Engineering Structures 70 (2014) 234245

Contents lists available at ScienceDirect

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

Numerical and hybrid analysis of a curved bridge and methods

of numerical model calibration
Adel E. Abdelnaby a,, Thomas M. Frankie b,1, Amr S. Elnashai b,2, Billie F. Spencer b,3, Daniel A. Kuchma b,4,
Pedro Silva c,5, Chia-Ming Chang d,6
a

Department of Civil Engineering, The University of Memphis, 3815 Central Avenue Room 106C, Memphis 38152, TN, USA
Department of Civil and Environmental Engineering, University of Illinois Urbana-Champaign, 205 N Mathews Avenue, Urbana 61801, IL, USA
Department of Civil and Environmental Engineering, George Washington University, 801 22nd Street NW, DC 20052, USA
d
Earthquake Engineering Research & Test Center, Guangzhou University, Guangzhou 510405, China
b
c

a r t i c l e

i n f o

Article history:
Received 22 August 2013
Revised 27 February 2014
Accepted 4 April 2014
Available online 8 May 2014
Keywords:
Curved bridge
Hybrid simulation
Numerical model calibration
Combined actions
Multi-directional loading

a b s t r a c t
Reinforced concrete (RC) bridge piers are subjected to combined loading conditions resulting from complex earthquake ground motions coupled with irregular geometry and asymmetry of the bridge structure.
Furthermore, the inuence of the assumptions and simplications made in modeling irregular and curved
bridges on the reliability of their resulting response data is still not fully known. For that purpose, in this
paper a hybrid simulation test is conducted on a curved four-span bridge. This test accounts for the threedimensional (3D) system-level interaction between the three experimental piers in two testing facilities
with the numerical models of the deck, restraints and abutments. Prior to the hybrid simulation, a
detailed numerical nite element, ber-based model of the whole bridge system is established. The analytical predictions of this model are then used for comparison with the hybrid simulation test results. Discrepancies between the numerical and experimental results of the bridge piers response are highlighted
and deciencies in the numerical model assumptions are discussed. A rigorous numerical model calibration procedure is then followed to adjust for the initial modeling assumptions and improve the bridge
model overall response. This study has proven that some modeling assumptions that are widely used
in seismic analysis of bridge structures are unrealistic and therefore may lead to inaccurate results.
Published by Elsevier Ltd.

1. Preamble
RC bridge piers are subjected to combined loading conditions
resulting from complex earthquake ground motions coupled with
irregular geometry and asymmetry of the bridge structure. The
technical challenge of assessing the risk posed to bridges with
irregular or curved geometry and subjected to multi-directional
loading is non-trivial. Additionally, the inuence of assumptions
and simplications made in previous experimental tests and
numerical analyses on the reliability of response data is not fully
Corresponding author. Tel.: +1 (901) 678 4633; fax: +1 (901) 678 3026.
E-mail addresses: bdelnaby@memphis.edu (A.E. Abdelnaby), frankie2@illinois.
edu (T.M. Frankie), aelnash@illinois.edu (A.S. Elnashai), bfs@illinois.edu (B.F.
Spencer), Kuchma@illinois.edu (D.A. Kuchma), silvap@gwu.edu (P. Silva), chang37@illinois.edu (C.-M. Chang).
1
Tel.: +1 (217) 714 3363.
2
Tel.: +1 (217) 265 5497.
3
Tel.: +1 (217) 333 8630.
4
Tel.: +1 (217) 333 1571.
5
Tel.: +1 (202) 994 6652.
6
Tel.: +1 (217) 333 1516.
http://dx.doi.org/10.1016/j.engstruct.2014.04.009
0141-0296/Published by Elsevier Ltd.

known. Therefore, the impact of accounting for or neglecting complex geometry, loading, and system level effects when assessing RC
bridge vulnerability remains unclear.
Extensive experimental tests of RC bridge piers subjected to
combined actions were conducted in literature [4,9,14,19
21,24]. In addition, numerous efforts have been done to invoke
numerical modeling to capture combined action effects
[11,23,24]. The testing and analysis results of piers subjected to
combined loading in previous studies were assessed on the pier
component level (i.e. without consideration of the response of
the entire structural system and its inuence on pier behavior).
Even when piers were subjected to complex actions that result
from combined loading of the structural system, the interaction
between the pier performance and the structural system was
not accounted for.
A summary of the key aspects of previous tests are provided in
Table 1. These aspects include the number of studied degrees of
freedom (DOFs) that lead to shear (V), bending (M), axial (P) and
torsional (T) deformations. In addition, the test type (analytical
or experimental) and the consideration of the interaction between
system response and pier behavior are included.

