You are on page 1of 21

Journal of Earthquake Engineering, 12:980–998, 2008

Copyright © A.S. Elnashai & N.N. Ambraseys


ISSN: 1363-2469 print / 1559-808X online
DOI: 10.1080/13632460701673019

Effect of Nonlinear Seismic Torsion on the


1559-808X
1363-2469
UEQE
Journal of Earthquake Engineering
Engineering, Vol. 0, No. 0, Dec 2007: pp. 0–0

Performance of Skewed Bridge Piers

PAIBOON TIRASIT and KAZUHIKO KAWASHIMA


Effect
P. Tirasit
of Nonlinear
and K. Kawashima
Seismic Torsion on Skewed Bridge Piers

Department of Civil Engineering, Tokyo Institute of Technology, Tokyo, Japan

This article presents an analytical investigation on the effect of seismic torsion on the performance
of a skewed bridge. A nonlinear torsional hysteretic model developed by the authors is applied to
idealize the torsional behavior of bridge piers. Deterioration of the torsional strength of piers due to
combined flexure is considered and deterioration of flexural strength due to torsion is not taken into
account. The effects of pounding between deck and abutments, cable restrainers, and damage of
bearing supports are also included in analysis. It is found that the eccentric impact force due to lock
of bearing movement results in extensive torsion in piers.

Keywords Bridges; Torsion; Seismic Design; Dynamic Response Analysis; Columns

1. Introduction
Irregular bridges are constructed in urban areas and the complex combination of internal
force components possibly occurs in such bridges during an earthquake because of their
unconventional structural response. Examples of these bridges include skewed bridges,
bridges supported by C-bent columns, curved bridges, etc., in which mass center is apart
from the rigidity center. During an earthquake, decks in a skewed bridge possibly rotate
around the vertical axis resulted from poundings and the action of restrainers between
decks and abutments or the adjacent spans [Priestley et al., 1996; Desroches and Fenves,
1997; Watanabe and Kawashima, 2004]. Consequently, this possibly results in twisting
moment coupled with other internal forces in piers. The combination of seismic torsion
and other internal force components—axial force, bending, and shear force—can result in
the complex flexure and shear failure in the bridge columns. Some evidence can be found
in the damage of bridge columns during the past earthquake as shown in Fig. 1 [Editorial
Committee for the Report on the Hanshin-Awaji Earthquake Disaster, JSCE, 1997]. This
bridge pier was used to support the deck with 52-degree skewed angle and damage in this
piers occurred due to torsion from the rotation of superstructure. This damage pattern was
completely different to that of cantilevered column under flexure which typically occurred
at the column base. Based on the available experimental studies, the flexural strength and
the ductility capacity of piers can be deteriorated remarkably if significant torsion exists
[Hsu and Wang, 2000; Hsu and Liang, 2003; Otsuka et al., 2003; Tirasit and Kawashima,
2007]. However, because an appropriate model on combined torsion and flexure for
reinforced concrete piers is not available, effect of torsion has been generally idealized by
linear model in the past analysis.

Received 8 October 2006; accepted 27 August 2007.


Address correspondence to Kazuhiko Kawashima, Department of Civil Engineering, Tokyo Institute of
Technology, Meguro, Tokyo 152-8550, Japan; E-mail: kawashima.k.ae@m.titech.ac.jp

980
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 981

FIGURE 1 Damage due to torsion in a pier of Kawaraginishi viaduct in Meishin express-


way during Hanshin-Awaji earthquake in 1995 (Courtesy of Editorial Committee for the
Report on the Hanshin-Awaji earthquake disaster, JSCE).

Seismic torsion of piers can also be induced by failure of bearing supports. It is often
that failure of bearings into several blocks causes strut action among failed blocks. Once
the upper bearings lock with the lower bearings, this obstructs the movement of upper
bearings relative to lower bearings.
This study shows an analysis for the effect of seismic torsion on the performance of a
skewed bridge. A nonlinear torsional hysteretic model developed based on the experimental
study [Tirasit and Kawashima, 2007] is applied to analyze the nonlinearity due to combined
torsion and bending moment in piers. The influence of pounding, cable restrainers, and the
failure of bearing supports is included in analysis under two near-field ground motions.

2. Structural Properties of Representative Skewed Bridge


A 40-degree 4-span continuous skewed bridge with a deck length of 4 × 40 = 160 m as
shown in Fig. 2 is analyzed in this study. A composite deck with a mass of 2,500 tons is
supported by 3 reinforced concrete piers and 2 reinforced concrete abutments. The longi-
tudinal and transverse directions of the bridge are defined as the directions parallel and
normal to the bridge axis, respectively, as shown in Fig. 2. Because the directions of prin-
cipal axes of the piers are not identical to the longitudinal and transverse directions of the
bridge axes, axis-1 and axis-2 which are the principal axes of the piers normal and parallel
to the skewed directions, respectively, are defined here as shown in Fig. 2. On the other
hand, the positive direction of the torsion is around the axis-3 orienting from the bottom to
the top of pier. The left, center, and right piers are referred hereinafter as P1, P2, and P3,
respectively. The left and right abutments are called as A1 and A2, respectively. As shown
in Fig. 2(b), the piers are 10 m high and have a 3 m x 3 m square section. A1 and A2 are
5.1 m high and have a 2 m x 15.66 m wall section. The longitudinal and tie reinforcements
are SD295 grade rebars with a yield strength of 295 MPa. The compressive strength of
concrete of the piers and the abutments is 21 MPa. Five steel bearings are used to support
982 P. Tirasit and K. Kawashima

