Beruflich Dokumente
Kultur Dokumente
a r t i c l e i n f o a b s t r a c t
Article history: This paper describes ambient vibration based seismic evaluation procedure of an isolated highway
Received 23 December 2010 bridge. The procedure includes finite element modeling, ambient vibration testing, finite element
Received in revised form model updating and time history analysis. Gülburnu Highway Bridge located on the Giresun–Espiye
30 May 2011
state highway is selected as a case study. Three dimensional finite element model of the bridge is
Accepted 31 May 2011
Available online 29 June 2011
created by SAP2000 software to determine the dynamic characteristics analytically. Since input force is
not measured, Operational Modal Analysis is applied to identify dynamic characteristics. Enhanced
Keywords: Frequency Domain Decomposition and Stochastic Subspace Identification methods are used to obtain
Ambient vibration test experimental dynamic characteristics. Analytical and experimental dynamic characteristic are
Finite element model updating
compared with each other and finite element model of the bridge is updated by changing of material
Highway bridge
properties to reduce the differences between the results. Analytical model of the bridge after model
Operational modal analysis
Seismic evaluation updating is analyzed using 1992 Erzincan earthquake record to determine the seismic behavior. EW, NS
Friction pendulum isolators and UP components of the ground motion are applied to the bridge at the longitudinal, transverse and
vertical directions, respectively. It is demonstrated that the ambient vibration measurements are
enough to identify the most significant modes of highway bridges. Maximum differences between the
natural frequencies are reduced averagely from 9% to 2% by model updating. It is seen from the
earthquake analyses that friction pendulum isolators are very effective in reducing the displacements
and internal forces.
& 2011 Elsevier Ltd. All rights reserved.
0267-7261/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.soildyn.2011.05.020
A.C. Altunis- ik et al. / Soil Dynamics and Earthquake Engineering 31 (2011) 1496–1510 1497
experimental measurements under ambient vibrations such as model updating. Zhao and DeWolf [8] performed the dynamic
traffic loads. Brownjohn et al. [7] studied on assessment of monitoring of steel girder highway bridge using modal flexibility
highway bridge upgrading by dynamic testing and finite element approach. It can be seen from the literature that there is no
enough studies about finite element analyses, experimental
measurements and seismic evaluation of base isolated and post-
Table 1 tensioned segmental highway bridges constructed with balanced
Classification of some relevant output-only identification algorithms [10]. cantilever method.
The objective of this study is to investigate seismic evaluation of a
Method Characteristics
base isolated and post-tensioned segmental highway bridge
Frequency Peak Picking (PP) Classical SDOF method
constructed with balanced cantilever method using operational
domain Frequency Domain MDOF method; application of SVD modal testing in service condition. Gülburnu Highway Bridge is
Decomposition (FDD) to reduce noise chosen as an application. Three dimensional finite element model
Enhanced Frequency MDOF method; application of SVD of the bridge is created by using SAP2000. Ambient vibration tests are
Domain Decomposition to reduce noise
performed and experimental dynamic characteristics are extracted
(EFDD)
Polimax MDOF method using Enhanced Frequency Domain Decomposition and Stochastic
Subspace Identification methods. Finite element model of the bridge
Time Random Decrement (RD) Operates on time domain series,
domain leading to a free decay curve is updated by changing of material properties to eliminate the
analysis differences between analytical and experimental dynamic character-
Recussive Techniques Time series modeling using istics. Seismic behavior of the bridge is determined using 1992
(ARMA) recursive algorithms Erzincan earthquake ground motion records after model updating.