235

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Table 1
Key aspects of RC pier testing programs.
Source

Test type

No. of cases

No. of DOFs

System response

Otsuka et al. [19]

Tirasit and Kawashima [21]
Zhiguo et al. [24]
Belarbi et al. [25]
Jeng [11]
Li et al. [14]
Hindi and Browning [9]
Zhang et al. [26]
Prakash et al. [27]

Expt.
Expt.
Expt. &Anly.
Expt.
Anly.
Expt.
Expt.
Anly.
Expt.

9
7
6
7
90
4
18
24
8

X
X
X
X

X
X
X
X

X
X
X
X

X
X

X
X

X
X

X
X

X
X
X
X
X
X

14
14
3
66
1
14
1
66
14

No
No
No
No
No
No
No
Yes
No

This study

Expt./Anly.

Yes

X: denotes included in the test.

In this paper, a hybrid experimental/analytical simulation is

conducted on a curved four-span bridge utilizing the Multi-Axial
Full-Scale Sub-Structuring Testing and Simulation (MUST-SIM)
facility of the George E. Brown Network for Earthquake Engineering Simulation (NEES) equipment site at the University of Illinois
at Urbana-Champaign. In this test, the bridge piers are experimentally tested while the rest of the bridge structure (including deck
and abutments) is tested numerically [7]. The test is a part of the
Combined Actions on Bridge Earthquake Research (CABER) project,
sponsored by the National Science Foundation. This project aimed
at investigating the combined actions of bridge piers subjected to
multi-directional ground motions.
The hybrid simulation approach is known to be a reliable costeffective approach for conducting performance assessment of
structures. In hybrid simulation, critical elements of a structure
(such as bridge piers) are physically tested while the remaining
elements are concurrently simulated computationally [12]. Hence,
hybrid simulation is capable of accurately capturing the interaction
effects between the entire structures and physically tested
components.
The hybrid simulation test conducted in this study considers
the 3D system-level interaction between the three experimental
piers in two testing facilities with the numerical models of deck,
restraints and abutments. In addition, a multi-directional earthquake loading is applied to the bridge system. The details and
capabilities of the hybrid test that is the origin of the data set utilized in this study is contrasted to the studies surveyed in the literature as highlighted in Table 1. This serves to highlight the
level of complexity and realism achieved in this hybrid test that
was not achieved elsewhere in the literature.
Prior to the hybrid simulation, a 3D numerical model of the
bridge is established using the ber-based open source Mid-America Earthquake Center analytical tool, Zeus-NL [6]. The analytical
predictions of this model are compared with the experimental
results of the bridge piers obtained from the hybrid test. Discrepancies between the experimental and analytical pier response are
highlighted and major deciencies in the numerical model
assumptions are discussed. Furthermore, an overview of the rigorous numerical model calibration procedure of the analytical model
based on the experimental data set is described. The numerical
model calibration process is essential, since commonly used modeling assumptions of bridges can result in misleading response,
especially when bridge piers are subjected to a high level of combined loading.

2. Bridge description
The overall geometry of the prototype bridge is based on a seismic design example from the National Cooperative Highway

Research Program (NCHRP) Project 12-49 [3]. The prototype bridge

in this example consists of ve continuous box girder spans of
100 ft. each, with four two-pier bents of unequal lengths. Modications are made to this design in order to increase the bridge irregularity as shown in Fig. 1. Most signicantly, removal of one span,
reduction to one-pier bents, varying spans, and introducing curved
geometry. The resulting geometry was selected based on an analytical parametric study aimed at generating high levels of combined
actions on the piers in all 6 DOFs.
The prototype bridge in this study is designed with the western
United States seismic requirements for a site with latitude of 47.0
degrees north, and longitude of 122.9 degrees west. The seismic
evaluation is performed for an earthquake level that corresponds
to the Maximum Considered Earthquake (MCE) with a 2500-year
return period. Non-liqueable soil conditions are assumed.
The ground motion record is synthetically generated using SIMQKE software [22] to match the MCE response spectrum. The
design response spectrum parameters including site coefcients
(Fa and Fv), short- (SS and SDS) and 1-s period (S1 and SD1) spectral
accelerations are summarized in Table 2.
The record has a peak ground acceleration (PGA) and duration
of 0.42 g and 30 s, respectively (Fig. 2). In order to reduce the run
time of the hybrid experiment, the synthetic earthquake record is
further modied by only considering the rst 10-s segment of
the full length record. The rst 10-s segment is selected such that
the response spectrum resulting from this segment represents the
best t when compared with the full length (30-s) record response
spectrum. The response spectrum of the full length record and the
rst 10-s segment are contrasted in Fig. 3.
The 10-s segment is scaled to four different performance levels.
Each performance level corresponds to estimates of the states of
structural response, namely cracking, yielding, design and failure,
based on an early numerical investigation. The response spectra
of the scaled records are plotted in Fig. 3. The used (40-s) earthquake record in the hybrid test consisted of the four scaled 10-s
segments applied sequence, as shown in Fig. 4.
The design performance level is the third of the four 10-s intervals of seismic loading applied to the bridge, and is therefore
applied as 1.0  MCE. The rst level is set to the cracking performance level, 0.08 MCE. The cracking limit state is determined from
a sensitivity analysis using Zeus-NL [6]. This is done by varying the
record scaling factor while monitoring concrete strains at critical
pier locations until the maximum strain at the critical ber reaches
the concrete cracking strain. Similarly, the second level, at 1020 s
of the overall (40-s) applied record, is the yielding performance
level, at 0.3 MCE. While, the nal 10 s (3040 s) reached a peak
of ground acceleration of 0.83 g (two times the PGA of the MCE
event) which represents the complete collapse limit state.
Multi-directional earthquake loading is applied to the curved
bridge. This is simulated through applying 100% of the earthquake

