Professional Documents
Culture Documents
Amplitude-Dependent Damping in Vibration Serviceability: Case of A Laboratory Footbridge
Amplitude-Dependent Damping in Vibration Serviceability: Case of A Laboratory Footbridge
Amplitude-Dependent Damping in Vibration Serviceability: Case of A Laboratory Footbridge
net/publication/295920414
CITATIONS READS
22 816
1 author:
Onur Avci
Iowa State University
85 PUBLICATIONS 1,932 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Onur Avci on 12 December 2017.
Abstract: Construction technology advancements in the last couple of decades have led to the use of lightweight and high-strength
materials in structural systems. Although longer spans and lighter materials result in floor systems with less mass, stiffness, and damp-
Downloaded from ascelibrary.org by Qatar University Library on 02/26/16. Copyright ASCE. For personal use only; all rights reserved.
ing, the trend toward a paperless office decreases damping and the amount of live load on the floors even more. Consequently, structures
have become more vulnerable to annoying vibrations, and vibration serviceability has become an area of serviceability concern. For
vibration serviceability calculations, the damping value of the structural systems is a critical parameter. Damping in structures has
proved to be dependent on the amplitude of the applied force on the structure. This condition is referred to as nonlinear damping, or
amplitude-dependent damping. Although damping is constant at low and high amplitudes, for in-between amplitudes, the damping value
increases with the levels of excitation amplitude. For wind and earthquake excitations, the amplitude-dependent characteristics of
damping have been studied extensively in the literature. For floor vibration serviceability applications, even though the nonlinear behav-
ior of damping has been accepted to exist and mentioned in some publications, it is not closely looked at or discussed in detail. The floor
vibration serviceability calculations are very sensitive to damping values, but vibration serviceability researchers and practicing engi-
neers are often uncomfortable with assigning a specific number as a damping ratio for a specific mode because of the inconsistency of
damping values obtained from different methods. This paper presents a closer look at the amplitude-dependent damping in vibration
serviceability and focuses on a laboratory footbridge with experimental and analytical studies. The laboratory footbridge was studied
extensively with static and dynamic tests. Three-dimensional finite-element (FE) models were developed, updated, and fine-tuned for
two bottom chord extension configurations for both static and dynamic tests. The amplitude-dependent damping behavior of the labora-
tory footbridge is shown for different amplitudes of sinusoidal excitations. The amplitude-dependent damping ratio values obtained
from effective mass calculations proved to be correct with the FE model acceleration predictions. The FE model predictions successfully
matched the test results with the nonlinear characteristic introduced for modal damping. One of the most difficult tasks in vibration serv-
iceability research is matching the measured acceleration responses with the FE models, and the success of this paper in matching the
acceleration responses for various levels of excitations (with corresponding amplitude-dependent damping values) with the FE model is
unique. Successful verification and clarification of the amplitude-dependent phenomenon and FE model matching of measured accelera-
tion responses reinforce the confidence in the FE models in vibration serviceability research by showing that the FE models are reliable
not only for natural frequency predictions but also for acceleration response predictions. DOI: 10.1061/(ASCE)AE.1943-
5568.0000211. © 2016 American Society of Civil Engineers.
tice, Wyatt’s model shows that stuck-friction models predict an tudes. Even though this was a very good idea, it has not been put in
amplitude-dependent damping characteristic. practice in floor vibration serviceability applications.
Damping has been known to be dependent on the amplitude of the This paper presents a closer look at amplitude-dependent damp-
applied force on the structure. De Silva (2005) defined amplitude- ing and focuses on the nonlinear behavior of damping by studying a
dependent damping as a “nonlinear form of damping where its laboratory footbridge without any nonstructural components on it.
value depends on the amplitude but not on the frequency of The laboratory footbridge was studied extensively with static and
motion.” Although damping is constant at low and high amplitudes, dynamics tests. Tests were conducted without any nonstructural
for in-between amplitudes, the damping value increases with the ex- components attached to focus only on the amplitude-dependent
citation amplitude. This nonlinear behavior of damping has been damping behavior of the bare structure. Three-dimensional finite-
studied by many researchers, such as Leonard and Eyre (1975), element (FE) models were developed, updated, and fine-tuned for
Jeary and Ellis (1981, 1983), Jeary (1986, 1996, 1997), Chang and different bottom chord extension configurations for both static and
Mohraz (1990), Zhang and Zhang (1994), Bachmann et al. (1995), dynamic tests. The amplitude-dependent characteristic of the foot-
Morita and Kanda (1996), Tamura and Suganuma (1996), Li et al. bridge is shown for different amplitudes of sinusoidal excitations.
