Sie sind auf Seite 1von 10

Response spectrum analysis of girder bridges

with seismic isolators using effective stiffness


Jzsef Simon, Lszl G. Vigh

Abstract In seismic design, the most commonly used analysis


method is the multi-mode response spectrum analysis (MMRSA) due
to its relative simplicity. The mechanical behavior of isolation
devices (laminated elastomeric bearings, elasto-plastic devices,
friction pendulums etc.) is usually non-linear, thus linear analyses do
not provide sufficient results to capture the real behavior of the whole
structural system. According to Eurocode 8, seismic isolators can be
modeled with equivalent linear characteristic by using the effective
stiffness and effective damping of each isolator unit. This results an
iterative method carried out by MMRSA. A three dimensional
numerical model is built in OpenSEES FEM software to study three
continuous girder bridges by MMRSA using effective stiffness values
and by non-linear time-history analysis (NTHA). The results are
compared and then the applicability and validity of the effective
stiffness method is presented.

Keywords Effective stiffness, effective damping, nonlinear


time-history analysis, response spectrum analysis, seismic isolator.

I. INTRODUCTION

RIDGES are key elements of the infrastructure. It is an


essential economical and social interest to protect and
maintain them.
According to previous studies [1], many of the girder
bridges are vulnerable to earthquake loads calculated in
accordance with Eurocode 8 [2], [3] even in moderate seismic
regions. Specific components such as piers, foundations and
bearings can suffer significant damage. Special care should be
taken in case of the assessment of existing structures because
of the possible lack of knowledge of material properties and
stiffness values. In some cases, merely strengthening the pier
is not an optimal solution [4]. This way we increase the
stiffness, and also the reaction forces, thus we obtain
uneconomical cross-sections. Seismic isolation systems can
mitigate the seismic forces by shifting the dominant natural
periods of the original, un-isolated bridge and through energy
dissipation as well. Thus expensive strengthening methods can
be avoided or in case of new bridges, piers and foundations
can be designed more economically.
Seismic isolators show non-linear cyclic behavior.
According to many design codes (e.g. EUROCODE,
AASHTO, CALTRANS, JRA etc.) this behavior can be
modeled with bi-linear characteristic and even can be
J. Simon is with Budapest University of Technology and Economics,
Department of Structural Engineering, H-1111 Budapest, Megyetem rkp. 3.,
Hungary (phone: +36-1-463-1825; e-mail: j.simon.bme@gmail.com).
L.G. Vigh is with Budapest University of Technology and Economics,
Department of Structural Engineering, H-1111 Budapest, Megyetem rkp. 3.,
Hungary (phone: +36-1-463-1779; e-mail: geri@vbt.bme.hu).

approximated with an equivalent linear stiffness and effective


damping by linearization methods. Undoubtedly, the second
approach using linear response spectrum analysis is preferred
among designers; however, its validity is limited within
certain range of damping. The paper illustrates the
applicability of the method through numerical examples using
non-linear displacement dependent seismic devices.
A numerical model is built in OpenSEES FEM [5] software
to study three different continuous girder bridges with seismic
isolators. First the equivalent linear method is used with
response spectrum analysis in accordance with Eurocode 8.
The effective damping is determined by different design
codes. Then non-linear time-history analysis is carried out
modeling the isolators with simplified bi-linear behavior. The
results then are evaluated and compared.
II. MODELING OF ISOLATION BEARINGS
A. Bi-linear Model
The real hysteretic behavior of different non-linear
displacement dependent seismic devices can be simplified to
bi-linear characteristic. EC8-2 [3] Section 7 defines the bilinear model by three parameters (Fig. 1): elastic stiffness at
monotonic loading (Ke), equal also to the unloading stiffness
in cyclic loading; yield force under monotonic loading (Fy);
post-elastic stiffness (Kp).

Fig. 1 Bi-linear model of isolation bearings according to EC8-2 [3].

Other parameters that will be used in the following sections


are: yield displacement (dy); design displacement (dbd);
dissipated energy per cycle at the design displacement (ED);
force at zero displacement under cyclic loading (F0);
maximum force, corresponding to the design displacement
(Fmax); secant stiffness, calculated from the design
displacement (Keff).

