Article

Advances in Structural Engineering

1–13
Modal parameter identification of a Ó The Author(s) 2017
Reprints and permissions:
long-span footbridge by forced sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/1369433217698322

vibration experiments journals.sagepub.com/home/ase

Q Wen1, XG Hua1, ZQ Chen1, JM Guo2 and HW Niu1

Abstract
Performing forced vibration tests on full-scale structures is the most reliable way of determining the relevant modal parameters in
structural dynamics, such as modal frequencies, mode shapes, modal damping, and modal masses. This study describes the modal iden-
tification of a double-level curved cable-stayed bridge with separate deck systems for pedestrians and vehicles via forced vibration
tests. The steady-state structural responses to sinusoidal excitations produced by an electrodynamic shaker are recorded under vary-
ing excitation frequencies, and the frequency response functions are established. The measured frequency response functions are
curve fitted to estimate the modal parameters. The numerical simulation of frequency response function–based modal parameter
identification of an elastically multi-supported continuous beam structure is carried out, and the emphasis has been placed on the eva-
luation of the effect of an additional shaker mass, excitation frequency step and range, multi-mode vibration, and noise on identifica-
tion results. Finally, the modal parameters for the first lateral mode of a double-level curved cable-stayed bridge are identified by
forced vibration experiments, and the results are compared with those from ambient vibration tests and free vibration tests. The
effect of the unmeasured wind excitation on identification is discussed. It is shown that the effect of ambient vibration is minor for
wind velocity of 3–5 m/s. The damping ratios identified by forced and free vibration tests are comparable, while those from ambient
vibration are subject to large variations. The modal mass obtained from forced vibration tests is in good agreement with finite element
prediction, which provides design basis for mass-type dampers.

Keywords
footbridge, forced vibration tests, frequency response function, modal identification

Introduction Performing actual vibration tests on full-scale struc-

tures is the most reliable way of determining the rele-
The human-induced vibration and serviceability prob- vant modal parameters in structural dynamics, such as
lem of lively footbridges have received wide attentions modal frequencies, mode shapes, modal damping, and
since the well-known example of the London modal masses. Dynamic tests for modal identification
Millennium Bridge (Dallard et al., 2001; Ingolfsson of full-scale structures can be generally classified
et al., 2012; Racic et al., 2009; Ricciardelli and according to the following three types: (1) ambient
Pizzimenti, 2007; Živanović et al., 2005). Increasing vibration tests, (2) free vibration tests, and (3) forced
strength of new structural materials and longer span of vibration tests. Ambient vibration tests based on mea-
new footbridges, accompanied by aesthetics require- surement of small vibrations induced by actions of
ments for greater slenderness, are now resulting in winds and vehicles provide an economical and conve-
light and slender footbridges that are more sensitive to nient way to identify the modal parameters since the
dynamic excitations. Thus, prediction of the responses
of lively footbridges to dynamic loadings induced by 1
Key Laboratory for Wind and Bridge Engineering of Hunan Province,
walking pedestrians has become an indispensable
College of Civil Engineering, Hunan University, Changsha, China
ingredient for dynamic design of footbridges, which 2
Guangzhou Municipal Engineering Design & Research Institute,
requires knowledge of the modal parameters of major Guangzhou, China
importance, such as modal frequency, mode shape,
modal damping, and modal mass. The information on Corresponding author:
XG Hua, Key Laboratory for Wind and Bridge Engineering of Hunan
the modal parameters is also crucial to development Province, College of Civil Engineering, Hunan University, 410082
and implementation of effective vibration control for Changsha, Hunan, China.
footbridges. Email: [email protected]
2 Advances in Structural Engineering