Axis-3
Axis-1
Transverse
direction Pier section
Longitudinal Axis-2
direction Left side

12
Right side
40°
Plan view

40 40 40 40
M M
l
5.1

5.1
F F F

10

10
10

A1 P1 P2 P3 A2
Elevation
(a)
176D29
3000
312.5 312.5
19@125 = 2375

120
312.5

100
D19@150
19@125 = 2375

D19@150
3000

100
312.5

120

120 100 100 120

(b)

FIGURE 2 Configuration of the representative skewed bridge: (a) plan view and eleva-
tion; and (b) square pier section and reinforcement details.

the deck at each pier and abutment. “F” and “M” in Fig. 2 represent fixed bearings and mov-
able bearings, respectively. The fixed bearings do not allow the deck movement in the longi-
tudinal and the transverse directions while the movable bearings permit the deck movement
only in the longitudinal direction. Cable restrainers are provided at both ends of the bridge.
They are installed along the longitudinal direction at the two outermost and the center girders.

3. Analytical Modeling

3.1. Idealization of Foundation, Deck, Abutments and Piers


A discrete analytical method is used to idealize the bridge as shown in Fig. 3. The founda-
tion stiffness was modeled as a set of six spring elements to simulate the response of
foundation in six degrees of freedom. The interaction between soil and foundation was
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 983

Pounding Pounding
spring spring

Restrainer
spring
Restrainer
Pounding spring
spring Pounding
A1 The rightmost A2 spring
fixed steel
bearing
P1 P2 P3
Fiber element Torsional spring

FIGURE 3 Finite element modeling of the skewed bridge.

already taken into account in the formulation of the foundation stiffness. The deck, abut-
ments, and piers above the flexural plastic hinge region are assumed to be elastic. The
fiber elements are used to idealize the nonlinear interaction between bilateral bending and
axial force in the piers at the flexural plastic hinge zone. In the fiber elements, the consti-
tute model of confined and unconfined concrete and its unloading and reloading paths are
idealized based on the studies of Hoshikuma et al. [2003] and Sakai and Kawashima
[2006], respectively. The hysteresis of longitudinal reinforcement is idealized by Mene-
gotto-Pinto model with slight modification [Sakai and Kawashima, 2003].
To idealize the nonlinear torsional behavior of a bridge pier, an empirical torsional
hysteretic model based on the authors’ experiment is used. Only brief explanation of this
model is presented here because details of experiment and hysteretic model were already
presented by Tirasit and Kawashima [2007]. This model is composed of the primary curve
and the unloading and reloading paths as shown in Fig. 4. A parameter named “rotation-
drift ratio,” r, is defined as

r=q Δ (1)

where q and Δ is the torsional rotation (rad) and the lateral drift at the effective height of
column. r is the governing parameter which takes account of the effect of combined bend-
ing and torsion.
Tri-linear model is used to idealize the torsional primary curve. The column is
assumed to be elastic until reaching the torsional strength T%c at the yield rotation q%y . The
torsional restoring force starts to deteriorate at the rotation q%d . The torsional primary
curve under combined action approaches the primary curve under pure torsion as r
increases. Based on the authors’ experimental results, initial torsional stiffness of a col-
umn under pure torsion (r = ∞) is assumed to be 68% of the gross concrete section’s.
On the other hand, the unloading and reloading paths are regulated by not only r but
also the maximum torsional rotation which column has ever experienced and the number
of loading cycles n. The important characteristic of the torsional hysteresis is the pinching
effect. The experimental results indicate that pinching becomes significant as r decreases.
Consequently, the unloading and reloading paths are idealized by a bilinear model when
pinching is less significant while two unloading stiffnesses and two reloading stiffnesses
are used when pinching is significant, as shown in Fig. 4.
984 P. Tirasit and K. Kawashima

Torsion Torsion
∼ ∼ ∼ ∼
Primary curve Tc Tc Primary curve Tc Tc
∼ ∼
Tu Tu

∼ ∼
∼ ∼ −θy ∼ ∼ −θy
−θu − θd −θu − θd
∼ ∼ ∼ ∼
θd θu Rotation ∼ θd θu Rotation
∼ θy
θy

∼ ∼
–Tu –Tu
∼ ∼ ∼ ∼
–Tc –Tc –Tc –Tc

(a) (b)

FIGURE 4 Primary curve and unloading and reloading paths of the torsional hysteretic
model: (a) less pinching range; and (b) significant pinching range.

It is noted that this torsional hysteretic model can be used in the columns subjected to
combined cyclic bending and torsion with r ranging from 0.5 to ∞, and the number of
loading cycles n ranging from 1 to 3. Because the effect of torsion on flexure at r ≤ 0.5 is
limited and can be negligible, r is assumed to be 0.5 in analysis when r becomes smaller
than 0.5. r sometimes becomes extremely large when q is very small. However, the effect
of torsion on the flexure is insignificant when the torsional rotation is small. Conse-
quently, r is assumed in analysis as