This non-linear model can capture the energy dissipation of


the isolation system but demands non-linear time-history
analysis which is still a time-consuming procedure and is not
supported by most of the commercial analysis softwares.
B. Equivalent Linear Model
Eurocode 8 and also other design specifications let the nonlinear forcedisplacement relationship of the isolating system
be approximated by the effective stiffness (Keff), i.e. the secant
value of the stiffness at the design displacement (see Fig. 1).
During the calculations the force, displacement and effective
stiffness values have to be compatible regarding the real bilinear model. In this case the energy dissipation of the
isolating system is modeled with an equivalent viscous
damping called effective damping (eff).
According to Eurocode 8 [2], [3], the effective stiffness
expressed by the basic parameters is:
,

(1)

and the effective damping can be described by the following


equation:
,

(2)

where F0 is the force at zero displacement:


.

(3)

In this case, curve fitting of the system identification results


for inelastic response spectra was carried out with constant
ductility ratios. The results were calculated without
considering the physical or mechanical significance of
isolation bearings. The constant ductility ratios were chosen
up to the value of 8. During an earthquake, the isolation
bearing may suffer higher deformation, and thus the curve
fitting in this region is possibly not valid.
The modification of the original AASHTO equation (6) is
suggested by Hwang et al. in 1996 [9] to refine the previous
Caltrans94 values. The inelastic displacement spectrum is
approximated by the equivalent elastic displacement spectrum
by some modification of the period shift and damping ratio.
The following equation is referred later as Caltrans 96 method:

(9)

Figs. 2~4 draw the effective damping ratios calculated by


the above-discussed methods as the function of deformation
ductility ratios and with constant strain hardening ratios (). In
the study, three different isolation bearings are used with
different parameters: 0.05, 0.10 and 0.15. The lower 0.05
value may represent simple elastomeric bearings, while higher
values characterize high damping or lead rubber bearings. It
has to be mentioned that the parameters of the bearings are
just estimated, they are not determined by any consideration of
the bearing types, since the study focuses mainly on the
comparison of the analysis methods.

In the study, the effective damping is calculated by four


other design codes, as well. AASTHO code [6] provides the
same equations as EC8, reformulated with ductility ratio:
,

(4)

and strain hardening ratio:


(5)
yielding to the following form:

Fig. 2 Effective damping ratios calculated from different design


codes with constant strain hardening ratio =0.05.

(6)

In the Japanese JPWRI manual [7] the design displacement


is reduced by a factor of 0.7:
.

(7)

A method suggested by Hwang et al. in 1994 [8] and then


applied in the Caltrans (California Department of
Transportation) design books proposes the following equation
to determine the effective damping:
0.0587

(8)

Fig. 3 Effective damping ratios calculated from different design


codes with constant strain hardening ratio =0.10.

not less than 2.5% of the total gravity load above


the isolation system.
During the analysis the difference between assumed and
calculated values of dbd does not exceed 5% of the assumed
value. This results in an iterative solution method.

Fig. 4 Effective damping ratios calculated from different design


codes with constant strain hardening ratio =0.15.

In every case, higher strain hardening ratio means lower


effective damping values. If the Kp stiffness is increased, the
elastic energy corresponding to the Keff stiffness is higher, thus
a lower damping value is obtained.
Comparing the different methods, one can see that with
lower ductility ratios the highest effective damping is provided
by the EC/AASTHO code, followed by the JPWRI, Caltrans
96 and Caltrans94 methods. The margin where this tendency
changes is around =10-30, and is lower if the strain
hardening ratio is higher. In the high deformation ductility
region the order of the methods from highest to lowest
effective damping is: Caltrans 94, Caltrans 96, JPWRI,
ASSTHO/EC, when > 0.05; otherwise the highest value is
obtained from the Caltrans 96 method.
The effective damping values show high deviation which
decreases in the most common deformation ductility ratio
range which is around 10-30. With increasing strain hardening
ratios this decrease is more intensive.
According to these observations and statements, high
deviation of the results is expected where lower reaction
forces are obtained in case of AASTHO/EC and JPWRI, and
higher ones from Caltrans94 and 96 methods.

