Vous êtes sur la page 1sur 10

Finite Element Model Updating of Canonica Bridge Using

Experimental Modal Data and Genetic Algorithm


Tao Liu, Dr.; Qilin Zhang, Prof. Dr.; Tobia Zordan, Prof. Dr.; Dept. of Structural Eng., Tongji Univ., Shanghai, China;
Bruno Briseghella; Prof., Dr., College of Civil Eng., Fuzhou Univ., Fuzhou, China. Contact: taoliu.liu@tongji.edu.cn
DOI: 10.2749/101686616X14480232444405

Abstract analysis. The outline of this study is as


follows: The second section presents
This paper calibrates the finite element model (FEM) of a tied-arch bridge a specifically developed program to
using an automatic model updating procedure developed within two softwares. perform FEM updating based on GA.
The former is used for sensitivity analysis and optimization analysis while the The next section shows the model of
latter is responsible for structural modeling and eigenvalue analysis. An inter- the investigated bridge before updat-
face connecting two softwares is first developed, as well as the procedure to ing. The fourth section describes the
calibrate the FEM. Based on the original drawing and topographic survey, the ambient vibration test of the bridge.
FEM of the investigated bridge is created. To obtain the experimental modal Calibration of the numerical model
parameters, eight global modes of the studied bridge are identified by ambient is conducted in the following section
vibration tests and the frequency domain decomposition technique. Then, the based on the developed program.
sensitivity analysis and FEM updating procedure is conducted and the opti- Finally, the updated numerical model
mal structural parameters are identified. Finally, the updated FEM is verified is verified by static tests in the sixth
by a series of static tests. Results show that the updated model could better section and conclusions are drawn in
represent the actual bridge. Therefore, the numerical model updated by the pro- the last section.
posed procedure could be further used for damage identification of the bridge
under service loads.
Finite Element Model Updating
Keywords: arch bridges; ambient vibration tests; FEM updating; sensitivity Based on Genetic Algorithm
analysis; genetic algorithm.
As iterative FEM updating is adopted
in this paper, the errors between the
experimental and numerical responses
Introduction on some codes and specifications that
should be minimized by varying the
are considered outdated today, struc-
Nowadays, with significant different parameters. In general, natu-
tural assessment of reinforced concrete
developments in the finite element ral frequencies and mode shapes can
arch bridges seems to be particularly
(FE) method and computational be obtained by conducting ambient
interesting.
resources, the finite element model vibration tests and used as experimen-
(FEM) is often used to simulate In general, FEM updating method- tal modal data to tune the FEM. The
the actual structure. However, the ologies can be classified into two procedure of FEM updating is car-
preliminary FEM generally cannot groups: direct methods2732 and itera- ried out by developing codes that per-
represent the true structure because tive methods.3336 The former updates form the iterative updating algorithm
of modeling uncertainties. Therefore, stiffness and mass matrices of the FEM interfaced with software38 as presented
in order to reflect the observed data directly without changing the physical in Fig. 1.
from actual structures more accurately, parameters, while the latter updates
At the beginning, the initial param-
FEM updating is often used to calibrate the physical parameters until the
eters are input in software.37 Based
the numerical model of structures,1,2 behavior of the FEM is close enough to
on these initial parameters, the FEM
especially different kinds of bridges, that of the actual structure. A detailed
is developed in software38 and then
such as footbridges,36 highway comparison between direct and itera-
eigenvalue analyses are performed.
bridges,79 cable-stayed bridges,1017 tive FEM updating methods is beyond
Finally, the computed natural frequen-
and arch bridges.1824 the scope of this study. Because of
cies and mode shapes are imported
their increased applications, iterative
It is important to know the dynamic to software37 again and are analyzed
methods are studied in this paper.
characteristics of arch bridges because further. In accordance with the
they can be sensitive to earthquakes Although several studies on FEM obtained experimental modal data
and, in particular, to asynchronous updating of arch bridges can be found, and modal assurance criterion (MAC)
motion.25,26 Since many reinforced few of them involve the application values, numerical modes are paired
concrete arch bridges in the first half of of automatic genetic algorithm in up to experimental ones in soft-
the 20th century were designed based combination with other software. This ware.37 Finally, a set of parameters
paper aims to present FEM updating are estimated using genetic algorithm
of a tied-arch bridge using experimen- to minimize the objective function
tal modal data and genetic algorithm residuals. It is found that the FEM
Peer-reviewed by international ex- (GA) developed within two softwares, updating procedure includes three
perts and accepted for publication
by SEI Editorial Board where the former37 is used for sensi- main contents: selection of variables to
tivity analysis and optimization analy- be updated, formulation of objective
Paper received: March 28, 2015 sis and the latter38 is responsible for function and application of optimiza-
Paper accepted: September 22, 2015 structural modeling and eigenvalue tion algorithm.

