You are on page 1of 18

www.sciencedirect.

com

1. Introduction
A great deal of research effort has been made to improve the design and retrofit methodology for
bridge structures in seismic active zones, especially after the 1994 Northridge, 1995 Kobe, and
1999 Chi–Chi earthquakes. Lessons [1] learned from the investigation of highway bridges
seriously damaged by these destructive earthquakes are highly valuable to the upgrading of many
bridges still in use, as these bridges probably are as seismically deficient as those that already
failed the test of nature.

General public would think girders, piers, and foundations receive the most attention in the
design and construction of a bridge structure. Nonetheless, secondary components such as rubber
bearings, hinge restrainers, side stoppers, and expansion joints if not well taken care of are likely
to result in undesirable overall performance of a bridge structure during extreme quake events
[2].

The rubber bearings commonly used in past decades have been seriously questioned as to their
seismic resistance. Expansion-type bearings are vulnerable to un-seating or toppling when deck
sliding is excessive. Fix-type bearings may break loose if their hinge-restrainer device is sheared
off by lateral seismic loading. New techniques in the form of an isolator or a damper device are
introduced to replace or enhance conventional bearings [2] and [3].

Previous studies have shown that the gap of joints between decks can have a significant effect on
the response of a bridge [3] and [4]. Using simplified model, it was found when sliding and
pounding occurs, accelerations of bridge decks may increase by a factor of ten, as compared to
that without pounding [3]. The abrupt increase of accelerations can result in severe impact forces
that damage structural members like the deck or pier.

Meanwhile, assessment of seismic vulnerability of common types of bridges based on techniques


such as fragility analysis is getting more attention. A study on the fragility of a bridge's inventory
shows that the most vulnerable bridge type is the multi-span simply supported concrete bridges
[5].

This study presents a case analysis involving deck sliding and pounding of a multi-span simply
supported concrete bridge damaged by the 1999 Chi–Chi earthquake. Its reference coordinate is
(120°46″E, 24°17″N). A brief description of its damages and related photos can be found in
public reports [6]. Bridges of the same kind are quite common on highway river-crossings all
www.sciencedirect.com

over Taiwan. Site investigation and survey data for this particular bridge collected following the
quake indicated that major structural responses that occurred on that day were the sliding of
several deck spans relative to pier cap beams and the pounding between adjoining decks, all
along the traffic direction. One extra pounding occurred between a special recessed pier cap
beam and a deck supported on this pier.

2. Bridge characteristics and observed damages


A brief description of the bridge of concern to this study is presented in this section, while more
details including bearings and gaps of joints are introduced along sections that follow.
Constructed in 1984, the bridge has thirteen segments with abutments at two ends. Fig. 1
provides an elevation view covering four segments on the south end. Each span of the bridge
superstructure (i.e. deck) is made of concrete slab cast onto pre-cast T-girders laterally braced by
diaphragms. Fig. 2 illustrates the transverse section view of a typical bent. Along the traffic
direction, T-girders are simply supported on top of a cap beam (i.e. wing beam) cast in one piece
with straight pier column, and embedded into a caisson foundation at the column base. In the
transverse direction, concrete side stoppers are installed to prevent excessive sidewise
movement. As the designer called for, each girder is simply supported with one end on a fix-type
rubber bearing having dowel bars (i.e. hinge) and another on an expansion-type rubber bearing
(i.e. roller).

Full-size image (54K)


Fig. 1. Elevation view of four segments at south end; the bridge has 13 segments:
13+11@35+13 m.

View Within Article


www.sciencedirect.com

Full-size image (54K)


Fig. 2. Transverse section view of a typical bridge bent (unit: cm).