2. Iteration procedure
Fig. 5 shows the flowchart for the iteration procedure. For
the analysis the elastic acceleration response spectrum should
be used according to EC8-1 [3]. First an initial Keff,i stiffness is
assumed (e.g. the Ke stiffness), the initial eff,i is set to 5%. The
MMRSA is then carried out and the force in the isolation
bearing (Fcalc,i) is determined. This force may not compatible
with the original bi-linear force-deformation characteristic.

Fig. 5 Flowchart for the iteration procedure of the equivalent linear


method.

The actual deformation is defined by the calculated force


and the assumed effective stiffness (see Fig. 6):

III. ANALYSIS METHODS


A. Multi Modal Response Spectrum Analysis

,
,

(10)

1. General conditions of application


Since multi modal response spectrum analysis (MMRSA) is
a linear analysis, linearization methods presented in Section II
shall be used. In accordance to Eurocode 8, equivalent linear
analysis can be carried out if the following conditions are met:

the effective stiffness of the isolation system is not


less than 50% of the effective stiffness at a
displacement of 0.2dbd;
the effective damping ratio of the isolation system
does not exceed 30%;
the force-displacement characteristics of the
isolation system do not vary by more than 10%
due to the rate of loading or due to the vertical
loads;
the increase of the restoring force in the isolation
system for displacements between 0.5dbd and dbd is

Fig. 6 Calculation of the actual deformation, actual force and


effective stiffness in each iteration step.

Once the actual deformation is known, the actual


corresponding compatible force (lying on the original bi-linear
curve) can be determined:
,

(11)

assuming that dbd,i is greater than dy. The updated effective


stiffness is defined by the following equation:
,

(12)

Meanwhile in each iteration step the effective damping ratio


(eff,i+1) should be calculated from the actual deformation (dbd,i)
and the effective damping of the whole system (eff,s,i+1) should
be defined. The elastic response spectrum curve should be
modified (see Fig. 7) with an eff,i factor calculated from the
new damping value:
,

, ,

(13)

The modified response spectrum curve and the new Keff,i+1


should be applied in the model and the procedure should be
started again from the beginning.

Fig. 7 Modification of the standard response spectrum curve during


the iteration procedure.

The iteration should be continued until the difference in dbd,i


and dbd,i+1 is less than 5%. In our investigation this condition is
set to 1%.
3. Effective damping of the whole system
In Section II, the determination of the effective damping for
isolation bearings is presented. The effective damping of the
whole bridge system (eff,s) is needed for the modification of
the elastic response spectrum in each iteration step. The
damping of the unisolated bridge can be taken into account
through material damping of the piers and the superstructure,
and can be estimated by the structural type and material
properties. In this case, the system is assumed proportionally
damped. Adding isolation bearing, the local damping of each
unit makes the system unproportionally damped, thus the
classical modal analysis cannot be carried out. There are
different approaches to solve this problem.
If the damping is classical (i.e. proportional), the modal
properties are real-valued, otherwise they are complex-valued.

With complex modal analysis (CMA) the damping values can


be determined even if the damping matrix is general [10].
Another approach is neglecting off-diagonal elements
(NODE) in the damping matrix. This way the system can be
analyzed as a series of SDOF sub-systems as it can be done in
case of proportionally damped systems.
The composite damping rule (CDR) was proposed by
Raggett in 1975 [11]. The basic principle of this method is that
the effective damping ratio of the whole system can be
calculated as the sum of the damping ratios of each component
weighted by the ratio of the components potential energy (e.g.
from the work of the springs representing the equivalent linear
behavior of the isolation bearings) to the total potential
energy.
While CMA is thought to be the exact solution, it demands
the most calculations. NODE and CMR methods also need
preliminary modal analysis and the evaluation of the modal
modes. These difficulties can be avoided if the effective
damping of the whole system is approximated based on
previous studies [12]. The theoretical maximum value of eff,s
is the sum of the effective damping of each isolation bearings
plus the assumed 5% material damping of the structure. In our
investigation the eff,s is calculated from the sum of the
effective damping of the isolator units, the material damping
of the structure is neglected.