Structural Engineering International Nr. 1/2016 Scientific Paper 27


Canonica Bridge with a longitudinal slope of 2.5%, is walkways. The girder is 1.23 m deep so
formed by two cells and two extrem- that a good transparency of the deck
The investigated bridge, shown in ity girders in a rectangular hollow was attained from an aesthetic stand-
Fig. 2, was built around 1950 between section (see Fig. 3); the width of the point. The two lateral cells suspend the
the twin towns of Vaprio and Canonica girder is 12.69 m, which includes deck by means of inclined ties made
(about 40 km from Milan). The deck, two traffic lanes and two pedestrian of conventional reinforcement bars
immersed in a cast-in-place grout. The
deck is not prestressed and is standard
reinforced concrete. The arches are
made by reinforced concrete, having
a T-shaped cross-section. An overall
view of the bridge can be seen in the
original drawing, as presented in Fig. 4.
A three-dimensional FEM of the
bridge is developed in software38 and
presented in Fig. 5. The structural
properties of the bridge are as follows:

1. unit weight of the concrete is set to


be 24.0 kN/m3 and that of the steel
is assigned to be 78.0 kN/m3;
2. the Poissons ratio of the con-
crete is 0.20 and that of the steel is
assumed to be 0.25;
3. Youngs modulus of the concrete
is 34 GPa (concrete strength class
C35/45) and that of the steel is 210
GPa (the reinforcement is in the
form of bars);
4. the concrete slabs are simulated by
four-node shell elements;
5. two-node beam elements are used
to model the lateral box stringers
and the transverse cross-beams of
the deck; connections between the
slabs and the grid are simulated
using rigid links;
6. the arches and bracing members
are modeled as beam elements;
7. the arch footings are considered as
fixed;
8. the ties are modeled as truss
elements;
9. the effect of the abutments, as
restraints to the horizontal move-
ments of the deck, is taken into
account by introducing a series
of uniaxial springs oriented in
the longitudinal direction and
Fig. 1: Flowchart of FEM updating based on genetic algorithm attached to both ends of the deck
along each node; to ensure that
the first natural frequency of the
built FEM is close enough to the
identified frequency, the stiffness
of these longitudinal springs is
assumed to be 5e8 N/m in the pre-
liminary FEM.
10. in the FEM, because beam ele-
ments are connected with other
types of elements, such as shell or
solid, rigid elements are used here
to simulate the overlapping parts
between them so as to accurately
compute the deformation of the
Fig. 2: Outline of the Canonica bridge whole bridge.

28 Scientific Paper Structural Engineering International Nr. 1/2016


Fig. 3: Cross-section of the deck (Units: [m])

Fig. 4: Original drawings of the Canonica bridge (Units: [m])

Fig. 6: Uniaxial piezoelectric accelerometer


used in the test

Fig. 5: Three-dimensional FEM of the Canonica bridge the natural frequencies and mode
shapes of structures. In general, the
modal testing procedure consists of
Overall, the FEM of the Canonica the ambient acceleration-time histo- two steps: data acquisition and data
bridge contains a total of 3096 nodes, ries were recorded for 3600 s at an processing. According to different
1856 beam elements, 36 truss elements interval of 0.005 s, so that the well- sources of vibration, experimental
and 1896 shell elements. known condition39 about the length modal analysis (or forced vibration
of the time windows acquired (which testing) and operational modal analysis
should be 10002000 times the period (or ambient vibration testing) can be
Ambient Vibration Test of the structures fundamental mode) performed, of which the latter is the
is largely satisfied. Figure 7 shows the more commonly used method.
The experimental investigation was
layout of measurement points in the
based on ambient vibration tests. The In the operational modal analy-
test. Installation of the accelerometers
full-scale tests were conducted on the sis, the dynamic characteristics of a
and connections between them and
bridge using a 16-channel data acqui- structure can be extracted using such
the data acquisition system can be seen
sition system with 14 uniaxial piezo- techniques as peak picking (PP), fre-
in Fig. 8.
electric accelerometers (WR model quency domain decomposition (FDD)
731A), each with a battery power unit, Modal testing is an important part in and stochastic subspace identification
as shown in Fig. 6. For each channel, model identification, which determines (SSI). In this paper, FDD is used as the