(2000, 2003), Butterworth et al. (2004), Živanovic et al. (2005), Initially, three-dimensional FE models were developed and
Brownjohn and Pavic (2007), Wu et al. (2007), Racic et al. (2009), updated for each bottom chord extension configuration to reflect the
Daoulatli (2010), Díaz and Reynolds (2010a, b), Middleton and static test results of the following:
Brownjohn (2010), Ingólfsson et al. (2011), Jang (2011, 2013), • Midspan point loading of bare joists (deflections and bottom
Hudson and Reynolds (2012), Sharma et al. (2012), Taillon et al. chord extension forces are matched with the FE model);
(2012), Aquino and Tamura (2013), Casini et al. (2013), Gromysz • Wet concrete loading on bare joists (deflections and bottom
(2013), Spence et al (2014), Zapico-Valle et al. (2013), Ho et al. chord extension forces are matched with the FE model);
(2014), Steinwolf et al. (2014), Zoghaib and Mattei (2014), and • Midspan point loading on cured concrete slab (deflections and
Salyards and Noss (2014). bottom chord extension forces are matched with the FE
According to Jeary (1996), damping is a result of the mobiliza- model); and
tion of imperfections within the vibrating structure. From this per- • Uniformly distributed loading on cured concrete slab (deflec-
spective, damping will increase linearly with the vibration ampli- tions and bottom chord extension forces are matched with the
tude but it will remain constant at very low and very high vibration FE model).
amplitudes. When a structure is excited, initially, the damping will After the static testing, three-dimensional FE models were
be constant until the large imperfections are mobilized (with low updated for each bottom chord extension configuration to reflect the
amplitudes of vibration). As the vibration amplitude increases, dynamic test results of the following:
damping will also increase as relatively smaller imperfections • Natural frequency of governing flexural bending mode.
within the structure are mobilized. With additional increase in the Being in excellent correlation with the previous test series, the
vibration amplitude, the smallest imperfections in the system will FE models were fine-tuned and detailed enough to reflect the
be mobilized and the damping will be constant for further increases changes in the footbridge structural system due to removal of
in the excitation amplitude. Even though some structures seem to bottom chord extensions from the footbridge structure. The results
conform to Jeary’s damping model (Jeary 1996), it is very difficult of the static testing are already shared by Avci and Murray (2012).
to quantify the damping values accurately (Middleton and The partial results of the dynamic testing (modal parameter estima-
Brownjohn 2010). Similarly, Smith (2001) noted the nonlinearity of tions) are also shared by Avci (2014).
damping and mentioned that “it is often difficult to obtain a precise The study presented in this paper focuses on the amplitude-
value for damping.” dependent damping characteristics of the laboratory footbridge.
For wind and earthquake excitations on structures, the amplitude- After the peak picking method, the natural frequency for the gov-
dependent characteristics of damping have been studied exten- erning bending mode had been determined; however, the nonli-
sively in the literature. However, for floor vibration serviceability nearity in modal damping values was unexplained. The damping
applications, even though the nonlinear behavior of damping has value discrepancies between the vibration software curve fits
been accepted and mentioned in some publications, it is not closely (MEScope) (Vibrant Technology, Inc. 2003), individual FRF
looked at or discussed in detail. The nonlinear damping behavior half-power method results, and time domain decay curve values
has been observed, and the phenomenon has been simplified in are explained and verified in this paper. The footbridge was put
such a way that sometimes the damping values from the frequency in resonance by applying sinusoidal excitations at the governing
domain (half-power method and curve fitting) and time domain bending mode frequency. From the effective mass calculations
(decay curves) do not perfectly match. A very simple explanation of the resonance conditions, the nonlinear behavior characteris-
for this is that the decay curve includes the response from several tics of the damping are identified and verified with the FE model
by showing that they are reliable not only for natural frequency
predictions but also for acceleration response predictions in vibra-
(b) tion serviceability research.
4
5
6
7
8 10
9 11 13
12 14
15
Fig. 8. Stage 1 FRF for center point chirp excitation and center point accelerometer
Fig. 9. Stage 2 FRF for center point chirp excitation and center point accelerometer
The shaker was placed on the force plate to measure the force
tuning the FE models. The footbridge was tested with two differ- applied on the footbridge (Fig. 4). The data from modal testing were
ent bottom chord extension configurations (stages) shown in then loaded, processed, and curve fitted in MEScope commercial
Fig. 3. The two stages of the footbridge are the bottom chord software (Vibrant Technology, Inc. 2003) for both stages, and natu-
extension configurations needed to determine the relative effect ral frequencies, modal damping ratios, and the corresponding mode
of installing bottom chord extensions after the concrete is placed.
shapes were approximated per the input data. MEScope is commer-
Static flexural stiffness testing protocol, FE modeling techni-
cial software used for observing and analyzing noise and vibration
ques, and stiffness test results are shown in detail by Avci and
problems in machinery and structures using either experimental or
Murray (2012), therefore, they will not be repeated in this manu-
analytical data. It can import or directly acquire multichannel time
script. The static stiffness test results (with and without concrete
or frequency data from a machine or a structure and postprocess it
slab) were verified with the FE models for the bottom chord
to come up with operating deflection shapes, resonant vibration,
extension configurations.
and mode shapes from acquired data.