Fig. 8 Applied standard response spectrum curves according to EC8


and an example of artificial acceleration record fitting.

Fig. 9 Artificial acceleration record created by fitting method.

4. Applied standard response spectrums


As mentioned in Section III, elastic acceleration response
spectrum curves should be applied for the analysis. Soil type C
and E is assumed to consider the effect of different soil
conditions, and 1.0 and 1.4 design ground acceleration (ag)
values to take into account the effect of intensity. Fig. 8 shows
the curves used in the iterations in initial state, with 5%
effective damping ratio.

TABLE I
PROPERTIES OF THE THREE ANALYZED BRIDGES

Girder type
Concrete
Composite
Steel

Global geometry
Spans
L
H

Girder properties
Iy
Iz

It

ap

bp

[no.]

[m]

[m]

[m2]

[m4]

[m4]

[m4]

[m]

[m]

[m2]

4
2
3

35
40
90

15
8
12

7.00
1.35
1.15

6.27
0.09
5.84

46.52
1.42
26.90

12.56
0.02
7.05

1.6
1.2
2.5

6.00
6.00
12.00

9.6
7.2
30.0

B. Non-linear Time-History Analysis


To capture the more realistic hysteretic behavior of the
isolation bearings, the bi-linear characteristic should be
applied in the numerical model and non-linear time-history
analysis should be carried out. This way the energy dissipation
is modeled directly and not through estimated damping effect.
A new algorithm [13] is developed by the authors which
creates artificial acceleration records for time-history analysis.
The records are determined to match the given acceleration
spectrum curves. In Fig. 9 an artificial record can be seen
created with the new algorithm, and Fig. 8 shows the
responses calculated from this record.
According to EC 8, at least 3 such records should be used.
In the study seven records are created and applied in case of
soil type C and E and 1.0 and 1.4 m/s2 design ground
acceleration values. With seven or more records the average of
the responses can be regarded as the final result.
IV. NUMERICAL MODEL
According to the comprehensive literature overview of [14]
on seismic numerical modeling of girder bridges a three
dimensional numerical model is developed in OpenSEES [5].
A schematic picture of the numerical model can be seen in
Fig. 10.
The main structure is modeled by beam elements. Elastic
beam element is used for the girder and the piers as well,
because it is expected that most of the dissipated energy is
concentrated to the isolation bearings. The beam elements are
placed in the center of gravity, the elements of the girder and
the piers are joined with rigid eccentricity elements.
The bearings on the piers and the abutments are modeled by
zero-length elements with non-linear spring characteristics
assigned to three degrees of freedom (ux, uy, uz), but rotations
are free to develop. For the seismic isolator bearings bi-linear
hysteretic material model is used.
The soil-structure interaction at the foundations of the piers
is taken into account with integrate linear springs, adjusted
with stiffness values calculated from foundation analysis.
The iteration procedure as well as the response spectrum
analysis (RSA) according to Eurocode 8 is programmed in
TCL language by the authors. The RSA module can be used
for any spatial model in OpenSEES.
V. ANALYZED BRIDGES
Three different continuous girder bridges are analyzed.
Different superstructure types, such as composite girder, RC

Pier properties
Iy

Iz

It

Dead Load
G2

[m4]

[m4]

[m4]

[kg/m]

2.05
0.86
15.63

4.80
3.60
30.00

6.76
2.71
54.19

7180
4650
10100

box and steel box girder, are supported by RC piers. The


global and cross-section dimensions are chosen to represent
realistic, common bridge layouts where the pier failure due to
seismic hazard is likely to occur.

Fig. 10 Numerical model for continuous girder bridges.

Previous studies have shown [1],[4] that the probability of


pier failure is high if the dead load of the superstructure is
significant and the pier height is lower than about 10-15 m.
This was taken into consideration when defining the main
parameters regarding the number of spans, the length of one
span, the girder type etc. The parameters are shown in Table I.
The stiffness of the linear springs representing the soilstructure interactions at the foundations is shown in Table II
and was calculated considering common soil properties and
pile layouts.