⎧q Δ ... q > q%y (r = 0.5) and r > 0.5


r=⎨ (2)
⎩ 0.5 ... otherwise

in which q%y (r = 0.5) is the yield torsional rotation at r = 0.5. Because q%y (r = 0.5) is the
smallest torsional rotation which can cause the nonlinearity in torsion in the model, r is
determined here by Eq. (2).
The value of number of loading cycles n is assigned independently for the positive
and the negative torsions. n is 0 when the column hysteresis moves along the envelope,
and becomes 1 as the column is unloaded from the envelope at the rotation larger than q%y ,
as shown in Fig. 5. The value of n increases at every time when the hysteresis which is
unloaded at q%m returns to a rotation which is 0.95q%m or larger after experiencing a loading
excursion. This excursion should be large enough to experience zero or opposite side rota-
tion. The effect of n on the unloading and reloading paths decreases as n increases.
To apply this torsional hysteretic model, the column shall have the similar aspect
ratio, amount of reinforcements, and axial load level with the specimen used in the exper-
iment. In addition, because the lateral drift and the torsional rotation were simultaneously
applied to the columns with the ratio r being constant during the experiment, the empirical
torsional hysteretic model can be applied for the column which is subjected to in-phase
lateral drift and torsional rotation.
The proposed torsional hysteresis is used in a rotational spring which connects
between the top and the bottom of a pier (refer to Fig. 3). However, the deterioration of
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 985

Torsion
~ ~
Primary curve θm ,Tm n = 0
n=1

Unloading at
~ ~
θ ≥ 0.95θm
n=2
Rotation

Hysteresis experiences zero


or opposite side rotation

FIGURE 5 Condition for counting the index n for the effect of number of loading cycles.

flexural strength of the pier due to the interaction with torsion is disregarded in this analy-
sis. Since r is not known before analysis, r is first assumed and iteration is conducted until
computed r corresponding to the maximum torsional rotation of pier becomes close
enough to the assumed r.
Generally, bridge piers oscillate bilaterally under a three-dimensional excitation.
Since the proposed torsional hysteretic model is obtained based on combined torsional
rotation and unilateral drift, the vector sum of bilateral displacement is used to evaluate r
for the hysteretic model.
Table 1 shows the parameters which are used to determine the torsional primary
curves under pure torsion. Torsional strength of a column can be estimated by the space
truss analogy theory which is available in many current design specifications
[New Zealand Concrete Design Committee (P3101), 1995; ACI Committee 318, 2002;
Japan Society of Civil Engineers, 2002]. By providing some modification, the torsional
capacity of column, Tc, can be approximated as

Tc = 2l Am qw ql (3)

in which

Am = bo do (4)

TABLE 1 Parameters of the torsional primary curves


Torsional properties P1& P3 P2
Torsional strength Tc, MNm 29.8 29.5
Initial torsional stiffness kT1, MNm/rad 7.80 × 103 7.80 × 103
Cracking torsion Tcr, MNm 12.1 11.9
Yield rotation qy, rad 0.00382 0.00379
Rotation when the torsion starts to deteriorate qd, rad 0.0240 0.0238
Rotation at the ultimate qu, rad 0.0458 0.0454
986 P. Tirasit and K. Kawashima

qw = Atw fwy s (5)

ql = ∑ Atl fly u (6)

l = 1 + 1.193sr . (7)

Am is the effective area for torsion, bo is the length of shorter side of transverse reinforce-
ment, do is the length of the longer side of transverse reinforcement, Atw is the area of a
single bar of transverse reinforcement, ∑ Atl is the area of the longitudinal reinforcement,
s is the longitudinal spacing of transverse reinforcement, u is the length of the centerline of
transverse reinforcement which is equal to 2(bo +do), and fwy and fly are the yield strengths
of the transverse and longitudinal reinforcement, respectively. The modification is given
to Eq. (3) by introducing an empirical coefficient λ which is based on the authors’ experi-
mental result in order to take the effect of axial compressive load. sr is the ratio between
the average axial compressive stress, which is evaluated from the axial loads at the top and
the bottom of column, and the concrete strength.
The Standard Specification for Concrete Structure-2002 [Japan Society of Civil
Engineering, 2002] provides a torsional strength limit in order to avoid the diagonal com-
pression failure in concrete before the yielding of torsional reinforcement. By removing
the safety factor, the limit of torsional strength, Tcu, can be estimated as

Tcu = K t fwc (8)

where

fwc = 1.25 fc′ (9)

K t = h x 2 y. (10)

fwc is the limit of shear stress to avoid the diagonal concrete crushing, fc′ is the concrete
compressive strength, Kt is the torsional constant related to the shorter and longer sides, x
and y, of the rectangular section, and h = 0.208 for the square section.
According to Table 1, P2 has slightly smaller torsional strength than P1 and P3 due to
smaller dead load. Other parameters are evaluated in accordance with the authors’ pro-
posed torsional hysteretic model [Tirasit and Kawashima, 2007].

3.2. Bearings
In reality, there is not any fixed bearing that fully connects bridge deck with substructure
and there is not any “friction-free” movable bearing neither. The behavior of bearings can
influence the overall response of the bridge and thus they should be simulated as close as
possible to the realistic. In this study, fixed and movable bearings are idealized, as shown
in Fig. 6. Fixed bearing is assumed to behave elastically before reaching its rupture
strength, Fy, at displacement, Δy, and the upper bearing slides relative to the lower bearing
as soon as the bearing fails. The lateral restoring force of a fixed bearing after failure is
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 987

Force Force Force


Fy –K Movement
gap, Δb kI
Δy
Fdf Fdf
K K
K
Displacement Displacement Displacement
–Fdf –Fdf
kI Movement
K
–Fy gap, Δb

(a) (b) (c)

FIGURE 6 Idealization of bearings: (a) fixed bearings; (b) movable bearings; and
(c) lock of bearings after failure.

