Engineering Structures 32 (2010) 2455–2465

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Finite element model updating in structural dynamics by using the response

surface method
Wei-Xin Ren ∗ , Hua-Bing Chen
Department of Civil Engineering, Central South University, Changsha, 410075, China
National Engineering Laboratory for High Speed Railway Construction, Changsha, 410075, China

article info abstract

Article history: Fast-running response surface models that approximate multivariate input/output relationships of time-
Received 3 February 2009 consuming physical-based computer models enable effective finite element (FE) model updating analyses.
Received in revised form In this paper, a response surface-based FE model updating procedure for civil engineering structures in
6 April 2010
structural dynamics is presented. The key issues to implement such a model updating are discussed such
Accepted 7 April 2010
Available online 4 May 2010
as sampling with design of experiments, selecting the significant updating parameters and constructing
a quadratic polynomial response surface. The objective function is formed by the residuals between
Keywords:
analytical and measured natural frequencies. Single-objective optimization with equal weights of natural
Response surface frequency residual of each mode is used for optimization computation. The proposed procedure is
Finite element illustrated by a simulated simply supported beam and a full-size precast continuous box girder bridge
Model updating tested under operational vibration conditions. The results have been compared with those obtained from
Optimization the traditional sensitivity-based FE model updating method. The real application to a full-size bridge has
Design of experiment demonstrated that the FE model updating process is efficient and converges fast with the response surface
Regression analysis to replace the original FE model. With the response surface at hand, an optimization problem is formulated
explicitly. Hence, no FE calculation is required in each optimization iteration. The response surface-based
FE model updating can be easily implemented in practice with available commercial FE analysis packages.
© 2010 Elsevier Ltd. All rights reserved.