Structural Engineering International Nr. 1/2016 Scientific Paper 29


where Ui is a unitary matrix holding
the singular vectors uij, Si is a diagonal
matrix holding the scalar singular val-
ues si and the superscript H denotes
the complex conjugate and transpose.
Near a peak corresponding to the rth
mode in the spectrum, the PSD matrix
approximates to a rank 1 matrix:
^ (w ) = s u uH
G wi wr (6)
YY i i i,1 i,1

and the first singular vector at the rth


resonance is an estimate of the rth
mode shape:

r = ui,1 (7)

The results of the modal identification


reveal eight global modes in the inves-
tigated frequency interval of 09 Hz.
In the following, the identified modes
Fig. 7: Plan and elevation of the bridge and measurement points in the test (Units: [m]) marked with B indicate the bending
modes of the deck and those with T
indicate the torsion modes of the deck.
GYY(w) = H*(iw)GXX(w)HT(iw) (2) For instance, B1 and T2 are represented
as the first bending mode and the sec-
where GYY() is the PSD function ond torsion mode of the bridges, and
of the output signal Y(t), GXX() is so on. The average of normalized sin-
the PSD function of the input sig- gular values is shown in Fig. 9, where
nal X(t), superscript * and T, respec- the resonant peaks are marked at 3.203
tively, denote the complex conjugate (B1), 3.438 (B2), 4.023 (T1), 4.805 (T2),
and transpose and is the angular 5.254 (B3), 5.957 (T3), 6.875 (B4) and
frequency. 7.813 Hz (T4). Accordingly, the iden-
After some mathematical manipula- tified mode shapes are presented in
tions, the output PSD can be reduced Fig. 10, where the dashed line is the
to a pole/residue form as follows: undeformed shape and the solid one is
the identified mode shape.
N
Ar Ar*
GYY ( ) = +
r =1 i r i r* Finite Element Model
Br B *
Updating of Canonica Bridge
+ + r

i r i r* Sensitivity Analysis
Fig. 8: Connection between accelerometers (3)
and data acquisition system where N is the number of modes, r
In general, sensitivity analyses convert
denotes the rth system pole and Ar is the
a physical system to a mathemati-
rth residual matrix of the output PSD.
modal identification method, which cal model and, based on this, try to
can be applied in the dynamic param- For a lightly damped system, the output understand how the model outputs
eter extraction of bridge structures PSD in Eq. (3) is an N by N Hermitian are affected by different sources
with good accuracy and computational matrix, which can be expressed in the of uncertainty in model inputs. In FEM
efficiency.20 The principles of the FDD following final form: updating, in order to choose the most
technique well known and are briefly d T d * * *T sensitive parameters, sensitivity analy-
described as follows: In the frequency GYY( ) = r r r + r r r * (4) sis needs to be performed.20
r =sub( ) i r i r
domain the relation between input In the common sensitivity analysis,
X(t) and output Y(t) of a system can where dr is the scaling factor of the only one variable is changed at a time.
be expressed as: rth mode and fr is the rth mode shape However, in this study, all considered
vector. At a certain frequency , only a input variables were varied simulta-
Y(iw) = H(iw)X(iw) (1) limited number of modes will contrib- neously. In this manner, global sen-
where X(i) and Y(i) are the Fourier ute significantly and this set of modes sitivities of the numerical outputs to
spectra of the input X(t) and output is denoted by sub(w). variation of different variables can be
Y(t), respectively, and H(i) is the Thus, by carrying out the singular value assessed. The Latin hypercube sam-
matrix of the frequency transfer func- decomposition (SVD) of the output pling was used to generate the multi-
tions of the system. Based on Eq. (1), PSD matrix at discrete frequencies variate samples stochastically, where a
the power spectral density (PSD) rela- w = wi, one can get: uniform distribution was assumed for
tion between X(t) and Y(t) can be different variables with the assigned
H
obtained as: GYY (wi) = UiSiU i (5) limits.