Girder type
Concrete
Composite
Steel

Kx
[N/m]

TABLE II
SOIL-SPRING STIFFNESS VALUES
Ky
Kz
Kx
Ky
Kz
[N/m]
[N/m]
[Nm/rad] [Nm/rad] [Nm/rad]

2.88E+09

1.88E+09

3.20E+11

5.60E+10

1.97E+10

1.25E+11

1.88E+09

1.80E+09

2.40E+11

4.20E+10

8.29E+09

4.99E+10

4.50E+09

3.75E+09

1.00E+12

3.50E+11

1.50E+11

1.00E+12

In this study only the longitudinal responses of the bridges


are analyzed, thus isolation bearings are placed only in this
direction. The original, unisolated bridges have only one
longitudinally fixed bearings and it is situated in every
example on the first pier next to one of the abutments. The
isolation is solved by replacing the fixed bearing with the
isolation bearings.
The properties of the isolation bearings (e.g. the yielding
force) have to be defined in a way that no plastic deformation

may occur in ultimate limit state (ULS). The design bearing


forces calculated from breaking forces and the bearing
properties are shown in Table III.

Girder type
Design force [kN]
Fy [kN]
Ke [kN/mm]
Kp [kN/mm]
(Kp/Ke)

TABLE III
ISOLATION BEARING PROPERTIES
Composite
Concrete
389
390
125
6.25
0.05

562
570
100
10
0.10

records and with 1.4 m/s2 design ground acceleration and C


soil-type. Both before and after isolation state is presented
here.

Steel
883
900
70
10.5
0.15

VI. RESULTS
As already mentioned above, the responses are studied only
in the longitudinal direction. In the following sub-sections five
different parameters are observed which characterize the local
behavior of the isolation bearing and the global behavior of
the whole structural system. These parameters are in the
longitudinal direction: bearing force, bearing deformation,
maximum pier moments, girder displacement and pier top
displacement.
A. Example Bridge 1 - Composite Bridge
The first example bridge is a composite girder with two
spans, with fixed bearing on the middle pier.
First modal analysis is carried out to capture the seismic
behavior. Table IV summarizes the dominant modal modes in
the longitudinal direction before and after the replacement of
fixed bearings. The first modal shape can be seen in Fig. 11.
TABLE IV
DOMINANT MODAL MODES IN THE LONGITUDINAL DIRECTION
Girder type
Composite
Concrete
Steel
Isolation
NO
YES
NO
YES
NO
Frequency [Hz]
1.15
0.96
0.45
0.42
1.37
Time period [s]
0.87
1.04
2.25
2.38
0.73
Effective modal mass [%] 95.5
95.0
78.6
78.4
73.5

Fig. 12 Longitudinal moments in the pier with fixed and isolation


bearing.

Due to the isolation system, the fundamental period of the


structure is elongated leading to decreased reaction forces.
Fig. 12 shows the longitudinal moments in the pier. The
substructure is well-isolated, the maximum moment is
drastically decreased by a factor of 2.5.
In Fig. 13 and 14 the longitudinal displacements of the
girder and of the pier top can be seen. With fixed bearings the
pier top and the superstructure move together, while in case of
isolation relative displacements develop. Despite the elongated
fundamental period the displacements of the girder are also
decreased with the isolation, due to the drastic decrease of pier
deformations.

YES
0.72
1.40
76.8

Fig. 13 Longitudinal displacements of the girder with fixed and


isolation bearing.

Fig. 11 Dominant longitudinal mode shape of the composite bridge


before (a) and after (b) the replacement of fixed bearing.

The effective modal mass exceeding 90% confirms that


single mode characterizes the longitudinal system behavior
before and even after the isolation and thus the bridge can be
represented as an SDOF system. The dominant natural period
is increased with the isolation bearing, while the effective
modal mass remains nearly the same.
Non-linear time-history analysis (NTHA) is carried out
first. The main observed parameters can be seen in Figs. 12-15
calculated by NTHA with one of the artificial acceleration

Fig. 14 Longitudinal displacements of the pier top with fixed and


isolation bearing.