equipment required during tests is only for data acqui- is briefly presented to introduce the definition and pro-
sition. Under the condition of Gaussian stationary cedure used in the study. The frequency responses
excitation, the modal parameters may be identified between a reference point of steady-state harmonic
using output-only modal identification procedures, excitation and a number of response measurement
such as peak-picking method, frequency domain points are collected and then fitted to single-mode or
decomposition (Brincker et al., 2000), the natural exci- multi-mode FRF by least square method to estimate
tation technique (James et al., 1995), complex modal the modal parameters. The accuracy of FRF-based
indication function (Ni et al., 2005), random decre- modal identification is explored through numerical
ment method (Wang et al., 2014; Wu et al., 2012), and examples of an elastically multi-supported continuous
stochastic subspace identification (SSI; Van Overschee beam with various exciter mass, excitation frequency
and De Moor, 1996). In free vibration tests, free oscil- step and range, multi-mode vibration, and noise.
lations are obtained by a large impulse or suddenly Finally, the modal parameters for the first lateral mode
releasing the structure from an initial disturbance, and of the Mianyang First Bridge are identified based on
natural frequencies and modal damping ratios for tar- forced vibration experiments, and the identification
get modes can be identified by least square method or results are compared with those from ambient and free
covariance-driven stochastic subspace identification vibration tests.
(SSI-COV) method (Magalhães et al., 2010). Both
ambient and free vibration tests have been widely used
in the modal parameter identification of footbridges Theoretical background
(Bayraktar et al., 2009; Caetano et al., 2010a; Gentile Modal identification using forced vibration tests is a
and Gallino, 2008). Free vibration tests are usually mature and widely used technique in mechanical engi-
performed under much more well-controlled condi- neering and aerospace engineering (Ewins, 2000). The
tions than ambient vibration tests and therefore pro- theoretical background of the FRF-based modal iden-
vide improved results of the modal parameters in tification is briefly described for completeness.
particular damping ratios. However, both methods
cannot provide the information on modal mass, which
is an important parameter for predicting the human- FRF of dynamic systems
induced vibration and designing the mass-type dam- The equation of motion for an n-degree-of-freedom
pers for vibration control. (DOF) viscous-damped linear system can be described
In forced vibration tests, the steady-state structural as
responses subject to mechanical exciters are recorded
under varying excitation frequencies, and the frequency € + CX(t)
MX(t) _ + KX(t) = F(t) ð1Þ
response functions (FRFs) are established. In general,
modal identification using forced vibration tests is a where M, K, and C are the mass, stiffness, and damp-
mature and standard technology in mechanical and _
ing matrices for the system, respectively; X(t), X(t),
aerospace engineering. However, it is rarely been used €
and X(t) represent the nodal displacement, velocity,
for engineering structures due to the requirement of and acceleration vectors, respectively; F(t) denotes the
heavy mechanical exciters, such as electrodynamic sha- nodal force vector.
ker. In the context of lightweight footbridges, it is prac- The response of the system can be expressed in the
ticable for forced vibration tests to identify the modal frequency domain as
parameters (Caetano et al., 2010b; Dallard et al., 2001;
Reynders et al., 2010; Živanović et al., 2006). X (v) = H(v)F(v) ð2Þ
The Mianyang Bridge is a double-level curved
where H(v) = [K + ivC 2 v2M]21.
cable-stayed bridge with a central span of 200 m and
Assuming that a harmonic force Fk(v) is applied at
separate deck systems for pedestrians and vehicles.
the kth DOF and the response of the jth DOF is Xj(v),
Due to large span of the pedestrian deck, it is prone to
the receptance FRF relating to the displacement
human-induced vibrations in the lateral direction at
response and input harmonic force is given as follows
relatively low pedestrian density. To implement struc-
(Ewins, 2000)
tural vibration control based on the tuned mass dam-
pers (TMDs), accurate estimation of the modal  
  X (v) X m
1 fjr fkr 
parameters for the lateral modes is essential. This study Hjk (v) = j  
= 3 2 
describes modal parameter identification of the Fk (v) r = 1 mr vr  v2 + 2izr vvr 
Mianyang First Bridge based on measurement of ð3Þ
FRFs by forced vibration experiments. The classical
FRF-based method of extracting the modal parameters or
Wen et al. 3

 
  X (v) X m Arjk  Modal parameter estimation based on FRFs
Hjk (v) = j  
=  ð4Þ
Fk (v) r = 1 v2r  v2 + 2izr vr v The modal parameters are identified by curve fitting of
equation (4) to the measured FRFs by nonlinear least
where vr, fr, and zr are the undamped frequency, square method. First, the steady-state structural
mode shape, and modal damping ratio for the rth responses to sinusoidal force produced by electrodynamic
mode, respectively; mr = frTMfr is the generalized shaker are recorded under varying excitation frequencies,
modal mass for the rth mode; Arjk is a modal constant and the recorded input and structural responses are pro-
determined by mode shape and modal mass for the rth cessed to establish the FRFs of the structure. Then, the
mode, that is, Arjk = fjr fkr =mr . There are two com- measured FRFs are employed to identify the modal
monly used schemes for normalization of mode shapes parameters based on nonlinear curve fitting.
in structural dynamics. The unity-scaled mode shape is In practice, the frequency bandwidth of excitation
used in this study, which also brings about a clear is limited, that is, v 2 [va, vb]. In this bandwidth, n*
physical meaning of TMD mass for the mass-type resonant responses are investigated, indicating that n*
vibration absorbers. Equation (4) is based on normal natural frequencies locate in this bandwidth. To reduce
mode assumption and therefore is valid for lightly the unknown parameters, equation (6) is simplified to
damped structures with proportional damping. equation (8), in which only n* modes are superposed
  