View Within Article

On 21 September 1999, shortly after midnight, a violent quake struck. What was left on the
bridge site a minute later were two spans of massive decks falling off at the south end and other
deck spans being displaced under South–North sliding motion. Fig. 3 presents the schematic
view of the damaged bridge. One particular pier appeared to have been pounded at its cap beam,
tilted 10° off the straight position, and left crushing marks at the cap beam and flexural cracks at
the column base (in Fig. 3 where star marks show). It is worth noticing in Fig. 1 and Fig. 3 that
the top of this pier cap beam is not built flush, in contrast to other piers, but has a special
recessed back-wall to accommodate girders of different depths from two adjoining deck spans.
Nevertheless, T-girders, piers, and caissons, all remained intact as individual members, except
one single pier severely damaged as mentioned. In terms of kinetics, the decks simply acted as
rigid masses, and piers as flexible columns fixed at the base in response to the quake.

Full-size image (53K)


Fig. 3. Observed damage condition of the bridge: two fallen decks and one severely damaged
pier.

View Within Article


www.sciencedirect.com

Based on damages data collected form public reports and several rounds of the author's personal
field trips, it is reasonably speculated that the fix–expansion bearings, as required by the design
engineer for each girder, were not duly in place as constructed. None of the dowel bars necessary
for fix-type bearings could be found, not even at the exposed end sections of the two fallen
decks. Fig. 4 provides a blowout view of the bearing details according to original design
drawings.

Full-size image (57K)


Fig. 4. Details of bearings at girder support.

View Within Article

3. Dynamic model of the bridge


To ensure setting up analysis models representative of bridge behaviors, as-built bridge plans
were collected and examined to determine dimensions of deck spans and pier cap beams and
columns, details of fix/expansion bearings, and material specifications of all material used. Each
of the eleven typical spans between piers is 35 m long. Each of the two end spans leading to
abutments is 13 m long. Each bent of the bridge has one single pier with cap beam and on top are
five girders simply supported using common rubber bearing pads, one pad at each girder end.
The girder end supports are classified as fix- or expansion-type, commonly known as hinge or
roller. The fix-type girder support on the pier cap beam consists of rubber bearing pads with
dowel bars projecting about 300 mm into the pier cap and 150 mm into the T-girder, as shown in
Fig. 4. The average column height of all piers is 5.9 m. Adjoining decks and abutments are
separated by expansion joints with gaps of 40 mm in between. Each deck superstructure consists
of a 13 m wide concrete slab and five 2-m-deep precast T-girders. A typical concrete pier
column having an oblong cross section 2 m deep by 4 m plus wide is built into a huge caisson
www.sciencedirect.com

having a circular cross section of 6 m in diameter and a height of 12 m, dug deep into the sandy
rock stratum under the riverbed.

Under extreme seismic actions, this type of bridge, just discussed, may exhibit highly nonlinear
behaviors, such as the yielding of dowel bars and sliding at rubber bearings, pounding between
decks at expansion joints, cracking and yielding of concrete pier columns, and soil–structure
interaction near the foundation. Ideally, an analytical model need incorporate all of these in the
dynamic analysis.

Damage conditions of the bridge indicated that its major response to seismic action took place in
the traffic direction. Except the tilted pier, all other pier columns remained intact. The ground
conditions at the site were normal, though heaved pavement approaching the south end abutment
was visible. It might be the outcome of severe pounding of the end deck onto the back-wall of
the abutment. Altogether, these indicated pier columns had a fairly rigid base at foundations. In
effect, the bridge responded mainly along the traffic direction, counting on the stiffness and
strength provided by bearings and pier columns.

The study adopts a practical but realistic approach to setting up analytical models of the bridge.
Nonlinear behaviors incorporated in models include sliding and friction at bearings, closing and
pounding at expansion joints. The stiffness of a bridge's deck bodies such as girders and slabs
have insignificant effect on the seismic response of the bridge. As shown in Fig. 5, the bridge is
idealized as a two-dimensional discrete model with deck bodies lumped as rigid masses,
supported by bearings, seated on top of pier columns, and subjected to violent ground motion
along both longitudinal and vertical directions. Its dynamic response is to be numerically solved
for in the traffic direction of the bridge.