30 Scientific Paper Structural Engineering International Nr. 1/2016


is divided into three regions, in each of
which the concrete Youngs modulus
and density are assumed as constant.
Furthermore, different supports of the
deck slab and the lower parts of the
arches were assigned in this study, that
is, support 1 and 2, respectively. Both
of them include three different degrees
of freedom. Youngs modulus and den-
sity of the steel were also considered as
updating variables.
Based on Latin hypercube sampling
and the bound limits listed in Table 1,
500 multivariate sample sets were gen-
Fig. 9: Identified natural frequencies based on average of normalized singular values erated and used to compute the modal
parameters, such as natural frequency
and mode shape. To pair up numeri-
cal modes with experimental modes,
the MAC coefficient is often applied,
which varies between 0 and 1. The
MAC value of two orthogonal mode
shapes is equal to 0, while the MAC
value of two identical mode shapes
is equal to 1. In fact, it is rare to have
orthogonal or identical correlations
between numerical and experimental
mode shapes. Under normal condi-
tions, an MAC value greater than 0.8
could be considered a good agree-
ment, while an MAC value less than
0.4 could be considered a poor match.
In this study, the recognition of correct
mode assignment was not only related
to the identified modal vectors but
also restricted by the order of the cor-
responding modes. The MAC matrix
between the initial numerical mode
and the experimental mode is shown
in Fig. 12. For instance, although the
fourth experimental mode has the high-
est MAC value with the first numeri-
cal mode, as seen in the figure, the
difference of frequency between them
is very large. Considering the limited
number of measurement points used
in the experiment, this restriction is of
great importance to avoid the poten-
tial abuse of MAC in mode pairing.
Because of the advantages of mono-
tonic correlation, the results of sen-
sitivity analyses are exhibited using
Spearmans rank correlation coef-
ficient, as shown in Fig. 13, where
Fig. 10: The experimental mode shapes identified using ambient vibration test (Units: []) absolute correlation coefficients less
than 0.2 are excluded from the graphi-
cal representation. In this figure, the
The Pearsons correlation coefficient a coefficient that indicates a general Spearmans correlation coefficient is
or Spearmans correlation coefficient monotonic correlation between model presented for each parameter and each
can be used to indicate the sensitiv- inputs and outputs, which is widely response (frequency or MAC). The
ity of modal properties of the bridge used in different fields4143 and was higher the Spearman correlation coef-
to different parameters. However, also adopted in this study. ficient, the higher the modal response
the Pearsons correlation coefficient To carry out sensitivity analysis, (frequency and MAC) affected by the
represents a linear correlation between different structural parameters of the parameters. It can be observed that
model inputs and outputs. To overcome bridge are selected and are presented in parameters Ec1, c1, Ec2, c2, Ec3, Es, Kh1
this drawback, Spearman40 proposed Table 1. As shown in Fig. 11, the bridge and Kh2 have a significant influence on

Structural Engineering International Nr. 1/2016 Scientific Paper 31


Number Parameter Designation Lower bound Upper bound Unit the modal parameters of the FEM. In
2 general, the dynamic characteristics of
1 Ec1 Youngs modulus of 2.0e10 4.5e10 N/m
concrete in region 1
the Canonica bridge are dominated by
the concrete elastic modulus in regions
2 rc1 Density concrete in 2.2e3 2.7e3 kg/m3 13 and the concrete density in regions
region 1
1 and 2. Also, the steel elastic modu-
3 Ec2 Youngs modulus of 2.0e10 4.5e10 N/m2 lus affects the vibration properties of
concrete in region 2 this bridge to some extent. In addition,
4 rc2 Density concrete in 2.2e3 2.7e3 kg/m3 the support of the deck and the arch
region 2 is dominated by the horizontal stiff-
5 Ec3 Youngs modulus of 2.0e10 4.5e10 N/m2 ness of the foundation. Since the other
concrete in region 3 investigated parameters, such as rc3, rs,
6 rc3 Density of concrete 2.2e3 2.7e3 kg/m3 Kv1, K1, Kv2 and K2, have little influ-
in region 3 ence on the modal properties of the
7 Youngs modulus of 2.0e11 2.3e11 N/m2
bridge, they will not be considered in
Es
steel the following FEM updating procedure.
8 rs Density of steel 7.7e3 8.0e3 kg/m3
9 Horizontal stiffness 1e8 1e9 N/m
Optimization
Kh1
of the supports 1 As seen in Fig. 13, parameters Ec1,
10 Kv1 Vertical stiffness of 1e8 1e9 N/m c1, Ec2, c2, Ec3, Es, Kh1 and Kh2,
the supports 1 have a significant sensitivity to the
11 Kq 1 Rotational stiffness 1e8 1e9 N/m modal properties of the bridge, and
of the supports 1 therefore they will be incorporated in
12 Kh2 Horizontal stiffness 1e8 1e9 N/m
the following optimization procedure.
of the supports 2 The optimization of the bridge model
is performed using the program devel-
13 Kv2 Vertical stiffness of 1e8 1e9 N/m
oped in the section. The generation
the supports 2
number is 50 in generic algorithm, with
14 Kq 2 Rotational stiffness 1e8 1e9 N/m 50 individuals in each generation. The
of the supports 2 crossover rate and the mutation rate
Table 1: Selection of the input parameters and their bound limits are set to be 0.7 and 0.08, respectively.
The roulette wheel selection method
was applied in this study. The objec-
tive function considers both the natu-
ral frequencies and the MAC values,
which can be expressed as:

N
| fi exp fi num |
f = a
i =1 fi exp

( )
N
+ b 1 MAC (iexp , inum ) (8)
i =1
Fig. 11: Different segments of the bridge used in sensitivity analysis
where fiexp and finum are the experimen-
tal and numerical natural frequencies
of mode i, respectively; fiexp and finum
are the experimental and numerical
mode shapes of mode i, respectively;
a and b are the weight factors of the
objective function and are assigned
to be 1.0. N is the total number of the
1.0 considered modes.
0.8
In this section, eight structural
0.6
MAC

parameters and 16 modal properties


0.4 are involved in the optimization process
20 to minimize the objective function.
0.2 19
18
17 The optimal parameters are then
0 16
15
13
14 determined using genetic algorithm.
1 12 Figure 14 shows the ratios of the
11
2 10 de
Ex
pe 3 4 8
9
ica l mo difference between optimal parameters
rim 7 mer and their lower bounds to that between
en 5 6 4
5
6 Nu
tal the upper bounds and lower bounds.
mo 7 8 2
3
de 1 Therefore, if the ratio is 0%, the optimal
parameter equals the lower bound,
Fig. 12: MAC matrix between the initial numerical mode and experimental mode (Units: []) while a ratio of 100% indicates that the

32 Scientific Paper Structural Engineering International Nr. 1/2016


optimal parameter coincides with the than Ec2 and Ec3. Table 2 presents the reference. As can be observed in this
upper bound. As can be observed, all comparison of modal data from the figure, all the errors are less than 5%
the optimal parameters are in between FEMs and the experimental results, and the maximum error is 4.95% for
the lower and the upper limits. It can be before and after calibration. mode T2. In addition, the average
expected that reciprocating traffic loads relative error decreases from 3.66 to
during the life cycle of the bridge will Presented in Fig. 15 are the relative 1.17% after calibration.
have more influence on the deck than errors between the numerical and
Figure 16 presents the MAC values
the arch. In other words, after long-term experimental frequencies, which
computed before and after calibration
fatigue damage, Ec1 will become smaller take the experimental frequencies as
of the FEM. It can be noted that the
MAC value is significantly improved
after updating the FEM, especially for
the first two modes. The average MAC
value changes from 0.7305 to 0.9750
after model updating. In addition, the
lowest value of MAC after calibration
is 0.9484, indicating that the updated
numerical model has a good agree-
ment with the true bridge.

Static Test
The verification of the updated FEM
is performed through static tests, as
shown in Fig. 17. The bridge is loaded
with a maximum of eight trucks, four
each of 350 kN (trucks numbers
2, 3, 6 and 7) and four each of 200
kN (trucks numbers 1, 4, 5 and 8)
weight. The trucks are positioned in
the middle of the bridge in order to
Fig. 13: Results of sensitivity analyses represented by Spearmans correlation coefficient simulate the maximum design loads
(Units: []) prescribed for the deck, as shown
in Fig. 18. Eight load sequences are
considered in the tests, including one
typical symmetric and seven asym-
metric conditions. The GS04 tester
and Linear variable differential
transformer (LVDT) displacement
sensor are used in the static test, as
presented in Fig. 19. The sensitivity
and linearity of the displacement
sensor are 0.002 mm and 99.6%,
respectively.
During the load tests, the displace-
ments in the middle of the bridge
are measured and compared with the
theoretical results. Table 3 shows the
comparison between numerical (ynum)
Fig. 14: The optimal structural parameters obtained using genetic algorithm and experimental (yexp) results based
on the updated numerical model.
As can be observed, the displacement
Experimental Initial model Updated model results of static analyses computed
Mode fexp (Hz) fnum (Hz) RE (%) MAC fnum (Hz) RE (%) MAC using the updated FEM show a good
B1 3.203 3.0839 3.72 0.0169 3.2030 0.00 0.9484 agreement with the results of static
B2 3.438 3.1576 8.16 0.0628 3.4382 0.01 0.9861
tests in all load sequences, and the
maximum relative deviation is less
T1 4.023 3.7205 7.52 0.9086 3.9655 1.43 0.9607 than 5%. Hence, the numerical model
T2 4.805 4.8621 1.19 0.9937 5.0430 4.95 0.9919 after being calibrated using the pro-
B3 5.254 5.1646 1.70 0.9800 5.2517 0.04 0.9870 posed updating approach represents a
T3 5.957 6.0264 1.17 0.9927 6.0220 1.09 0.9932 true approximation of the real bridge
and can be further used to evaluate the
B4 6.875 7.0241 2.17 0.9107 6.8689 0.09 0.9536
dynamic performance of the bridge
T4 7.813 8.0965 3.63 0.9787 7.9478 1.73 0.9787 and the comfort of the passengers
Table 2: Comparison of modal data from the FEMs and the experimental results under the service loads.