In Fig. 15 the calculated hysteresis curve of the bearing is


presented. The maximum bearing force is 1.5 times higher
than the yielding force because of the strain hardening effect.

THA

THA

22

8237

38

14

-5

-17

-7

-12

-7

651

45

10527

65

18

22

72

19

52

19

566

31

9159

49

16

19

14

535

26

8876

43

15.4

6.9

0.15

14.3

0.26

9.9

THA

Pier top disp. [mm]


-14

437

10

7132

25

12

-8

-38

-13

-16

-12

545

28

8836

45

15

15

65

53

10

487

18

7903

34

14

10

-4

16

475

16.7

8197

29.5

14

*Note: 1 - EC/AASHTO; 2 - JPWRI; 3 - CALTRANS94; 4 - CALTRANS96

[mm]

656

46

10613

66

18

10

26

11

22

11

570

32

9225

50

16

-5

-12

-3

-8

-4

597

36

9527

54.5

16.6

6.9

0.39

7.1

0.15

14.6

0.26

10.1

Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Average

Pier top disp. [mm]

[mm]
Girder disp.

426

6951

22

12

-21

-69

-26

-46

-25

439

11

7163

25

12

-18

-60

-24

-40

-22

549

28

8897

46

15

-5

10

-3

489

19

7945

35

14

-9

-29

-15

-17

-14

537

26.6

9394

42

16

0.35

2.7

0.32

3.4

0.13

9.1

0.21

6.0

*Note: 1 - EC/AASHTO; 2 - JPWRI; 3 - CALTRANS94; 4 - CALTRANS96

[-]

Girder disp.

12

-25

[mm]

Pier moment [kNm]

22

-16

[mm]

6915

-52

Ductility rat.

THA

-11

0.39

Iteration num. [-]

423

-13

Pier moment [kNm]

0.0

Damping rat. [-]

Isolator def.

Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Average

Isolator force [kN]

14

-29

[mm]

TABLE VI
RESULTS FOR THE COMPOSITE BRIDGE
(soil type C, ag=1.0 m/s2)

39

-13

Isolator def.

Ductility rat. [-]

0.40

*Note: 1 - EC/AASHTO; 2 - JPWRI; 3 - CALTRANS94; 4 - CALTRANS96

8305

-38

Isolator force [kN]

Iteration num. [-]

Pier top disp. [mm]

508

22

-14

TABLE VIII
RESULTS FOR THE COMPOSITE BRIDGE
(soil type E, ag=1.0 m/s2)

Damping rat. [-]

Girder disp.

-8

[mm]

Pier moment [kNm]

14

-14

[mm]

37

-8

Isolator force [kN]

8163

-21

512

0.39

Ductility rat. [-]

21

-6

-14

Iteration num. [-]

503

14

-30

Damping rat. [-]

Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Average

Isolator def.

38

-13

*Note: 1 - EC/AASHTO; 2 - JPWRI; 3 - CALTRANS94; 4 - CALTRANS96

TABLE V
RESULTS FOR THE COMPOSITE BRIDGE
(soil type C, ag=1.4 m/s2)

8241

-41

Ductility rat. [-]

21

-15

Iteration num. [-]

508

Damping rat. [-]

After the NTHA, the iterative equivalent linear analysis is


carried out by MMRSA presented in Section III. The effective
damping is calculated with the presented four methods. The
results of the MMRSA and the average results of the seven
NLTHAs with artificial records are summarized in Table VVIII, for the various seismic scenarios.

Pier top disp. [mm]

Girder disp.

Fig. 15 Hysteretic force-deformation diagram of the isolation bearing.

Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Calc.
Diff. [%]
Average

Pier moment [kNm]

Isolator force [kN]

Isolator def. [mm]

TABLE VII
RESULTS FOR THE COMPOSITE BRIDGE
(soil type E, ag=1.4 m/s2)

0.34

2.6

0.31

3.3

0.13

8.9

0.21

5.9

Figs. 16~19 also compare the various analysis methods. In


this case an isolator with 0.05 strain hardening ratio is used.
According to Fig. 2 EC/AASHTO and the Japanese JPWRI
may overestimate the effective damping ratio if the ductility
ratio is under around 10. This is proved by the results here,
EC/AASHTO and JPWRI always provide lower values than
the NTHA. Besides, the results from the EC method should
not be used, because the condition on the 0.30 limitation of the
effective damping ratio is not met.
General tendency is that the biggest difference is between
the isolation bearing deformations and also between the girder
displacements. Lower difference values are obtained if the soil
conditions are better, but higher values if the intensity of
earthquake is lower. This is because of the decreasing ductility
ratios and thus the overestimation of effective damping ratios.
In case of soil type C, Caltrans methods give reasonable
results, although Caltrans96 underestimates the pier moments
if the design ground acceleration is lower.

If the soil condition is worse (e.g. soil type is E), only


Caltrans94 method provide conservative results. The ductility
ratios are higher, but still the effective damping is estimated
well that the solution is on the safe side.
In Table V-VIII the number of iteration is also shown. It
can be stated that 6-8 iterations are normally sufficient for the
acceptable accuracy.

B. Example Bridge 2 - Concrete Bridge


The second example bridge is a five-span concrete box
section girder.
Here an isolation bearing with higher stain hardening ratio,
= 0.10 is used. This means that the same force causes less
deformation in the bearing, thus the ductility ratio is lower
where overestimation of the effective damping is already
observed. The ductility ratios are around 2-5 in case of soil
type E.

Fig. 20 Dominant longitudinal mode shape of the concrete bridge


before (a) and after (b) the replacement of fixed bearing.
Fig. 16 Difference between NTHA and MMRSAs
(composite bridge, soil type C, ag=1.4 m/s2).

Table IV indicates that with the isolation the natural period


is slightly changed and the seismic behavior in the
longitudinal direction can be described with the first dominant
mode. The effective modal mass belonging to this mode is
around 78%, which is the whole mass of the fixed pier and the
girder. The dominant mode shape can be seen in Fig 20.
Comparison between the NTHA and MMRSA is presented
in Figs. 21~24. Similar tendencies as for the previous example
can be observed. Caltrans94 method is the most conservative
one; however, it also leads to unconservative results if soil
type E is considered.

Fig. 17 Difference between NTHA and MMRSAs


(composite bridge soil type C, ag=1.0 m/s2).

Fig. 21 Difference between NTHA and MMRSAs


(concrete bridge, soil type C, ag=1.4 m/s2).
Fig. 18 Difference between NTHA and MMRSAs
(composite bridge, soil type E, ag=1.4 m/s2).

Fig. 22 Difference between NTHA and MMRSAs


(concrete bridge, soil type C, ag=1.0 m/s2).
Fig. 19 Difference between NTHA and MMRSAs
(composite bridge, soil type E, ag=1.0 m/s2).

Fig. 23 Difference between NTHA and MMRSAs


(concrete bridge, soil type E, ag=1.4 m/s2).

Fig. 27 Difference between NTHA and MMRSAs


(steel bridge, soil type C, ag=1.0 m/s2).

Fig. 24 Difference between NTHA and MMRSAs


(concrete bridge, soil type E, ag=1.0 m/s2).

Fig. 28 Difference between NTHA and MMRSAs


(steel bridge, soil type E, ag=1.4 m/s2).

C. Bridge 3 - Steel Bridge


The third example is a three-span steel box girder. With the
isolation ( = 0.15 is used) the fundamental period is doubled
as Table IV indicates. This is reached by the relatively stiff
fixed pier, the soft isolation bearing and the high dead load on
the girder. Fig. 25 shows the dominant mode shape of the
structure.
Fig. 29 Difference between NTHA and MMRSAs
(steel bridge, soil type E, ag=1.0 m/s2).
Fig. 25 Dominant longitudinal mode shape of the steel bridge before
(a) and after (b) the replacement of fixed bearing.

The differences in the main parameters are shown in Figs.


26~29. Soil type C, ag 1.4 m/s2 is the first case where nearly
all the methods are on the safe side. The effective damping
here is under 30% in case of all the methods.
The pier moment is critical again. Caltrans94 method is safe
in every case except that the pier moments are underestimated
if lower design ground acceleration (ag=1.0 m/s2) is taken into
consideration.