For modal testing, the shaker and force plate assembly were
Modal Testing of the Footbridge Structure placed at different locations on the footbridge, and chirp signals
were used to sweep a frequency range of 4–20 Hz for the two bot-
The aim of modal testing is to apply dynamic excitations to the tom chord extension configurations. The first bending frequency
structure and update the FE model based on the test results with the was the main interest because it is the only mode that falls in the
intent of minimizing the discrepancies between them for improved range that governs the acceleration response of the structure for
Table 1. First Bending Mode Natural Frequencies and Damping Ratios for the Footbridge
20
0
dB = 20*log10(abs(XferDat))
-20
-40
5.4
4.9988
5.0 Xfer Actual
3.0
2.6
Half Power Method:
2.2 β = [8.125- 8.079] / (2x8.10)
β = 0.284 %
1.8
1.4
8.05 8.06 8.07 8.08 8.09 8.10 8.11 8.12 8.13 8.14 8.15
Frequency (Hz)
Fig. 11. FRF for Stage 1: zoomed in for half-power method calculations
0.045
0.035
Filtered:
0.025 above 10 Hz
below 6 Hz
Acceleration (g)
0.015
0.005
-0.005
-0.015 Curve Fit
β = 1.3%
-0.025 Filtered
-0.035
-0.045
1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0
Time (s)
Fig. 12. Stage 1 heel drop excitation decay curve and the fitted curve
ap
p0
System Response
Input Excitation
Time Time
Fig. 13. Input excitation at resonant frequency and the corresponding system response for damped systems
Force Plate Time History and Autospectrum Time History and Autospectrum of Accelerometer Data
30 0.2
Accelerometers (g)
Force Plate (lbs)
0.1
20 Sinusoidal excitation
Downloaded from ascelibrary.org by Qatar University Library on 02/26/16. Copyright ASCE. For personal use only; all rights reserved.
0 −0.2
0 5 10 15 20 0 5 10 15 20
Time (sec) Time (sec)
0.08 0.2
Autospectrum FP
Autospectrum
Channel 1 x 1.414 x (224.3 lb/Volt)= 12.4 lbs Peak Values: (Peak)x(1.414)=Time Domain Amplitude
Peak Ch 2 = 0.099425 (Peak Ch 2 x 1.414) = 0.14059 g
0.04 0.1 Peak Ch 3 = 0.016365 (Peak Ch 3 x 1.414) = 0.02314 g
X: 8
Y: 0.03902
Peak Ch 4 = 0.019753 (Peak Ch 4 x 1.414) = 0.02793 g
Peak Ch 5 = 0.098821 (Peak Ch 5 x 1.414) = 0.13973 g
0.02 0.05 Peak Ch 6 = 0.10041 (Peak Ch 6 x 1.414) = 0.14198 g
Peak Ch 7 = 0.040583 (Peak Ch 7 x 1.414) = 0.057384 g
Peak Ch 8 = 0.04239 (Peak Ch 8 x 1.414) = 0.05994 g
0 0
0 5 10 15 20 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 14. Sinusoidal excitation at 8.0 Hz, amplitude = 56.5 N (12.7 lbs), acceleration response amplitude = 0.14 g: (a) input; (b) output
Force Plate Time History and Autospectrum Time History and Autospectrum of Accelerometer Data
40 0.4
Accelerometers (g)
Force Plate (lbs)
0.2
20 Sinusoidal excitation amplitude Sinusoidal
= 26.7 lbs acceleration
0 response
amplitude
0 = 0.26 g
−0.2
−20 −0.4
0 5 10 15 20 0 5 10 15 20
Time (sec) Time (sec)
0.1 0.2
0.08
Autospectrum FP
Y: 0.08371
Fig. 15. Sinusoidal excitation at 8.0 Hz, amplitude = 118.8 N (26.7 lbs), acceleration response amplitude = 0.26 g: (a) input; (b) output
Accelerometers (g)
Force Plate (lbs)
Maximum Absolute Force: 0.2 Sinusoidal
Channel 1 = 44.7 lbs
acceleration
0 0 response amplitude
= 0.36 g
−0.2
−50 −0.4
0 5 10 15 20 0 5 10 15 20
Time (sec) Time (sec)
Downloaded from ascelibrary.org by Qatar University Library on 02/26/16. Copyright ASCE. For personal use only; all rights reserved.
0.2 0.4
(Peak)x(1.414)=Time Domain Amplit.