1. Introduction model updating is the process of using experimental results to re-

Nowadays the finite element (FE) method has become an
important and practical numerical analysis tool. It is commonly
used in almost all areas of engineering analysis. However, the
FE model of a structure is normally constructed on the basis of
highly idealized engineering blueprints and designs that may not
truly represent all the aspects of an actual structure. As a result,
the analytical predictions from a FE model often differ from the
results of a real structure. These discrepancies originate from the
uncertainties in simplifying assumptions of structural geometry,
materials as well as inaccurate boundary conditions. It is often
required to update or calibrate the uncertain parameters of a FE
model that leads to the better predictions of the responses of an
actual structural.
Finite element model updating is such a procedure that mod-
ifies or updates the uncertainty parameters in the initial finite
element model based on the experimental results so that a more
realistic or refined model can be achieved [1]. In other words, FE
∗ Corresponding author at: Department of Civil Engineering, Central South
University, Changsha, 410075, China. Tel.: +86 731 82654349; fax: +86 731
85571736.
E-mail addresses: renwx@mail.csu.edu.cn, renwx@yahoo.com (W.-X. Ren).
0141-0296/$ – see front matter © 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2010.04.019
fine a mathematical model of a physical structure. Basically, FE
model updating is an inverse problem to identify or correct the
uncertain parameters of FE models. It is usually posed as an op-
timization problem. Setting-up of an objective function, selecting
updating parameters and using robust optimization algorithm are
the three crucial steps in FE model updating. In a model updating
process, not only the satisfactory correlation is required between
analytical and experimental results, but also the updated param-
eters should preserve the physical significance. The updated FE
models are used in many applications for civil engineering struc-
tures such as damage detection, health monitoring, structural con-
trol, structural evaluation and assessment.
Finite element model updating is a topic of significant interest
in the field of structural dynamics. A number of FE model updating
methods in structural dynamics have been proposed. The direct
updating methods compute a closed-form solution for the global
stiffness and/or mass matrices using the structural equations
of motion and the orthogonality equations. These non-iterative
methods that directly update the elements of stiffness and mass
matrices are one-step procedures [2,3]. The resulting updated
matrices reproduce the measured structural modal properties well
but do not generally maintain structural connectivity and the
corrections suggested are not always physically meaningful.
2456 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465
The iterative parameter updating methods involve using the
sensitivity of the parameters to find their changes. The sensitivity-
based FE model updating methods are often posed as optimization
problems. These methods set the errors of the structural response
features between analytical and experimental results as an
objective function and try to minimize the objective function by
making changes to the pre-selected set of physical parameters of
the FE model. Link [4] gave a clear overview of the sensitivity-
based updating methods in structural dynamics. It is noted that
the mathematical model used in the model updating is usually
ill posed and the special attention is required for an accurate
solution [5]. Jaishi and Ren [6–8] used either single-objective or
multi-objective optimization technique to update the FE models of
civil engineering structures in structural dynamics. However, the
sensitivity-based method involving in the determination of local
gradients at points may cause not only computational intensive,
but also convergence difficulty.
If the structure of interest is represented by, e.g. a large finite
element model, the large number of computations involved can
rule out many approaches due to the expense of carrying out
many runs. For such a large FE model where so many elements
are involved, there are huge of both geometric and physical
possible parameters to be updated. In addition, there are now
many commercial finite element analysis packages available such
as ANSYS, ABAQUS and SAP2000 et al.. The structural FE models
are often constructed by using these packages. In all the iterative
parameter updating methods, each iteration needs to go back
to run the finite element analysis package with any parameter
updated, which limits the popular applications of structural FE
model updating in practice.
One way of circumnavigating the time-consuming and FE
analysis package-related problems during the sensitivity-based
model updating is to replace the FE model by an approximate
surrogate/replacement meta-model that is fast-running and less
parameters involved [9]. Such a meta-model is easier to compute
with, because it is controlled only by a few explanatory variables.
The FE model updating is carried out on the meta-model instead
of the analytical FE model. Response surface is one of the
commonly used meta-models. Response surface methodology is
originally an experimental design approach for selecting design
parameters for experiments with the objective of optimizing
some function of a response [10–12]. It provides a mechanism
for guiding experimentation in search of optimal settings for
design parameters or optimal values of unknown response. Many
additional applications (largely a consequence of the increased
use of computational analyses) have broadened the range of
application of response surface methods in the statistical and
engineering literature. Recent literature has addressed more
flexible functional forms for modeling the response, new methods
of response surface construction [13], alternate approaches to
updating the surface estimate [14], new sampling methods [15],
etc.
In many fields of engineering, the term ‘‘response surface’’
is used synonymously with ‘‘meta-model’’ or ‘‘surrogate model’’,
which refer to any relatively simple mathematical relationship
between parameters and a response, often based on limited
data [16]. In the case of structural finite element model updating,
once the response surface of a structure has been constructed,
updating the model is reduced to the task of finding the
smallest value on the response surface. The parameter values that
correspond to this smallest value are those that are used to update
the model. Recently, the response surface that is represented by
a simple least-squares multinomial model has been adopted in
structural FE model updating, verification and validation [17–20].
The response surface method for damage detection and
reliability analysis is not quite new [21,22]. However, the response
surface method for structural finite element model updating is
somewhat new, especially with the civil engineering communities.
This paper is intended to present a response surface-based
finite element model updating procedure in structural dynamics
and to take advantages of response surfaces for the FE model
updating of civil engineering structures in practice. Its purpose
is to estimate the values of structural parameters (moment of
inertia, cross-sectional area and modulus of elasticity) based
on the measured response quantities (natural frequencies). The
proposed procedure is based on constructing quadratic response
surfaces. Those surfaces represent an estimate for the relation
between the unknown parameters of the finite element model
and response quantities of interest. With the response surface at
hand, an optimization problem, whose solution is the estimate for
the values of the structural parameters, is formulated explicitly.
Hence, no further finite element simulations are required. The
objective function is formed by the residuals between analytical
and measured natural frequencies. Single-objective optimization
with equal weights of each natural frequency is implemented for
