplitude of Response. The models were unable to reproduce the amplitudes of the

measured response to loading with sufficient consistency to allow their use as a reliable

predictive tool. Because the structural models were able to reproduce structural

frequencies with good accuracy, the difficulties in reproducing the measured response

amplitudes were attributed largely to poor definition of the loading function.

• Duration of Heel Drop Pulse. In the heel drop tests, the measured and predicted times of

the initial positive and negative peaks in the acceleration response correlated well.

Although the magnitudes did not always match up, this finding indicates that the duration

of the model pulse was representative of the duration of the actual heel drop pulse.

• Heel Drop Pulse Representation. The double triangular pulse matched the heel drop

pulse most effectively. However, this conclusion was reached by matching measured and

predicted responses, not loads. The choice was also partly driven by the fact that, in many

cases, the first upward acceleration peak in the response was larger than the first

downward one. Even with the double triangular shape, the match between the measured

and predicted response was only moderately good.


148

• Walking Test Response. The acceleration response of the model to the walking forcing

function remained consistent along the entire acceleration record for the case where the

two walkers walked out of phase with one another. This was similar to the measured

data. However, the magnitudes of the accelerations were slightly larger in the models’

response. The variables of the walking function (shoe type, walker’s weight, step

frequency) could account for this disparity in the magnitudes. Furthermore, in the

interests of simplicity, the dynamic loading in the model was applied at the center of the

slab, whereas in the tests the walkers moved randomly over the bay. In both cases, the

acceleration was detected at mid-span. The model loading was therefore effectively more

intense than the test loading, and that may account for some of the difference between the

results.

5.3 Recommendations
5.3.1 Recommendations for Practice
The aim of this research was to produce a set of modeling recommendations that both predict the

modal characteristics of a floor and accurately determine the floor’s response to human

excitation. In engineering practice, the modeling recommendations discussed in Section 5.2

should be followed in order to accurately determine the mode shapes and frequencies of a

particular floor. The use of these techniques will generally lead to models that can generate

accurate modal characteristics of a floor in a building from a 2-D floor model. In particular, they

may allow an area smaller than a complete floor to be modeled, thereby saving engineering

effort. Neglecting to apply these principles in the modeling process may lead to incorrect modal

properties, which in turn will adversely affect the predicted responses of the floor to human

excitation.
149

5.3.1 Recommendations for Further Research


Though these recommendations can be used to determine the modal information for a particular

floor, the response analysis that was conducted indicated the modeling techniques and chosen

loading functions were not adequate to correctly reproduce the actual response of the floor. To

improve the model response, additional research must be conducted. Some of these areas of

research are included below:

• Modeling Prior to Measurements. Analyzing the building prior to a site visit for

measuring the response would allow the load locations to be chosen to coincide with the

antinodes of specific pre-selected modes. This would allow more critical comparisons to

be made.

• Standardized Loading Functions. Realistic forcing functions associated with the heel

drop pulse and walking excitation must be determined. They can then be applied to the

floor models and could produce responses which are more representative of the actual

floors’ response. Previous work on the subject suggests that this might be difficult,

because step loads vary widely with individual walkers, shoe type, floor properties, etc.

However, if the loading functions could contain these features as variables, the response

to families of different loading patterns could be determined.

• Standardized Loading. Replacing the heel drop test with one in which a standard weight

is dropped from a specific height would generate the same impulse for every test and

would help to provide more reliable damping and modal information. Introducing a

metronome into the walking tests would force the pulse from each step to be at the same

set interval, creating another forcing function that would be the same for every test. The

implementation of more automated testing would allow for more consistent loading from
150

test to test, thereby improving the comparisons between the measured and model

response.

• Proximity to Loading. The effects of the proximity to the loading should be explored.

The response of the floors and bays adjacent to the loaded area should be monitored to

determine the extent of the affected areas.

• Additional Testing. The use of an array of accelerometers, rather than a single one,

would allow the mode shapes, as well as the frequencies, to be identified, creating one

less unknown variable for the analysis. This would allow the mode shapes from the

model to be compared with those from the actual building. It would increase the

complexity of the data gathering and analysis process, but that cost may be necessary to

improve the present modeling guidelines.


151

APPENDICES
Appendix A: Methods of Determining Damping of Structures
The inability to measure clean acceleration responses for most of the measured buildings created

difficulty in determining the amount of damping that was present. To simplify the analysis, the

flooring systems of the buildings in this study were assumed to be SDOF systems, when in fact

they are Multi-Degree of Freedom systems. The decay of motion for a damped SDOF system

follows the decay as shown in Figure A1 (a), with the natural period of damped vibration, TD, as

the interval between successive peaks. Alternatively, Figure A1 (b) shows the measured

acceleration record of a heel drop test conducted in the center of a slab in the West Campus

Parking Deck. (Note that the time axis measures t/Td, not t). The intervals between peaks for the

measured data in this case range from 0.00267 seconds to 0.112 seconds. These intervals are

much less than the natural period of the building, which, determined from fundamental

frequency in the Power Spectrum generated from this acceleration response, was found to be

approximately 0.138 seconds. Therefore, multiple peaks are present within each period interval.

The disagreement in the period of the response implies the contribution of multiple modes to the

response of the actual building.

Figure A1: (a) decay of motion for a damped SDOF system; (b) Decay of motion for acceleration
record in West Campus Parking Deck heel drop test
152

Figure A2: Power Spectra for heel drop tests performed at (a) the center
of the slab and (b) at the beam mid-span in West Campus Parking Deck

The power spectra of this heel drop test and the heel drop test performed at the midspan of the

beam adjacent to this slab confirm this multiple modal contribution. The spectrum for the test in

the middle of the slab (Figure A2a) indicates that the predominant mode has a frequency of 19

Hz, while the fundamental frequency’s contribution to the overall response is approximately half

that of the 19 Hz mode. Figure A2b provides a different story for the modal contributions to the

beam’s response. The fundamental frequency is the primary contributor to the response in this

case, while frequencies of 11 Hz and 15 Hz also make significant contributions. The power

spectra not only indicate that multiple modes are being excited, but that different modes become

predominant depending upon the location of the loading. Since the simplistic SDOF model has

been chosen to analyze these systems and because previous studies have indicated that the

fundamental mode accounts for the majority of the response, the fundamental frequency was

chosen to characterize the mode of the assumed SDOF systems. Since this analysis focuses on

excitation due to walking, which has a frequency of about 2Hz, this also makes sense from a

resonance standpoint. Given that this excitation frequency is low, the fundamental frequency, as
153

the lowest frequency of the system, is the mode most susceptible to resonance. Therefore, this is

the mode and corresponding period should be the basis of the analysis.