Structural Engineering International Nr. 1/2016 Scientific Paper 33


Conclusions
This paper presents the FEM updating
of a reinforced concrete tied-arch
bridge using operational modal
analysis and genetic algorithm. The
FEM updating procedure was first
developed within two softwares. Based
on the original drawing and prelimi-
nary estimation of the structural prop-
erties, the FEM of the bridge was first
built. Then, eight global modes of the
bridge were identified by ambient
vibration tests, including four bending
Fig. 15: Relative errors between numerical and experimental frequencies (Units: []) modes and four torsion modes. After
this, eight updating variables were
selected by a multivariate sensitivity
analysis from 14 uncertain structural
parameters. Finally, calibration of the
FEM was performed and verified
using a series of static tests.
Based on the results obtained in this
study, the following conclusions can be
drawn:
1. From sensitivity analysis, it is found
that global modes of the bridge are
significantly affected by the con-
crete properties involved in regions
1 (the deck) and 2 (the arches above
the deck) as well as the Youngs
modulus of the steel. In addition,
Fig. 16: MAC values before and after FEM calibration (Units: []) the Youngs modulus of concrete in
region 3 (the arches below the deck)
and the horizontal stiffness of both
supports also influence the global
modes to some extent.
2. The optimal structural parameters
obtained are in between the lower
and the upper limits specified in this
study. It is found that the modulus of
deformability of the deck is much less
than that of the arch, which is also the
main reason that the first measured
mode is anti-symmetric while the
second one is symmetric. From an
engineering standpoint, the modu-
lus of deformability of the arches is
seldom affected under service loads
because of the compression-domi-
nant internal force, but the stiffness
of the deck is significantly influenced
Fig. 17: The static test by the potential cracked section
caused by reciprocate traffic loads
during the life cycle of the bridge.
3. After calibration, the frequency
errors between the numerical and
experimental results are less than
5% and the average error of natural
frequencies decreases from 3.66 to
1.17%. Regarding the MAC values
after calibration, they are found to
be always above 0.9484. In addition,
the average MAC value increases
from 0.7305 to 0.9750 after FEM
Fig. 18: Scheme of the position of the tracks during the static loads updating.