TABLE 2 Rupture strength of bearings


Transverse Longitudinal
Location direction, kN direction, kN
Piers P1, P2, P3 5.75 × 102 5.75 × 102
Abutments A1, A2 2.23 × 102 –

controlled by the friction force, Fdf. On the other hand, movable bearing can slide only in
the longitudinal direction as shown in Fig. 6(b). Table 2 shows the rupture strengths of the
bearings assumed in analysis. Rupture displacement, Δy, and friction coefficient of the
bearings after failure are assumed as 1 and 0.05 mm, respectively.
If sliding between the upper and lower bearing takes place smoothly after failure, the
behavior of a bridge supported by damaged bearings is similar to a bridge supported by
isolators. However, it was often the case in past earthquakes that bearings which were bro-
ken into several blocks locked relative movement between the upper and lower bearings.
Furthermore, if the upper bearings dislodged from the lower bearings, the upper bearings
could not return to the rest position which forced the relative displacement between upper
and lower bearings to accumulate only in one direction. Once bearings are locked, a large
lateral force is transferred from the deck to the substructures. Therefore, the effect of lock
of bearings is included in analysis by idealizing the force-displacement hysteresis as
shown in Fig. 6(c). Lock of bearing is assumed to occur when the relative displacement
between upper and lower bearings exceeds a certain gap, Δb, in positive or negative direc-
tion. The stiffness of a lock element is assumed to be equal to the impact spring stiffness
shown below.

3.3. Restrainer System and Pounding


The force-displacement hysteresis of the cable restrainers is idealized, as shown in
Fig. 7(a). A cable restrainer begins to resist tension when the relative displacement
between the deck and the abutment is larger than the restrainer gap. The residual deforma-
tion of restrainer accumulatively occurs after yielding and the restrainer begins to resist
tension again when its elongation reaches the residual deformation. The restrainer gap is
assumed to be 150 mm. The design force of cable restrainers is evaluated based on the
Seismic Design Specification for Highway Bridges [Japan Road Association, 2002] so
that it is equal to 1.5 times the reaction force due to the dead load of superstructure. Hence,
988 P. Tirasit and K. Kawashima

Force Force
Yield
displacement Gap, xg
Fy

Displacement
Collision Separation
Displacement kI
Restrainer
gap
(a) (b)

FIGURE 7 Idealization of cable restrainers and pounding: (a) a cable restrainer; and
(b) pounding.

3 restrainers with 4-f17.8mm SWPR19N (3@1.32 = 3.96 MN yield force) are used at
each end of the bridge.
The elastic pounding between the corners of the deck and the abutments is included in
the study. Force-displacement hysteresis for pounding is written as (refer to Fig. 7(b))

⎧⎪ 0 x > − xg
Fimp = ⎨
( )
(11)
− k − xg − x x ≤ − xg
⎩⎪ I

where Fimp is the impact force, kI is the stiffness of impact spring, and x and xg are the rel-
ative displacement and the gap width between the corners of the deck and the abutment,
respectively. Poundings occur along axis-1 (refer to Fig. 2(a)). The pounding gap is
assumed to be 100 mm. The stiffness of impact spring, kI, is given as [Kawashima and
Penzien, 1976; Watanabe and Kawashima, 2004]

g nEA
kI = (12)
L

where EA and L are the axial stiffness and the length of a bridge deck, respectively; n is
the total number of elements; and g is the stiffness ratio which is recommended to be
nearly 1.

3.4. Input Ground Motions


To investigate the seismic behavior of the bridge, NS, EW, and UD components of the JR
Takatori and JMA Kobe ground motions which were measured during 1995 Kobe, Japan
earthquake are imposed to the bridge in the longitudinal, transverse, and vertical direc-
tions, respectively. Figure 8 shows the two ground accelerations. The time history analysis
is conducted using the Newmark-b method. The constant acceleration method is used in
the numerical integration of equations of motion with the time interval of 2x10-4 s.

4. Seismic Response of the Skewed Bridge


Because a skewed bridge oscillates bilaterally and rotates around the vertical axis under a
three-dimensional excitation, the outermost bearings are generally subjected to larger
deformation than the inner bearings. As the deformation of the bearing increases, the
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 989

Acceleration (m/s2)

Acceleration (m/s2)
10 10
NS component NS component
5 5
0 0
–5 –5
–10 –10
Acceleration (m/s2)

Acceleration (m/s2)
10 10
EW component EW component
5 5
0 0
–5 –5
–10 –10
Acceleration (m/s2)

Acceleration (m/s2)
10 10
UD component UD component
5 5
0 0
–5 –5
–10 –10
0 5 10 15 20 25 30 0 5 10 15 20
Time (s) Time (s)
(a) (b)

FIGURE 8 Ground accelerations: (a) JR Takatori; and (b) JMA Kobe.