Maximum Likelihood Stochastic methods based on the
Methods minimization of a covariance matrix
Stochastic Subspace Stochastic methods based on the
Identification Methods project of state vector on a vector of
(SSI-DATA) past realizations 2. Formulation
The mathematical background of nearly all these methods is 2.2. Stochastic Subspace Identification (SSI) method
quite similar, but a few different parts are available in imple-
mentation aspects such as data reduction, type of equation The Stochastic Subspace Identification method shares the
solvers and sequence of matrix operations. In this study, two advantage that stabilization diagrams can be constructed
different methods, which are Enhanced Frequency Domain by identifying parametric models of increasing order. These
Decomposition (EFDD) in the frequency domain and Stochastic diagrams are very valuable in separating the true system poles
Subspace Identification (SSI) in the time domain, are used for from the spurious numerical poles. SSI is an output-only time
modal parameter extraction. domain method that directly works with time data, without the
need to convert them to correlations or spectra. The method is
2.1. Enhanced Frequency Decomposition Domain (EFDD) Method especially suitable for operational modal parameter identification,
but it is difficult procedure to explain in detail in a short way for
EFDD method is an extension to FDD method which is a basic civil engineers. The model of vibration structures can be defined
technique that is easy to use. In this method, modes are simply picked by a set of linear, constant coefficient and second-order differ-
locating the peaks in Singular Value Decomposition plots calculated ential equations [16]:
from the spectral density spectra of the responses. As FDD method is
based on using a single frequency line from the Fast Fourier Trans- MU€ ðtÞ þC2 U_ ðtÞ þ KUðtÞ ¼ FðtÞ ¼ B2 uðtÞ ð2Þ
form analysis, the accuracy of the estimated natural frequency
depends on the FFT resolution and no modal damping is calculated.
However, EFDD gives an improved estimation of both the natural
frequencies and the mode shapes and also includes damping [11].
In EFDD, the single degree of freedom (SDOF) Power Spectral
Density (PSD) function, identified around a peak of resonance, is
taken back to the time domain using the Inverse Discrete Fourier
Transform. The natural frequency is obtained by determining the
number of zero-crossing as a function of time, and the damping
by the logarithmic decrement of the corresponding SDOF normal-
ized auto correlation function [11].
In EFDD method, the relationship between the unknown input
and the measured responses can be expressed as [11,12]
Gyy ðoÞ ¼ ½HðoÞn ½Gxx ðoÞ½HðoÞT ð1Þ
where Gxx is the r r Power Spectral Density (PSD) matrix of the
input, r is the number of inputs, Gyy is the m m PSD matrix of
the responses, m is the number of responses, HðoÞ is the m r
Frequency Response Function (FRF) matrix, and n and superscript
T denote complex conjugate and transpose, respectively. Solution
of Eq. (1) is given detail in the literature [13–15]. Fig. 5. Dimensions of box girder cross section (all dimensions are in cm) [21].
where M, C2 , K are the mass, damping and stiffness matrices, F(t) suspension and viaduct were evaluated from 1998 to 2000 and it
is the excitation force, and U(t) is the displacement vector was decided to construct segmental box girder bridge using
depending on time t. Observe that the force vector F(t) is balanced cantilever method. Manufacturing of cross-cutting began
factorized into a matrix B2 describing the inputs in space and a in October 2001, but this work was stopped in November 2001
vector u(t). Although Eq. (2) represents quite closely the true because of the geological features of sea floor. According to the
behavior of a vibrating structure, it is not directly used in SSI additional drilling reports, it was started to prepare alternative
methods. So, the equation of dynamic equilibrium (2) will be projects in January 2002. Finally on 31 December 2004, the project
converted to a more suitable form: the discrete-time stochastic is revised and it was decided to make a concrete highway bridge
state-space model. Solution of Eq. (2) is given detail in the using balanced cantilever method with |200 cm diameter bored
literature [17–19]. pile. The construction of the bridge was started in November 2005
Comparison of the advantages and disadvantages of these
methods according to the modal identification of the highway
bridge are given in Table 2.
Fig. 7. Schematic views of the abutments (all dimensions are in cm) [21].
Fig. 8. Some view of the friction isolator before and after placing.
1500 A.C. Altunis- ik et al. / Soil Dynamics and Earthquake Engineering 31 (2011) 1496–1510
and the bridge was opened to the traffic in May 2009. Some views of 3.1. Bridge deck
Gülburnu Highway Bridge are given in Fig. 1.
Gülburnu Highway Bridge was constructed with balanced The superstructure of this bridge is a continuous single cell
cantilever method using cast-in-place construction technique. box girder constructed with cast-in-place which is the post-
This method is the best and optimum method for the passing of tensioned. The bridge’s deck has a 330 m total span and 30 m
large and long valley with reinforced concrete highway bridges total width. The traffic on the each deck is going on two lines.
using maximum span and minimum columns. The use of this
method is recommended especially where scaffolding is difficult
or impossible to erect as over deep valleys, wide rivers, traffic
yards or in case of expensive foundation conditions. In this
method, firstly, columns and small part of bridge deck are
constructed over substructure using suitable framework. Then,
segments (3–5 m length) are erected on opposite sides of each
column to balance the loads by using a movable form carrier.