  € 2 X n Arjk 
Dynamic tests using electrodynamic shakers Hjk (v) = X j (v)=v =  
 ð8Þ
m0 A0 v2 r = 1 v2r  v2 + 2izr vr v
Some kind of shakers is necessary to generate the
desired excitation forces when performing forced vibra- The mode shape can also be estimated by placing N
tion tests. The electrodynamic shaker is employed in accelerometers at different DOFs including the DOF
this study. The sinusoidal excitation force at the kth of exciting, and N modal constants Arjk for each mode
DOF generated by the electrodynamic shaker equals to are obtained by the FRF-based method. Then, the
the inertial force of the moving mass, which can be unity-scaled mode shape amplitudes jfr j are estimated
expressed in frequency domain as follows by equation (9), as follows

Fk (v) = m0 A0 v2 ð5Þ 1
jfr j =
max (Arjk ) ð9Þ
where m0 (kg) and A0 (m) are the moving mass of the
½ Ar1k ... Arjk ... ArNk T (j = 1, . . . , N)
shaker and vibration amplitude of the moving mass,
respectively; v (rad/s) is the excitation frequency. The phases of fr will be 0° or 180° for lightly
The recorded response is usually the structural damped structures or proportional damping cases,
acceleration at the jth DOF. So the frequency response which can be easily obtained by comparing the sign of
amplitude function of receptance FRF is given as the response at kth DOF with that at jth DOF at the
  same time when the excitation frequency is nearest to
  X (v) € j (v)=v2
X X m Arjk 
Hjk (v) = j   the resonant frequency.
= =  
Fk (v) m0 A0 v2 r = 1 v2r  v2 + 2izr vr v After obtaining the mode shapes, the modal mass
ð6Þ can be estimated by the following equation

The acceleration of the moving mass of the shaker is fjr fkr

mr = ð10Þ
relative to the structure, so the motion equation of the Arjk
coupled shaker–structure system can be written as Numerical simulation
qr (t) + cr q_ r (t) + kr qr (t) = fkr m0 A0 v2 sin vt
(mr + m0 )€ In this section, several numerical simulations are per-
ð7Þ formed to demonstrate and verify the FRF-based
method with emphasis placed on studying the effects
From equation (7), it is seen that the mass of the of additional shaker mass, the excitation frequency
shaker can influence the dynamic properties of the step and range, multi-mode vibration, and noise.
structure under test. In practice, m0 is far less than mr,
for example, Caetano et al. (2010b), m0 = 1200 kg
and mr = 165,880 kg. Hence, the mass of the shaker Elastically multi-supported continuous beam
has a small effect on the modal parameter An elastically multi-supported beam, which has the
identification. characteristics of sinusoidal-type mode shapes and
4 Advances in Structural Engineering

Figure 1. Elastically multi-supported continuous beam (mm).

Table 1. Modal parameters obtained by finite element analysis.

Mode Frequency Damping Modal

(Hz) ratio (%) mass (kg)

1 0.702 0.50 31.413

2 0.788 0.50 31.444
3 1.086 0.54 31.496

closely spaced frequencies for the first several modes,

Figure 2. Frequency response functions for two mass ratios:
is designed, as shown in Figure 1. The beam spans
m = 0 and m = 0.01.
10 m and is supported by a total of 19 elastic springs
evenly distributed along the span with the same stiff-
ness of 60 N/m. The use of elastic springs is intended
to simulate the closely spaced modal frequencies possi- Table 2. Identified modal parameters under different
bly encountered in real cable-supported or suspension excitation mass ratios.
bridge structures. The girder is made of a steel rectan-
Case Mass Frequency Damping Modal
gular cross section with width of 0.2 m and depth of
ratio (Hz) ratio (%) mass (kg)
0.004 m. The material properties are as follows:
Young’s modulus E = 2.06 3 1011 Pa and density 1 0 0.702 0.50 31.431
r = 7850 kg/m3. 2 0.005 0.700 0.50 31.314
The damping is modeled by Rayleigh damping. 3 0.01 0.698 0.50 31.232
4 0.02 0.695 0.50 31.068
Rayleigh damping coefficients are derived by deter-
mining the first and second modal damping ratios as
0.5%. An analytical modal analysis is carried out to shown in Figure 2. The results are listed in Table 2.
estimate the modal parameters, as shown in Table 1. When m = 0, the identified results are exactly the same
The numerical simulations are performed by frequency as the analytical results, indicating that the FRF-based
domain analysis or time domain analysis. The excita- method is viable to accurately estimate the modal para-
tion force is given as F(t) = F0 sin(vt), where F0 is meters. The accuracy of the estimated parameters
unity, that is, F0 = 1 N. degrades with the increase in mass ratio. Therefore,
provided that the excitation force is sufficiently large to
Shaker mass set the structure in vibration, a smaller mass is prefer-
able and lead to more accurate results.
The shaker fixed on the structure has two effects: addi-
tional mass and excitation force. The additional mass
induces changes in the dynamic characteristics of the
Excitation frequency step and range
structure. The shaker mass is given by mass ratio m, The minimal frequency step is limited by the design
where m = m0/m1, in which m1 is the modal mass for and manufacture of the exciter. For example, in
the first mode. Harmonic response analyses, with dif- Caetano et al. (2010b), the frequency step is 0.01 Hz.
ferent mass ratios, shaker fixed at the 1/2 span, exciting In theory, the smaller the frequency step, the higher
frequency range f 2 [0.670, 0.730] Hz, and frequency the accuracy of the estimated parameters.
step Df = 0.004 Hz, are performed to obtain fre- The excitation frequency step and range determine
quency response characteristics. Then the modal para- the number of the repetitive vibration exercise. Too
meters are estimated by fitting frequency response wide range and too small step are uneconomical and
characteristics to a single-mode FRF. The analytical time-consuming. Therefore, it is necessary to select a
and identified results for m = 0 and m = 0.01 are suitable range and step. The numerical identification
Wen et al. 5