(Peak Ch 2 x 1.414) = 0.35861 g
Autospectrum FP
Autospectrum
Force Amplitude
(Peak Ch 4 x 1.414) = 0.072783 g
Channel 1 x 1.414 x (224.3 lb/Volt)= Peak Ch 2 = 0.25361
(Peak Ch 5 x 1.414) = 0.35976 g
0.1 44.1 lbs 0.2 Peak Ch 3 = 0.044877 (Peak Ch 6 x 1.414) = 0.35911 g
Peak Ch 4 = 0.051473
(Peak Ch 7 x 1.414) = 0.16109 g
Peak Ch 5 = 0.25443
(Peak Ch 8 x 1.414) = 0.17858 g
0.05 0.1 Peak Ch 6 = 0.25397
Peak Ch 7 = 0.11393
Peak Ch 8 = 0.1263
0 0
0 5 10 15 20 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 16. Sinusoidal excitation at 8.0 Hz, amplitude = 198.8 N (44.7 lbs), acceleration response amplitude = 0.36 g: (a) input; (b) output
Table 2. Stage 1: Sinusoidal Excitations Test Data Sinusoidal Excitations for Resonance
Sine excitation amplitude
Acceleration response amplitude Neither the frequency domain nor the time domain methods pro-
N lbs (test) (%g) vided a consistent damping ratio value for the footbridge. The
56.5 12.7 0.14 damping ratios from the half-power method, MEScope curve fits,
118.8 26.7 0.26 and decay curves do not provide clear information. The most reli-
198.8 44.7 0.36 able method for achieving more consistent modal damping ratios
remains the effective mass calculations when the footbridge
structure is put in resonance. This can be done by applying excita-
1.5% of each other for Stages 1 and 2). It is also observed from tions at the structure’s natural frequency. Therefore, sinusoidal
Table 1 that the frequency domain methods for damping ratio pre- excitations were applied to the footbridge through a shaker in an
dictions (half-power method and MEScope curve-fitted damping attempt to put the structure in resonance. The resonance condition
ratios) do not consistently match each other. The time domain would allow the use of the effective mass procedure to calculate
method of the decay curve yields a much higher damping ratio than modal damping ratios. A sinusoidal excitation input will result in
the frequency domain predictions for the two stages (even though a sinusoidal acceleration response on the footbridge at the reso-
the decay curve was filtered to contain the first bending mode con- nant frequency (Fig. 13).
tent only).
For the FE model acceleration predictions, damping ratio values Resonant Condition for Stage 1
need to be entered for the footbridge models. However, based on
The footbridge was sinusoidally excited with three different lev-
the scatter observed in Table 1, it is not clear which damping ratio
els of excitation amplitudes (56.5, 118.8, and 198.8 N) for this
should be used in the FE models. The damping ratio for the first
stage. The shaker was placed at the center of the footbridge, and
bending mode cannot be confidently used with the inconsistency
sine waves with an 8.0-Hz frequency were applied to the struc-
shown in Table 1.
ture. The footbridge was put in resonance. The sinusoidal excita-
The FRF in the logarithmic plot is shown in Fig. 10. Fig. 11
tion time history and acceleration responses were measured. The
shows the zoomed-in view for the peak of the FRF plot for half- force amplitudes and the resulting acceleration responses are
power method calculations. Fig. 12 shows the time domain heel shown in Figs. 14–16. The resonant test results for Stage 1 are
drop decay curve for a heel drop conducted at the center of the struc- tabulated in Table 2.
ture. The curve fit to the decay curve yields a damping ratio of
1.3%.
To clarify the modal damping ratio inconsistency, it is decided Resonant Condition for Stage 2
to apply sinusoidal excitations to the footbridge and put the struc- The bottom chord extensions were removed from the system
ture in resonance. When the footbridge is in resonance, the effective and dynamic testing continued. The footbridge was sinusoi-
mass procedure can be used to get the modal damping ratios. The dally excited with four different levels of excitation ampli-
excitation force amplitudes and the resulting acceleration ampli- tudes (33.8, 53.4, 120.1, and 195.7 N) for this stage. The
tudes will be used for the effective mass calculations. shaker was placed at the center of the footbridge, and sine
Accelerometers (g)
65 Sinusoidal
amplitude= 0.1 acceleration
7.6 lbs response
60 0 amplitude=
0.12g
55
−0.1
50
−0.2
5 10 15 5 10 15
Time (sec) Time (sec)
Downloaded from ascelibrary.org by Qatar University Library on 02/26/16. Copyright ASCE. For personal use only; all rights reserved.
0.2 0.1
Autospectrum FP
Autospectrum
AUTOSPECTRUM (Peak)x(1.414)=Time Domain Amplit.