After the concreting, prestress tendons are inserted in the Fig. 10. Finite element model of Gülburnu Highway Bridge.
segments and stressed with post-tension. The prestressing ten-
dons are arranged according to the moment diagram of a
cantilever, with a high concentration above the column. Towards Table 4
the mid-span or the abutment the number of tendons gradually Material properties considered in finite element analyses [21].
decreases. Finally, form carrier is moved to the next position and a
Elements Class Modulus of Poisson ratio Density
new cycle starts. This sequence is completed in a week and going
elasticity (N/m2) (dimensionless) (kg/m3)
on until bridge decks meet at midspan. At the midspan, closure
segment is established to complete one span. Fig. 2 schematically Deck C40 3.6E10 0.2 2500
shows the balanced cantilever method, in which segments are Columns C35 3.5E10 0.2 2500
symmetrically added to the columns. Piles C30 3.4E10 0.2 2500
Foundation C30 3.4E10 0.2 2500
Seismic isolation system with Friction Pendulum isolators was
Steela S420 2.1E11 0.3 7850
used in the construction of Gülburnu Highway Bridge. This
system can be an effective tool for the earthquake resistant a
Yield stress ¼1600 MPa, ultimate stress ¼ 1860 MPa.
design of bridges that can be used in both new construction and
retrofit. The Friction Pendulum System is a seismic isolation
bearing, with a mechanism based on its concave geometry and
surface friction properties. The supported structure is adminis-
tered into a pendulum motion as the housing plate simulta-
neously glides on the concave dish and dissipates energy with
friction [20]. Also, main period of the structures increases,
horizontal forces and moments are reduced significantly. The
cross section of friction pendulum system used in the Gülburnu
Highway Bridge is shown in Fig. 3. f1 = 1.086Hz f2 = 1.642Hz
The Gülburnu Highway Bridge is twin prestressed concrete 1st Vertical Mode 1st Torsional Mode
box girder structures. Each of two bridges carries two lanes of
East Black Sea Coast Road. The bridge deck consists of a main span
of 165 m and two side span of 82.5 m each. The total bridge
length is 330 m and width of bridge is 30 m. In the construction of
the bridge, 3800 ton building iron, 738 ton prestress steel and
25,000 m3 concrete were used. Basic configuration of the bridge is
shown in Fig. 4. The structural system of the bridge consists of
deck (superstructure), columns, side supports. f3 = 2.355Hz f4 = 3.102Hz
1st Transverse Mode 2nd Vertical Mode
Table 3
Modeling properties of friction isolators.
U1 U2 U3
f6 = 4.524Hz
f5 = 3.318Hz 1st Longitudinal Mode
Properties used for linear analysis cases
2nd Transverse Mode
Effective stiffness 14E7 72,280 72,280
Effective damping 0 0 0
Shear deformation location
Distance from End-J – 0.225 0.225
Table 5
Measurement tests setups and accelerometers locations.
Test setup Accelerometers Frequency span (Hz) Total duration (min) Step
Points Directions
3.3. Base isolators The first eight mode shapes obtained from analytical solutions
of the bridge is given in Fig. 11. From the modal analysis, a total of
Eight friction pendulum isolators exist on the four columns. eight natural frequencies are attained analytically, which range
According to the project design criteria, earthquake isolation between 0 and 6 Hz. The analytical mode shapes can be classified
system has the ability of movement of 50 cm. This movement is into vertical, transverse, longitudinal and torsional modes.
restricted by octagon inhibitor structure constructed under the
main segments. The isolators are produced in California and some
experimental measurements performed using some special test 5. Ambient vibration tests and experimental dynamic
methods in there to test of movement capacity of isolators. Some characteristics
views of the friction isolator are shown in Fig. 8.