236

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Curved bridge
plan view
Pier 3
Pier 1
Pier 2

Box Girder

22.5

48"

28.5

Pier 1

Pier 2

37.5
Pier 3
28#10

Curved bridge elevation

#5 stirrup

Pier cross section

Typ.

Cap Beam
Typ.

Fig. 1. Conceptual drawing of the curved bridge.

Table 2
MCE response spectrum parameters [2].
Parameter

Ss (g)

Fa

S1 (g)

Fv

SDS (g)

SD1 (g)

Value

1.17

0.90

0.41

2.40

1.06

0.99

Fig. 3. Response spectra of the four levels of input ground motions.

Fig. 2. Full length of synthetic record.

record in the transverse direction of the bridge, while a 25% scaled

version of the record is applied in the longitudinal direction.
3. Hybrid test
This section describes the analytical modeling and numerical
verication of the 3D hybrid simulation framework that lead to
the successful execution of the full-scale hybrid experimental/analytical test. It is worth noting that the hybrid simulation framework, using UI-SimCor, has been validated in previous studies for
2D sub-structural models only [8,12,15].

Fig. 4. The used record in the test, which consists of four levels of input ground
motions applied in sequence.

3.1. Whole numerical model

A detailed nite element model (FEM) is constructed using
Zeus-NL, as shown in Fig. 5 [1]. This model is referred to as the

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

spring model is shown in Fig. 7. This gure also shows gap widening with intense ground shaking.
The Newmark time integration algorithms are employed in the
inelastic response history analysis [18] and a 0.005 s time step is
used. Rayleigh proportional damping is applied with 0.54% and
0.74% damping ratios for modes 1 and 2, respectively.

shear key
with gap

pier
cap beam

237

deck

abutment

Fig. 5. Zeus-NL whole numerical model.

initial un-calibrated whole numerical model of the bridge. The

deck was divided into sixteen segments; each segment is 25 ft.
long. Piers cross sectional dimensions, shown in Fig. 1, were modeled using the reinforced concrete circular sections (rccs) built in
Zeus-NL section library [6].
Prior to starting the simulation, gravity (dead and live) loads
were applied to the system as vertical concentrated forces at deck
structural nodes. Gravity forces are calculated per their respective
tributary areas and their effect (PD) is incorporated in the simulation. Ground accelerations are applied at pier supports and abutments to simulate earthquake loading.
The abutment-deck interface, at both deck ends, is simulated by
two nonlinear springs (Fig. 6). The rst spring simulates the gap
between the bridge deck and the abutment and can therefore
account for the pounding effects during earthquake shaking. The
second spring model represents the hysteretic behavior of the
shear key element. The two springs are connected together in series to yield an equivalent spring model that was used in the analysis. The hysteretic response of the equivalent gap and shear key

3.2. Numerical verication of hybrid simulation framework

It is worth noting that prior to this work UI-SIMCOR was limited
to model two-dimensional structures only. Therefore, in this study
UI-SIMCOR was upgraded to handle space structures, hence the
verication of the framework is sought.
The bridge model was sub-structured into three numerical
modules. The rst module comprised of the bridge deck, transverse
beams (pier caps) and abutments, as shown in Fig. 8. The second
and third modules comprised of the two outer piers and the inner
pier, respectively. The mass and damping were modeled in the
simulation coordinator (UI-SimCor) platform, which integrates
the bridge sub-structured modules in order to provide a systemlevel simulation of the structure [12].
Similar to the whole model, gravity loads are applied prior to
the simulation. However, the a-OS integration technique is used
for the dynamic analysis [5]. Same time step and damping ratios
are used for the whole and sub-structured models.
The results obtained from the initial whole model are compared
with the numerical sub-structured model. The relative deformations at the top of left pier are plotted in Fig. 9. The good match
of the response between the two models demonstrates the success
of running 3D hybrid simulation using UI-SimCor to communicate
between three modules that are representative of an ultimately
more complex combined experimental/analytical hybrid test.
3.3. Experimental/analytical hybrid simulation
In this test, the hybrid bridge structure is composed of one
small-scale (module 3) and two large-scale (module 2) experimental piers. The curved deck, abutments and restraints are modeled
analytically in Zeus-NL (Fig. 10). The testing of the experimental

Fig. 6. Illustration of the formulation of the equivalent spring model.