The acceleration record is not purely sinusoidal, again making it evident that multiple modes are

participating in the floor’s response. Therefore, finding the damping from the decay of the

acceleration response would not give a good estimate of the damping in the fundamental

frequency. To reduce the effect of the higher modes, the displacement response curve, which is

simply the second integral of the acceleration curve function, was selected to be used to

determine the damping. The displacement response, shown in Figure A3, is a much cleaner

response, indicating a response more closely related to a SDOF system. Furthermore, most of

the peaks are self-evident, and are spaced at intervals close to 1/7.23 seconds, i.e. the fundamental

frequency.

0.0015

0.001
Displacement (in)

0.0005

-0.0005

-0.001

-0.0015
0 0.5 1 1.5 2
Time (sec)

Figure A3: Displacement record for heel drop tests performed at the center of the slab

For SDOF systems, the relationship between two successive peaks in an acceleration record is

𝑑𝑑(𝑡𝑡) 𝑑𝑑
𝑑𝑑(𝑡𝑡+𝑇𝑇𝐷𝐷
= 𝑑𝑑 = 𝑒𝑒 2𝜋𝜋𝜋𝜋 (𝐸𝐸𝐸𝐸. 𝐴𝐴1)
) 𝑖𝑖+1
154

where ζ is the damping ratio, di is the initial peak response, and di+1 is the peak following the

initial peak. Given the response of a SDOF system, the amount of damping in the system can be

determined from this relationship. This relationship cannot be applied to the West Campus

Parking Deck displacement response data, however, because the relationship for a SDOF system

is based on the fact that successive peaks are separated by an interval of TD, and is independent

of the time, t, at which the peaks occur. As mentioned previously, this is not the case for the

West Campus Parking Deck or any other MDOF system. In order to maintain the SDOF

assumption for this building which will enable this relationship to be applied to find the damping

of the system, an effective period between successive peaks must be applied to the measured

acceleration response.

Two different methods which account for the fact that successive peaks do not occur at intervals

of TD were chosen to scale the relationship in Eq. A1 to fit the displacement response and

provide an upper and lower bound for the system damping. The results of the application of the

two methods, referred to as the Envelope Method and the Least Squares Method, to the

displacement record s of each building are listed in Table A1.

Table A1: Damping Values for Tested Buildings


Damping, ζ (%)
Building Floor
Envelope LSM min
West Campus Parking Deck 3 3 4.3 3
UW Faculty Club 2 2.5 4.33 2.5
K1 18 2.0 4.5 2
34 5 6.1 5
K2 10 4.5 6.1 4.5
24 7 8.2 7

While the SDOF relationship between successive peaks is independent of the time at each peak,
155

the first method generates an displacement curve that is time-dependent. This envelope method

uses the following equation to determine the peak at each time, t, along the response curve,

creating an upper bound envelope for the curve:

𝑡𝑡
−2𝜋𝜋𝜋𝜋 � �
𝑑𝑑(𝑡𝑡) = 𝑑𝑑𝑜𝑜 𝑒𝑒 𝑇𝑇 𝐷𝐷 (𝐸𝐸𝐸𝐸. 𝐴𝐴2)

where do is the initial peak displacement of the measured response. The damping in this

equation is then manually manipulated until it meets any of the peak values from the measured

displacement. As seen in Table A1, this envelope method generates 3% damping.

0.0012

0.001
Peak Displacments (in)

0.0008

0.0006
Measured Peaks
0.0004 Envelope

0.0002

0
0 1 2 3
Time (sec)

Figure A4: Envelope method for damping determination in West Campus Parking Deck heel drop test

The function utilized in the second method is also time-dependent. However, instead of

depending upon the time at each peak to calculate the response, this method uses the ratio

between the period between each successive peak and the damped period of the structure, TD, to

provide an alternate method of determining the damping. The function to calculate the

displacement at each time step is

∆𝑡𝑡
−2𝜋𝜋𝜋𝜋 � �
𝑑𝑑𝐿𝐿𝐿𝐿𝐿𝐿,𝑖𝑖 = 𝑑𝑑𝐿𝐿𝐿𝐿𝐿𝐿,(𝑖𝑖−1) 𝑒𝑒 𝑇𝑇 𝐷𝐷 (𝐸𝐸𝐸𝐸. 𝐴𝐴3)
156

where dLSM,i is the peak displacement at a given time step, dLSM,(i-1) is the previous peak

displacement, and ∆t is the period between the two successive peaks. Multiplying the damping

by the factor ∆t/TD basically adjusts the period for each time step, creating a separate effective

period for each interval between peaks. The variance between the measured displacement and

the displacement using this method is then tabulated using the least squares equation (Eq. A4) for

each time step, and the sum of this error is recorded.

𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐸𝐿𝐿𝐿𝐿𝐿𝐿 = �(𝑑𝑑𝑖𝑖 − 𝑑𝑑𝐿𝐿𝐿𝐿𝐿𝐿,𝑖𝑖 )2 (𝐸𝐸𝐸𝐸. 𝐴𝐴4)

where di is the measured peak displacement at the given time step.

Using the Solver feature within Microsoft Excel enables the optimal amount of damping to be

determined to minimize the sum of the calculated error. While the envelope method depends

upon the peak initial displacement chosen, the displacement at each time step in the least squares

method is a function of the displacement value determined in the previous time step.

Additionally, the initial displacement is introduced as a variable within the analysis to produce a

curve that results in the smallest possible error between the calculated curve and the measured

record. Making this value a variable as opposed to simply setting it equal to the initial peak from

the displacement record creates a method that provides a curve which follows the displacement

record curve more completely, rather than at only meeting the curve at the initial peak

displacement and one other value. Thus, the damping and initial peak displacement were varied

in order to optimize the error.

Applying this method to the West Campus Parking Deck data produced the curve shown in

Figure A5 along with the peak values and the two envelope methods. The initial peak
157

acceleration determined from the solver analysis was 14.98 in/s2 and the sum of the error was

93.48, resulting in 4.5% damping.

0.0012

0.001
Acceleration (in/s2)

0.0008

0.0006 Measured Peaks


Envelope
0.0004
LSM
0.0002

0
0 1 2 3
Time (sec)

Figure A4: Least Squares Method for damping determination in West Campus
Parking Deck heel drop test

The final step of this process was to determine which damping to set for the system. Using these

two methods provided an objective approach to determining how much damping is actually in

the system. Overestimating the amount of damping can lead to an underestimate in a system’s

dynamic response. Selecting the lowest damping value from these two methods prevents this

overestimation from occurring. Therefore, the smallest damping value, which is 3% from the

Envelope Method, was set as the damping of the West Coast Parking Deck.
158

Appendix B: Exterior Beam Effective Width Factor


For the models of the buildings that had a beam and slab configuration, modifications were

necessary for the exterior beams of the modeled section of the building. These exterior beams

must behave like interior beams because in the actual building, they are not exterior beams, but

part of a continuous system. Therefore, they should not only support their own self-weight and

share the weight of the slab attached to it and its neighboring interior beam, but must also carry

the weight of the slab that is not modeled, but is located outside of it in the actual building. To

account for the presence of additional slabs outside of these beams in a continuous system,

weight and mass factors were applied to these outer beams. However, from the analysis manuals

provided by SAP, it was unclear how the effective flange width of T-beams is determined for

analysis. An analysis was run on the three bay model of West Campus Parking Deck to

determine the effective flange width used for the analysis. Table B1 shows the results of this

analysis.