Full-size image (43K)


Fig. 5. Analytical model for a bridge showing masses and nonlinear elements.
www.sciencedirect.com

View Within Article

The analytical model consists of four spans: three typical spans together with a shorter span
connecting to the abutment. For bridges with more spans under seismic motions in the
longitudinal direction, the maximum possible one-way sliding toward the river centerline is a
variable value bounded by the total sum of gaps in a succession of expansion joints. Consider a
13-span bridge with all decks seated on 2000-mm-wide cap beams and separated by 40 mm
nominal gaps. The maximum one-way sliding for any one of the 13 decks is 520 mm, still short
of the deck-seating length of 980 mm. Although a more conservative design may call on the
remaining 460 mm to be reserved for opposite movement in deformed piers, it is still unlikely to
form a 460 mm gap, counting all piers. This is because each pier is strong and deforms
elastically with lateral movement below 30 mm. The above is true prior to the failure of Pier 3
under pounding by Deck 2. In the study, the maximum gap for one-way sliding at the north end
of the four-span model is set at 160 mm, sufficing to cover three more spans further up north. In
fact, responses of the pounded pier are not significantly affected by the number of spans used in
the model, as illustrated later.

The backfill material is assumed to be granular soils with 30° angle of internal friction. A set of
soil springs along the height of the abutment can be constructed to model the lateral force–
displacement behavior of the backfill. The summed total stiffness is estimated at 874 000 kN/m.
Since soil heaving just behind the abutment was observed after the quake, the passive earth
pressure (Pp) of backfill soils is used as the yield strength for soil springs; Pp is estimated at
5300 kN. Though these soil data may seem less deterministic than those of solid materials, they
have only localized and limited effects on the backfill and do not significantly change final
results of this study.

4. Comparative study of models


As stated in the previous section, models used for failure study shall match the site survey data
and reflect the mechanical behavior of each element and member. The cause-finding process can
then be carried out by varying key parameters in the dynamic analyses of models and relating the
results to information disclosed by site survey. Therefore, models are organized in groups. Model
group I and II in Table 1 are set up to account for the variation of design-specific parameters, i.e.
www.sciencedirect.com

bearing types and arrangement at ends of each deck span. Fig. 1 provides the identity of each
bearing. The bearing at each deck end may be chosen as either fix type or expansion type. Model
groups subdivided in Table 2 are needed to determine the sensitivity of analyzed results to the
variation of uncertain parameters, i.e. the friction coefficient used at the interface where the
bearing pad comes into contact with the T-girder as well as the cap beam (Fig. 4).

Table 1.

Bearing type (fix or expansion) for bridge models

Model group Bearing IDs

a b c d e f g h
I Fix Exp. Fix Exp. Fix Exp. Fix Exp.
II Exp. Exp. Exp. Exp. Exp. Exp. Exp. Exp.
Full-size table

View Within Article

Table 2.

Coefficient of friction at bearings used in comparative models

Model ID I1 I2 I3 II1 II2 II3


Coefficient of friction 0.1 0.2 0.3 0.1 0.2 0.3
Full-size table

View Within Article


www.sciencedirect.com

5. Ground motions
The quake had a magnitude of 7.3 and a focal depth around 10 km [6]. A set of two ground
motion data recorded at the Taiwan Weather Bureau TCU068 observation station during the
quake is employed in combination as ground waveform inputs to the bridge model. This
particular station is located inside a nearby public school. In the dynamic analysis, the primary
ground motion input is the N–S component of the strong motion because the major response and
subsequent deck falling of the bridge took place in the longitudinal direction which runs nearly
parallel to the N–S direction. The peak ground acceleration of the N–S ground motion recorded
is 3.61 m/s2. The secondary ground motion, needed to determine the variation of friction force
with respect to the vertical reaction at the bearing, is the Z (vertical) component of the quake
having a peak ground acceleration of 5.19 m/s2. Fig. 6 presents these two records of ground
accelerations.