optimization computation. The presented procedure is illustrated
by a simulated simply supported beam and a full-size precast
continuous box girder bridge tested under operational vibration
conditions. The results have been compared with those obtained
from the traditional sensitivity-based FE model updating method.
The real application to a full-size bridge has demonstrated that the
response surface-based FE model updating procedure is simple and
fast so that it can be easily implemented in practice.
2. Response surface-based finite element model updating
Response surface-based finite element model updating is an
approach to achieve the global approximations of the structural
response feature objectives and constrains based on functional
evaluations at various points in the design space. It often involves
experimental strategies, mathematical methods, probability and
statistical inference that enable an experimenter to make efficient
empirical exploration of the structure of interest. The flowchart of
response surface-based finite element model updating is shown in
Fig. 1. The main steps include the following.
• The selection of the structural parameters and the definition of
a number of ‘‘level’’ for each selected parameters by using the
design of experiments (DOE) techniques.
• In design space, the response features are calculated by carrying
out finite element analysis (FEA).
• Performing the final regression followed by a regression error
analysis to create the response surface model of the structure.
• The response features of the structure are measured and
corresponding objective functions (feature residuals) to be
minimized are constructed.
• The finite element model updating (iteration) is carried
out within the established response surface model. Updated
parameters are obtained and transferred to the original finite
element model.
2.1. Sampling and parameter selection
To create a response surface that will serve as a surrogate for
the FE simulation model, the basic process is one of calculating
predicted values of the response features at various sample points
in the parameter space by performing a experiment at each of
those points. A number of feature values from the experiment
ran across the parameter domain are fit with a response surface.
The key is to select the parameters carefully, to minimize the
number of dimensions in the parameter space, and then to select
 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465
Fig. 1. Flowchart of response surface-based FE model updating.
the combinations of parameter values where the experiment is
performed.
The term experiment herein refers to either physical experi-
ments or computer experiments. The planning of experimentation
is further referred to design of experiments (DOE). The selection of
sample points is related to the accuracy and cost of a response sur-
face to be constructed. Less sample points may reduce the surface
accuracy, while more sample points may improve the surface accu-
racy but increase the work load. In the real application, the sample
points mainly depend on the problem to be solved, the response
feature values of interest and the selected method of DOE.
DOE plays an important role in constructing a response surface.
Two efficient DOE methods are commonly used. They are central
composite design (CCD) method and D-optimal design method.
There are several other DOE methods that can be used in
constructing a response surface such as box-behnken design, Latin
Square design, fractional factor design [12]. For the purpose of
structural finite element model updating in structural dynamics,
Guo and Zhang [18] found that both CCD and D-optimal design
methods can achieve almost the same accuracy in the creation of
polynomial surfaces. In the current study, the CCD method in DOE
is used in the paper as it is simple in constructing the response
surfaces of a polynomial type.
Central composite design uses the orthogonal table to perform
the experimentation to determine the sample points of selected
parameters. It contains a fractional factorial design 2k (levels
are ±1 and k is quantity of factors) with central points that
are augmented with a group of 2k star points that allow for
the estimation of curvature. 2k star points are (±α, 0, . . . , 0),
(0, ±α, . . . , 0), . . ., (0, 0, . . . , ±α ). For the purpose of constructing
high-order surfaces, another 2k star points (±α1 , 0, . . . , 0),
(0, ±α1 , . . . , 0), . . ., (0, 0, . . . , ±α1 ) are appended. The precise
values of α and α1 rely on certain properties such as orthogonality
and rotatability desired for the design and on the number of factors
involved [12].
2457
The selection of updating parameters to calculate the response
features of a structure is an important issue in FE model updating.
The structural parameters selected for updating should be able to
clarify the ambiguity of the model, and in that case it is necessary
for the model response to be sensitive to these parameters.
The problem always arises: How many parameters should be
selected and which parameters from many possible parameters
are used in the FE model updating? If too many parameters
are included in the FE model updating in structural dynamics,
the optimized problem may appear ill-conditioned because only
limited vibration modes can be effectively identified from the
field vibration measurements. The parameter selection requires
a considerable physical insight into the target structure, and
trial-and-error approaches are often used with different set of
selected parameters for complicated structures. In the frame of
response surface-based FE model updating, the selected updating
parameters can be evaluated from the sampled data by performing
a parameter effect analysis (hypothesis testing) based on statistical
variance (square of the standard deviation) analysis.
Compared with tests on means for a hypothesis testing, tests
on variances are rather sensitive to the normality assumption.
Variance analysis formulates all individual square of deviations of
sample data. The basic idea of variance analysis is to divide the total
square deviation of sampling features into two parts: SA —a square
of deviation caused by design parameter A (system deviation) and
Se —a square of deviation caused by the experiment.
The F-test method [12] can be used to carry out the hypothesis
testing to check the significance of parameter A:
SA /fA
FA = ∼ F (fA , fe ) (1)
Se /fe
where fA and fe are the degrees of freedom of SA and Se respectively.
For a given significance level p, if FA ≥ F1−p (fa , fe ), the effect of
parameter A is significant.
2.2. Response surface regression
The family of response surface forms selected to represent
a response can have a substantial impact on the results of an
analysis. The selected response surface form should be capable
of attaining surfaces that meet specific smoothness requirements
of an application. There is often a balance between assumptions
and data requirements. Different surface families may be preferred
for different applications. In the case of structural finite element
model updating in structural dynamics, polynomials are popular
forms representing a response surface because the calculations
are simple and the resulting function is closed-form algebraic
expression. For example, a quadratic polynomial response surface
has the form:
k k X
k
X X
y = β0 + βi xi + βij xi xj (2)
i =1 i=1 j=1
where β0 , βi , βij are the regression coefficients to be estimated
from the experimental data. In such a way, a response surface
y is represented as a function of the parameters or variables x
of the design space (generally some subset of Euclidian k-space,
Rk ). In the case of k = 2, the vector of surface parameters is six
dimensional.
The number of sampling points n must be greater than or
equal to the number of terms in the polynomial. When there are
more sampling points, the equation is over-determined and the
regression techniques are required to fit the response surface to the
sampled data. In this case, the surface does not, in general, exactly
match the response values at every sample points. The method
2458 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465