Table 3. Modal parameters with different frequency ranges and steps.

Case Parameters of excitation Identified modal parameters

Range (Hz) Step (Hz) Number Frequency (Hz) Damping ratio (%) Modal mass (kg)

5 0.670–0.730 0.004 16 0.702 0.50 31.431

6 0.680–0.720 0.005 9 0.702 0.51 31.254
7 0.670–0.730 0.005 13 0.702 0.51 31.283
8 0.650–0.750 0.005 21 0.702 0.50 31.313
9 0.680–0.720 0.01 5 0.702 0.52 31.114
10 0.670–0.730 0.01 7 0.702 0.52 31.136
11 0.650–0.750 0.01 11 0.702 0.52 31.164
12 0.600–0.800 0.02 11 0.700 0.56 31.075

Table 4. Identified modal parameters based on multi-mode FRF.

Case Mode Frequency (Hz) Damping ratio (%) Modal constant Arjk Mode shape fkr Modal mass (kg)

13 1 0.702 0.50 0.0159 0.706 31.482

2 0.788 0.50 0.0318 1 31.415

results using different excitation frequency steps and

ranges are shown in Table 3. It demonstrates that
choosing suitable range and small step is important to
obtain the modal parameters accurately.

Multi-mode FRF
As shown in equation (6), the measured FRFs may be
contributed by a number of modes. When the modes
are sparse, the modal parameters can be estimated
accurately by fitting the frequency response character-
istic to a single-mode FRF. Because the displacement Figure 3. Frequency response functions with multi-mode
(Hkk(v)).
of the mid-span for the second mode vanishes, the
first-order modal parameters are estimated accurately
by fitting to single-mode FRF when the excitation
force acts at the mid-span, as shown in Table 3. When
the excitation force is applied at the 1/4 span, the FRF
will be closely spaced frequency components.
Therefore, the frequency response characteristic should
be fitted to a multi-mode FRF.
Harmonic response analyses are performed to obtain
the frequency response functions by applying the sinusoi-
dal excitation forces at the 1/4 span whose frequency
range is from 0.67Hz to 0.83Hz with an increment of
0.005Hz, the results are shown in Figure 3. Then the
modal parameters (Table 4) are identified by two-mode Figure 4. Identified mode shapes compared with the analytical
FRF. The identified mode shapes are shown in Figure 4. results.
For multi-mode FRF, the method is remarkably accurate.
unmeasured ambient vibration during forced vibration
tests may be regarded as noise or imperfect excitation.
Effect of noise Therefore, two types of noise are considered in simula-
In practice, the noise is unavoidable. In addition to the tion analysis: (1) zero-mean Gaussian white noise in
commonly encountered measurement noise, the excitation forces, which is simplified to act at the 1/4,
6 Advances in Structural Engineering

Table 5. Identified modal parameters with various noise levels.

Modal R1 = 0, R1 = 5% R1 = 10% R1 = 15% R1 = 20%

parameters R2 = 0
R2 = 10% R2 = 20% R2 = 10% R2 = 20% R2 = 10% R2 = 10%

m1 (kg) 32.18 31.742 6 1.114 31.613 6 1.639 30.882 6 2.861 30.657 6 2.928 29.368 6 2.759 27.112 6 4.528
f1 (Hz) 0.699 0.699 6 0.000 0.699 6 0.000 0.699 6 0.000 0.699 6 0.000 0.699 6 0.001 0.699 6 0.001
z1 (%) 0.49 0.50 6 0.03 0.50 6 0.04 0.52 6 0.07 0.52 6 0.08 0.52 6 0.10 0.54 6 0.19

Figure 5. Elevation, plan, and side view of the bridge (m).