possibility of failure and lock of the bearing increases. Because the outermost bearings at
each support have the equal possibility to fail and lock, it is assumed here that the right-
most bearing at P2 locks in the longitudinal direction. The movement gap, Δb, is assumed
to be 10 mm in both positive and negative directions (refer to Fig. 6(c)).
The seismic response of the skewed bridge with lock of the rightmost bearing at P2 is
analyzed under JR Takatori ground motion. The effects of pounding between the deck and
the abutments as well as the cable restrainers are considered in the analysis. It is found that
r of P2 converges to 1.09.
Figure 9 shows the deck accelerations at A1 and impact forces resulted from pound-
ing between the deck and A1. Poundings result in large lateral force (7.95MN) which is
32% of the total deck weight. High spike accelerations occur at the deck where pounding
takes place. As shown in Fig. 10, several acceleration spikes are seen in the deck response
at P2 in the longitudinal direction. This is resulted from not only the pounding between the
deck and the abutments but also the impact at the locked bearing which is shown later.
The peak acceleration of the deck at P2 is 1.16 g in the longitudinal direction and 0.92 g in
the transverse direction. The maximum deck displacements are 0.165 and 0.159 m in the
longitudinal and transverse directions, respectively.
In-plane deck rotation at P2 accumulatively increases in the counter-clockwise direc-
tion every time when pounding between the deck and A1 or A2 takes place. The maxi-
mum peak rotation of the deck reaches 0.00242 rad. As a result of the in-plane deck
rotation, large response displacement of the deck occurs in the transverse direction at both
ends of the bridge. The maximum transverse displacements of the deck at A1 and A2
reach 0.273 and 0.314 m, respectively.
From the analysis it is found that all bearings fail at the early stage of excitation. The lat-
eral restoring forces of the center and the rightmost bearings at P2 in the longitudinal direc-
tion are shown in Fig. 11. Five bearings at P2 reach their rupture strength of 0.58 MN and
990 P. Tirasit and K. Kawashima

10 8

Pounding (MN)

Acceleration (g)
Obtuse corner 5.28 g Longitudinal

0 0

7.95 MN
–10 –8

10 8

Acceleration (g)
Pounding (MN)

Acute corner Transverse

0 0

7.42 MN 5.90 g
–10 –8
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) (b)

FIGURE 9 Response of bridge deck at A1 under JR Takatori ground motion: (a) pounding
force between deck and A1; and (b) deck response acceleration at A1.
Displacement (m)

1.5 0.2 0.165 m


Acceleration (g)

Longitudinal Longitudinal

0 0

1.16 g
–1.5 –0.2
Displacement (m)

1.5 0.2
Acceleration (g)

Transverse 0.159 m Transverse


0.92 g

0 0

–1.5 –0.2
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) (b)

0.003 0.00242 rad


Inplane rotation
(rad)

–0.003
0 5 10 15 20 25 30
Time (s)
(c)

FIGURE 10 Response of bridge deck at P2 under JR Takatori ground motion: (a) accel-
eration; (b) displacement; and (c) inplane deck rotation.

fail in both the longitudinal and transverse directions. The leftmost bearing fails at 1.00 s,
followed by the failure of the center and rightmost bearings at 1.01 and 1.04 s, respectively,
in the transverse direction. Soon after this, the rightmost bearing fails at 1.09 s, followed by
the failure of the center and leftmost bearings at 1.37 and 1.41 s, respectively, in the longitu-
dinal direction. After the failure, the upper bearings slide on the lower bearings and only the
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 991

0.58 MN

Bearing force (MN)

Bearing force (MN)


1 6 5.30 MN
0.5 0.58 MN
0 0
–0.5
–1 –6
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) (b)

FIGURE 11 Lateral restoring force of bearings at P2 in the longitudinal direction under


JR Takatori ground motion; (a) center bearing; and (b) rightmost bearing.

friction forces are transferred from the deck to the substructure. Lock first occurs at the right-
most bearing on P2 at 1.61 s in the longitudinal direction. The maximum force induced by
the lock reaches 5.30 MN at 4.84 s which is about 21% of the superstructure weight.
Figure 12 shows the moment responses of P1 and P2 along axes 1 and 2. The maximum
moment at P2 along axis-2 is 62.1 MNm at 4.83 s. It is noted that this maximum bending
moment is induced by the above-mentioned lock force at the rightmost bearing. On the other
hand, the maximum moment at P1 along axis-2 is 58.5 MNm. Moment-curvature hystereses
of the piers along axis-1 and axis-2 are shown in Fig. 13 It is seen that P2 has larger ductility
demand in both axes-1 and 2 than P1 due to the lock of rightmost bearing. The maximum
curvatures of P2 along axes-1 and 2 are 3.51 × 10−3 and 4.66 × 10−3 1/m, respectively, which
are about 4.4 times and 1.7 times larger than those of P1 along axes-1 and 2, respectively.
Torsion responses of P1 and P2 are shown in Fig. 14. The torsion response of P3 is not
shown here since it is similar to that of P1. The peak torsion of P2 is 20.1 MNm which is
5.4 times larger than the peak torsion of P1 (3.75 MNm). However, the maximum rotation of
P2 is 0.00689 rad which is 13.0 times larger than that of P1 (0.000531 rad). A small residual
rotation of 0.00043 rad takes place in P2. Figure 15 shows the torsional hysteresis of P2. The
torsional hysteresis of P1 is not shown here since it remains in the elastic range. To evaluate
the response ductility of torsion in the pier, torsion ductility ratio, μT, is defined here as

q%max
mT = (13)
q%y
Moment (MNm)

80 80
Moment (MNm)

Axis-1 Axis-1

0 0
– 29.9 MNm – 42.6 MNm
– 80 – 80
80 80
Moment (MNm)

Moment (MNm)

Axis-2 Axis-2

0 0

– 58.5 MNm – 62.1 MNm


– 80 – 80
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) Pier P1 (b) Pier P2

FIGURE 12 Bending moment responses of P1 and P2 under JR Takatori ground motion.