Table 1
Natural frequency differences of simply supported beam before and after updating.
Mode Undamaged beam (Hz) Damaged beam (Hz) Difference (%) Updated (Hz) Error (%)

1 8.99 7.78 15.63 7.78 0.01

2 35.92 33.24 8.04 33.25 0.04
3 80.63 79.14 1.89 79.15 0.01
4 142.93 128.04 11.63 128.09 0.04
5 149.14 146.22 2.00 146.07 −0.10

of least-square fitting is usually used in the coefficient estimation subjected to

process to create a response surface.
Before the regressed response surface is put forward to be used
in structural FE model updating, it should be verified to check
whether the regressed surface has enough accuracy. If not, the
parameters are adjusted (based on the response data) to achieve
a balance between variability and bias of the regressed surface. R2
(ranged from 0.0 to 1.0) criterion and empirical integrated squared
Error (EISE) criterion can be used in response surface verification:
N 3. Numerical verification
[yRS (j) − y(j)]2
P
i=1
R2 = 1 −
N an assumed damage element is considered to demonstrate the
[y(j) − ȳ]2
P
j =1
1
N
X elements as shown in Fig. 2. The density and elastic modulus of
EISE = ∗ [yRS (j) − y(j)]2
N j =1
where yRS is the response value of the confirmation samples; y is
the true value of the confirmation samples; ȳ is the mean of all true
values. The larger the value of R2 , the more accurate the regressed
response surface. On the contrary, the smaller the value of EISE, the
closer the fit is to the data.
Another criterion is the root mean squared error (RMSE). It is
the square root of the mean square error (MSE) where the distance
vertically from the point is taken to the corresponding point on
the curve fit (the error). The squaring is done so negative values
do not cancel positive values. The RMSE is thus the distance, on
average, of a data point from the fitted line, measured along a
vertical line. The smaller the value of root mean squared error, with
0 corresponding to the ideal, the closer the fit is to the data. In the
real application, the R2 criterion and EISE or the RMSE criterion
can be used in a complementary way to check the accuracy of the
regressed response surface.
2.3. Model updating
In the finite element model updating in structural dynamics, the
structural response features of interest are often eigen solutions
related to such as natural frequencies and mode shapes. In this
study, the structural natural frequency is employed as a response
feature. Therefore, the optimized objective function is formulated
in terms of the residuals between analytical and measured natural
frequencies
m
X E10 while parameter x2 refers to I10 . Fig. 3 illustrates the regressed
Π (x) = wi (λai − λei ) 0 ≤ wi ≤ 1
i =1
where λai and λei are the analytical (finite element calculated)
and measured natural frequencies of the ith mode respectively.
wi is the weight factor to impose to the different order of natural
frequencies. m is the number of modes involved and x is the design
set.
The FE model updating can then be posed as a constrained
minimization problem to find the satisfied design set such that
Min kΠ (x)k22
xlk ≤ xk ≤ xuk , k = 1, 2, 3, . . .
where xlk and xuk are the upper and lower bounds on the kth
design variable to be set. N is the total number of design variables
(updated parameters). Single-objective optimization with equal
weight of each natural frequency is used in this paper for
optimization computation.
A simulated simply supported beam without damage and with
(3)
procedure of response surface -based FE model updating. The beam
of 6 m length is equally divided into 15 two-dimensional beam
(4) the beam are 2500 kg/m3 and 3.2 × 1010 Pa, respectively, while
the area and moment of inertia of cross-section are 0.05 m2 and
1.66 × 10−4 m4 , respectively.
FE modal analysis is first carried out on the beam without
damage to get the analytical natural frequencies. To simulate the
actual (target) beam, one damage location is assumed at beam
element 10 where the element bending stiffness is reduced by 50%.
The FE modal analysis is again carried out on this damaged beam to
get the simulated measured modal parameters. The initial values
of the first 10 natural frequencies selected and corresponding
differences are shown in Table 1. The maximum and minimum
errors that appeared in the natural frequency are 15.63% and 1.89%
respectively.
Updating of the FE model of the undamaged beam is to achieve
a goal to correlate the natural frequencies with the damaged beam.
In this study, moment of inertia I10 and elastic modulus E10 of
individual element 10 are chosen as the updating parameters that
affect the independent variable, natural frequencies. The range
selected for each parameter should reflect the change that one
expects to observe for the domain of the prediction of interest with
1.920 ≤ E10 ≤ 3.465(×1010 Pa) and 0.863 ≤ I10 ≤ 1.797
(×10−4 m4 ).
The central composite design (CCD) method in design of
experiments is then used to get the sample points of the parameter
values selected. A total of 9 runs of experiments are carried out. The
sampled parameter values and corresponding natural frequencies
calculated from FE model are listed in Table 2. By using least-square
fitting with these sample values, a quadratic polynomial response
surface (Eq. (2)) can be regressed where parameter x1 refers to
(5) response surfaces of the first and second natural frequencies. The
horizontal axes are parameters selected while the vertical axis
gives the response (natural frequency) at any point or location.
To evaluate the above-selected parameters of the FE model, a
statistics-based parameter effects analysis is performed. Parts of
the general regression significance of selected parameters (x1 and
x2 ) calculated from Eq. (1) on each mode of natural frequencies
comparing with the significance level of p = 0.05 are shown
in Fig. 4. The results show that both elastic modulus E10 and
moment of inertia I10 are high significance on the structural natural
(6) frequencies. It is observed that the cross-quadratic term x1 x2 in
 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465 2459

Fig. 2. Simulated simply supported beam.

Table 2
Central composite design and sample values.
Run E10 (1010 Pa) I10 (10−4 m4 ) Freq 1 (Hz) Freq 2 (Hz) Freq 3 (Hz) Freq 4 (Hz) Freq 5 (Hz)

1 1.920 1.000 8.22 34.10 79.67 132.17 147.19

2 1.920 1.660 8.67 35.12 80.23 137.86 147.19
3 3.200 1.000 8.68 35.13 80.24 137.91 149.14
4 3.200 1.660 8.99 35.92 80.63 142.93 149.14
5 1.655 1.330 8.36 34.39 79.83 133.72 146.41
6 3.465 1.330 8.92 35.72 80.54 141.62 149.37
7 2.560 0.863 8.36 34.40 79.85 133.77 148.40
8 2.560 1.797 8.91 35.72 80.53 141.60 148.40
9 2.560 1.330 8.72 35.24 80.29 138.57 148.40

9.5

8.5
Frequency

38
8
36
7.5
Frequency

34
7

6.5 32
3.5

3 30
1.5
2
2.5
x1

2
1.5
2 0 x1 2.5
×1

1
0.5 ×10 1 –4
0

×10
10

1 0 3
1.5 1.5 0.5
2 x2 ×10–4 3.5 0
x2

(a) Surface of first natural frequency. (b) Surface of second natural frequency.