1/2, and 3/4 span, respectively and (2) zero-mean

Gaussian white measurement noise, which is modeled
by superposing it to the analytical response. The level
of noise is defined by the noise-to-signal ratio R = sy/
sx. For noise excitation R1, sy and sx are the root-
mean squares of the responses induced by noise excita-
tion and the shaker at the resonant frequency, respec-
tively. For measurement noise R2, sy and sx are the
root-mean squares of measurement noise and the
steady-state response induced by the shaker at the Figure 6. View of the Mianyang First Bridge from a
resonant frequency, respectively. neighboring building.
The steady response of the first 200 s, induced by
excitation forces acting at the 1/2 span and noise, is
analyzed. The frequency range and step are [0.670, Mianyang, China. It is a curved cable-stayed bridge
0.730] Hz and 0.005 Hz, respectively. The identified with a central span of 200 m and two side spans of
modal parameters are listed in Table 5. For each level 100 m, as shown in Figure 5. Figure 6 shows the view
of noise, 200 cases with different noise levels were con- of the Mianyang First Bridge from a neighboring
ducted. The results are represented by the mean values building. The distinctive feature of the bridge is the
and the distribution range with amplitude up to the separate box-girder systems designed for the vehicles
standard deviation. It demonstrates that the FRF- and pedestrians, respectively. Figure 7 illustrates the
based method is of acceptable accuracy when R1 and general cross sections for the roadway deck and pedes-
R2 are less than 10%. With the increase in the level of trian deck. The roadway deck carrying three lanes in
noise, the distribution range increases. each direction is a concrete box-girder with 28 m width
and a central depth of 3 m; the pedestrian deck is
formed by an orthotropic steel box-girder of 6 m wide
Case study and a depth of 2 m. The weight of the bridge deck
Description of the Mianyang First Bridge including the pavement, crash barrier, and handrail is
about 60 ton/m for the roadway deck and 5.0 ton/m
The Mianyang First Bridge is located across the for the pedestrian deck. The pedestrian deck is located
Fujiang River in the Chengnan New District of above the roadway deck in elevation, and it roughly
Wen et al. 7

Figure 7. General cross sections of road deck and pedestrian deck (mm).

Figure 8. Layout of steel braces and TMDs: (1) HB, OB, and VB donate horizontal, oblique, and vertical brace, respectively and (2)
1# to 5# present the TMDs.

follows a shallow letter ‘‘S’’ in plan such that it goes pedestrian dynamic loads. Field dynamic tests on the
through the roadway bridge from one side of its deck full-scale bridge confirm that the synchronization of
to the opposite side at the central span. The bridge gir- lateral excitation occurs when the central span is occu-
ders resting atop pylons’ cross-beams and concrete pied with a walking pedestrian density of 0.125 person
piers are supported by their individual cable systems. per square meter, which is considerably less than the
Each inverted-Y pylon reaches 102.5 m high above its design value, 1.0 person per square meter. (Wen et al.,
pile foundations, as given in the side view of Figure 5. 2016). After a comparison with vibration control
The pylon is composed of a steel segment for ancho- scheme using only damping systems, a hybrid control
rage of stay cables, a concrete segment, and a strategy of stiffness upgrading and damping enhance-
concrete–steel composite segment between them. The ment is used for the present bridge. The lateral and ver-
longitudinal viscous dampers are attached between tical stiffness of the pedestrian deck is strengthened by
bridge girders and cross-beams of pylons mainly for three pairs of steel braces linking the roadway deck and
seismic purposes. pedestrian deck at central span and both side spans.
Due to high flexibility and curved appearance of the Then, the eddy current TMDs will be used to control
pedestrian deck, vibration severability has been taken the lively modes for the bridge with the steel braces.
into consideration at the design phase. A finite element The layout of steel braces and TMDs is illustrated in
model of the original bridge design scheme (without Figure 8. Figure 9 shows the prototype of the vertical
steel braces as will be discussed later) has been estab- brace and the oblique brace at the central span. The use
lished for analyzing the human-induced vibrations. of the steel braces is intended to reduce the number of
The numerical simulations show that the fundamental lively modes of the original bridge, therefore reducing
natural frequency for the lateral mode is only 0.60 Hz, the cost of structural vibration control system.
which implies that the pedestrian deck may experience
the synchronization of lateral excitation as found in
the London Millennium Bridge (Dallard et al., 2001). Finite element model of the bridge with steel braces
Additionally, there are seven vertical modes whose nat- The lengths of steel braces are 8.65 and 9.53 m for ver-
ural frequencies fall in the forcing frequency of tical and oblique braces at central span and 7.06 m for
8 Advances in Structural Engineering