Full-size image (229K)


Fig. 6. (a) Ground acceleration (N–S component), (b) ground acceleration (Z component).

View Within Article

6. Parameters for dynamic analysis


www.sciencedirect.com

Each of the deck and pier bodies is modeled as discrete mass supported by flexible pier column,
through interfaces of bearings and gaps of joints. For a typical deck having mass M and damping

C under the longitudinal ground acceleration , the equation of motion can be expressed as

(1)

Fi (also referred to in Eq. (5) later) is the impact spring force initiated by a condition

(2)

X2-X1-Gp>0,
where X2 and X1 denote, respectively, displacements of two adjoining decks, or displacements
of Deck 2 and Pier 3; Gp is the gap at expansion joints.

Fs (also referred to in Eq. (8) later) is the friction force at bearings governed by a condition

(3)

where and denote, respectively, velocities of a deck and its supporting pier.

6.1. Mass

Each typical 35 m deck span has a total mass of 633 ton (metric unit), counting in pavement,
slabs, girders, diaphragms, and railing. The shorter 13 m deck span connecting to the abutment
has 182 ton. Each typical pier has 162 ton, including the cap beam and column. Using Rayleigh's
procedure and the first mode shape of a laterally loaded cantilever, we can determine the lumped
mass to be 71 ton for a typical pier.

6.2. Pounding

When the bearings slide, pounding between deck spans can significantly affect the response of
the bridge and thus must be considered in the bridge model.

In view of the 40 mm gap at the expansion joint between decks, pounding action is modeled
using a compression-only impact spring when the gap is closed. In case of fully elastic impact,
www.sciencedirect.com

the stiffness of an impact spring shall be several orders of magnitude of the pier column stiffness
[3], [4] and [5]. The following equation is used as an estimate:

(4)

Ki=EdAd/Ld,
in which Ed, Ad, and Ld are the modulus of elasticity, transverse sectional area, and length of the
deck body, respectively. The chosen value for the impact spring stiffness shall not produce
inconsistent results. On the basis of numerical tests on a range of values, the stiffness of the
impact spring used in the analysis is 4.56×106 kN/m. The impact spring force in Eq. (1) is then
determined by

(5)

Fi=(X2-X1-Gp)Ki.

When two bodies in motion collide with each other, a situation called elastic impact arises,
assuming that the plastic deformation occurring in the contact zone is negligible. The usage of an
impact spring element at the interface of a deck and a pier merely enables the transfer of
mechanical energy from the rigid deck to the flexible pier.

At the instant of contact, the action of impact produces an impulsive force that affects local
stresses and strains of colliding bodies near the contact zone, given that the duration of contact
can be measured. In this study, it is the internal force (shear) developed in the deformed pier that
is of major concern. Test cases to illustrate this point are provided later.

6.3. Behaviors of rubber bearing

Referring to the details in Fig. 4, the bearing pads mounted on pier cap beams and abutments are
composed of interlacing layers of neoprene rubber and steel sheets to provide horizontal
flexibility and vertical supports. Ideally, this device is modeled as a spring–dashpot. Under
violent quake action, the capability of a dashpot to absorb detrimental earthquake energy is
practically small and ignored, when compared with other factors.
6.3.1. Bearing stiffness

Each bearing pad has a total thickness of 61 mm, consisting of five layers of 11-mm-thick
neoprene rubber bonded to six layers of 1-mm-thick steel sheet. The steel sheet covers of the pad
come into contact with steel plates pre-embedded in the concrete T-girder and the pier cap beam.
The neoprene rubber has a hardness of 60 as per ASTM D412. Under severe dynamic loading,
www.sciencedirect.com

the sliding friction induced at the interface of the bearing pad and the girder/pier depends on the
surface condition of the steel material. The horizontal stiffness against sliding is provided by the
rubber material.