Fig. 3. Typical regressed response surfaces.

Table 3
RMSE values for each mode.
Mode 1 Mode 2 Mode 3
RMSE 0.0028 0.0002 0 0.0090
the surface expression is less significant on the structural natural
frequencies. For the simplification of a quadratic polynomial
response surface, the cross-quadratic term can be omitted in
practice to increase the efficiency of structural FE model updating.
To check the accuracy of the regressed surface, the RMSE is
calculated for each mode as shown in Table 3. It is demonstrated
that all RMSE values are close to zero, which indicates that the
created response surface has a high regression accuracy.
Now it is time to carry out the FE model updating where the
FE model is replaced by the regressed quadratic polynomial. The
Mode 4 Mode 5 residuals between analytical (undamaged beam) and measured
0.0028
(damaged beam) natural frequencies are used as the optimized
objective function. Single-objective optimization algorithm with
equal weight of each natural frequency is implemented to
achieve the best minimization of natural frequency residuals. The
optimization algorithm used to minimize the objective function is
a standard Trust Region Newton method in MATLAB. The tuning
minimization process is over when the tolerances are achieved or
pre-defined number of iterations is reached. The updated natural
frequencies and differences of the simulated simply supported
beam are also summarized in Table 1. It can be observed that a
good agreement of natural frequencies has been achieved after
carrying out the response surface-based FE model updating. The
2460 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465

(a) Parameter significance on the first natural frequency. (b) Parameter significance on the second natural frequency.

Fig. 4. The regression significance of selected parameters on the natural frequencies.

Fig. 5. Convergence of objective function of response surface-based updating. Fig. 6. Convergence of objective function of sensitivity-based updating.

final updated results on the parameters are E10 = 1.57 × 1010 Pa

(true value E10 = 1.60 × 1010 Pa) and I10 = 0.85 × 10−4 m4 (true
value I10 = 0.83 × 10−4 m4 ).
Fig. 5 illustrates the convergence of the objective function
during each optimization iteration. The horizontal axis is the
number of iteration while the vertical axis is the square sum of
natural frequency residuals of each mode. To demonstrate the
efficiency of current response surface-based FE model updating,
the objective function convergence of the traditional sensitivity-
based FE model updating is shown in Fig. 6. For such a numerical
simple beam, it is shown that the convergence of the objective
function is fast and iteration number is dramatically reduced to
reach the same residual level of the objective function by using the
Fig. 7. Potion of Hongtang bridge.
response surface method.

4. A precast concrete bridge tested in the field
A real case study is carried out on the Hongtang Bridge, located
in Fuzhou city, the capital of Fujian province, China. The bridge is a
multi-span continues-deck precast concrete highway bridge. The
construction was completed in December 1990. The total length
of the bridge is 1843 m, with a span layout of (16 + 27 + 4 ∗
30 + 60 + 120 + 60 + 31 ∗ 40 + 8 ∗ 25) m. It includes three
types of spans: simply supported spans, precast truss supported
spans and precast continuous girder spans. The photograph of
the part of bridge at present condition is shown in Fig. 7. For
the purpose of dynamics-based condition assessment, one portion
of six continuous girder spans with a span length of 40 m were
tested on the site under operational vibration conditions. On-site
dynamic testing of a structure provides an accurate and reliable
description of its current dynamic characteristics.
 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465
(a) Plan and longitudinal elevation of bridge showing sensor placement for the ambient vibration test.
(b) Details near the connection of the abutment and piers with deck
showing the bearing.
Fig. 8. Details of the Hongtang bridge and bearings with measurement points for the ambient vibration test.
The equipments used for the operational vibration measure-
ments include accelerometers, signal cables, and a 32-channel
data acquisition system with signal amplifier and conditioner. The
force-balance (891-IV type) accelerometers and INV306 data ac-
quisition system were used. Accelerometers convert the ambient
vibration responses into electrical signals. Cables are used to trans-
mit these signals from sensors to the signal conditioner. A signal
conditioner unit is used to improve the quality of the signals by
removing undesired frequency contents (filtering) and amplifying
the signals. The amplified and filtered analog signals are converted
to digital data using an analog to digital (A/D) converter. The sig-
nals converted to digital form are stored on the hard disk of the
data acquisition laptop computer.
Forty-nine measurement points were chosen at one side of the
bridge as shown in Fig. 8 (a). The accelerometers were directly
installed on the surface of the bridge deck in the vertical direction.
Five test setups were conceived to cover the planned testing
locations of the bridge. One reference location was selected near
one side of abutment. The sampling frequency on site was 300 Hz.
An operational modal identification procedure will need to base
itself on output-only data. In this study, the natural frequencies
of the bridge are obtained by picking the peaks on the graphs of
the average normalized power spectral densities (ANPSDs) from all
acceleration measurements. More details about the field ambient
vibration measurements and modal parameter identification can
be found in Jaishi and Ren [8].
The initial finite element model has been developed according
to the blue print of the bridge aimed at simulating the dynamic
behavior in the vertical direction. The deck of the considered
spans has the form of the hollow-core precast concrete girder with
a overall width of 11 m. The deck and piles of the bridge are
modeled by two-dimensional beam elements without the shear
deformation consideration. The equivalent values for the cross-
sectional area and inertia moment of the girder and piers are pre-
calculated and given as the inputs for the beam elements. The
2461
(c) Details of the bearings used in support.
concrete is considered to be homogeneous with an initial value for
Young’s modulus of 3.50 × 104 MPa and density of 2500 kg/m3
(C50 grade of concrete).
It is well known that the modeling of bridge boundary condi-
tions will heavily influence the calculated dynamic properties. The
deck of tested bridge is connected to the supporting piers and abut-
ments by neoprene bearings that allow both vertical and rotational
displacements of the bridge deck as shown in Fig. 8(b) and (c). To
simulate the behavior of the translational and rotational springs,
each neoprene support is modeled by means of additional con-
nected beam elements. These elements consisted of small length
(0.01 m) beams that connect the deck to abutments or piers. The
roller supports can be simulated in the numerical model by choos-
ing a cross-section of the connected beam element with a large sec-
tion area and small inertia moment, whereas the rigid supports can
be simulated in the numerical model by choosing a cross-section of
the connected beam element with a large value of both section area
and inertia moment. In this work, the translational and rotational
behavior of the neoprene supports are simulated by choosing the
suitable values of the section areas and inertia moments of these
connected elements. Each neoprene bearing as shown in Fig. 8(c)
has seven layers reinforced with steel plates between them. The
equivalent Young’s modulus of a single-layered composite element
can be calculated with the following formula:
Ecomp = 0.1(530S − 418) (MPa) (7)
in which
A
S= (8)
2t (b + c )
where S is the shape coefficient of the neoprene bearing; b, c , t are
the width, length and thickness of one layer of neoprene bearing
respectively; and A is the area of bearing. The vertical spring
stiffness of the multi-layered composite bearing is then equal to
Ecomp A Ecomp bc
Kv = P = P N/m. (9)
t t
2462 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465