34 Scientific Paper Structural Engineering International Nr. 1/2016


(a) (b) changes in cable forces and model updating. J.
Struct. Eng. 2009; 135(9): 10931106.
[11] Bayraktar A, Altunisi AC, Sevim B, Trker
T. Modal testing, finite-element model updat-
ing, and dynamic analysis of an arch type steel
footbridge. J. Perform. Constr. Facil. 2009; 23(2):
8189.
[12] Park W, Kim HK, Jongchil P. Finite element
model updating for a cable-stayed bridge using
manual tuning and sensitivity-based optimiza-
tion. Struct. Eng. Int. 2012; 22(1): 1419.
[13] Park W, Park JY, Kim HK. Candidate model
construction of a cable-stayed bridge using
parameterised sensitivity-based finite element
Fig. 19: Measurement equipments used in the static test: (a) GS04 tester and (b) LVDT model updating. Struct. Infrastruct. Eng. 2015;
11(9): 11631177.
displacement sensor
[14] Shabbir F, Omenzetter P. Particle swarm
optimization with sequential niche technique
Load sequences Upstream Downstream for dynamic finite element model updating.
yexp (m) ynum (m) RE (%) yexp (m) ynum (m) RE (%) Comput. Aided Civ. Infrastruct. Eng. 2015;
30(5SI): 359375.
(m) (m) (%) (m) (m) (%)
[15] Briseghella B, Busatta F, Gentile C, Zordan
8 0.0006 0.00062 3.33 0.0004 0.00042 5.00 T. Finite Element Modelling of a curved deck
8+7 0.0027 0.00266 1.48 0.0018 0.00178 1.11 cable stayed bridge in Venice from operational
8+7+6 0.0047 0.00479 1.91 0.0030 0.00299 0.33 modal analysis. Proceedings of the 17th Congress
of IABSE, Chicago, US, 2008.
8+7+6+5 0.0054 0.00555 2.78 0.0033 0.00338 2.42
[16] Briseghella B, Siviero E, Lan C, Mazzarolo
8+7+6+5+4 0.0058 0.00595 2.59 0.0040 0.00408 2.00 E, Zordan T. Safety monitoring of the cable
8+7+6+5+4+3 0.0070 0.00721 3.00 0.0061 0.00624 2.30 stayed bridge in the Commercial Harbor
of Venice, Italy, bridge maintenance, safety,
8+7+6+5+4+3+2 0.0083 0.00851 2.53 0.0081 0.00843 4.07
management and life-cycle optimization.
8+7+6+5+4+3+ 0.0087 0.00901 3.56 0.0087 0.00910 4.60 Proceedings of the 5th International Conference
2+1 on Bridge Maintenance, Safety and Management,
Table 3: Comparison between experimental and numerical displacements measured in the Philadelphia, PA, US, 2010, pp. 19351942.
middle of the bridge [17] Briseghella B, Chen A, Li X, Zordan T,
Lan C, Mazzarolo E. Analysis on applica-
bility of health monitoring techniques on a
4. Verified by a series of static tests, bridge: the complete process. J. Sound. Vib. 2007; curved cable stayed bridge. Proceedings of
the responses of the updated FEM, 301(12): 126145. the 6th International Conference on Bridge
specifically the vertical deformation, [4] Moaveni B, Behmanesh I. Effects of changing Maintenance, Safety and Management, Stresa,
are in a fairly good agreement with ambient temperature on finite element model Lake Maggiore, Italy, 2012, pp. 26172624.
the experimental results. Hence, the updating of the Dowling Hall Footbridge. Eng. [18] Zordan T, Briseghella B, Liu T. Finite ele-
calibrated FEM could be further Struct. 2012; 43: 5868. ment model updating of a tied-arch bridge using
used for damage identification of [5] Behmanesh I, Moaveni B. Probabilistic iden- Douglas-Reid method and Rosenbrock optimi-
the bridge under transient traffic tification of simulated damage on the Dowling zation algorithm. J. Traffic Transp. Eng.(English
Hall footbridge through Bayesian finite element Edition) 2014; 1(4): 280292.
loads.
model updating. Struct. Cont. Heal. Monit. 2015;
[19] Gentile C. Modal and structural identifica-
22(3): 463483.
tion of an R.C. arch bridge. J. Struct. Eng. Mech.
Acknowledgements [6] Cismasiu C, Narciso AC, Amarante dos Santos 2006; 22(1): 5370.
F. Experimental dynamic characterization and
The authors would like to thank Prof. [20] Ribeiro D, Calada R, Delgado R, Brehm
finite-element updating of a footbridge structure.
Carmelo Gentile and the technical personnel M, Zabel V. Finite element model updating of a
J. Perform. Constr. Facil. 2015; 29(4): 04014116,
(Antico M. and Cucchi M.) of the VIBLAB bowstring-arch railway bridge based on experi-
DOI: 10.1061/ (ASCE)CF. 1943-5509.0000615.
Laboratory of Vibrations and Dynamic mental modal parameters. Eng. Struct. 2012; 40:
Monitoring of Structures, Polytechnic of [7] Bayraktar A, Altunisik AC, Sevim B. Finite 413435.
Milan, for the valuable help in conducting the element model updating of Komurhan highway
bridge based on experimental measurements. [21] Schlune H, Plos M, Gylltoft K. Improved
field tests.
Smart. Struct. Syst. 2010; 6(4): 373388. bridge evaluation through finite element model
updating using static and dynamic measure-
References [8] Altunisik AC, Bayraktar A. Finite element ments. Eng. Struct. 2009; 31(7): 14771485.
model updating effect on the structural behavior
[1] Simoen E, De Roeck G, Lombaert G. Dealing of long-span concrete highway bridges. Comput. [22] Xie KZ, Chen, GQ, Qin LQ. An updat-
with uncertainty in model updating for dam- Concr. 2014; 14(6): 745765. ing model for CFST arch ribs based on the
age assessment: a review. J. Mech. Syst. Signal improved ant colony algorithm. J. Optoelectron.
Process. 2015; 5657: 123149. [9] Mosavi AA, Sedarat H, OConnor SM, Adv. Mater. 2014; 16(910): 11531158.
Emami-Naeinia A, Lynch J. Calibrating a high-
[2] Marwala T. Finite-Element-Model Updating fidelity finite element model of a highway bridge [23] Liu Y, Li Y, Wang DJ, Zhang SY. Model
Using Computational Intelligence Techniques: using a multi-variable sensitivity-based optimi- updating of complex structures using the combi-
Applications to Structural Dynamics. Springer, sation approach. Struct. Infrastruct. Eng. 2014; nation of component mode synthesis and kriging
2010. 10(5): 627642. predictor. Sci. World J. 2014; 2014: 113.
[3] Zivanovic S, Pavic A, Reynolds P. Finite ele- [10] Hua XG, Ni YQ, Chen ZQ. Structural dam- [24] Li X, Zhang DY, Yan WM, Xie WC.
ment modelling and updating of a lively foot- age detection of cable-stayed bridges using Effects of model updating on the estimation of