992 P. Tirasit and K. Kawashima

80 80
Axis-1 Axis-2
0.79 × 10–3 1/m
40 40

Moment (MNm)

Moment (MNm)
0 0

– 40 – 40

– 2.69 × 10–3 1/m


– 80 – 80
–4 0 4 –4 0 4
Curvature × 10–3 (1/m) Curvature × 10–3 (1/m)
(a)

80 80
Axis-1 Axis-2
3.51 × 10–3 1/m
Moment (MNm)

40 40

0
Moment (MNm) 0

– 40 – 40

– 4.66 × 10–3 1/m


– 80 – 80
–4 0 4 –4 0 4
Curvature × 10–3 (1/m) Curvature × 10–3 (1/m)
(b)

FIGURE 13 Moment-curvature hystereses in the piers under JR Takatori ground motion:


(a) pier P1; and (b) pier P2.

4 3.75 MNm 30
Torsion (MNm)

Torsion (MNm)

20.1 MNm
2
0 0
–2
– 20.1 MNm
–4 – 30
Torsional rotation
Torsional rotation

0.001 0.008 0.00689 rad


0.00531 rad
(rad)
(rad)

0 0

– 0.001 – 0.008
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) (b)

FIGURE 14 Torsional response of piers under JR Takatori ground motion: (a) pier P1;
and (b) pier P2.

where q%max is the maximum torsional rotation of the column and q%y is the yield torsional
rotation of column under combined bending and torsion. Because q%y is 0.0027 rad, μT
is 2.55 at P2. After experiencing the peak rotation of 0.00689 rad, the unloading stiffness
deteriorates to 3.67 × 103 MNm/rad which is about half of the initial torsional stiffness.
The reloading paths direct to the point of deteriorated torsion at the maximum unloading
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 993

30
0.00689 rad
20

Torsion (MNm)
10
kunload
0

–10

–20

–30
–0.008 0 0.008
Rotation (rad)

FIGURE 15 Torsional hysteresis of P2 under JR Takatori ground motion.

rotation in both positive and negative torsions. The pinching in the torsional hysteresis has
not yet developed and the maximum index n for counting the number of loading cycles is 1.
Similar seismic responses are obtained under JMA Kobe ground motion. Table 3
compares the peak responses of P2 under JR Takatori and JMA Kobe records. r of P2 con-
verges to 0.8 under JMA Kobe ground motion. The effect of torsion in P2 under JMA
Kobe ground motion is less significant than that under JR Takatori ground motion. P2 also
yields in torsion under JMA Kobe ground motion and results in μT of 1.67. The yield tor-
sion is 17.5 MNm under JMA ground motion that is 87% of that under JR Takatori ground
excitation. This is resulted from the smaller lock force in bearing at P2 under JMA Kobe
earthquake.
In order to study torsional response under more significant excitation, an analysis is
conducted for the same bridge by only increasing the intensity of JR Takatori ground
motion twice. The torsional hysteresis of P2 is shown in Fig. 16. r obtained by Eq. (2) con-
verges to 0.65. The maximum torsional rotation is 0.0199 rad which results in μT of 9.27.
This maximum rotation is 2.9 times larger than that of bridge under normal level of JR
Takatori ground motion. Pinching occurs significantly in the torsional hysteresis in the
positive side after P2 experiences the maximum rotation. The maximum n is still 2 which
remains in the range of the model applicability.

TABLE 3 Peak responses of bridge


Ground motion JR Takatori JMA Kobe
Idealization for pier torsion Linear Nonlinear Linear Nonlinear
Deck response at P2
- Acceleration (longitudinal direction), g 1.19 1.16 1.28 1.26
- Displacement (longitudinal direction), m 0.166 0.165 0.116 0.118
- Inplane rotation, rad × 10–3 2.45 2.42 1.03 1.07
P2
- Torsion, MNm 28.7 20.1 22.8 17.5
- Torsional rotation, rad × 10–3 3.69 6.89 2.93 3.97
- Moment (axis-1), MNm 43.2 42.6 43.3 43.0
- Curvature (axis-1), 1/m × 10–3 3.77 3.51 2.41 2.33
- Moment (axis-2), MNm 62.2 62.1 63.3 63.3
- Curvature (axis-2), 1/m × 10–3 4.85 4.67 4.56 4.54
Lock force of bearing, MN 5.69 5.30 5.14 4.85
994 P. Tirasit and K. Kawashima

20

0.0199 rad
10

Torsion (MNm)
0

– 10

– 20
– 0.01 0 0.01 0.02
Rotation (rad)

FIGURE 16 Torsional hysteresis of P2 under 200% JR Takatori ground motion.

80 80
Axis-1 Axis-2
40 40
Moment (MNm)

Moment (MNm)

0 0

– 40 – 40
– 8.94 × 10–3 1/m
– 30.2 × 10–3 1/m
– 80 – 80
– 10 0 10 – 40 – 20 0 20 40
Curvature × 10–3 (1/m) Curvature × 10–3 (1/m)
(a) (b)

FIGURE 17 Moment-curvature hystereses in the piers under 200% JR Takatori ground


motion: (a) axis-1; and (b) axis-2.

The moment curvature hystereses of P2 under 200% JR Takatori ground motion are
shown in Fig. 17. The maximum curvatures of P2 along axes-1 and 2 are 8.94x10-3 and
30.2x10−3 1/m, respectively, which are about 2.5 times and 6.5 times larger than those of
P2 along axes-1 and 2 under normal level of JR Takatori ground motion, respectively.