and pavement on the decks are considered in finite ele-

ment modeling (MASS21), and their stiffness is consid-
ered negligibly small. The steel braces are modeled by
the truss elements with the equivalent axial stiffness.
Figure 11 shows the finite element model which con-
sists of 646 elements and 469 nodes. The natural fre-
quency and unity-scaling modal mass for the first
lateral mode are 0.735 Hz and 470 ton, respectively,
and the associated mode shape is shown in Figure 12.
The natural frequency for the second lateral mode is
well above 1.2 Hz and is not sensitive to human-
induced vibrations.
Figure 9. Prototype of the vertical and oblique steel braces at
central span.
Wind-excited ambient vibration and aerodynamic
damping
horizontal braces at both side spans. Figure 10 shows
the structural design drawings of steel braces. Each A series of dynamic tests are carried out to identify the
steel brace consists of a steel pipe with a diameter of modal parameters of the bridge for fine tuning of
426 mm and a thickness of 10 mm and a high damping TMDs as well as to evaluate the lateral and vertical
rubber pad of 80 mm thickness at both ends. The rub- vibration responses to different pedestrian group
ber pad is designed such that the connections between streams. The latter tests are described in detail by Wen
steel braces and girders behave like hinge, and it is also et al. (2016). The dynamic tests for modal identifica-
helpful to isolate the vibration of the roadway deck tion include ambient vibration test, free vibration test,
induced by passing traffic. The steel brace is pre- and most importantly forced vibration test.
stressed using the high-strength strands inside steel As the bridge has not been open to traffic and
pipe to guarantee it operates in compression. The pedestrian during the dynamic tests, natural wind
brace is modeled as a series system of steel pipe and therefore becomes the main excitation source for ambi-
two 80-mm-thick rubber pads for accounting for the ent vibration. Due to the lightweight pedestrian deck,
limited axial stiffness of rubber pads. As the axial stiff- a question at the first glance is whether the wind could
ness of the rubber pad is significantly less than that of produce the obvious aerodynamic damping and there-
steel pipe, the equivalent stiffness is predominantly fore alter the total damping. Based on the force mea-
determined by the rubber pad. The equivalent axial surement results on the section model of the pedestrian
stiffness is almost the same for three steel braces with deck, this can be approximately assessed with the
different lengths and is calculated approximately as quasi-steady theory (Scanlan and Simiu, 1996). For
16.6 M N/m, which is only a fraction of the stiffness deck with constant depth of cross section of the pedes-
of steel pipe. trian deck, the aerodynamic damping factor is given
A spatial finite element model for the bridge with by
steel braces is established using the ANSYS FE pack-
rUCD H
age based on the design drawings for determining the zae = ð11Þ
modal properties. The bridge pylons and girders are 2mv
modeled by the Timoshenko beam elements where the numerator is the aerodynamic damping coef-
(BEAM44), and the stay cables are modeled by two- ficient, U is the mean wind velocity, CD is the drag
node cable elements (LINK8). The mass of handrail coefficient of the pedestrian deck measured in the wind

Figure 10. Structural design drawings of steel braces.

Wen et al. 9

Figure 11. Finite element model of the bridge with steel

braces. Figure 13. The section model of pedestrian deck in wind
tunnel.

Figure 12. The first lateral mode of the bridge with steel
brace.

tunnel, H is the depth of pedestrian deck cross section,

r is the air mass density, m is the mass per unit length
of the pedestrian deck, and v is the circular frequency
of the mode under investigation. Figure 13 shows the
1:20 section mode of the pedestrian deck in the wind Figure 14. Acceleration responses of pedestrian deck at the
middle of central span.
tunnel, and the measured drag coefficient in smooth
flow is 1.97 at 0° wind attack angle. For a mean wind
velocity of 5 m/s, the aerodynamic damping factor for
the first lateral mode is evaluated as 0.05%. Therefore,
it is concluded that aerodynamic damping is insignifi-
cant when the mean wind velocity is not high during
ambient vibration tests. It is also implied that the var-
iation in the measured total damping will not be signif-
icant due to the variation in incoming wind velocities.
Several ambient vibration tests are also performed.
The measurement acceleration responses of the pedes-
trian deck for the vertical and lateral directions at the
middle of central span before performing forced vibra- Figure 15. Stabilization diagram of SSI for ambient vibration
tion testing are shown in Figure 14. SSI is employed to responses.
identify modal frequencies and damping ratios from
those ambient vibration responses. Figure 15 shows
the stabilization diagraph of SSI for one typical ambi- properties are identified by curve fitting. It is accom-
ent vibration data. The mean and standard deviation plished by the development of a self-made electrody-
of damping ratio for the lateral modes are 0.76% and namic horizontal shaker, as shown in Figure 16. The
0.21, respectively. The damping ratios obtained from shaker consists of a moving platform with additional
ambient vibration tests have large variations. mass and servomotor numerically controlled by a com-
puter. It is capable of delivering sinusoidal force or
random force, and the former is employed in this
FRF measurement and modal identification study. The excitation force equals to the product of
Forced vibration test is carried out to establish the the moving mass and its acceleration, and the moving
FRF for the first lateral modes, and the modal mass includes the weight of the platform and the extra
10 Advances in Structural Engineering