238

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

longitudinal steel. Transverse reinforcement is provided by

annealed wires of 0.0348 in. diameter. A micro-concrete mix is prepared using appropriate scaled and graded aggregate, along with
Type III cement to achieve the desired characteristics, including
an early-age compressive strength of 5 ksi.

200
150

50
0
-50
-100
-150
-200
-250
-5

-4

-3

-2

-1

Displacement (in)
Fig. 7. Hysteretic response of shear key with gap model.

components is enabled by the unique 6 DOFs Load and Boundary

Condition Boxes (LBCBs) at the Illinois MUST-SIM facility. Additional information on the existing testing capabilities at the
MUST-SIM facility can be found at [7].

3.3.1. Properties of the bridge piers

At full scale, the three RC bridge piers share the same cross-sectional properties (Fig. 1), and are of varying lengths. The piers have
a diameter of 4 ft., with longitudinal reinforcement provided by
28#10 bars distributed evenly. Transverse reinforcement is a #5
continuous spiral stirrup. The pier heights are 28.5, 37.5, and
22.5 ft., respectively, when viewing from left to right in Fig. 1.
The outer piers are constructed and tested at 1:3 scale, while
the inner pier is tested at 1:20 scale. The 1:3 scale piers are constructed with a diameter of 16 in., with 17#4 longitudinal bars
and a #3 spiral at 2 in. pitch and 3=4 in. cover. Resulting heights
are 9.5 and 7.5 ft. respectively for the left and right outer piers.
The constructed dimensions of the small-scale pier include a
length of 22.5 in. and diameter of 2.4 in. The scale down of the longitudinal reinforcement consisted of 348 threaded rods to achieve
the desired reinforcement ratio with an appropriate number and
distribution of rods. These rods are annealed at 1000 F in order
to achieve the target stressstrain response of the 1:3 scale

3.3.2. Similitude relationships

The similitude relationships for developing a level of certainty
in the test specimens are based on extensive studies done by
[10]. In addition, a validation test of the small-scale specimen
similitude relationships was conducted prior to the hybrid simulation. The test consisted of a large- (1:3) and a small-scale (1:20)
test specimens. The small scale specimen was scaled down based
on [10] relationships. The two specimens were subjected to the
same cyclic drift patterns using the large-scale and small-scale
LBCBs. The force-drift relationships of the two specimens are contrasted in Fig. 11. The excellent match not only in the location and
extent of damage of the two piers but also in the hysteretic loaddrift loops proves the realism of the small-scale specimen
response.
3.3.3. Test results
The highlights of the test results from the hybrid experiment
are provided below:
Piers 1 & 3 deformation response is controlled by predominantly exural behavior, while pier 2 exhibits some contribution to deformations and actions from shear response.
Extensive hinge formation in the piers, particularly pier 1, leads
to softening of the piers and period elongation.
The resulting period elongation shifts the dynamic response
during the application of the nal seismic scaling level.
Stiffness degradation, particularly for pier 1, causes greater
deformation response demands than was obtained from initial
analytical predictions.
Torsional loading contribute signicantly to the formation and
propagation of cracks for both piers.
Higher torsion-to-bending ratios are observed in the outer
bridge piers closest to the abutments as compared to the inner
pier.
response

response

command

deck (module 1)

pier 3 (module 3)

piers 1 & 2 (module 2)

Fig. 8. Illustration of the analytical hybrid simulation.

command

Force (kips)

100

239

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Zeus-NL

y (vertical)

UI- SimCor
x (transverse)

z (longitudinal)

100

x (rad)

10

y (mm)

-100

10

20

30

0.01

0.005

2
0
-2

10

20

30

10

20

30

40

10

20

30

40

10

20
time (sec)

30

40

-0.01

40

0.02

z (rad)

z (mm)

-0.005

50

-50

x 10

0
-5

40

y (rad)

x (mm)

-3

200

10

20
time (sec)

30

40

0
-0.02
-0.04

Fig. 9. Comparison of the numerical results with the whole model (Zeus-NL) and the numerical sub-structured model (UI-SimCor) for the middle pier (Pier 1).

response

command

response

command

deck (module 1)

7.5 ft

22.5 in
9.5 ft

pier 3 (module 3)

piers 1 & 2 (module 2)

Fig. 10. Illustration of the large scale hybrid simulation.

240

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Fig. 11. Force-drift relationships of large-scale (left) and small-scale (right) specimens.

Flexural stiffness and capacity both appear to have been

reduced in the large-scale piers due to the inuence of torsional
softening.
Response shape of the piers in the transverse direction of the
bridge is single curvature, while in-plane stiffness of the deck
creates restraints that push the piers into double curvature in
the longitudinal direction.
Transverse lateral drift at yield for the three piers is approximately 2%, 1%, and 3% respectively.