Finite element method can produce a good representation of a
true structure. However, the prediction from this method is not
always accurate. Inaccuracies and errors in a finite element model
4. Finite element analyses and analytical dynamic may arise due to:
characteristics
inaccurate estimation of material and geometric properties;
Three dimensional finite element model of the bridge is con- deterioration due to environmental hazards such as wind,
structed using SAP2000 [22] software. Deck, columns and bored piles earthquakes, and higher serviced loads after so many years
are modeled by frame element having three translational DOFs and of use;
three rotational DOFs at each node. Raft foundations are modeled as poor approximation of boundary conditions and inadequate
shell element. In the finite element model, each of the bridge deck modeling of joints;
consists of 65 frame elements. Each frame represents the each faulty assumptions in individual element shape functions and
segment. Automatic mesh frame option is considered at the inter- poor quality mesh;
mediate joints. Maximum station spacing is considered as 0.5 in the nonlinearities, damping mechanisms, and coupling effects that
frame output stations. In the additional output and design stations, are not taken into account in the model.
intersections with other elements and concentrated load locations
including bridge lane loading points is selected. Abutments (expan-
sion joints) are modeled using restricted boundary conditions and
only longitudinal translational freedoms are released. Boundary
conditions at the base of bored piles are defined using very rigid
springs. The values of spring stiffness are assumed as 1E9 N/m at the
longitudinal and transverse directions, 6E9 N/m at the vertical direc-
tion based on the design project [21]. Post-tension cables are modeled
using frame elements constrained to rotation and fixed to end of the
each segments. Post-tension loads are considered as strain. Friction
pendulum isolators are modeled using link elements. The view of the
location and working principles of the friction isolators is shown in
Fig. 9. Each isolator has a 50 cm displacement capacity on the
longitudinal and transverse directions. Modeling properties of the
isolators are given in Table 3. Fig. 10 shows the full three dimensional
view of the finite element model of Gülburnu Highway Bridge.
Material properties used in the analyses are given in Table 4.
Some kinds of load cases are considered in the 3D modal
analyses of the bridge:
Dead load: Weight of all elements. They are calculated from the
finite element software directly.
Additional mass: Weight of the asphalt, cobble, pipeline and its
supports, scarecrow. 50.90 kN/m distributed load is added to
each segment.
Table 6
Experimental modal parameters attained from the first test setup.
Testing is performed to increase the knowledge and under- thought that dynamic characteristics of these parts should be
standing of the behavior of a structure. This is accomplished by determined and compared with each other. Therefore, Giresun
observing the response of a structure to a set of known condi- part of the bridge is measured from the box girder on
tions. Currently, the most popular dynamic testing technique is reciprocal points.
modal testing [17,23]. For the modal testing of structures, there
are two main types of dynamic tests: forced vibration and
ambient vibration tests. In the first method, the structure is
excited by known input forces and correlated input–output
measurements are performed. Impulse hammers, drop weights
and electro-dynamic shakers are the main excitation equipments.
The successes of forced vibration tests are limited for relatively
small structures. In case of large and flexible bridges like highway,
cable-stayed or suspension bridges, it often requires heavy
equipments and involves important resources to provide a con-
trolled excitation at enough high levels, which becomes difficult
and costly. In the forced vibration test, the traffic has to be shut
down. This could be a serious problem for intensively used
bridges. During the past few years, ambient vibration testing
(Operational Modal Testing) proved to be a valuable alternative
according to classic forced vibration testing. Instead of using one
or more artificial excitation devices, in operation modal testing
makes use of the freely available ambient excitation caused by
natural excitation sources on or near the test structure. Also, the
test structure remains in its operating condition during the test.
In the experimental measurements, the responses of Gülburnu
Highway Bridge are measured by using B&K 8340 and B&K 4507
type uni-axial accelerometers. The minimum frequency span and
sensitivity of these accelerometers are 0.1–1000 Hz and 10 v/g for
B&K 8340, 1–25,600 Hz and 1 v/g for B&K 4507, respectively. The
signals are acquired in the B&K 3560 type data acquisition system
and then transferred into the PULSE Lapshop software [24]. For
parameter estimation from the Ambient Vibration Survey data,
the Operational Modal Analysis software is used [25].