The excitation frequency of the horizontal shaker

varies from 0.68 to 0.77 Hz. The steady-state accelera-
tion responses induced by different excitation frequen-
cies and their power spectral densities are shown in
Figure 18. There is a remarkable peak at each power
spectral density (PSD) plot, and the frequency of the
peak is the same as the excitation frequency rather than
the natural frequency of the footbridge. Moreover,
there are small peaks concentrated at integral multiples
of the excitation frequency. The higher order harmo-
nics present in the acceleration responses are due to
Figure 16. The horizontal vibration electrodynamic shaker. nonlinear excitation force output from the shaker
which is induced by harmonic distortion in sinusoidal
signal generator. Therefore, the following procedure is
mass placed on the platform. The total moving mass is used to obtain the FRFs: (1) the raw measurement data
about 310 kg and the stroke is set to be 5.2 cm. The are processed using band-pass filters with the fre-
driving frequency of the shaker is automatically con- quency range of 0.5–0.9 Hz and Hilbert transform, (2)
trolled by the computer, and it varies from 0.68 to the standard deviations of the processed data of integer
0.77 Hz after a first examination of ambient vibration number of cycles are estimated and the peak ampli-
data. The shaker is fixed on the middle of central span tudes of the accelerations are givenpffiffiby
ffi multiplying the
of the pedestrian deck where the displacement is the standard deviation by a factor of 2, and (3) the peak
largest for the first later mode. DH610 transducers accelerations are finally converted into displacement
suitable for measurement of the low-frequency accel- amplitudes. The obtained peak accelerations and dis-
eration are used to record the mid-span lateral accel- placement amplitudes under different excitations are
eration response of the pedestrian deck. Due to the summarized in Table 6.
curved pedestrian deck, it is worthy of noting that Figure 19 shows the FRF obtained from forced
both the shaker and DH610 lateral transducer are vibration experiments and its curve fitting based on
oriented in perpendicular to the roadway deck, which the receptance FRFs. The natural frequency and the
is consistent to the global coordinate system used in damping ratio for the first lateral modes are 0.723 Hz
the finite element model. Figure 17 shows data acquisi- and 0.63%, respectively. The modal mass based on
tion system and the transducers on the pedestrian unity-scaled mode shape is 519 ton. As a comparison,
deck. the modal mass computed from finite element model is

Table 6. Peak accelerations and displacements under different excitation frequencies.

Excitation frequency (Hz) 0.680 0.690 0.700 0.710 0.720 0.730 0.740 0.750 0.760 0.770
Peak acceleration (mm/s2) 4.496 5.703 8.670 14.64 45.13 25.05 17.33 11.13 9.96 7.94
Peak displacement (mm) 0.246 0.303 0.448 0.736 2.205 1.191 0.802 0.501 0.437 0.339

Figure 17. Data acquisition system and transducers on pedestrian deck.

Wen et al. 11

Figure 18. The steady-state responses and their power spectral densities.

modal mass for the lateral mode provides an improved

design of TMDs, allowing for an accurate frequency
tune of mass dampers.
During field forced vibration testing, the wind velo-
city is somewhere between 3 and 5 m/s, and the ratio
of noise to signal is about 8% and less than 10%,
which is calculated by the root-mean square of ambi-
ent vibration before forced vibration testing to that of
forced vibration with excitation frequency of 0.72 Hz.
It is believed that the effect of unmeasured wind exci-
tation is not significant.
Figure 19. Frequency response functions.
Comparison with ambient vibration and free
513 ton for the footbridge with TMDs as extra weight. vibration tests
It is seen that the experimentally estimated modal mass The identification results from forced vibration tests
from forced vibration tests is in good agreement with are compared with those obtained from free vibration
that from finite element prediction. The estimated tests and ambient vibration tests. The modal frequency
12 Advances in Structural Engineering

Table 7. Modal parameters for the first lateral mode estimated by different methods.

Frequency (Hz) Damping ratio (%) Modal mass (103 kg)

Analytical analysis 0.735 – 513

Free vibration 0.715 0.72 –
Ambient vibration 0.721 0.76 6 0.29 –
Forced vibration 0.723 0.63 519

The FRF-based method is used for a lively foot-

bridge to estimate the modal parameters. The esti-
mated modal mass allows for an accurate frequency
tune of mass dampers.