Ultimate drift reached in each pier during the hybrid simulation

is 5.5%, 3.3%, and 6.7% respectively.
4. Numerical model calibration
A rigorous calibration procedure of the analytical model based
on the experimental data set is performed. This procedure adjusts
for initially incorrect modeling assumptions affecting stiffness and
strength degradation, hysteretic damping, and local mechanisms.

6000
4000

x (transverse)

My (k-ft)

y (vertical)

z (longitudinal)

2000
0
-2000

top node
of pier 1

-4000
-0.02

Analytical
Experimental
0

-0.01

0.01

0.02

0.03

(rad)
y
400

500

F (kips)

F (kips)

200

-5

(in)
x

10

15

-200
-400

Analytical
Experimental
-500
-10

20

-600
-4

Analytical
Experimental
-2

(in)
z

Fig. 12. Comparison of experimental results and numerical predictions of the initial un-calibrated model (at Pier 1).

241

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Fig. 13. Comparison of 6DOF Krypton data (blue) with 6DOF cap (left) and base (right) deformations (red).

The effects of these initial shortcomings are noted, and a series of

nonlinear springs are introduced to the Zeus-NL model in order
to represent the behavior of the bridge observed in the hybrid
simulation.

model steel and concrete constitutive relationships. The invalidity

of the other assumptions is discussed in the next section as model
calibration methods are introduced to account for their undesired
consequences.

4.1. Comparison of experimental results with analytical predictions

4.3. Model calibration procedures

Variation between the analytical and experimental response of

the outer left pier, Pier 1 in the transverse (Fx, dx), longitudinal (Fz,
dz) directions as well as the torsional (My, hy) response can be
viewed in Fig. 12. These results compare the corrected experimental data with the numerical results from the initial un-calibrated
model. The results demonstrate the potential for signicant
improvements in model calibration.
In the transverse and longitudinal directions, higher stiffness
values in the numerical model are observed. Higher contrast in
stiffness is revealed in the longitudinal direction because the stiffness of the deck contributes to the response of the piers, which
drives the piers into double curvature.
Variation of torsional stiffness between the analytical and
experimental results is the most signicant inconsistency in terms
of the accuracy of stiffness values assumed in individual DOFs. The
reason is the modeling shortcoming of the elastic torsional behavior assumed in Zeus-NL. Signicant stiffness degradation is also
observed in addition to torsional capacity softening under
increased levels of load combinations.

A series of steps are taken in model calibration including adjusting for the assumptions made in the initial model regarding behavior at the pier-cap interface, torsional response, inclusion of shear
exure interaction, and the ability to capture hysteretic damping
effects.

4.2. Un-calibrated numerical model assumptions

4.3.1. Flexural rotational spring model in transverse and longitudinal

directions
This step is taken to perform the calibration procedure involves
the examination of the behavior of the pier-base and pier-cap
interface. Relative rotations observed at the interface observed
clearly contradict the xation and rigid body motion assumptions
at (Fig. 13). In addition, excessive cracking at these locations
resulted in reducing the stiffness and increasing the ductility of
the bridge piers.
Fig. 14 shows plot of the relative rotations between the base of
pier 1 and bottom end of the pier against the moment calculated at
the base of the pier from the LBCB data. Based on the rotational
response at the pier-base and pier-cap junctures, a nonlinear rotational spring model is developed to account for stiffness reduction
as well as stiffness and strength degradation. The spring model is

The following assumptions were made in the initial un-calibrated model. These assumptions are commonly used in modeling
bridge systems using FEM.

x 10

experimental
spring model backbone curve
spring model unloading curve

4
2
0

M (N-m)

Elastic deck and pier caps response.

Inelastic response of pier materials (Nonlinear constant connement Mander model for concrete [16]; and Menegotto-Pinto
steel model for reinforcing bars [17] are used).
Rigid body motion and xation at pier-cap interface.
Fixation at the base of pier.
Linear torque-twist response of pier cross section.
Flexural dominant piers (shear deformations neglected).
Rayleigh damping added to account for non-hysteretic damping
effects.

-2
-4

The rst and second assumptions are proven to be relatively

accurate through a series of preliminary studies conducted on
the superstructure components of the bridge and the piers. The
strains at critical locations along the deck and transverse beams
were monitored and were shown to remain in the elastic range.
Similarly, normal strain values measured at the concrete cover
and reinforcing bars of the physically tested piers in the hybrid
simulation, were close to the values predicted by the numerical

-6
-8
-0.02

-0.015

-0.01

-0.005

0.005

0.01

0.015

0.02

(rad.)
z
Fig. 14. Response at pier-base interface (pier 1) and features of spring model used in
the calibrated numerical model.