The total horizontal stiffness provided at each deck end is

(6)

Kr=(GrAr/Tr)Nr,
where Gr, Ar, Tr are shear modulus, planar area, and thickness of the rubber pad, respectively. Nr
(=5) is the number of bearings at each deck end. Table 3 presents data used in the analysis.

Table 3.

Stiffness calculation of rubber bearing at each deck end

Shear modulus Gr 1650 kN/m2


Thickness Tr 5.5 cm
Bearing area at girder end Ar 2200 cm2
Number of bearings Nr 5 Piece
Translation stiffness Kr 33000 kN/m
Full-size table

View Within Article

6.3.2. Dowel bar

One dowel bar is planned for each fix-type bearing at one end of each girder. Each bar material,
as per AASHTO M183, has a specified tensile strength 4.02×105 kN/m2 and a circular cross
section of 60 mm in diameter. The dowel bar shear strength is

(7)

Vb=AbFsbNb.
www.sciencedirect.com

Nb (=5) is the number of dowel bars at each deck end. Fsb is the shear strength and taken as 0.577
times the tensile strength, referring to von Mises yield criterion. The calculated result is

, to be used as a conservative estimate since similar tests reported that Fsb might
be the same order of magnitude as the tensile strength [7].
6.3.3. Bearing friction

Upon sliding of deck bodies under the horizontal quake motion, the behavior of a metal sheet-
covered bearing is captured by Coulomb friction. Mechanism of friction is associated with
molecular adhesion, surface roughness, and other factors. Experimental data of un-cleaned
metal-on-metal contact gave the friction coefficient ranging from 0.15 to 0.30. In railway terms,
typical wheel-on-rail friction coefficient is 0.2, a threshold value used in the study. Referring to
Fig. 4, for each bearing pad, the friction force in Eq. (1) is

(8)

where μ is the coefficient of friction and Wn the normal force exerted on the interface of bearing
pad with T-girder and pier cap beam. The total weight of a typical deck body is 6210 kN.
Usually, Wn is half the weight, i.e. 3105 kN. In view of the violent vertical quake motion, up to
5.19 m/s2, this may not be the case. Instead, reactive forces exerting on the interfaces with T-
girder and pier cap beam are determined first in response to the vertical motion, and substituted
for Wn, simultaneously, for use in response to the horizontal motion.

6.4. Behaviors of concrete pier column

Data and calculations described next refer to the original design plans and are judged reasonable
for use in the study. The laboratory testing of concrete core samples taken from the site by a
local investigation team indicated that the quality of concrete material is quite normal as
constructed.
6.4.1. Flexure

Each pier column in Fig. 2 has a longitudinal reinforcement ratio 1.3% and tie reinforcement
ratio (volumetric ratio) 0.2%. The specified yield strength of reinforcing bars and specified
compressive strength of concrete are 275.6 and 20.6 MPa, respectively. The behavior of a
reinforced concrete column under flexural–axial loading depends on the stress–strain model used
www.sciencedirect.com

[8]. The effect of properly designed tie reinforcement is to confine and then enhance the concrete
strength and ductility. Typical calculation for the complete moment–curvature relationship of a
column section involves slicing the concrete sections into fiber layers. Each layer then relates to
the specified stress–strain relationship, dependent on whether the layer is confined or
unconfined. An in-house code of such a kind has been developed to generate the moment–
curvature curve for concrete column sections. The moment and curvature at yield are
24 900 kN m and 1.56×10−3 /m, respectively. Effective flexural rigidity (EIe) up to the yield point
is 2.12×107 kN m2. This value is about half that calculated for the un-cracked column section,
taking into account the nonlinear behavior of concrete material. With a main reinforcement 1.3%
the gross section area and a compressive stress around 3% the concrete strength, these data are
on the conservative side. The effective lateral stiffness of the pier column is thus

(9)

where Hp (=5.9 m) is pier column height. The typical pier column stiffness used in the analysis is

3.1×105 kN/m. The shear strength of the pier at yield .