Channel 1 x 1.414 x (224.3 lb/Volt)= 7.59lbs Peak Values: (Peak Ch 2 x 1.414) = 0.11958 g
0.06 Peak Ch 2 = 0.084572 (Peak Ch 3 x 1.414) = 0.013773 g
0.1 Peak Ch 3 = 0.0097404 (Peak Ch 4 x 1.414) = 0.013526 g
Peak Ch 4 = 0.0095656 (Peak Ch 5 x 1.414) = 0.12287 g
0.04 Peak Ch 5 = 0.086896 (Peak Ch 6 x 1.414) = 0.12006 g
0.05 X: 6.95
Y: 0.02393
Peak Ch 6 = 0.08491 (Peak Ch 7 x 1.414) = 0.015746 g
0.02 Peak Ch 7 = 0.011136 (Peak Ch 8 x 1.414) = 0.019461 g
Peak Ch 8 = 0.013763
0
0
5 10 15 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 17. Sinusoidal excitation at 6.95 Hz, amplitude = 33.8 N (7.6 lbs), acceleration response amplitude = 0.12 g: (a) input; (b) output
Force Plate Time History and Autospectrum Time History and Autospectrum of Accelerometer Data
80
Accelerometers (g)
0.2
Force Plate (lbs)
70 Sinosoidal Sinusoidal
excitation 0.1 acceleration
60 amplitude = response
12 lbs 0
50 amplitude
−0.1 = 0.17g
40
−0.2
30
5 10 15 5 10 15
Time (sec) Time (sec)
0.2
0.15 Maximum Absolute Force:
AUTOSPECTRUM
Autospectrum FP
Time Domain force amplitude 0.15 Peak Values: (Peak)x(1.414)=Time Domain Amplitude
= 0.03904 x 224.3 x 1.414 = Peak Ch 2 = 0.11958 (Peak Ch 2 x 1.414) = 0.16909 g
0.1 Peak Ch 3 = 0.01376 (Peak Ch 3 x 1.414) = 0.019456 g
12.4 lbs X: 6.95
Y: 0.1196(Peak Ch 4 x 1.414) = 0.018602 g
0.1 Peak Ch 4 = 0.013156
(Peak Ch 5 x 1.414) = 0.1739 g
Peak Ch 5 = 0.12298
0.05 Peak Ch 6 = 0.1201 (Peak Ch 6 x 1.414) = 0.16982 g
X: 6.95
0.05 Peak Ch 7 = 0.016026 (Peak Ch 7 x 1.414) = 0.02266 g
Y: 0.03904
Peak Ch 8 = 0.019963 (Peak Ch 8 x 1.414) = 0.028227 g
0
0
5 10 15 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 18. Sinusoidal excitation at 6.95 Hz, amplitude = 53.4 N (12.0 lbs), acceleration response amplitude = 0.17 g: (a) input; (b) output
waves with a 6.95-Hz frequency were applied to the structure. Effective Mass Calculations for Resonant Conditions
The footbridge was put in resonance. The sinusoidal excitation
time history and acceleration responses were measured. The When a system is vibrating at a natural frequency, its deforma-
force amplitudes and the resulting acceleration responses are tion follows a specific pattern, which is called a mode shape.
shown in Figs. 17–20. The resonant test information for Stage For each mode shape the amount of mass contributing to the
2 is tabulated in Table 3. system response is different. Effective mass is an indication of
Accelerometers (g)
Force Plate (lbs)
0.2 Sinusoidal
80 Sinusoidal acceleration
excitation response
60 amplitude 0 amplitude =
= 0.24g
40 27.0 lbs −0.2
20
0 5 10 15 20 0 5 10 15
Time (sec) Time (sec)
0.4 0.2
Downloaded from ascelibrary.org by Qatar University Library on 02/26/16. Copyright ASCE. For personal use only; all rights reserved.
Autospectrum FP
AUTOSPECTRUM
Autospectrum
X: 6.95
0.3 0.15 Peak Values:
Y: 0.1725
(Peak)x(1.414)=Time Domain Amplitude
(Peak)x(1.414)= Time Domain Force AmplitudeMaximum Absolute Force:
Channel 1 = 28.7393 lbs Peak Ch 2 = 0.17246 (Peak Ch 2 x 1.414) = 0.24386 g
Channel 1 x 1.414 x (224.3 lb/Volt)= 27.3 lbs Peak Ch 3 = 0.0203 (Peak Ch 3 x 1.414) = 0.028704 g
0.2 0.1 Peak Ch 4 = 0.019474 (Peak Ch 4 x 1.414) = 0.027536 g
Peak Ch 5 = 0.12298 (Peak Ch 5 x 1.414) = 0.1739 g
Peak Ch 6 = 0.1201 (Peak Ch 6 x 1.414) = 0.16982 g
0.1 0.05 Peak Ch 7 = 0.016026 (Peak Ch 7 x 1.414) = 0.02266 g
X: 6.95 Peak Ch 8 = 0.019963 (Peak Ch 8 x 1.414) = 0.028227 g
Y: 0.0862
0 0
0 5 10 15 20 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 19. Sinusoidal excitation at 6.95 Hz, amplitude = 120.1 N (27.0 lbs), acceleration response amplitude = 0.24 g: (a) input (b) output
Force Plate Time History and Autospectrum Time History and Autospectrum of Accelerometer Data
0.4
Accelerometers (g)
80
Force Plate (lbs)
0
−0.4
0 5 10 15 20 0 5 10 15 20
Time (sec) Time (sec)
0.2 0.4
Autospectrum FP
0 0
0 5 10 15 20 0 5 10 15 20
(a) Frequency (Hz) (b) Frequency (Hz)
Fig. 20. Sinusoidal excitation at 6.95 Hz, amplitude = 195.7 N (44.0 lbs), acceleration response amplitude = 0.32 g: (a) input; (b) output
mass sensitivity of each mode to an excitation. Mode shapes Table 3. Stage 2: Sinusoidal Excitation Test Data
can be calculated experimentally using modal data. Each ac-
celerometer location is assigned a lumped mass of the tested Sine excitation amplitude
Acceleration response amplitude
structure (Fig. 21), and collected chirp data for each lumped N lbs (test) (%g)
mass location can then be loaded into commercial vibration
33.8 7.6 0.12
analysis software (MEScope) and the mode shapes can be
53.4 12.0 0.17
determined (Fig. 22).