242

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

y (vertical)

calibrated
model
shear
spring

z (longitudinal)

flexural
spring
experimental
module

analytical
module

torsional
spring

Fig. 15. Model calibration in the longitudinal direction (shear spring included).

y (vertical)

calibrated
model

x (transverse)
analytical
module
experimental
module

torsional
spring

flexural
spring

Fig. 16. Model calibration in the transverse direction (shear spring not included).

employed in the calibrated model and its hysteretic response characteristics are shown in Fig. 14.
4.3.2. Torsional model
As discussed, the torsional stiffness of the initial un-calibrated
model was based on an anticipated pure torque-twist response of
the RC cross section of the pier. In addition to the development of
two rotational springs to represent the torsional stiffness degradation at the pier-cap and pier-base interfaces, a nonlinear rotational
spring model is developed to model the reduction of stiffness and
accelerated degradation of the pier stiffness and strength due to
combined action effects (Figs. 15 and 16). It is worth noting that
in addition to the improved torque-twist relationship, there is a
signicant improvement in the agreement between the analytical
and experimental displacements and rotations in other degrees
of freedom.
4.3.3. Shear spring model
Examination of the structural response in the longitudinal
direction revealed that a component of the displacements at higher
loads is due to shear deformations in the piers. This is caused by

the rotational stiffness restraint provided by the stiffness of the

deck in the longitudinal direction that drives the pier into double-curvature and hence shortening the pier effective height and
reducing the dominance of pure exural response.
The obtained experimental data is employed to derive the primary curve of the shear model of the bridge piers in the longitudinal direction (Fig. 15). To account for both the exural and shear
deformation of the bridge, a exure-shear interaction model is
adopted [13], in which a ber-based beam element are connected
in series to model the exural and shear behaviors, respectively.
The hysteretic model of the shear spring is dened by a quadri-linear primary curve that accounts for cracking, yielding, and ultimate
states. For this model, the obtained experimental data is utilized to
nd the primary curve of the pier which is then idealized as the
quadri-linear primary curve. It is noted that the shear deformations in the transverse direction are negligible because the rotational deck stiffness in this direction is minimal (Fig. 16).
4.3.4. Damping
The trend of higher analytical actions and deformations that
increase over the length of the record was observed in the numer-

243

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

1000

y (vertical)
M (k-ft)

500

x (transverse)

z (longitudinal)

-500
Analytical
Experimental

top node
of pier 1

-1000
-0.02

-0.01

0.01

0.02

0.03

(rad)
300

300

200

200

100

100

Fz (kips)

Fx (kips)

0
-100
Analytical
Experimental

-200
-300
-20

-10

10

20

(in)

0
-100
Analytical
Experimental

-200
30

-300
-4

-2

(in)

Fig. 17. Comparison of experimental results with the calibrated model predictions (at pier 1).

ical model, resulting in the greatest differences of response at signicant load levels. This is attributed to that the Zeus-NL model
employed was not able to capture the nonlinear hysteretic damping at higher levels of loading. Therefore, the distributed Rayleigh
damping for the bridge model is increased based on a parametric
study that was performed to identify the appropriate damping
ratios to be used to obtain deformations and actions that most
agree with the experimental results.
The improvement in the overall response of the piers, following
the implementation of the aforementioned calibrations, in the
transverse and longitudinal directions as well as torsional behavior
is shown in Fig. 17.

4.4. Simplied model

spring as shown in Fig. 18. Similarly, the equivalent transverse

and torsional springs are developed.
The results of this procedure in the simplied model with three
equivalent nonlinear springs is determined to be appropriate for
use under slight and moderate levels of ground motions (cracking,
yielding and design levels), i.e. for the rst thirty seconds of the
record (Fig. 19). This decision is made to save computational time
and resources, since each analysis runs approximately ve times
faster using the simplied model.
However, the use of the simplied model is not appropriate for
performing analyses under severe earthquake loads. It is worth
noting that a much less desirable match between the numerical
results and experimental data is observed for nal scaling of the
hybrid simulation record (3040 s), when the simplied model is
used.

The simultaneous implementation of 8 nonlinear linear springs

for each pier (four rotational springs for exure at the pier-cap and
pier-base interfaces, three torsional rotational spring, and shear
spring) resulted in a computationally expensive analysis. Therefore, efforts are made to produce a smaller set of equivalent springs
capable of representing an approximation of the effective stiffness
of the nonlinear springs developed above. This procedure results in
one lateral spring in each direction, transverse and longitudinal, as
well as one effective torsional spring. The effective stiffness of the
equivalent spring in the longitudinal direction in the simplied
model is therefore calculated as follows.

1
1
1
1

keq ka  L kb  L kc

where ka is the rotational spring modeling of pier-base rotation, kb

is the pier-cap rotation, while kc is the shear spring model, and L is
length of the pier that is used to convert rotational springs to an
equivalent displacement spring. Through this procedure, the fully
calibrated pier models undergo transformation to a simplied

Fig. 18. Conversion from fully calibrated to simplied equivalent spring model.