Structural Engineering International Nr. 1/2016 Scientific Paper 35


stochastic seismic response of a concrete-filled model updating in structural dynamics. J. Mech. [37] MathWorks. MATLAB 7.8.0.347 [Computer
steel tubular arch bridge. Struct. Infrastruct. Eng. Syst. Signal Process. 2012; 29: 262283. software]. MathWorks: Natick, MA.
2014; 10(12): 16201637.
[31] Modak SV. Direct matrix updating of vibro- [38] Mazzoni S, McKenna F, Scott MH, Fenves
[25] Nuti C, Vanzi I. Influence of earthquake acoustic finite element models using modal test GL. OpenSees Command Language Manual.
spatial variability on differential soil displace- data. AIAA J. 2014; 52(7): 13861392. Pacific Earthquake Engineering Research
ments and SDF system response. Earthq. Eng. Center, University of California: Berkeley, CA,
[32] Zhang JF, Ouyang HJ, Yang J. A new direct
Struct. Dyn. 2005; 34(11): 13531374. 2007.
method for updating mass and stiffness matri-
[26] Carnevale L, Lavorato D, Nuti C, Vanzi ces with no spillover. In Vibration Engineering [39] Cantieni R. Experimental methods used
I. Response of continuous deck bridges to and Technology of Machinery, Mechanisms and in system identification of civil engineering
non-synchronous seismic motion. Proceedings Machine Science, vol. 23, 2014, 609618. structures. Proceedings of the 1st International
of Sustainable Development Strategies for Operational Modal Analysis Conference
[33] Boscato G, Russo S, Ceravolo R, Fragonara
Constructions in Europe and China Conference, (IOMAC), 2005; 249260.
LZ. Global sensitivity-based model updating-
Roma, Italy, 2010.
for heritage structures. Comput. Aided Civ. [40] Spearman C. The proof and measure-
[27] Yang YB, Chen YJ, Hsu TW. Direct updating Infrastruct. Eng. 2015; 30: 620635. ment of association between two things. Int. J.
method for structural models based on orthogo- Epidemiolog. 2010; 39(5): 11371150.
[34] Osmancikli G, Bayraktar A, Trker T, Uak
nality constraints. J. Mech. Syst. Signal Process.
S, Mosallam A. Finite element model calibration [41] Mottershead JE, Mares C, James S, Friswell
2009; 16(5): 390401.
of precast structures using ambient vibrations. MI Stochastic model updating: part 2-application
[28] Yuan YX, Guo YQ A direct updating Constr. Build. Mater. 2015; 93: 1021. to a set of physical structures. J. Mech. Syst.
method for damped gyroscopic systems using Signal Process. 2006; 20(8): 21712185.
[35] Wan HP, Ren WX. Parameter selection
measured modal data. Appl. Math. Model. 2010;
in finite-element-model updating by global [42] Mares C, Mottershead JE, Friswell MI.
34(6): 14501457.
sensitivity analysis using Gaussian process Stochastic model updating: part 1-theory and
[29] Lei Y, Wang HF, Shen WA. Update the finite metamodel. J. Struct. Eng. 2015; 141(6): 04014164. simulated example. J. Mech. Syst. Signal Process.
element model of Canton Tower based on direct DOI:10.1061/(ASCE)ST.1943-541X.0001108. 2006; 20(7): 16741695.
matrix updating with incomplete modal data.
[36] Wan HP, Ren WX. A residual-based [43] Brehm M, Zabel V, Bucher C. An automatic
Smart Struct. Syst. 2012; 10(45): 471483.
Gaussian process model framework for finite mode pairing strategy using an enhanced modal
[30] Jacquelin E, Adhikari S, Friswell MI. A sec- element model updating. Comput. Struct. 2015; assurance criterion based on modal strain
ond-moment approach for direct probabilistic 156: 149159. energies. J. Sound. Vib. 2010; 329(25): 53755392.

More on SED 14:

Safety and Sustainability -


the Structural Engineers Role
Integration Concept of Sustainable Engineering
A Sustainable Approach to Structural Design
Sustainability and Cultural Heritage Buildings
Measuring Sustainability and Lif-Cycle Assessment
Asset Management
Sustainability and Bridges
Stuctural Reassessment for Lifetime Extension
Strustainability through Disaster Risk Reduction
Green Materials for Concrete Production

Members: Download SED 14 (free) in the online


Members Area, buy hardcover copy at the onlineshop.

www.iabse.org/onlineshop

Hardcover e-book (ePDF)


Members: CHF 40 free
Non-Members: CHF 70 CHF 20

36 Scientific Paper Structural Engineering International Nr. 1/2016

Vous aimerez peut-être aussi