Table B1: Output for 3-bay model of West Campus Parking Deck
Exterior Beam Beam Mid-span Displacement
Cardinal (in) Fundamental
Mass Factor
point Frequency (Hz)
Approach Exterior Interior
Top Center None 0.6007 0.8666 3.605
Top Center 1 0.919 0.8863 3.421
Top Center 2 0.703 0.8729 3.566
Centroid None 0.6976 0.7976 3.747
Centroid 1 1.03 0.8322 3.435
Centroid 2 0.8044 0.8087 3.672

Two different factors were applied to the outer beams in the model. The first factor, labeled

Exterior Beam Mass Factor #1 in Table 1, included the self-weight of the beam and accounted

for half of the entire weight of one bay of slab. Using this approach is slightly unrealistic

because the beam, which is modeled as a rectangular section, must carry the additional weight of

half of the adjacent slab; however, as a result of the slab not actually being placed in the model,
159

the beam does not gain stiffness from the slab. Thus, the outer rectangular behaving beams do

not have the stiffness of the interior T-beams, but still must support the same weight.

The second weight and mass factor, Exterior Beam Mass Factor #2, was employed to remedy

this problem by maintaining the same ratio of mass to stiffness for the outer beams as is present

for the interior beams. The second factor also includes the self-weight of a portion of the

adjacent slab, but only includes half of the effective width of the T-beam as specified by ACI

8.10.2. In ACI, the effective width is taken as the minimum of the following:
𝐿𝐿
(𝐴𝐴𝐴𝐴𝐴𝐴 8.10.2)
4

𝑏𝑏𝑤𝑤 + 16𝑡𝑡𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 (𝐴𝐴𝐴𝐴𝐴𝐴 8.10.2𝑎𝑎)

𝐶𝐶𝐶𝐶 𝑡𝑡𝑡𝑡 𝐶𝐶𝐶𝐶 (𝐴𝐴𝐴𝐴𝐴𝐴 8.10.2𝑏𝑏)

where L is the beam span length, bw is the web thickness, tslab is the slab thickness, and CL to CL

is the distance between the centerlines of two consecutive beams. The results of the static

analysis using this factor are encouraging, as the error between the static deflections of each of

the beams, interior and exterior, is less than 1%. Therefore, the Exterior Beam Mass Factor that

was applied to each exterior beam was the factor which includes half of the effective width of the

T-beam.

These approaches address the mass that is absent from the model at the artificial boundary.

However, they do not fully address the stiffness there. In particular, the transverse bending

stiffness of the missing slab is excluded and, if only half the beam is included, its torsional

stiffness will not be correct. Fixing the edge against rotation about the longitudinal axis was

considered, but, while this is representative of modes in which the transverse slope is indeed
160

zero, at the edge beam, it is clearly wrong for others. Other options include: including more

bays in the model, developing a special “boundary” element that replicates the effects of the

missing slab bay, or doing nothing. Clearly none of these options are perfect.
161

Appendix C: Column Stiffness Determination


Each of the building models is a two-dimensional floor model; therefore, no columns are present.

Since the floor models are 2D representation of part of a 3D building, the effects of these

columns and the rest of the building’s frame must be accounted for. The axial stiffness of the

columns effectively restrain the beam-column connection joints from deflection in the z-

direction. Although the column also provides stiffness at the supports against deflection in the x-

and y-directions, these joints are fully fixed against such translations in the models.

Additionally, the presence of the columns adds rotational stiffness at these joints. The rigidity of

the column against these translations and rotations must be determined in order to sufficiently

satisfy the geometric constraints of each of the buildings’ flooring systems. The analysis

involved in this process and its implementation to the West Campus Parking Deck model are

outline here.

C.1 Column Bending Models

It would be inefficient to model the entire structure to determine the effect that the columns have

on the beam-column connections. Thus, an alternative approach was adopted in which only a

short section of column was connected to either side of the beam at the supports. These columns

will provide the translational and rotational stiffness similarly to the actual building. The major

issue that needed to be addressed was what the length of the columns in the model should be and

what the ‘free’ end conditions for these columns should be. This issue was addressed by

implementing a point of inflection analysis, which is usually applied to beams, to determine the

translational and rotational stiffness at the location of the column free end. Figure C1.1 displays

the analysis approach used, with the column oriented as if it were a continuous beam, which is
162

the most common application of this analysis. The locations of the rollers in the figure are where

the beams connect to the column.

z M

y -αM

αθ θ
-αM M

θ3 θ4
M -αM
θ2 θ1

Figure C1.1: Column Point of Inflection Analysis

Using this type of analysis, a moment of some magnitude M would be applied to the beam-

column joint at the floor above the beam-column joint for which the stiffness values are desired.

Applying this end moment to the “continuous beam” will create an effective moment at the next

support, which is a factor, α, of the end moment, M. Since the curvature of the column will

change along the length of this segment, there is a point of inflection where the direction of

curvature changes, resulting in a moment of zero at this point. Applying the following equations

for the angles due to these rotations enable the location of the point of inflection to be

determined.