Acknowledgements
The collaboration of the Bureau of Municipal and Rural
Construction of Mianyang, in particular, Dr Xiao-xian
Kang, is of special mention.
Figure 20. Free decay vibration of the first lateral mode.
Declaration of Conflicting Interests
and damping ratio for the first lateral mode are esti- The author(s) declared no potential conflicts of interest with
mated by least square method from free vibration tests respect to the research, authorship, and/or publication of this
and are identified by SSI from ambient vibration tests. article.
The acceleration time history of free decay vibration is
shown in Figure 20. Funding
In those tests, the maximum vibration amplitudes The author(s) disclosed receipt of the following financial sup-
of the footbridge are variable. The dynamic character- port for the research, authorship, and/or publication of this
istics of the footbridge are of nonlinear relating to article: This study was supported by the National Science
vibration amplitude. As a consequence, the modal fre- Foundation of China (nos 51278189, 51422806, 91215302).
quencies and damping ratios estimated by these three
methods are of slight difference (Table 7). The modal References
mass estimated by forced vibration test is identical to Bayraktar A, Altunisxik A, Sevim B, et al. (2009) Modal test-
the modal mass predicted by analytical modal analysis. ing, finite-element model updating, and dynamic analysis
The result is helpful to predict human-induced vibra- of an arch type steel footbridge. Journal of Performance of
tion and design TMD optional parameters. Constructed Facilities 23(2): 81–89.
Brincker R, Zhang L and Anderson P (2000) Modal identifi-
cation from ambient response using frequency domain
decomposition. In: Proceedings of the international modal
Conclusion
analysis conference (IMAC), San Antonio, TX, 7–10
An FRF-based method is implemented to modal para- February.
meter identification of a curved double-level cable- Caetano E, Cunha Á, Magalhães F, et al. (2010a) Studies for
stayed bridge. The purpose is to identify the modal controlling human-induced vibration of the Pedro e Inês
parameters, especially the modal mass, for predicting footbridge, Portugal. Part 1: assessment of dynamic beha-
viour. Engineering Structures 32(4): 1069–1081.
and suppressing the potential human-induced vibra-
Caetano E, Cunha Á, Moutinho C, et al. (2010b) Studies for
tions. The results from the simulation analyses show controlling human-induced vibration of the Pedro e Inês
that (1) a small shaker mass leads to a more accurate footbridge, Portugal. Part 2: implementation of tuned
result; (2) choosing suitable frequency range and small mass dampers. Engineering Structures 32(4): 1082–1091.
step of excitation force is important to obtain the Dallard P, Fitzpatrick T, Flint A, et al. (2001) The London
modal parameters accurately; (3) the FRF-based millennium footbridge. Structural Engineer 79(22): 17–33.
method is of acceptable accuracy for modal identifica- Ewins DJ (2000) Modal Testing: Theory, Practice and Appli-
tion when the noise-to-signal ratio is less than 10%. cation. London: Research Studies Press.
Wen et al. 13

Gentile C and Gallino N (2008) Ambient vibration testing Ricciardelli F and Pizzimenti AD (2007) Lateral walking-
and structural evaluation of an historic suspension foot- induced forces on footbridge. Journal of Bridge Engineer-
bridge. Advances in Engineering Software 39(4): 356–366. ing 12(6): 77–88.
Ingolfsson ET, Georgakis CT and Jonsson J (2012) Pedes- Simiu E and Scanlan RH (1996), Wind Effects on Structures:
trian-induced lateral vibrations of footbridges: a literature Fundamentals and Applications to Design, Wiley, New York.
review. Engineering Structures 45(1): 21–52. Van Overschee P and De Moor B (1996) Subspace Identifica-
James GH, Carne TG and Lauffer J (1995) The natural tion for Linear Systems: Theory, Implementation and
excitation technique (NExT) for modal parameter Applications. Dordrecht: Kluwer Academic Publishers.
extraction from operating structures. International Jour- Wang Z-C, Zhang D, Ren W-X, et al. (2014) Structural
nal of Analytical and Experimental Modal Analysis modal parameter identification from forced vibration
10(4): 260–277. with analytical mode decomposition. Advances in Struc-
Magalhães F, Cunha Á, Caetano E, et al. (2010) Damping tural Engineering 14(8): 1129–1143.
estimation using free decays and ambient vibration Wen Q, Hua XG, Chen ZQ, et al. (2016) Control of human-
tests. Mechanical Systems and Signal Processing 24(5): induced vibrations of a curved cable-stayed bridge: design,
1274–1290. implementation and field validation. Journal of Bridge
Ni YQ, Fan KQ, Zheng G, et al. (2005) Automatic modal Engineering 21: 04016028.
identification and variability in measured modal vectors Wu W-H, Chen C-C and Liau J-A (2012) A multiple random
of a cable-stayed bridge. Structural Engineering and decrement method for modal parameter identification of
Mechanics 19(2): 123–139. stay cables based on ambient vibration signals. Advances
Racic V, Pavic A and Brownjohn JMW (2009) Experimental in Structural Engineering 15(6): 969–982.
identification and analytical modeling of human walking Živanović S, Pavic A and Reynolds P (2005) Vibration servi-
forces: a literature review. Journal of Sound and Vibration ceability of footbridges under human-induced excitation:
326(1–2): 1–49. a literature review. Journal of Sound and Vibration
Reynders E, Degrauwe D, De Roeck G, et al. (2010) 270(1–2): 1–74.
Combined experimental-operational modal testing of Živanović S, Pavic A and Reynolds P (2006) Modal testing
footbridges. Journal of Engineering Mechanics 136(6): and FE model tuning of a lively footbridge structure.
687–696. Engineering Structures 28(6): 857–868.