120.1 27.0 0.24
When a mode shape of a structure is known, the effec-
195.7 44.0 0.32
tive mass corresponding to that mode can be calculated.
Fig. 21. Mode shape determination where ap = peak acceleration; p0 = excitation amplitude of the sinu-
soidal forcing function; b = modal damping ratio; and Meff = effec-
tive mass for a mode.
Effective mass of a mode in a two-way structural system is Eq. (5) can also be written as the following:
defined as ap p0
¼ (6)
g 2 b Weff
Lðx Lðy
Meff ¼ Mðx; yÞð f ðx; yÞÞ2 dxdy (1) where Weff = the effective weight for a mode.
0 0 The total weight of the footbridge structure is 6,344 kg
(13,986 lbs).
Considering Eq. (6), the ap and p0 values are known from the
Effective mass of a mode in a one-way structural system (like
sinusoidal excitation test results for Stages 1 and 2 (Tables 2 and
the footbridge discussed in this paper) is defined as
3). When the effective weight values of 0.470 and 0.502 W are
used for Stages 1 and 2, respectively, the corresponding damping
ðL
ratios are found for each sinusoidal test run, using Eq. (6). Table
Meff ¼ MðxÞð f ðxÞÞ2 dx (2) 5 shows the calculated damping ratios for Stage 1, and Table 6
0 shows the calculated damping ratios for Stage 2. It is very clear
from Tables 5 and 6 that as the sine excitation amplitude
where MðxÞ = mass per unit length of the system (a constant for increases, the damping ratio value also increases. This is the am-
floor systems); and f ðxÞ = mode shape normalized with respect to plitude-dependent behavior of the damping as plotted in Fig. 23.
maximum midspan deflection. The damping values for both Stages 1 and 2 are very similar and
Eq. (1) can also be written as show the same trend for the nonlinear damping phenomenon. It
is observed that for very low excitations, the damping ratio value
Meff ¼ f T M f (3) can go as low as 0.0045, and for increased excitations, the damp-
ing ratio value is around 0.0098.
where Meff = effective mass; M = lumped mass matrix caused by
the tributary area of each accelerometer point on the footbridge; Verification of Amplitude-Dependent Damping with
and, f = mode shape vector normalized with respect to maximum Acceleration Predictions in the FE Model
midspan deflection.
Accelerometer locations and the first bending mode shape were For vibration researchers, the natural frequency test data have been
shown in Fig. 21. The corresponding mass matrix is relatively easier to match with FE models than the acceleration
Table 4. Effective Mass Results via Mode Shapes for the First Bending Table 6. Stage 2: Calculated Damping Ratios via Effective Mass
Mode Procedure
Stage Effective mass calculations via MEScope mode shapes Sine excitation
Calculated damping
amplitude
1 0.470 M Acceleration response ratio ( b ) using
2 0.502 M N lbs amplitude (test) (%g) Meff = 0.502 M
33.8 7.6 0.12 0.00451
53.4 12.0 0.17 0.00503
Table 5. Stage 1: Calculated Damping Ratios via Effective Mass 120.1 27.0 0.24 0.00801
Procedure 195.7 44.0 0.32 0.00979
Sine excitation
Calculated damping
amplitude
Acceleration response ratio ( b ) usring
N lbs amplitude (test) (%g) Meff = 0.470 M FE models have always proved to be more difficult (Barrett 2006;
Davis 2008). Because the acceleration response predictions require
56.5 12.7 0.14 0.00690 the user to input the damping value for the FE model, the results are
118.8 26.7 0.26 0.00781 highly dependent on the often nonlinear damping value (Van
198.8 44.7 0.36 0.00944 Nimmen et al. 2014). Researchers usually use curve-fitted damping
ratios for acceleration prediction (Rainer 1979) and FE modeling
purposes (Živanovic et al. 2005, 2006).
response test data. Because the natural frequency and the mode The level of nonlinearity in damping has been shown in this pa-
shapes of a structure are independent of the loading and damping of per thus far. Yet, when the damping ratios for different levels of the
the structure, natural frequency predictions are easier to match with excitations are known, the FE models can be checked for the accel-
the FE models. In vibration serviceability research, the success rate eration responses.
of matching the acceleration response is not as high as matching the For that purpose, the sinusoidal excitations of Stages 1 and 2 are
natural frequencies with FE models. Acceleration predictions with run in the SAP2000 software with the calculated modal damping
for the loading amplitude of 198.8 N (44.7 lbs). The resulting accel- while reporting damping values in floor vibration testing, the ex-
eration response for this loading is also sinusoidal with an ampli- citation levels of the tests also need to be reported. Such future
tude of 0.348g in Fig. 24. The test data for this case are also shown studies would result in a data set that would offer greater insight
in Table 7. This is an excellent match because the test value for the into amplitude-dependent damping phenomenon in floor vibra-
acceleration response was about 0.36 g. The ratio for the SAP2000 tion applications.
acceleration prediction and the test acceleration are 0.348/0.36 =
0.967, within 3.5%.