𝜃𝜃 = 𝜃𝜃1 + 𝜃𝜃4 𝑎𝑎𝑎𝑎𝑎𝑎 α𝜃𝜃 = 𝜃𝜃2 + 𝜃𝜃3 (Eq. C1.1)

𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀 −α𝑀𝑀𝑀𝑀 −α𝑀𝑀𝑀𝑀


where 𝜃𝜃1 = 3𝐸𝐸𝐸𝐸 , 𝜃𝜃2 = 6𝐸𝐸𝐸𝐸 , 𝜃𝜃3 = 3𝐸𝐸𝐸𝐸
, 𝑎𝑎𝑎𝑎𝑎𝑎 𝜃𝜃4 = 6𝐸𝐸𝐸𝐸
.
163

Figure C1.2: Point of Inflection Analysis. a) Moment Diagram;


b) Dissected Section

Solving this set of equations for α and using the geometry shown in Figure C1.2a, the value of

βL, which is the distance from the beam-column joint to the point of inflection, can be

determined. The values determined from this method are as follows:

α = 0.2679, β = 0.211
𝑀𝑀𝑀𝑀 𝑀𝑀𝑀𝑀
𝜃𝜃 = 0.2887 = 3.464𝐸𝐸𝐸𝐸 (Eq. C1.2)
𝐸𝐸𝐸𝐸

(Eq. C1.1)Dissecting the column at the point of inflection, as shown in Figure C1.2b, is the next

step in determining the translational and rotational stiffness at this location. At this point, there

is an internal shear force, labeled P in the figure, and a negligible axial load. However, since it is

the point of inflection, there is no internal moment present at this location. The deflection at this

point has two components which are displayed in Figure C1.3: the deflection due to the moment

created at the support and the deflection due to the cantilever that is now present at the point of

inflection.
164

Figure C1.3: Point of Inflection Analysis Deflection Contributions

Solving for the deflection,

𝑃𝑃𝐿𝐿3 𝑃𝑃𝐿𝐿3
∆ = 0.01598 𝐸𝐸𝐸𝐸
= 62.56𝐸𝐸𝐸𝐸 (Eq. C1.3)

The translational stiffness for the x- and y-directions, determined from the elastic stiffness

relationship, k = P/∆, is solved next:

𝑃𝑃 𝑃𝑃 62.56𝐸𝐸𝐸𝐸
𝑘𝑘 = ∆ = 𝑃𝑃 𝐿𝐿 3
= (Eq. C1.4)
�0.01598 � 𝐿𝐿 3
𝐸𝐸𝐸𝐸

Similarly, the rotational stiffness about the x- and y-axes can be solved:
𝑀𝑀 𝑃𝑃α𝐿𝐿×3.46𝐸𝐸𝐸𝐸 3.464𝐸𝐸𝐸𝐸
𝑘𝑘𝜃𝜃 = = = (Eq. C1.5)
𝜃𝜃 𝑃𝑃α𝐿𝐿2 𝐿𝐿

The translational and rotational stiffnesses are both a function of the modulus of elasticity of the

column, its moment of inertia, and the length of column between levels. Since this is a parking

deck, there are half levels between each main level. Therefore, the effective column lengths of

the interior columns, shown in Figure C1.4a, are actually half the height between each main

level. This short, stiff column contributes significant stiffness to the interior beam-column
165

connection because the stiffnesses are inversely related to the column length. The half levels do

not affect the exterior columns, however, so the exterior beam-column joints do not have this

additional contribution from the shorter columns. The stiffness values at the points of inflection

for the interior and exterior columns and the input values to determine these stiffnesses are

displayed in Table C1.1.

Lvl 3 24’’

4.96’
24’’

Lvl 21/2
a) b)

Figure C1.4: (a) Column height between levels for interior beam-column joints; (b) Column cross-section

Table C1.1: Column point of inflection stiffness values for West Campus Parking Deck
Column Properties
Column Translation Stiffness, Rotation Stiffness,
Location Effective Column Moment of Inertia k (k/in) kθ (k-in)
Height (ft) (in4)
Interior 4.96 27,648 24,159 8,045,419
Exterior 9.92 27,648 3,020 4,022,710

The rotational stiffness for a fixed-fixed member is 4EI/L, compared to 3EI/L for a pinned-end

member. This indicates that this column, which has a stiffness nearly splitting the values for

these two cases, behaves semi-rigidly. This behavior can be exemplified further when

comparing the rotational stiffness of the interior column to the flexural stiffness of the beam,

EIb/Lb:
166

27648𝑖𝑖𝑖𝑖4
3.46 � �
𝑘𝑘𝜃𝜃𝜃𝜃 3.46𝐸𝐸𝐼𝐼𝑐𝑐 /𝐿𝐿𝑐𝑐 4.96𝑓𝑓𝑓𝑓 × 12 𝑖𝑖𝑖𝑖�𝑓𝑓𝑓𝑓
= = = 35.4
𝑘𝑘𝑓𝑓𝑓𝑓 𝐸𝐸𝐼𝐼𝑏𝑏 /𝐿𝐿𝑏𝑏 31098.67𝑖𝑖𝑖𝑖4
� 684𝑖𝑖𝑖𝑖 �

Doubling this ratio to account for the rotational stiffness contributions from the columns above

and below this connection creates a ratio of 70.8 and a rotational stiffness of 16,090,838.72 kip-

in. According to Finite Element Modeling Continuous Improvement (FEMCI) group at NASA,

𝑘𝑘
when 𝑘𝑘 𝜃𝜃𝜃𝜃 <1.0, the rotational stiffness of the column is less than that of the beam and therefore,
𝑓𝑓𝑓𝑓

𝑘𝑘 𝜃𝜃𝜃𝜃
the beam-column connection exemplifies pinned behavior. ratios greater than 100 describe
𝑘𝑘 𝑓𝑓𝑓𝑓

fully fixed behavior. Therefore, the beam-column connection in the West Campus Parking

𝑘𝑘 𝜃𝜃𝜃𝜃
Deck, with a ratio nearing 100, behaves like a nearly fixed connection.
𝑘𝑘 𝑓𝑓𝑓𝑓

The model can now be modified with a column with the length, βL = 1.047 ft, connected to both

the top and bottom of the supports with rotational and translational springs with the stiffnesses

determined above. Table C1.2 lists the various models run and displays the resulting

fundamental frequency. These frequencies are compared to the fundament frequency of 7.23 Hz,

which was determined from the measured data. For these models, various support options for

the columns and beams were explored. Three support types for the beams were used. The first,

labeled Type 1 in Table 3, is a simply-supported beam with ux, uy, and uz restrained at one

support and ux and uz at the opposite support. Type 2 consisted of ux supports at both beam

ends, while both ux and rx are fixed at both beam ends for Type 3. Five column end support

options were used. Type 1 has x and y translational springs, while leaving translation in the z-

direction free. Type 2 is the same as Type 1; however, uz is now restrained. Type 3 adds fixed
167

rotations to Type 1, while Type 4 adds fixed rotations to Type 2. Finally, Type 5 replaces the

fixed rotations in Type 4 with rotational springs.

Table C1.2: Frequency response of West Campus Parking Deck SAP models with columns

C.2 Frame Model

The boundary conditions for this model can become more efficient while maintaining the

stiffness created by the beam-column connection. This can be achieved by transferring the

rotation springs from the column ends directly to the beam-column joint. The stiffness at this

location will not be identical as the column point of inflection, so the frame SAP model shown in

Figure C2.1 was created to determine the stiffness at this location. This frame model is a 2-D

representation of the columns and beams in the parking deck. A moment of varying magnitudes

was applied at the beam-column joint on the half story between the third and fourth floor. The

resulting rotation at this location indicates the rotational stiffness provided at this location by the

column and the entire frame.