Fig. 26 Difference between NTHA and MMRSAs


(steel bridge, soil type C, ag=1.4 m/s2).

VII. CONCLUDING REMARKS


Application of linear response spectrum analysis for seismic
isolated continuous girders is investigated in the paper through
three numerical examples. Non-linear time-history analysis is
invoked and referred as the accurate solution. Linear response
spectrum analysis, using effective stiffness and effective
damping of the isolation units is also carried out. Different
methods available in the codes are used for the calculation of
effective damping ratio.
Based on the comparison of the results, it is concluded that
none of the effective damping calculation methods provides
exact solution. Among the various investigated design
parameters, Caltrans 94 method is usually on the safe side;
however it is often too conservative. It is also stated that all
the methods are applicable to predict the internal forces of the
system components (isolator unit and pier) with an accuracy of
25%.
The observed dispersion of the results induces the need of
further investigation on the applicability of the effective
stiffness method and on the calculation methods of effective
damping.
It must be emphasized that this investigation is carried out
using isolator bearings only in the longitudinal direction. The

study should be extended to spatial analysis where higher


dispersion of the results is expected.
ACKNOWLEDGMENT
The research has been supported in the framework of the
project Talent care and cultivation in the scientific workshops
of BME" project by the grant TMOP 4.2.2.B-10/120100009.
REFERENCES
[1]

[2]

[3]
[4]
[5]

[6]

[7]
[8]

[9]

[10]

[11]

[12]

[13]

[14]

. Zsarnczay,"Seismic performance analysis of common bridge types


in Hungary" (in Hungarian), Master Thesis at Budapest University of
Technology and Economics, 2010
CEN: EN 1998-1:2008 Eurocode 8: Design of structures for earthquake
resistance - Part 1: General rules, seismic actions and rules for buildings,
2008
CEN: EN 1998-2:2008 Eurocode 8: Design of structures for earthquake
resistance - Part 2: Bridges, 2008
J. Simon, "Seismic retrofit methods of bridges" (in Hungarian),
ptmnyeink vdelme konferencia, Rckeve, Hungary, 2013
F. McKenna, G.L. Feneves, "Open system for earthquake engineering
simulation", Pacific earthquake engineering research center, version
2.3.2., 2012
Guide specifications for seismic isolation design. (2000). American
Association of State Highway and Transportation Officials, Washington,
D.C.,2000
Manual for Menshin design of highway bridges. Public Works Research
Institute. Tsukuba City, Japan, 1992
J.S. Hwang, L.H. Sheng, J.H. Gates, Practical analysis of bridges on
isolation bearings with bi-linear hysteresis characteristics. Earthquake
Spectra, Vol. 10, No. 4, pp.705-727.,1994
J.S. Hwang, J.M. Chiou, L.H. Sheng, J.H. Gates, A refined model for
base-isolated bridges with bi-linear hysteretic bearings. Earthquake
Spectra, Vol. 12, No. 2, pp.245-272.,1996
A.S. Veletsos, C.E. Ventura, Model Analysis of Non-Classically
Damped Linear Systems, Earthquake Engineering and Structural
Dynamics, Vol. 14, pp 217-243, 1986
J.D. Raggett, Estimation of damping of real structures, Journal of
Structural Division, Proceedings of the ASCE, Vol. 101, No. ST9, pp.
1823-1835.,1975
M.Q. Feng, S.C. Lee, Determining the effective system damping of
highway bridges, Final report, Department of Civil & Environmental
Engineering University of California, 2009
J. Simon, L.G. Vigh, "Seismic assessment of Hungarian highway
bridges A case study", Proc. First. Intl. Conf. PhD Students in Civil
Engineering, pp. 155-162, Cluj-Napoca, Romania, 2012
J. Simon, " Numerical model development for seismic assessment of
continuous girder bridges", Proceedings of the Conference of Junior
Researchers in Civil Engineering, pp. 216-224., Budapest, Hungary,
2012

Das könnte Ihnen auch gefallen