As mentioned previously, one of the most difficult tasks in vibra- Conclusions
tion serviceability research is matching the test acceleration
responses with the FE model predictions. On top of the previous This paper presented a closer look into the amplitude-dependent
verifications of the FE model with static and dynamics test results, damping behavior in vibration serviceability. The work described
in the paper focuses on the nonlinear nature of damping by studying
0.012 a laboratory footbridge experimentally and analytically. The labora-
tory footbridge structure was studied extensively with static and
Calculated Modal Damping Ratio
Sine excitation
amplitude
Acceleration response Calculated damping ratio ( b ) Acceleration response Acceleration response amplitude
N lbs amplitude (test) (%g) using Meff = 0.470 M amplitude SAP2000 (%g) ratio: SAP2000/test
56.5 12.7 0.14 0.00690 0.135 0.964
118.8 26.7 0.26 0.00781 0.252 0.969
198.8 44.7 0.36 0.00944 0.348 0.967
Sine excitation
amplitude
Acceleration response Calculated damping ratio ( b ) Acceleration response Acceleration response amplitude
N lbs amplitude (test) (%g) using Meff = 0.502 M amplitude SAP2000 (%g) ratio: SAP2000/test
33.8 7.6 0.12 0.00451 0.1226 1.022
53.4 12.0 0.17 0.00503 0.1736 1.021
120.1 27.0 0.24 0.00801 0.2453 1.022
195.7 44.0 0.32 0.00979 0.3271 1.022
System Response
Input Excitation
Input Output
Time Time
Daoulatli, M. (2010). “Rate of decay of solutions of the wave equation with of Structures, 243–252.
arbitrary localized nonlinear damping.” Nonlinear Anal. Theory Salyards, K., and Noss, N. (2014). “Experimental evaluation of the influ-
Methods Appl., 73(4), 987–1003. ence of human-structure interaction for vibration serviceability.” J.
Davis, D. B. (2008). “Finite element modeling for prediction of low fre- Perform. Constr. Facil., 10.1061/(ASCE)CF.1943-5509.0000436,
quency floor vibrations due to walking.” Ph.D. dissertation, Virginia 458–465.
Polytechnic Institute and State Univ., Blacksburg, VA. Sharma, A., Patidar, V., Purohit, G., and Sud, K. K. (2012). “Effects on the
De Silva, C. W., ed. (2005). Vibration and shock handbook, CRC Press, bifurcation and chaos in forced Duffing oscillator due to nonlinear
Boca Raton, FL. damping.” Commun. Nonlinear Sci. Numer. Simul., 17(6), 2254–2269.
Díaz, I. M., and Reynolds, P. (2010a). “Acceleration feedback control of Smith, R. (2001). “Changing the effective mass to control resonance prob-
human-induced floor vibrations.” Eng. Struct., 32(1), 163–173. lems.” Sound Vib., 35(5), 14–17.
Díaz, I. M., and Reynolds, P. (2010b). “On-off nonlinear active control of Spence, S. M. J., Bernardini, E., Guo, Y., Kareem, A., and Gioffrè, M.
floor vibrations.” Mech. Syst. Sig. Process., 24(6), 1711–1726. (2014). “Natural frequency coalescing and amplitude dependent damp-
Ewins, D. J. (2000). Modal testing: Theory, practice, and application, 2nd ing in the wind-excited response of tall buildings.” Probab. Eng. Mech.,
Ed., Research Studies Press Ltd., Baldock, Hertfordshire, U.K. 35, 108–117.
Gromysz, K. (2013). “Verification of the damping model vibrations of rein- Steinwolf, A., Schwarzendahl, S. M., and Wallaschek, J. (2014).
forced concrete composite slabs.” Procedia Eng., 57, 372–381. “Implementation of low-kurtosis pseudo-random excitations to compen-
Ho, C., Lang, Z.-Q., and Billings, S. A. (2014). “A frequency domain analy- sate for the effects of nonlinearity on damping estimation by the half-
sis of the effects of nonlinear damping on the Duffing equation.” Mech. power method.” J. Sound Vib., 333(3), 1011–1023.