168

Figure C2.1: Frame model of West Campus Parking Deck

Rigid end offsets for both the columns and the beams must be set to ensure that the beam and

column do not overlap one another at the joint locations and to maintain the rigidity of the joint.

In SAP, this is performed by setting the Rigid Zone factor to 1 when creating the end offsets.

The various tests performed on this frame model are reviewed in Table C2.1. The variables in

these tests were the end offset information, the magnitude of the moment applied, and the beam-

column joint boundary conditions. The rotational stiffness is not the same for the interior and

exterior joints since the interior joints are connected to shorter, stiffer columns. Therefore, the

moment was applied separately at both the interior beam-column joint (labeled point 1 in Figure

C2.1) and the exterior beam-column joint (point 2) for the final two tests.

Table C2.1: Test Summary for West Campus Parking Deck frame models
Beam Column Angle of Rotation, θ kθ (k-in/rad)
Moment
Test End Offset End Offset
RZF RZF (k-in) Interior Exterior Interior Exterior
(in) (in)
1 - - - - 100 0.000018 - 5,555,556 -
2 12 1 - - 100 0.000017 - 5,882,353 -
3 - - 14 1 100 0.000012 - 8,333,333 -
4a 12 1 14 1 100 0.000012 - 8,333,333 -
4b 1000000 0.117908 - 8,481,189 -
5f 12 1 14 1 1000000 0.06738 0.07243 14,841,199 13,806,434
6 f* 12 1 14 1 1000000 0.07202 0.07784 13,885,032 12,846,865
fFixed horizontal translation for exterior beam-column joints
* Ibeam set to 0.01in4
169

The final test is the most realistic representation of the actual structure. In this test, the beam and

column offsets were both set with a rigid zone factor (RZF) of 1. Additionally, the exterior

beam-column joints at each half floor were fixed against translation in the y-direction because

horizontal translation of the frame is not expected to be significant for walking loads and the

modes that are anticipated to be excited by this loading. Finally, the moment of inertia for the

beam at Level 3.5 was set to be very small, thereby removing the beam’s contribution to the

stiffness rotation at the beam-column joint. This step is essential because the springs that will be

placed at the supports in the floor model must only account for the stiffness contributed by the

columns and frame, but must not include the contribution of the beam on this level since it is

already present in the floor model.

C.3 SAP Model with Rotation Springs at Supports

Using the rotational stiffnesses determined in the previous sections, the final geometric

component of the modes can be put into place. The floor model was analyzed using rotational

springs at each beams’ support, and the modal output for various stiffness values for these

springs is shown in Table C3.1. The final test uses the interior and exterior spring stiffness

values determined in Test 6 of the frame model analysis. Therefore, the geometric constraints of

this model most closely coincide with those of the actual parking deck.
170

Table C3.1: Frequency response of West Campus Parking Deck SAP models
with rotational springs at both supports

C.4 Column Axial Stiffness

The assumption that the columns provide essentially rigid support against vertical translation can

be investigated approximately. This can be done by treating one story of the column as a spring
𝐴𝐴𝐴𝐴
of stiffness = 𝐿𝐿
, where L is the story height. Let the weight of one floor be w. The natural

frequency of the floor on the spring is

1 𝑘𝑘 1 𝐴𝐴𝐴𝐴 × 𝑔𝑔 1 𝐸𝐸 × 𝑔𝑔
𝑓𝑓𝑛𝑛 = � = � = � (Eq. C4.1)
2𝜋𝜋 𝑤𝑤 2𝜋𝜋 𝐿𝐿 ×𝐴𝐴×𝑓𝑓 2𝜋𝜋 𝐿𝐿 ×𝑓𝑓
𝑐𝑐 𝑐𝑐

where fc is the stress in the column due to the weight, w, of one floor. For typical values of fc =

0.200 ksi and Ec = 5300 ksi, fn = 45 Hz. It is unlikely that any mode would consist of column

shortening along, but the fact that this approximate “column” frequency is so much higher than

the typical “floor” frequencies of 5-15 Hz suggests that the rigid support model is adequate.
171

Appendix D: Cardinal Point Analysis


When area elements, which are used to model slabs, and beam elements are connected, the center

of the area element runs through the cardinal point of the beam element. The beam cardinal

point is the point at which the gravity center of the beam is taken and is also the point on the

cross-section at which supports and other elements are attached. In SAP 2000, any of the points

shown in Figure D1 can be selected as the beam cardinal point, but the default is the centroid of

the beam. Therefore, the slab is placed in the center of the beam by default.

Figure D1: Possible cardinal point locations

To model the beam and slab correctly, the slab must be offset from the cardinal point so that it

rests on top of the beam. This can be done one of three ways. In the first method, the default

cardinal point is selected and the slab offset distance is half of the height of the beam plus half of

the slab thickness. The second method uses the top center as the cardinal point, making the slab

offset equal to half of its thickness. The cardinal point is at the bottom center for the third

method, and the slab offset is adjusted accordingly. Though each of these creates the same effect

of the slab on top of the beam, the response of each method varies. The reasons are unknown,
172

and the program support staff was unable to offer a timely explanation. Thus, a short parameter

study was conducted to investigate the consequences of the different choices.

D.1 Single Beam and Slab Strip Analysis

In order to determine which method produces the optimal results, the static and dynamic

response of a model was analyzed. Before performing this analysis on an entire floor model, an

analysis was initially performed at the most basic level: a single W21x62 beam with an 18 inch

wide strip of slab on top of it shown in Figure D1.1. This simply supported beam was 352

inches long. It is the same beam that exists in the UW Faculty Club, but the slab connectivity is

different.

Figure D1.1: W21x62 beam with concrete slab

The density of the 18 inch slab strip was multiplied by 12, effectively making the beam support

the weight of an 18 ft wide slab. Also, the modulus of elasticity of the concrete was multiplied

by 12, as well, to adjust the stiffness of the strip.


173

This beam and slab strip was modeled using the three cardinal point methods outlined

previously. Each model is shown in Figure D1.2. A static and modal analysis was performed on

each model using 16 elements longitudinally and shear deformations switched off. The results of

this analysis are shown in Table D1.1.