6.4.2. Shear

For concrete members designed in the 1980s, such as the pier column just discussed, the cross
sectional shear strength depends mainly on the quality of concrete and amount of tie
reinforcement used [9]. The shear capacity of each pier column (Vp) is derived from concrete (Vc)
and from tie reinforcing (Vs) together as

(10)

Vp=Vc+Vs,
with

(11)

in which fvc, Ae and dc denote the shear strength, effective area, and effective depth of the
concrete cross section, respectively. Ash, fsy and s are the sectional area, tensile strength, and
www.sciencedirect.com

spacing of lateral ties. For the pier column here, , , and

7. Solution scheme
Analytical models are set up using a multiple-degree-of-freedom dynamic analysis program
GENDYN, developed for research use at NKFUST since 2001 [10].

By means of direct time stepping of the governing differential equations based on the fourth-
order Runge–Kutta scheme, nonlinear transient results of motions and forces can be obtained.
Solution accuracy is ensured by varying the time stepping size ranging from 1/10 000 to
1/20 000 s and checking for convergent results. Since only localized and limited plastic
deformation were observed for this rather massive concrete structure, a uniform damping ratio of
2% is deemed proper and used in the dynamic analysis.

8. Analyzed responses
Based on design requirements, Group I models arrange for each deck span the bearings to be fix
type at one end and expansion type at the other, in an alternating pattern. Group II models
assume all bearings to be expansion type. Fig. 1 has shown the identity of each deck span or pier
column referred to in the following.

8.1. Bridge model Group I

Here, bearings are of expansion- and fix-type, alternating over the deck spans. To support a full
35 m deck span on a typical pier, such as Pier 1, a fix-type bearing develops large restraining
force to carry most of the seismic loading. As depicted in Fig. 7, the maximum restraining force

required for the fix-type bearing has exceeded the dowel shear strength ( ). In
case the dowel bars were not overcome, the maximum restraining force would exert on Pier 1

and cause plastic cracking at the pier column due to yield by flexure ( ), as
indicated in Fig. 8. At Pier 3 in particular, where a fix-type bearing undertakes seismic loading
from a lighter 13 m deck span to the right, any damage to the pier or the bearing is deemed
unlikely.
www.sciencedirect.com

Full-size image (25K)


Fig. 7. Shear force at fix bearing (<0).

View Within Article

Full-size image (21K)


Fig. 8. Maximum shear force of Pier 1 (<0).

View Within Article

8.2. Bridge model Group II

Here, all bearings are of the expansion type. The largest seismic load exerted on Pier 1 is
2130 kN, well below its shear capacity. Pier 3, supporting the 35 m Deck 2 to the left, is under
the threat of this deck sliding from left onto the special back-wall at the pier cap beam. The
quake intensity is indeed excessive enough to shake loose whatever frictional resistance the
bearing can provide, and the deck pounds and leans on Pier 3. The pounding produces severe
impact load and transmits shear load to Pier 3, as illustrated in Fig. 9 and Fig. 10. Depending on
the friction coefficient μ chosen for the bearing, the pounding-induced shear force is able to
cause damage to the Pier 3. For a μ value ranging from 0.2 to 0.3, the pier suffers local plastic
www.sciencedirect.com

cracks due to yield by flexure ( ). When the μ value dips below 0.2, it first cracks

by yield and fails eventually by direct shear ( ). Prior to the failure of Pier 3 by
pounding, the analyzed maximum sliding movement of the deck is 100 mm to the left. Photos
from personal site trips revealed that several successive deck spans had actually moved off their
designed positions by 100–200 mm. Fig. 11 illustrates motions of Deck 2 and Pier 3, when they
close in on a 40 mm gap and collide at the high time of quake motions.