5. Influence of Linear and Nonlinear Torsional Hystereses of Piers


Because an appropriate hysteretic model for torsion was not available, piers which are
subjected to combined flexure and torsion were generally assumed to be linear elastic in
torsion in the past seismic response analysis. This assumption, however, is applicable
only when the effect of torsion is less significant. Hence, a response analysis on the
same bridge is conducted here to see how different the responses of the bridge are
between linear and nonlinear idealizations on the torsional hystereses of the piers. The
elastic torsional stiffnesses of the piers are determined based on the proposed model by
assuming r = ∞.
Figure 18 shows the torsion responses of P2 with linear and nonlinear torsional hys-
tereses under JR Takatori ground motion. It is seen that the peak torsion of 28.7 MNm in
P2 with linear torsional hysteresis is about 1.4 times larger than that of 20.1 MNm with
torsional nonlinearity. Obviously, the linear idealization for torsion overestimates the
torsional strength of the pier and the energy dissipation during unloading and reloading
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 995

Torsional rotation
30 28.7 MNm 0.008

Torsion (MNm)
Nonlinear 0.00369 rad
Linear

(rad)
0 0

– 30 – 0.008
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time (s) Time (s)
(a) (b)
30
0.00369 rad
20
Torsion (MNm) 0.00689 rad
10

– 10

– 20

– 30
– 0.008 0 0.008
Rotation (rad)
(c)

FIGURE 18 Torsional responses of P2 with linear and nonlinear torsional hystereses:


(a) torsion; (b) torsional rotation; and (c) torsional hysteresis.

hysteresis cannot be considered in analysis. The maximum rotation of P2 with nonlinear


torsion, 0.00689 rad, is about 1.9 times of the maximum rotation of P2 with linear torsion,
0.00369 rad. Table 3 summarizes the peak responses of P2 with linear torsion under JR
Takatori ground motion.
Effect of the linear and the nonlinear torsional hystereses on the moment-curvature
hystereses of P2 is shown in Fig. 19. P2 with linear torsion experiences slightly larger
maximum curvatures of 3.77×10-3 1/m and 4.85×10-3 1/m along axis-1 and axis-2, respec-
tively. This is attributed to the larger bearing lock force at P2 with linear torsion. It is
noted that the effect of torsional hysteresis of piers is not significant on the deck
responses. The peak deck displacements and inplane deck rotations of both linear and non-
linear torsion are very close to each other as shown in Table 3. This is because the flexural

80 80
Nonlinear
Linear 3.77 × 10–3 1/m
40 40
Moment (MNm)

Moment (MNm)

0 0

– 40 – 40

– 4.85 × 10–3 1 / m
– 80 – 80
–4 0 4 –4 0 4
Curvature × 10–3 (1/m) Curvature × 10–3 (1/m)
(a) (b)

FIGURE 19 Effect of linear and nonlinear torsional hystereses on the moment-curvature


hystereses of P2 under JR Takatori ground motion: (a) axis-1; and (b) axis-2.
996 P. Tirasit and K. Kawashima

deformations of substructure have larger contribution to the global responses of the deck
than their torsional deformations. Similar responses are also seen under JMA Kobe ground
motion as shown in Table 3.
It is worth noting that the deck movement depends on not only the torsional stiffness
but also the flexural stiffnesses of the piers. However, the contribution of the in-plane
deck rotation to the torsion in the pier may be more significant in a single or two span con-
tinuous skewed bridge supported by fewer piers. Furthermore, the effect of torsion in the
pier is possibly more severe if the pier is rigidly connected to the deck because the torsion
due to the in-plane deck rotation can be fully transferred to the pier. Therefore, idealiza-
tion of nonlinear torsional hysteresis of the pier in such bridges may be more important.

6. Expected Damage of P2 due to Combined Torsion and Bending


According to the bridge responses under JR Takatori ground motion, peak torsional
rotation of P2 is 0.00689 rad and the corresponding flexural drift is 0.631% (r = 1.09).
Figure 20 shows the damage developed in a model column which was loaded in the experi-
ment under combined cyclic loading of q = 0.005 rad and Δ = 0.5% (r = 1), and q = 0.01 rad
and Δ = 1% (r = 1) [Tirasit and Kawashima, 2007]. Therefore, it is anticipated that simi-
lar pattern and degree of damage within the range shown in Fig. 20 may be developed in
P2 under JR Takatori ground motion.
Based on the experimental study, the level of combined bending and torsion of r = 1
results in about 10 and 15% deterioration of the flexural strength and the ultimate dis-
placement, respectively. Hence, the effect of torsion in the pier of this bridge is not negli-
gible under JR Takatori ground motion.
On the other hand, under 200% JR Takatori ground motion, the maximum torsional
rotation of P2 is 0.0199 rad and the corresponding flexural drift is 3.06% (r = 0.65).
Figure 21 shows the damage developed in the model columns under combined cyclic
loading of q = 0.015 rad and Δ = 3% (r = 0.5), and q = 0.03 rad and Δ = 3% (r = 1).
Hence, the comparable pattern and degree of damage within the range shown in Fig. 21
may occur in P2 under 200% JR Takatori ground motion.
This clearly shows how important the seismic torsion is and that the effect of seismic
torsion should be included in the design of skewed bridges. The effect of bearing locking
after failure is important since it can produce large eccentric force which results in signifi-
cant torsion to the substructure. Therefore, bearing shall be carefully designed. Moreover,
eccentric bearing locking probably takes place in straight bridge that may give rise to the
similar results to this analysis. Since the torsional hysteretic model here can be applied to
the columns which have the similar aspect ratio and the amount of reinforcement to the
specimens in the experiment and the deterioration of flexural strength of column has not

N W S E N W S E
900

600

300

0
(a) (b)

FIGURE 20 Expected damage at the maximum torsional rotation of P2 under JR Takatori


ground motion: (a) 0.005 rad - 0.5% drift (r = 1); and (b) 0.01 rad - 1% drift (r = 1).
Effect of Nonlinear Seismic Torsion on Skewed Bridge Piers 997

N W S E N W S E
900

600

300

0
(a) (b)

FIGURE 21 Expected damage at the maximum torsional rotation of P2 under 200% JR


Takatori ground motion: (a) 0.015 rad - 3% drift (r = 0.5); and (b) 0.03 rad - 3% drift (r = 1).

been considered yet, there is a plenty room to extend this model to be applicable to the
general cases.