Figure D1.2: Three methods of modeling slab on top of beam; Cardinal point location is
(a) centroid, (b) top center, and (c) bottom center.
174

Table D1.1: Beam and slab strip model static and dynamic analysis output
Closed-Form Solutions
Effective Moment of
Mid-Span Deflection Fundamental Frequency
Inertia
Ie = 3918.24 in4 ∆max = 0.1410 in ω1 = 58.98 rad/sec
f1 = 9.39 Hz

Static Output
Beam
Slab Midspan
Cardinal Static EIeff Static Ieff ∆model/∆actual
Offset (in) Disp (in)
Point
Top Center 2 0.1398 114,629,418 3,953 0.9912
Centroid 12.5 0.1398 114,629,418 3,953 0.9912
Bottom
23 0.1398 114,629,418 3,953 0.9912
Center
Dynamic Output
Beam
Slab Dynamic Dynamic Dynamic Error Static/Dynamic
Cardinal f1 (Hz) ω1 (rad/sec)
Offset (in) EIeff Ieff (%) Error (%)
Point
Top Center 2 9.406 59.10 114,078,368 3,934 -0.396 0.481
Centroid 12.5 9.276 58.28 110,946,813 3,826 2.360 3.213
Bottom
23 8.92 56.05 102,594,259 3,538 9.711 10.499
Center

Because the width of strip of slab is so narrow, the beam-slab model can be assumed to behave

like a simply supported beam. Therefore, the closed form solution for both the mid-span

deflection and the fundamental frequency can be determined. These values are both functions of

the effective moment of inertia of the composite section.

From the deflections under self-weight, the moment of inertia can be determined. This moment

of inertia is compared to the calculated moment of inertia for each model and with the values

from the other tests. Table D1.1 indicates that there is no difference in the static response for the

different models. The error between these values and the true moment of inertia is 1%.

For the modal analysis, the fundamental frequency from each model can be used to calculate the

dynamic moment of inertia. These values should be consistent with the static moments of
175

inertia. While the static moments of inertia were constant for each model, the dynamic moments

of inertia were not. The difference in the moment of from the static value inertia for the top

center model is approximately 0.5%, while the centroid model is slightly larger, and the bottom

center model is larger than 10%. Compared to the closed form solution, the errors drop slightly

for each model, but the top center model maintains the smallest error. Because the error for the

bottom center model was so large in comparison with the two other cardinal point options, this

option was left out of consideration for the final modeling method. The error for the remaining

methods was low; therefore, further comparison of these two methods was conducted.

D.2 Three Bay System

Building up from the most basic beam and slab strip model, the two remaining methods were

implemented in a beam and slab model with multiple bays. This three bay model was the initial

model of the West Campus Parking Deck shown in Figure D2.1. Applying the Exterior Beam

Mass Factors 1 and 2, which are described in Appendix B, the floor was modeled using both the

top center and centroid cardinal point methods. The static and modal information from these

models was recorded and is shown in Table D2.1.

Figure D2.1: Three bay model of West Campus Parking Deck


176

Table D2.1: Output for 3-bay model of West Campus Parking Deck
Beam Mid-span Displacement
Cardinal Exterior Beam (in) Fundamental
dext/dint
point Mass Factor Frequency (Hz)
Exterior Interior
Top Center None 0.6007 0.8666 0.693 3.605
Top Center 1 0.919 0.8863 1.037 3.421
Top Center 2 0.703 0.8729 0.805 3.566
Centroid None 0.6976 0.7976 0.875 3.747
Centroid 1 1.03 0.8322 1.238 3.435
Centroid 2 0.8044 0.8087 0.995 3.672

The dext/dint ratio is a gage of how near the displacements of the exterior beams are to that of the

interior beams. This ratio is the nearest to one for the final test with a centroid cardinal point,

indicating the exterior and interior beams have nearly identical displacements. This model is

performing most like a continuous system.

D.3 Cardinal Point Conclusions

The variance between the two methods’ fundamental frequencies was 0.13 Hz for the beam and

slab strip model and 0.11 Hz for the three bay model. The bandwidth of the Power Spectra

Density generated in RT Pro from the measured acceleration data is 0.09 Hz. The measured

fundamental frequency may vary up to two bandwidths, meaning that the error may be up to 0.18

Hz. Because the frequency error between the two cardinal point methods is within the error of

the measuring equipment, either cardinal point method can be used effectively to model the

measured buildings.

The results of models using the top center and centroid beam cardinal point methods are very

similar. From the basic beam and slab strip model results, the top center method had slightly

smaller error, while the behavior of the centroid method was better in a model of a floor with

entire bays of slab instead of just a strip. Because this method will be implemented in models of
177

entire floor systems, the centroid method was chosen to model the buildings with beam-slab

floors.
178

Appendix E: Heel Drop Load Simulation


The loading function associated with the heel drop tests consists of a pulse with a finite duration.

The vibration measurements taken in each building do not provide information about the

duration of the pulse or the magnitude of the force at any point in time during the duration of the

pulse. Therefore, the parameters of the heel drop loading function had to be created and applied

to the model of each building. The response of this model was then compared to the measured

acceleration response, and the pulse was modified until these responses matched. Measured data

was necessary for this exercise, and the West Campus Parking Deck was used for this purpose.

Compared to the other options, a triangular pulse seemed like the most practical pulse to

represent the pulse created by the heel drop. A rectangular pulse was unreasonable because the

force of the heel drop is not constant during the duration of the load. Additionally, modifying

the properties of the triangular pulse function is a much easier than for a sinusoidal pulse, which

made the triangular pulse the optimal choice. As mentioned in Section 4.2.1 Heel Drop Load

Simulation, for pulse durations less than Tn/4, the pulse shape is not important to the response of

the structure as long as the impulse is constant. Therefore, if the pulse duration proved to be in

this range, the choice to use a triangular pulse would be sufficient. If the pulse duration

exceeded Tn/4, then other pulse shapes would have to be explored.

The impulse of the heel drop is closely related to the heel drop forcing function. Since velocity

is the derivative of acceleration, Newton’s 2nd Law of Motion, which states force equals mass

times acceleration, can be rewritten in terms of the velocity. Integrating these terms displays that

the impulse of the load (mass times velocity) is equal to the integral of the forcing function.

These steps are displayed below.


179

𝑑𝑑
𝐹𝐹 = 𝑚𝑚 × 𝑎𝑎 = 𝑑𝑑𝑑𝑑 (𝑚𝑚 × 𝑣𝑣) (𝐸𝐸𝐸𝐸. 𝐸𝐸1)

𝑚𝑚 × 𝑣𝑣 = ∫ 𝐹𝐹𝐹𝐹𝐹𝐹 (𝐸𝐸𝐸𝐸. 𝐸𝐸2)

Figure E1: (a) SDOF system; (b) Triangular pulse forcing


function [Chopra]

As seen from Figure E1, the integral of the triangular pulse function is simply,
1
∫ 𝐹𝐹𝐹𝐹𝐹𝐹 = 2 𝑝𝑝𝑜𝑜 × 𝑡𝑡𝑑𝑑 = 𝑚𝑚 × 𝑣𝑣 (𝐸𝐸𝐸𝐸. 𝐸𝐸3)

where po is the peak force and td is the duration of the load. This relationship indicates that the

peak force and load duration are not as important independently as their product. Thus, any

number of values can be selected for the peak force and load duration, as long as the impulse

remains constant. Since this is true, either the value for peak force or the time duration must be

assumed and the other value determined from the impulse accordingly. For this analysis, the

pulse duration was assumed in all cases, and the corresponding peak force was determined.