Syst. Sig. Process., 45(1), 49–67. Sun, C. T., and Lu, Y. P. (1995). Vibration damping of structural elements,
Hudson, M. J., and Reynolds, P. (2012). “Implementation considerations for Prentice Hall, Upper Saddle River, NJ.
active vibration control in the design of floor structures.” Eng. Struct., Taillon, J., Legeron, F., and Prud'homme, S. (2012). “Variation of damping
44, 334–358. and stiffness of lattice towers with load level.” J. Constr, Steel Res., 71,
Ingólfsson, E. T., Georgakis, C. T., Ricciardelli, F., and Jönsson, J. (2011). 111–118.
“Experimental identification of pedestrian-induced lateral forces on Tamura, Y., and Suganuma, S. (1996). “Evaluation of amplitude-dependent
footbridges.” J. Sound Vib., 330(6), 1265–1284. damping and natural frequency of buildings during strong winds.” J.
Inman, D. J. (2000). Engineering vibration, 2nd Ed., Prentice Hall, Upper Wind Eng. Ind. Aerodyn., 59(2–3), 115–130.
Saddle River, NJ. Tilly, G. P., Cullington, D. W., and Eyre, R. (1984). “Dynamic behaviour of
Jang, T. S. (2011). “Non-parametric simultaneous identification of both the footbridges.” IABSE Surveys S-26/84, No. 2/84, International Association
nonlinear damping and restoring characteristics of nonlinear systems of Bridge and Structural Engineering, Zurich, Switzerland, 13–24.
whose dampings depend on velocity alone.” Mech. Syst. Sig. Process., Van Nimmen, K., Lombaert, G., De Roeck, G., and Van den Broeck, P.
25(4), 1159–1173. (2014). “Vibration serviceability of footbridges: Evaluation of the cur-
Jang, T. S. (2013). “A method for simultaneous identification of the full non- rent codes of practice.” Eng. Struct., 59, 448–461.
linear damping and the phase shift and amplitude of the external harmonic Vibrant Technology, Inc. (2003). ME’ScopeVES online help, Scotts
excitation in a forced nonlinear oscillator.” Comput. Struct., 120, 77–85. Valley, CA.
Jeary, A. P. (1986). “Damping in tall buildings—A mechanism and a pre- Wu, J. R., Liu, P. F., and Li, Q. S. (2007). “Effects of amplitude-dependent
dictor.” Earthquake Eng. Struct. Dyn., 14(5), 733–750. damping and time constant on wind-induced responses of super tall
Jeary, A. P. (1996). “The description and measurement of nonlinear damp- building.” Comput. Struct., 85(15–16), 1165–1176.
ing in structures.” J. Wind Eng. Ind. Aerodyn., 59(2–3), 103–114. Wyatt, T. A. (1977). “Mechanisms of damping.” Symp. on Dynamic
Jeary, A. P. (1997). “Damping in structures.” J. Wind Eng. Ind. Aerodyn., Behaviour of Bridges, Transport and Road Research Laboratory,
72, 345–355. Crowthorne, U.K., 10–21.
Jeary, A. P., and Ellis, B. R. (1981). “Vibration tests of structures at varied Zapico-Valle, J. L., García-Dieguez, M., and Alonso-Camblor, R. (2013).
amplitudes.” Proc., 2nd Specialty Conf. on Dynamic Response of “Nonlinear modal identification of a steel frame.” Eng. Struct., 56,
Structures: Experimentation, Observation, Prediction and Control, 246–259.
ASCE, Reston, VA, 281–294. Zhang, W., and Zhang, H. (1994). “Modeling and analysis of nonlinear
Jeary, A. P., and Ellis, B. R. (1983). “On predicting the response of tall build- damping mechanisms in vibrating systems.” Int. J. Mech. Sci., 36(9),
ings to wind excitations.” J. Wind Eng. Ind. Aerodyn., 13(1–3), 173–182. 829–848.
Leonard, D. R., and Eyre, R. (1975). “Damping and frequency measure- Živanovic, S., Pavic, A., and Reynolds, P. (2005). “Vibration serviceability
ments on eight box girders.” Rep. No. LR682, Transport and Road of footbridges under human-induced excitation: A literature review.” J.
Research Laboratory, Dept. of the Environment, Crowthorne, U.K. Sound Vib., 279(1–2), 1–74.
Li, Q., Liu, D., Fang, J., Jeary, A., and Wong, C. (2000). “Damping in build- Živanovic, S., Pavic, A., and Reynolds, P. (2006). “Modal testing and FE
ings: its neural network model and AR model.” Eng. Struct., 22(9), model tuning of a lively footbridge structure.” Eng. Struct., 28(6),
1216–1223. 857–868.
Li, Q. S., Yang, K., Wong, C. K., and Jeary, A. P. (2003). “The effect of am- Zoghaib, L., and Mattei, P.-O. (2014). “Time and frequency response of
plitude-dependent damping on wind-induced vibrations of a super tall structures with frequency dependent, non-proportional linear damping.”
building.” J. Wind Eng. Ind. Aerodyn., 91(9), 1175–1198. J. Sound Vib., 333(3), 887–900.