244

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

9

M (N.mm)

r (rad)

0.01
0.005

-0.005
-0.01

10

15

20

25

0.5
0
-0.5
-1
-1.5

30

x 10

10

15

20

25

30

10

15

20

25

30

10

15

20

25

30

1
Analytical
Experimental

0
-200

x 10

0.5
x

200

F (N)

x (mm)

400

0
-0.5

10

15

20

25

-1

30

80

60

0.5

F (N)

z (mm)

40
20

0
-0.5

-1

-20

-1.5

10

15

20

25

30

x 10

time (sec)

time (sec)

Fig. 19. Comparison of experimental results and numerical predictions from the simplied model at Pier 1.

5. Summary and conclusion

The achievement of the hybrid simulation of a curved four-span
bridge of different span lengths and unequal pier heights is discussed. The validity of the hybrid test results used in the numerical
model calibration is highlighted. Model calibration procedures are
described, focusing on piers response during the hybrid test and
the inadequacy of the numerical tools in capturing this response.
Model calibration is performed to obtain a high degree of agreement between numerical results and the experimental data set.
Corrections for assumptions made in the initial un-calibrated
model are adjusted for through the inclusion of several nonlinear
springs. The main ndings of this study are summarized below.
The initial assumption of xed end conditions and rigid body
motion at the pier-cap interface proves inaccurate and overestimates the exural stiffness of the pier response. Nonlinear rotational springs are provided in the exural and torsional degrees
of freedom at the top and bottom of each pier to appropriately
model the stiffness degradation and hinge formation at these
locations.
There is a severe reduction in stiffness and accelerated stiffness
degradation in the torsional response under combined loading
when compared to pure torque-twist response. A nonlinear torsional spring is developed and implemented, resulting in
improved matching of lateral displacements in the calibrated
model as well.
Flexural dominant behavior of piers is observed in the bridge
transverse direction. However, in the longitudinal direction,
shear deformations are not negligible. Therefore, a shear spring
is developed and implemented in the calibrated numerical
model.
The numerical model was shown to be incapable of fully capturing hysteretic damping. Actions and deformations in the analytical model are therefore overestimated, particularly at higher
levels of load. Distributed Rayleigh damping is provided to
account for these damping effects, and effectively reduces piers
resulting actions and deformations.

Initial and degrading stiffness behavior is adjusted in each

degree of freedom using the derived 6DOF data coupled with
global pier response. Adjustments in the stiffness of each degree
of freedom typically impacted other degrees, displaying some
inuence of combined actions and inuence of torsional
response on lateral actions.
Other adjustments made in model calibration reveal that significant reductions in stiffness occur in the torsional response.
Softening in the torsional degree of freedom is also shown to
in turn inuence the lateral deformation and load capacities,
displaying signicant interaction.
The nonlinear rotational spring models developed in this study are
purely empirical; they do not also consider coupling effects.
Therefore, these models are specic only for the earthquake loading applied in this hybrid simulation test, and cannot be generalized for other bridge congurations. The model calibration
assumptions in this paper leave scope for future studies which will
consider developing general spring parameters on physical basis.
The conclusions listed above are derived from one hybrid test
and limited specically to the studied bridge case. However, the
disparity between the initial analytical predictions and experimental results shown in this paper highly indicates the inadequacy of
the existing analytical solutions in accurately modeling bridge
piers behavior under high levels of combined loading. Most studies
conducted on seismic analysis and evaluation of bridges do not
account for important modeling features including a) appropriate
torsional behavior, b) exibility at pier-cap and pier-base interfaces and c) shear-exure interaction. The lack of these features
in modeling bridge piers subjected to combined loading conditions
can result in highly misleading bridge numerical response.
Acknowledgments
The work presented in this paper was supported through a
NEESR-SG project (Seismic Simulation and Design of Bridge Columns under Combined Actions and Implications on System
Response) funded by the National Science Foundation under Award