The next step of this analysis was to determine the impulse of the load. A heel drop test has a

few variables, such as the height from which the heel is dropped and the duration of the load.

One constant for each separate load test, however, is the weight of the person performing the

drop test, which in this case, was approximately 185 pounds. The other component of the
180

impulse, the velocity, depends upon the height of the drop. Using an initial drop height of three

inches and applying the equations of uniformly accelerated linear motion, the velocity of the heel

of the foot at impact with the floor was determined. The impulse could then be calculated. It is

given by

𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼 = 𝑚𝑚 × 𝑣𝑣 = 𝑚𝑚�2𝑔𝑔ℎ (𝐸𝐸𝐸𝐸. 𝐸𝐸4)

Both the pulse duration and the peak force are unknown values. Therefore, for a given pulse

duration, the resulting value for the peak force can be determined.

A simply-supported beam was the initial model used to determine the HD pulse because the

closed form response could be determined and compared to the SS beam model. The impulse

generated by a 185 lb person dropping from three inches was used for each of the trials. Starting

with a pulse duration of 0.00625 seconds and a peak force of 7.38 kips, the pulse duration was

increased and the peak force adjusted accordingly to maintain the same impulse. The minimum

and maximum peaks of acceleration and displacement and the times at which each occurred were

noted. The closed form response to a triangular pulse without damping described in Chopra

Chapter 4.9 was used as a comparison for the model’s output. The peak values for both the

model and closed-form solution and the error between the two are provided in Tables E1 and E2.

As seen from the error calculations, the displacement error does not exceed 1.26% for all pulse

durations. The acceleration error is less than 5% for pulses longer than 0.05 seconds

(td/Tn>0.25); however, for shorter pulses (td/Tn<0.25), the error begins to increase, reaching

nearly 30% for td = 0.00625 seconds. This indicates that the model is sufficiently accurate for

longer pulses but is not effective at predicting the response to shorter duration pulses.
181

Table E1: SS beam peak accelerations due to triangular pulse excitation


td = 0.20 sec td = 0.15 sec td = 0.1 sec td = 0.05 sec
Model Exact Model Exact Model Exact Model Exact
Min -8.90 -8.80 -13.13 -12.79 -17.23 -16.41 -24.17 -22.64
Max 8.96 8.80 13.13 12.79 17.31 16.41 22.76 20.41
Min Error 0.011 0.026 0.048 0.063
Max Error 0.017 0.026 0.052 0.103

td = 0.025 sec td = 0.0125 sec td = 0.00625 sec


Model Exact Model Exact Model Exact
- -
Min -107.00 -100.93 257.60 301.80 -715.90 -674.13
Max 110.60 100.93 254.00 234.38 777.60 552.55
Min Error 0.057 -0.172 0.058
Max Error 0.087 0.077 0.289

Table E2: SS beam peak displacements due to triangular pulse excitation


td = 0.20 sec td = 0.15 sec td = 0.1 sec td = 0.05 sec
Model Exact Model Exact Model Exact Model Exact
Min -0.0109 -0.0095 -0.0140 -0.0136 -0.0176 -0.0173 -0.0202 -0.0199
Max 0.0095 0.0095 0.0138 0.0136 0.0176 0.0173 0.0202 0.0199
Min Error 0.126 0.030 0.015 0.016
Max Error 0.004 0.010 0.014 0.017

td = 0.025 sec td = 0.0125 sec td = 0.00625 sec


Model Exact Model Exact Model Exact
Min -0.0218 -0.0215 -0.0231 -0.0227 -0.0238 -0.0235
Max 0.0218 0.0214 0.0229 0.0226 0.0234 0.0234
Min Error 0.017 0.018 0.016
Max Error 0.017 0.014 0.0001

The next step was to apply the same triangular pulses to a computer model of one of the tested

buildings so that the results could then be compared to measured heel drop data. The West

Campus Parking Deck model was chosen. The model consists of five nominally identical bays,

each separated by a beam. The beams were 18 ft on centers, and the beam span was 57 ft. The

fundamental period of the floor in this building is 0.138 seconds, which corresponds to a

frequency of 7.23 Hz. The damping for this model is 3%, as outlined in Appendix A. Additional
182

details for this model are provided in Chapter 4: Modeling Principles. A unit force was applied

in the z-direction to the central node of the slab in the center bay. Using a time history analysis,

the forcing function then applies a value for the force applied to that node at each time step.

12,000 time steps were assigned to the analysis, with a time step size of 0.0001 seconds, creating

a response record of 1.2 second. The output from these tests compared with the measured data

for the West Campus Parking Deck heel drop test is shown in Tables E3, E4, and E5. The peak

minimum and maximum accelerations, displacements, and the time at which these values

occurred were recorded from the response history and compared to the peak values from the

measured data. For each test, the pulse was modified, while maintaining a constant impulse, and

the process was repeated in an effort to generate a response that most closely mirrored the

measured response.

Table E3: Measured Peak Acceleration and


Displacement values for WCPD
Acceleration (in/s2) Displacement (in)
min -19.349 -4.63E-03
max 19.289 3.54E-03

Table E4: Peak Acceleration values (in/s2) for WCPD SAP model under pulse loading
Triangular Pulse
td = 0.05 0.1 0.125 0.15 0.175 0.25
min -44.45 -7.710 -8.441 -7.055 -6.138 -1.152
max 50.89 9.700 9.548 6.749 6.171 1.987

Double Triangular
Sin Pulse Skewed Triangular Pulses
Pulse
td = 0.125 0.125 0.15 0.1 0.125 0.15
min -10.426 -14.512 -15.091 -15.657 -11.021 -6.871
max 11.286 22.010 16.783 19.630 14.216 8.515
183

Table E5: Peak Displacement values (in) for WCPD SAP model under pulse loading
Triangular Pulse
td = 0.05 0.1 0.125 0.15 0.175 0.25
min -0.01252 -0.007 -0.006 -0.005 -0.005 -0.003
max 0.00781 0.004 0.003 0.002 0.002 0.000

Double Triangular
Sin Pulse Skewed Triangular Pulses
Pulse
td = 0.125 0.125 0.15 0.1 0.125 0.15
min -0.007 -0.005 -0.004 -0.008 -0.006 -0.005
max 0.004 0.002 0.003 0.004 0.003 0.002