Table 4
Natural frequency differences of six span bridge before and after updating.
Mode Initial FE model (Hz) Field test (Hz) Error (%) Updated FE model (Hz) Error (%)

1 3.200 3.070 −4.24 3.101 −1.01

2 3.352 3.291 −1.86 3.188 3.13
3 3.957 3.542 −11.72 3.730 −5.31
4 4.780 4.149 −15.21 4.431 −6.80
5 5.650 4.611 −22.52 5.173 −12.18

Table 5
Parameters and their bounds.
Parameter Min. Mid. Max.

x1 (1010 Pa) 2.450 3.500 4.550

x2 (10−3 m2 ) 1.050 2.099 3.149
x3 (10−3 m2 ) 3.240 6.480 9.720

Table 6
R2 values with cross-quadratic terms.
Mode 1 Mode 2 Mode 3 Mode 4 Mode 5

0.998 0.998 0.998 0.998 0.997

The success of finite element model updating largely depends

Fig. 9. Diagram to calculate the equivalent rotational stiffness of support.
One bridge support consists of several neoprene bearings
arranged as shown in Fig. 8(b). The equivalent spring stiffness in
the vertical direction at one support is given by the summation
of all individual bearing springs stiffness obtained from Eq. (9)
at that support. The equivalent rotational spring stiffness of the
support around the transverse direction of the bridge is derived
with reference to Fig. 9 using the value of vertical stiffness. Due
to the application of bending moment M1 at support, both vertical
springs with stiffness K1 and K2 , which are separated L distances
apart, are compressed with deflection ∆1 and ∆2 respectively.
Using simple statics, it can be obtained that
 