A.E. Abdelnaby et al. / Engineering Structures 70 (2014) 234245

Number CMMI-0530737. This study is also supported by the funding from National Science Council in Taiwan under Grant No. NSC095-SAF-I-564-036-TMS.
References
[1] Abdelnaby A, Frankie TM, Spencer Jr BF. Numerical hybrid simulation
modeling verication for a curved 3-pier bridge. J Syst, Cybern Infor
2013;11(5):4851.
[2] American Association of State Highway and Transportation Ofcials. AASHTO
LRFD Bridge Design Specications, Washington, DC; 2009.
[3] Applied Technology Council and the Multidisciplinary Center for Earthquake
Engineering Research. NCHRP Report 472: Comprehensive Specication for the
Seismic Design of Bridges, Transportation Research Board of the National
Academies, Washington, DC; 2002.
[4] Belarbi A, Prakash S, Silva PF. Incorporation of decoupled damage index models
in the performance-based evaluation of RC circular and square bridge columns
under combined loadings. ACI Special, Publication, SP271-05; 2010. p. 79102.
[5] Combescure D, Pegon P. a-Operator splitting time integration technique for
pseudo-dynamic testing error propagation analysis. Soil Dyn Earthquake Eng
1997;16:42743.
[6] Elnashai AS, Papanikolaou V, Lee D. ZEUS NL a system for inelastic analysis of
structures, Mid-America Earthquake Center, University of Illinois at UrbanaChampaign, Urbana, IL; 2002.
[7] Frankie TM. Impact of complex system behavior on seismic assessment of RC
bridges. Ph.D. Dissertation, University of Illinois at Urbana-Champaign, IL, USA;
2013.
[8] Gencturk B, Elnashai A. Hybrid simulation of RC and ECC frames with an
experimental module at small-scale. ASCE Structures Congress, Chicago, IL,
March 2931; 2012.
[9] Hindi RA, Browning BJ. Torsionally loaded circular concrete members conned
with spirals. ACI Struct J 2011:13947.
[10] Holub C. Similitude considerations for small scale distributed hybrid
simulation of reinforced concrete structures. MS Thesis, Department of Civil
and Environmental Engineering, University of Illinois at Urbana-Champaign,
Urbana, Illinois, USA; 2005.
[11] Jeng CH. Simple rational formulas for cracking torque and twist of reinforced
concrete members. ACI Struct J 2010:18998.
[12] Kwon O, Nakata N, Elnashai AS, Spencer Jr BF. A framework for multi-site
distributed simulation and application to complex structural systems. J
Earthquake Eng 2005;9(5):74153.
[13] Lee DH, Elnashai AS. Seismic analysis of RC bridge columns with exure-shear
interaction. J Struct Eng 2001:54653.

245

[14] Li Q, Belarbi A, Prakash SS. Seismic performance of square RC bridge columns

under combined loading including torsion with low shear. Earth and space
2010: engineering, science, construction, and operations in challenging
environments. Honolulu, HI: ASCE; 2010. p. 30118.
[15] Mahmoud HN, Elnashai AS, Spencer BF, Kwon OS, Bennier DJ. Hybrid
simulation for earthquake response of semi-rigid steel frames. ASCE J Struct
Eng 2013;139:113448 (Special Issue: NEES 1: Advances in, Earthquake
Engineering).
[16] Mander JB, Priestley MJN, Park R. Theoretical stress engineering strain model
for conned concrete. J Struct Div Am Soc Civil Eng 1988;114:180426.
[17] Menegotto M, Pinto PE. Method of analysis for cyclically loaded reinforced
concrete plane frames including changes in geometry and nonelastic behavior
of elements under combined normal force and bending. In: Proc. IABSE
symposium on resistance and ultimate deformability of structures acted on by
well-dened repeated loads, Lisbon; 1973.
[18] Newmark N. A method of computation for structural dynamics. ASCE, J Eng
Mech Div 1959;85(EM3).
[19] Otsuka H, Takeshita E, Yabuki W, Wang Y, Yoshimura T, Tsunomoto M. Study
on the seismic performance of reinforced concrete columns subjected to
torsional moment, bending moment and axial force. In: Proceedings of the
13th world conference on earthquake engineering, Paper No. 393. Vancouver,
Canada; 2004.
[20] Prakash S, Belarbi A, You YM. Seismic performance of circular RC columns
subjected to axial force, bending, and torsion with low and moderate shear.
Eng Struct 2010;32(1):4659.
[21] Tirasit P, Kawashima K. Seismic performance of square reinforced concrete
columns under combined cyclic exural and torsional loadings. J Earthquake
Eng 2007;11:42552.
[22] Vanmarcke E, Fenton G, Heredia-Zavoni E. SIMQKE-II, conditioned earthquake
ground motion simulator: users manual, version 2; 1997.
[23] Xu SY, Zhang J. Axial-shear-exure interaction hysteretic model for RC bridge
columns under combined actions. Eng Struct 2012;34(1):54863.
[24] Zhiguo S, Bingjun S, Dongsheng W, Xun G. Experimental research and nite
element analysis of bridge piers failed in exure-shear modes. Earthquake Eng
Eng Vib 2008;7(4):40314.
[25] Belarbi A, Prakash S, You YM. Effect of Spiral Reinforcement on Flexural-ShearTorsional Seismic Behavior of Reinforced Concrete Circular Bridge Columns.
Struct Eng Mech 2009;33(2):13758.
[26] Zhang J, Xu SY, Tang Y. Inelastic Displacement Demand of Bridge Columns
Considering Shear-Flexure Interaction. Earthquake Eng Struc 2011;40:
73148.
[27] Prakash SS, Li Q, Belarbi A. Behavior of Circular and Square Reinforced Concrete
Bridge Columns under Combined Loading Including Torsion. ACI Struct J
2012:31727.