Comparing the results of these tests to the measured response indicates that a triangular pulse in

which td/Tn<0.25 will not generate a response that is near the measured response. This can be

displayed by the results of the pulse with td = 0.05sec, which corresponds to td/Tn = 0.36. This

pulse generates peak accelerations and displacements which are much larger than the measured

values, and since decreasing the pulse duration will result in an increase in these peak values, a

pulse with td/Tn<0.25 is insufficient. Therefore, the shape of the pulse proves to be important,

and if none of the pulses with longer durations generate a sufficient response, additional pulse

types must be explored. Though different pulse types must be considered, the response of the

models to these longer pulses should prove to be more reliable since the simply-supported beam

tests indicated that the response of the FE model to shorter pulses (td/Tn<0.25) is less accurate.
184

Figure E2: (a) Displacement and (b) Acceleration response of WCPD model to triangular pulse with pulse
duration of 0.125 seconds

By visual inspection of the triangular pulse responses, none of the triangular pulses applied to the

model matched the measured data. This is especially true for the displacement response. Each

of the model tests generated a response in which the first positive and negative peaks are the

overall peaks in the response and the successive peaks decrease thereafter due to damping. This

is different from the measured response, which has its overall peaks in the second cycle of

oscillation. The displacement response that most closely matches the measured data was the

pulse with td = 0.125 seconds, which is shown along with the measured displacement response in

Figure E2(a). The acceleration response for this pulse, however, did not closely match the

acceleration response, as seen in Figure E2(b). Therefore, additional pulse types were explored

to determine if a better response could be generated. One type of pulse used was a sinusoidal

pulse using identical pulse durations and impulses as the triangular pulses. Two additional pulse

types were chosen as an attempt to replicate the delay of the overall peaks displayed in the

measured displacement response. The first was a skewed triangular pulse with its peak force

occurring at 0.75td instead of 0.5td. An example of this is shown in Figure E3. The second pulse

consists of two separate triangular pulses separated by a time gap in which no force is applied.

As seen in Figure E4, the duration of both pulses is the same, but the second pulse has a peak
185

force and impulse that is arbitrarily set at half of the first pulse. Pulses with multiple durations

were used and the results from the tests using these three additional pulse types are shown along

with the triangular pulse results in Tables E4 and E5.

Figure E3: Skewed Triangular Pulse Figure E4: Double Triangular Pulse with time gap

The method of determining the optimal pulse to represent the heel drop consisted of ranking four

different variables. Each of the pulses were ranked for separately in each of these four

categories, and the average ranking for each pulse was then determined. The pulse with the best

overall average ranking was then determined to be the optimal pulse. A least squares error was

calculated for both the displacement and acceleration data sets between each of the pulse tests

and the measured data. These were the first two categories used in the ranking process. The

other two categories consisted of the error calculated between the peak measured values and

peak simulated values of both the displacement and acceleration. These rankings are shown in

Table E6.
186

Table E6: Ranking of pulse responses


Category
Average
Pulse Type Duration Displacement Acceleration D(peak) A(peak)
Ranking
Triangle 0.1 1 7 8 7 5.25
Triangle 0.25 2 2 11 11 6.5
Triangle 0.125 3 9 3 6 5
Triangle 0.15 4 8 4 8 6.25
Triangle 0.225 5 3 10 10 7.25
Triangle 0.175 6 6 7 9 7.25
Sinusoidal 0.125 7 10 6 5 7.25
Double Tri 0.125 9 1 2 3 3.75
Double Tri 0.15 11 4 1 2 4.5
Tri-skew 0.1 8 12 9 1 7.5
Tri-skew 0.125 10 11 5 4 7.75

As indicated in Table E6, the top three ranking pulses are the double triangular pulses with pulse

durations of 0.125 seconds and 0.15 seconds and the triangular pulse with duration of 0.125

seconds, with average rankings of 3.75, 4.5, and 5, respectively. The displacement and

acceleration records for the td = 0.125 second triangular pulse were shown previously in Figures

E2 (a) and (b). The responses for the two double triangular pulses are compared to the measured

responses in Figures E5 and E6. Though the response is not identically replicated by any of

these pulses, the peak displacements and accelerations can be closely predicted with these pulses.

Since it has the best overall ranking and has an acceleration and displacement history which is

relatively consistent to the measured response, the double triangular pulse with td = 0.125

seconds was chosen as the pulse to represent the heel drop test.
187

Figure E5: (a) Displacement and (b) Acceleration response of WCPD model to double triangular pulse with pulse
duration of 0.125 seconds

Figure E6: (a) Displacement and (b) Acceleration response of WCPD model to double triangular pulse with pulse
duration of 0.15 seconds
188

BIBLIOGRAPHY
Aguinaldo, Arnel, Mahar, Andrew, Litavish, Michael, and Morales, Abigail. Ground Reaction
Forces in Running Shoes with Two Types of Cushioning Column Systems. International
Symposium on Biomechanics in Sports, 2002.

Allen, David E., Onysko, Donald M., and Murray, Thomas. ATC Design Guide 1: Minimizing
Floor Vibrations., 1999

Alwan, Majd, Dalal, Siddharth, Kell, Steve, and Felder, Robin. Derivation of Basic Human Gait
Characteristics from Floor Vibrations. 2003.

Ebrahimpour, A., Hamam, A., Sack, R.L., and Patten, W.N. Measuring and Modeling Dynamic
Loads Imposed by Moving Crowds. Journal of Structural Engineering, December 1996.

Glover, Brian H. Structural Design for Vibration Isolation at Benaroya Hall. The Journal of the
Acoustical Society of America, Volume 103, Issue 5, May 1998.

Gordon, Colin G. Generic Criteria for Vibration-Sensitive Equipment. Vibration Control in


Microelectronics, Optics, and Metrology, SPIE Vol. 1619, 1991.

Hanagan, Linda M. Floor Vibration Serviceability Tips and Tools for Negotiating a Successful
Design. Moder Steel Construction, April 2003.

Neville, Adam M. Properties of Concrete. 4th Edition, 1996.

Murray, Thomas, Allen, David E., and Ungar, Erik E. Floor Vibration Due to Human Activity,
AISC Design Guide Series No. 11, 1997.

Pan, Tso-Chien, You, Xuting, and Lim, Chee Leong. Evaluation of Floor Vibration in a
Biotechnology Laboratory Caused by Human Walking. ASCE Journal of Performance of
Constructed Facilities, May/June 2008.

Willford, M.R., and Young, P. A Guide for Footfall Induced Vibration of Structures: A tool for
designers to engineer the footfall vibration characteristics of buildings or bridges. The Cement
Centre, 2006.

Personal References

Taylor, Andy, and Webster David. KPFF Consulting Engineers, Seattle, WA.

S-ar putea să vă placă și