M1 1 1
θ= + .
L2 K1 K2
In the case of K1 = K2 = Kv , the equivalent rotational stiffness
of the support can be written as
M1 Kv L 2
Krot = = N m/rad.
θ 2
The equivalent section area and second moment of area of
the connected beam elements calculated from these vertical and
rotational stiffness are used as the initial values in the FE model
updating. Subsequently, the established FE model of tested six
continuous spans of the bridge, as shown in Fig. 10, is composed
of 87 nodes, 159 elements (79 beam elements, 73 mass elements,
7 connected beam elements). The natural frequencies calculated
from the initial FE model and identified from operational vibration
measurement are shown in Table 4.
on the selection of updating parameters. In practice, the discrep-
ancies between the experimental and analytical models are due
to many parameters and uncertainties. The updating parameter
selection should be made with the aim of considering the uncer-
tainties in the model. Physical and geometric properties of finite
elements as well as the stiffness of the connection are normally
chosen as the updating parameters. In this study, three important
parameters are selected for updating on the basis of the created fi-
nite element model. These parameters are elastic modulus of con-
crete material of bridge deck standing for x1 , cross-sectional area
of connection element at abutment bearings standing for x2 and
cross-sectional area of connection element at interior pier bear-
ings standing for x3 . The initial values of selected parameters and
their low and upper bounds are listed in Table 5 where the support
stiffness of bearings is represented by varying the cross-sectional
area of corresponding connection elements. Although it is very dif-
ficult to accurately estimate the variation bound of the parameters
during updating, a quite large of variation to the parameters of con-
nection elements (bearings) is expected due to more possible un-
certainties involved in the bridge bearings.
Once the model structural parameters and their bounds are
selected, the sampled parameter points can be determined by
design of experiments (DOE). Again, the central composite design
(CCD) method is used to obtain the sample points of the parameter
values. With these sampled parameter values, corresponding
(10) structural natural frequencies (responses) are calculated from
FE modal analysis. By using least-square fitting, a quadratic
polynomial response surface (Eq. (2)) can be constructed with
regard to parameters x1 , x2 and x3 .
For the construction of a quadratic polynomial response surface
in FE model updating in structural dynamics, as discussed in the
(11) case study of a simulated simply supported beam, the cross-
quadratic term (interaction effect) in the surface expression is
less significant on the structural natural frequencies. To check
the correlation of the regressed surfaces, the R2 criterion is
used. For the first five natural frequencies, the R2 values of the
regressed quadratic polynomial surfaces with and without the
cross-quadratic terms are listed in Tables 6 and 7 respectively.
It can be seen that the correlation accuracies of both quadratic
polynomial surfaces with and without the cross-quadratic terms
are almost the same. Therefore, the cross-quadratic terms in the
quadratic polynomial can be omitted to increase the efficiency
 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465
Fig. 10. Finite element model of six continuous spans.
Fig. 11. Created response surfaces for the first natural frequency.
Table 7
R2 values without cross-quadratic terms.
Mode 1 Mode 2 Mode 3 Mode 4
0.996 0.997 0.997 0.997
Table 8
Surface regression coefficients for first natural frequency.
β0 β1 β2 β3 β11 β22
0.698 0.742 0.041 0.061 −0.041 −0.006
of structural FE model updating. The final regression coefficients
of the created quadratic polynomial surface without the cross-
quadratic terms are shown in Table 8. Fig. 11 illustrates the created
response surfaces of the first natural frequency. The horizontal
axes are the parameters (x1 , x2 and x3 ) selected, while the vertical
axis gives a value of the first natural frequency (response) at any
point or location when one parameter is changed and other two
parameters are fixed.
To demonstrate the significance of three selected parameters
on the structural responses (natural frequencies), the test on
variances (F-test) is carried out to check the significance of selected
parameters. The regression significance of selected parameters
calculated from Eq. (1) on the first mode of natural frequency
compared with the significant level of p = 0.05 is shown in Fig. 12.
It can be observed that the parameter x1 , the elastic modulus
of concrete material of bridge deck, has the most significant
effect on the structural natural frequencies among three selected
parameters.
Subsequently, an optimization procedure for model updating is
carried out on the created quadratic polynomial response surface.
The first five natural frequency residuals between analytical
and measured natural frequencies are used as the optimized
objective function. Single-objective optimization algorithm with
equal weight of each natural frequency is implemented to achieve
the best minimization of the natural frequency residuals. The
optimization algorithm used to minimize the objective function
is a standard Trust Region Newton method in MATLAB. The final
correlation of natural frequencies after the FE model updating is
shown in Table 4. This shows that the overall correlations between
2463
Mode 5
0.995
β33
−0.002
Fig. 12. Parameter significance on the first natural frequency.
FE calculated and experimental natural frequencies have been
improved after performing the response surface-based FE model
updating. However, the error of the second mode natural frequency
increases after the model updating process. That implies that it
is not always guaranteed that the updated finite element model
gives an improved prediction for every order of frequencies as the
natural frequencies reflect the overall structural stiffness.
After updating, the changes in the selected three updating
parameters of the bridge are shown in Table 9. It can be seen that
the initial value for Young’s modulus of the concrete is quite good
such that a small correction is needed, but a significant correction
on the initial estimation of the stiffness of the neoprene supports
turns out as the stiffness of neoprene supports depends much on
the hardness of the rubber that is difficult to determine exactly.
Therefore, the simulation of support conditions in the FE model
updating of a real bridge is very important, and replacing the
support simply with rollers cannot actually simulate the dynamic
behavior of the bridge. The neoprene supports in general provide a
significant value of the rotational stiffness.
Fig. 13 illustrates the convergence of the objective function
during each optimization iteration. The horizontal axis is the
2464 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465

Table 9 5. Concluding remarks

Selected parameters of Hongtang bridge before and after updating.
Selected parameters
Elastic modulus of deck
(1010 Pa)
Connection element area
at side bearings (10−3 m2 )
Connection element area
at interior bearings
(10−3 m2 )
Fig. 13. Convergence of objective function of response surface-based updating.
Initial values Updated values Change (%) The paper presents an application of the response surface
3.500 3.118 −10.91 method for finite element model updating of civil engineering
structures in structural dynamics. A quadratic polynomial re-
2.099 2.686 27.96
sponse surface is constructed using the finite element model sim-
6.480 12.909 99.10 ulation data and then employed to improve the computation
efficiency in the model updating. The proposed procedure is il-
lustrated on a simulated simply supported beam and a full-size
precast continuous box girder bridge tested under operational vi-
bration conditions. With the response surface to replace the origi-
nal FE model, the model updating process becomes efficient and
converges fast compared with the traditional sensitivity-based
model updating method. Once the response surface is constructed,
there is no finite element calculation involved in each optimization
iteration. As a result, the FE model updating can be easily imple-
mented in practice with available commercial finite element anal-
ysis packages.
It is demonstrated that updating parameter selection from all
possible physical or geometrical parameters in the FE model and
a good approximation of response surface construction are the
keys to the success of the method. A small number of structural
parameters chosen in the examples and an estimation of the
response surfaces by a quadratic function may not be far from the
truth for more general problems. It is still a challenge to test the
response surface method for the finite element model updating
of more complex civil engineering structures where the relation
between the unknown parameters and the response quantities of
interest is more complex and a much large number of parameters
of the finite element model are involved.