7. Conclusions
A numerical analysis was conducted to investigate the effect of seismic torsion on the per-
formance of a skewed bridge. An empirical nonlinear torsional hysteretic model based on
the experimental study was applied to take account of the nonlinearity in torsion of the
piers. The fiber element was used to consider the interaction of two principal bending
moments and axial load. Based on the result presented herein, the following conclusions
may be deduced:
1. The effect of bearing lock results in large impact force transferred from the superstruc-
ture to P2. P2 yields in torsion when lock of bearing occurs at the rightmost bearing.
2. Under JR Takatori and JMA Kobe ground motions, the torsional hystereses of P2
remain in the range of contant torsion of the primary curve. Even though the effect
of pinching is less significant under this range, sharp deterioration of torsional
stiffness occurs in the unloading paths of hystereses.
3. The eccentric impact force due to the lock of bearing results in not only the sharp
increase of seismic torsion but also the increase of flexural ductility demand in P2.
4. P2 with consideration of nonlinear torsion gives the larger maximum torsional
rotation than that with linear torsion. This is because use of linear torsion hystere-
sis overestimates the torsional strength of the pier and the energy dissipation dur-
ing unloading and reloading cannot be included.
5. Based on the rotation-drift ratio, the maximum torsional rotation and the corre-
sponding lateral drift of P2, damage patterns and the performance of the piers can
be evaluated based on the experimental results. The effect of torsion in P2 is sig-
nificant and cannot be neglected. Therefore, the effect of seismic torsion shall be
taken into account in the design of bridge columns.

References
ACI Committee 318 [2002] Building Code Requirements for Structural Concrete (ACI 318–02) and
Commentary (ACI 318R–02), American Concrete Institute, Michigan, USA.
Desroches, R. and Fenves, G. L. [1997] “Evaluation of recorded earthquake response of a curved
highway bridge,” Earthquake Spectra 13, 363–386.
Editorial Committee for the Report on the Hanshin-Awaji Earthquake Disaster [1997] Report on the
Hanshin-Awaji Earthquake Disaster Vol. 4 (in Japanese), JSCE, Tokyo, Japan.
998 P. Tirasit and K. Kawashima

Hoshikuma, J., Kawashima, K., Nagaya, K. and Taylor, A. W. [1997] “Stress-strain model for con-
fined reinforced concrete in bridge piers,” Journal of Structural Engineering 123(5), 624–633.
Hsu, H.-L. and Wang, C.-L. [2000] “Flexural–torsional behavior of steel reinforced concrete
members subjected to repeated loading,” Earthquake Engineering and Structural Dynamics 29,
667–682.
Hsu, H.-L. and Liang, L.-L. [2003] “Performance of hollow composite members subjected to cyclic
eccentric loading,” Earthquake Engineering and Structural Dynamics 32, 443–461.
Japan Road Association [2002] Specifications for Highway Bridges – Part V Seismic Design,
Tokyo, Japan.
Japan Society of Civil Engineers [2002] JSCE Guidelines for Concrete No.3 - Standard Specifica-
tions for Concrete Structures-2002 “Structural Performance Verification,” Maruzen, Tokyo.
Kawashima, K. and Penzien, J. [1976] “Correlative investigation on theoretical and experimental
dynamic behavior of a model bridge structure,” Report No. UCB/EERC-76/26, Earthquake
Engineering Research Center, University of California, Berkeley, USA.
New Zealand Concrete Design Committee (P3101) [1995] Concrete Structures Standard: Part 1 –
The Design of Concrete Structures (NZS3101:Part1:1995), Standards New Zealand, Wellington,
New Zealand.
Otsuka, H., Takeshita, E., Yabuki, W., Wang, Y., Yoshimura, T., and Tsunomoto, M. [2004] “Study
on the seismic performance of reinforced concrete columns subjected to torsional moment, bend-
ing moment and axial force,” 13th World Conference on Earthquake Engineering, Vancouver,
Canada, Paper No. 393.
Priestley, M. J. N., Seible, F., and Calvi, G. M. [1996] Seismic Design and Retrofit of Bridges, John
Wiley & Sons, Inc., New York.
Sakai, J. and Kawashima, K. [2003] “Modification of the Giuffre, Menegotto and Pinto model for
unloading and reloading paths with small strain variations,” Journal of Structural Mechanics and
Earthquake Engineering, 738/I-64, 159–169 (in Japanese).
Sakai, J. and Kawashima, K. [2006] “Unloading and reloading stress – strain model for confined
concrete,” Journal of Structural Engineering, ASCE, 132(1), 112–122.
Tirasit, P. and Kawashima, K. [2007] “Seismic performance of square reinforced concrete columns
under combined cyclic flexural and torsional loadings,” Journal of Earthquake Engineering
11(3), 425–452.
Watanabe, G. and Kawashima, K. [2004] “Effectiveness of cable–restrainer for mitigating rotation
of a skewed bridge subjected to strong ground shaking,” 13th World Conference on Earthquake
Engineering, Vancouver, Canada, Paper No. 789.

You might also like