Full-size image (52K)


Fig. 9. (a) Maximum response of a pier under pounding, (b) maximum response of pier under
pounding (coefficient of friction=0.1).

View Within Article

Full-size image (24K)


Fig. 10. Maximum shear force of Pier 3 (>0).

View Within Article


www.sciencedirect.com

Full-size image (22K)


Fig. 11. Displacements of Deck 2 and Pier 3 involving sliding and pounding (at time=15 s, 4 cm
gap closed and first pounding began).

View Within Article

Referring again to Fig. 9a, as the number of spans for the analytical model goes from two, three,
to four, shear forces developed in Pier 3 do not change much. For lower friction at bearings
(μ=0.1), the fast-moving decks are bound to shear off Pier 3 in all models. For higher friction
(μ=0.2), the corresponding shear force is a bit higher in the four-span model than in others.

Regarding the impact force under pounding, Fig. 9b shows results from three test cases for the

four-span model (μ=0.1), in which the value of impact spring stiffness is deliberately set at
Ki, and 3Ki,respectively. The shear forces developed in Pier 3 in these cases are nearly the same,
though forces in the impact spring vary significantly.

9. Discussions
Group I model is the designer's choice, but is bound to suffer damages due to either the shear-off
of dowel bars, or the yielding of pier columns. Such damage to the bridge is meant to be
extensive and common all over the deck spans. Group II model, assuming that the dowel bars
necessary for the fix-type bearings are missing, simply cannot sustain the sliding and pounding.
Such damage is most likely at the most vulnerable Pier 3. The impact load hit this pier hard
enough to inflict concrete spalling at pier cap beam and serious flexural cracking at pier column
base. The 35 m Deck 2 to the left of Pier 3, pinning on its cap beam, and knocking its column off
www.sciencedirect.com

to the right, thus fell off the support on the next pier to the left. Further to the left, the 35 m Deck
1, losing its contact with the fallen Deck 2, and sliding freely to the right, eventually fell off too.

10. Conclusions
The cause-finding process based on the comparative study of analytical models has presented
quantifiable information in good match with the field observation of a bridge structure damaged
by the 1999 Chi–Chi earthquake. The models used for failure study reflect the behavior of each
element and member by referring to the design specifications, as-built drawings, and damage
survey data. In particular, the modeling for sliding and pounding between deck bodies proves to
be essential to a realistic dynamic analysis of the bridge under the violent earthquake.

Without analytical models and the comparative study, it might be quite convenient to have the
nature (the earthquake) to blame for the loss of a bridge. Instead, the technical judgments derived
from this study are: (1) The bridge constructor somehow had missed out on installing the dowel
bars necessary for restraining the deck bodies, and failing such, the desirable flexural actions in
the pier columns did not take effect. Thus things turned otherwise and the free sliding and
pounding among deck bodies happened. (2) The bridge designer had arranged a back-wall at the
cap beam of one particular pier, resulting in a weak spot. As the pounding and sliding began, it
then led to the falling-off of two deck spans near the hardest hit pier. The study demonstrates that
details at bearings can significantly affect a bridge's behavior. In fact, since bridges of the similar
kind are still around, this study is presented in the hope to help identify sensible changes to
specifications used by the designer and constructor alike.

In particular, the design of a special back-wall atop the cap beam is to be avoided so as to make
way for sliding and pounding among deck spans, when dowel bars are overcome during extreme
earthquakes. Retrofitting such cap beams can be done at a feasible cost, compared to what had
happened to the case bridge. The point is that a bridge with sliding decks pounding each other
still has chance to survive the same earthquake without decks falling off.

The analysis also demonstrates that adding dowel bar sizes/numbers to form a hinge at bearings
simply transmits even more seismic loading from decks to pier columns, resulting in damages
such as the plastic cracking of concrete. Instead, considerations may be given to the use of
energy-absorbing bearings such as lead-in-rubbers or viscous dampers.

You might also like