The bridge was opened to traffic on May 2009. During the test in
June 2009, normal traffic over the bridge was used as a source of
ambient vibration. All kind of vehicles such as car, van, truck and bus
can pass over the bridge. Cars usually pass over the bridge almost
every minute. But trucks rarely pass. Since input force was not
measured, the use of Operational Modal Analysis to identify modal
parameters was indispensable. Five ambient vibration tests were
carried out in the box girder. Due to the limited availability of
accelerometers and data acquisition equipment, maximum of 16
accelerometers for each test step could be monitored simultaneously.
Among these accelerometers, 4507 type uni-axial was used as
reference accelerometer and its location unchanged throughout the
test. The 8340 types were used as roving accelerometers and were
moved in order to cover all accelerometer locations. Explanation of
the selected measurements points are given below:
Table 7
Experimental modal parameters attained from the second test setup.
Table 8 Second test setup: In the second test setup, Trabzon part of the
Experimental modal parameters attained from the third test setup. bridge is measured on reciprocal points in the box girder.
Accelerometers are placed at the symmetry point of the first
Frequency EFDD Method Average of SSI Method
number auto spectral
Frequencies Damping densities Frequencies Damping
Fig. 16. Dynamic characteristics attained from fourth test using EFDD and SSI
methods: (a) singular values of spectral density matrices, (b) average of auto spectral
densities of the first step, and (c) stabilization diagram of estimated state space model.
Table 9
Experimental modal parameters attained from the fourth test setup.
setup to compare the dynamic characteristics of Giresun and 5.1. The first test setup
Trabzon parts.
Third test setup: The third test setup is performed in the box Singular values of spectral density matrices, average of auto
girder and only main span between two columns is measured spectral densities and stabilization diagram of estimated state
using vertical and lateral located accelerometers. space model of the first test setup attained from vibration signals
Fourth test setup: In the fourth test setup, some additional using EFDD and SSI methods are shown in Fig. 13. Natural
accelerometers to third setup are placed along the bridge deck frequencies and damping ratios obtained from the first test setup
(only main span) to obtain longitudinal modes. are given in Table 6.
Fifth test setup: In the fifth measurement, the bridge is
measured from end to end in the box girder.
5.2. The second test setup
Accelerometers locations on the 2D schematic view of the Singular values of spectral density matrices, average of auto
bridge are given in Fig. 12. Table 5 summarizes details of five spectral densities and stabilization diagram of estimated state
group of measurement together with accelerometer locations. space model of the second test setup attained from vibration
signals using EFDD and SSI methods are shown in Fig. 14. Natural
frequencies and damping ratios obtained from the second test
setup are given in Table 7.
Table 10
Experimental modal parameters attained from the fifth test setup.
Frequency EFDD Method Average of SSI Method 5.3. The third test setup
number auto spectral
Frequencies Damping densities Frequencies Damping Singular values of spectral density matrices, average of auto
1 0.991 2.377 1.000 0.992 2.512
spectral densities and stabilization diagram of estimated state
2 1.508 1.018 1.520 1.507 0.342 space model of third test setup attained from vibration signals
3 2.235 0.693 2.230 2.226 0.624 using EFDD and SSI methods are shown in Fig. 15. Natural
4 2.854 1.004 2.860 2.832 0.657 frequencies and damping ratios obtained from the third test setup
5 3.142 0.644 3.170 3.175 0.556
are given in Table 8.
6 4.315 0.621 4.310 4.325 0.424
7 4.790 0.453 4.780 4.832 0.857
8 5.599 0.634 5.590 5.612 1.026
Table 12
Analytical and experimental dynamic characteristics after model updating.
6.0 40
Acceleration (m/s2)
Displacement (cm)
Before updating After updating
3.0
0.0 35
-3.0
Max=4.86 m/s2
-6.0 30
0 5 10 15 20 0.0 82.5 165.0 247.5 330.0
Time (s) Distance (m)
6.0
Acceleration (m/s2)
0
Displacement (cm)
3.0
0.0 -20
-3.0
Max=5.05 m/s2
-6.0 -40
0 5 10 15 20 0.0 82.5 165.0 247.5 330.0
Time (s) Distance (m)
6.0 5
Acceleration (m/s2)
Displacement (cm)
3.0
0
0.0
-3.0 -5
Max=2.43 m/s2
-6.0 -10
0 5 10 15 20 0.0 82.5 165.0 247.5 330.0
Time (s) Distance (m)
Fig. 19. Time histories of ground motion accelerations of 1992 Erzincan earth-
Fig. 20. Changing of maximum displacements along the bridge deck: (a) longitudinal
quake: (a) ERZ/EW component, (b) ERZ/NS component, and (c) ERZ/UP component.
displacements, (b) transverse displacements, and (c) vertical displacements.