Acknowledgements

Supports from the China ‘‘863’’ Project under grant number

2009AA11Z101 and Natural Science Foundation of China (NSFC)
under grant number 50678193 are greatly acknowledged.

References

Fig. 14. Convergence of objective function of sensitivity-based updating.
number of iteration, while the vertical axis is the square sum
of relative natural frequency residuals of each mode. For the
comparison as the first example, the convergence of the objective
function using the traditional sensitivity-based finite element
model is shown in Fig. 14. It is clearly shown that the convergence
error of the objective function after the same 30 iterations is less
by using the current response surface-based finite element model
updating method.
It should be noted that the real bridge problem has serious
scouring and erosion at the submerged part of the structure while
the parameters selected are those at the bearings. An equivalent
damage at the supports of the bridge deck is assumed with this
assumption and this would lead to non-realistic results though
the final natural frequencies may have good match with the
measured ones. Hidden damages cannot be noticed by engineers
and they may be unaccounted for the application of the proposed
approach.
[1] Friswell MI, Mottershead JE. Finite element model updating in structural
dynamics. The Netherlands: Kluwer Academic Publishers; 1995.
[2] Baruch M, Bar-Itzhack IY. Optimal weighted orthogonalization of measured
modes. AIAA J 1978;16(4):346–51.
[3] Berman A, Nagy EJ. Improvement of large analytical model using test data.
AIAA J 1983;21(7):1168–73.
[4] Link M. Updating of analytical models-Review of numerical procedures and
application aspects. In: Structural Dynamics forum SD2000. Los Alamos;
1999.
[5] Fritzen CP, Jannewein D, Kiefer T. Damage detection based on model updating
methods. Mech Syst Signal Process 1998;12:163–86.
[6] Jaishi B, Ren WX. Structural finite element model updating using ambient
vibration test results. J Struct Eng, ASCE 2005;131(4):617–28.
[7] Jaishi B, Ren WX. Damage detection by finite element model updating using
modal flexibility residual. J Sound Vib 2006;290(1–2):369–87.
[8] Jaishi B, Ren WX. Finite element model updating based on eigenvalues and
strain energy residuals using multiobjecive optimization technique. Mech Syst
Signal Process 2007;21(5):2295–319.
[9] Lewa TL, Spencera AB, Scarpaa F, Wordena K, Rutherfordb A, Hemezb F.
Identification of response surface models using genetic programming. Mech
Syst Signal Process 2006;20:1819–31.
[10] Khuri AI, Cornell JA. Response surface designs and analysis. Marcel Dekker, Inc;
1987.
[11] Myers RH, Montgomery DC. Response surface methodology: process and
product in optimization using designed experiments. New York: John Wiley
& Sons, Inc; 1995.
[12] Montgomery DC. Design and analysis of experiments. 3rd ed. New York: John
Wiley & Sons, Inc; 2006.
[13] Romero VJ, Swiler LP, Giunta AA. Construction of response surface based
on progressive-lattice-sampling experimental designs with application to
uncertainty propagation. Struct Saf 2004;26:201–19.
[14] Batmaz I, Tunali S. Small response surface designs for metamodel estimation.
European J Oper Res 2003;145:455–70.
 W.-X. Ren, H.-B. Chen / Engineering Structures 32 (2010) 2455–2465
[15] Simpson TW, Williams BJ, Notz WI. Sampling strategies for computer
experiments: design and analysis. Int J Reliab Appl 2002;2(3):209–40.
[16] Rutherford BM, Swiler LP, Paez TL, Urbina A. Response surface (meat-model)
methods and applications. In: Proceedings of the 24th international modal
analysis conference. 2006. Paper No. 184.
[17] Doebling SW, Hemez FM, Schultze JF, Cundy AL. A metamodel-based approach
to model validation for nonlinear finite element simulations. In: Proceedings
of the 20th international modal analysis conference. 2002. p. 671–8.
[18] Guo QT, Zhang LM. Finite element model updating based on response surface
methodology. In: Proceedings of the 22nd international modal analysis
conference. 2004. Paper No. 93.
2465
[19] Hemez FM, Doebling SW, Anderson MC. A brief tutorial on verification
and validation. In: Proceedings of the 22nd international modal analysis
conference. 2004. Paper No. 250.
[20] Steenackers G, Guillaume P. Finite element model updating taking into
account the uncertainty on the modal parameters estimates. J Sound Vib 2006;
296:919–34.
[21] Faravelli L, Casciati S. Structural damage detection and localization by
response change diagnosis. Struct Saf Reliab 2004;6:104–15.
[22] Rutherford AC, Inman DJ, Park G, Hemez FM. Use of response surface
metamodels for identification of stiffness and damping coefficients in a simple
dynamic system. Shock Vib 2005;12:317–31.