6. Finite element model updating finite element model updating. But, measure 1–5 had a maximum
percentage difference equal to 2.1%, 2.2%, 1.9%, 2.0% and 2.3% after
When the analytically and experimentally identified dynamic finite element model updating. A good agreement is found between
characteristics of the Gülburnu Highway Bridge are compared natural frequencies and mode shapes after model updating.
with each other, it is seen that there is a good agreement between
mode shapes but some differences between natural frequencies.
So, finite element model of the bridge is updated by changing of 7. Earthquake behavior
material properties to eliminate these differences. Changing of
the material properties (modulus of elasticity) for finite element Earthquake behavior of Gülburnu Highway Bridge before and
model updating is given in Table 11. after finite element model updating is performed using 1992
Comparison of the analytical and experimental dynamic charac- Erzincan earthquake ground motion (Fig. 19). This earthquake has
teristics of the Gülburnu Highway Bridge before and after finite a magnitude value as 6.9 (M ¼6.9). EW, NS and UP components of
element model updating is given in Table 12. According to Table 12, it the ground motion are applied to the bridge at the longitudinal,
is seen that maximum differences in the natural frequencies are transverse and vertical directions, respectively. Also, experimen-
reduced averagely from 9% to 2%. Measure 1–5 had a maximum tally identified damping ratios for each mode are considered in
percentage difference equal to 8.5%, 8.3%, 8.5%, 8.6% and 8.8% before the updated analytical finite element model.
A.C. Altunis- ik et al. / Soil Dynamics and Earthquake Engineering 31 (2011) 1496–1510 1507
40 40
Max = 34.61cm Max = 32.21cm
20 20
Displacements (cm)
Displacements (cm)
0 0
-20 -20
20
Displacements (cm)
-20
Max = 7.60cm
-40
0 5 10 15 20
Time (s)
Fig. 21. Time histories of the maximum displacements for bridge deck before and after finite element model updating.
Table 13
Maximum displacements occurred on the bridge columns.
Fig. 22. Changing of maximum bending moments along the bridge deck.
2E+4 updating is given in Fig. 20. It is seen from Fig. 20 that displace-
ments after model updating are bigger than the other. Long-
itudinal displacements have a constant value along to the bridge
0E+0 deck. But, transverse and vertical displacements have an increas-
ing trend along to middle of the bridge deck.
The maximum longitudinal displacements are occurred as 32
and 34.6 cm, the maximum transverse displacements are
-2E+4
occurred as 29.8 and 32.2 cm, and the maximum vertical dis-
placements are occurred as 7.6 and 8.8 cm before and after
Max = 2.2E4kNm model updating, respectively. The time histories of maximum
-3E+4 displacements before and after model updating are presented in
0 5 10 15 20
Fig. 21.
Time (s)
Distribution of the bending moments along the bridge deck
Fig. 23. Time histories of maximum bending moment for bridge deck before and before and after finite element model updating is given in Fig. 22.
after finite element model updating. It is seen from Fig. 22 that bending moments have an increasing
1508 A.C. Altunis- ik et al. / Soil Dynamics and Earthquake Engineering 31 (2011) 1496–1510
5
Before updating
After updating
4
Height (m)
3
0
0 1 2
Displacements (mm)
2 2
1 1
Displacements (mm)
Displacements (mm)
0 0
-1 -1
Fig. 24. Changing of longitudinal displacements along the bridge columns (a) and time histories of maximum longitudinal displacements before (b) and after (c) modal
updating: (a) changing of longitudinal displacements, (b) before updating, and (c) after updating.
4
Column Height (m)
2E+5 2E+5
0E+0 0E+0
-1E+5 -1E+5
Fig. 26. Time histories of maximum bending moment for bridge columns: (a) before model updating and (b) after model updating.
Table 14
Variation of the displacement and internal forces along to the friction isolators.