ArticlePDF Available

A Cost-effective Integral Bridge System with Precast Concrete I-girders for Seismic Application


Abstract and Figures

To promote accelerated bridge construction in seismic regions, a large-scale experimental investigation was conducted to examine the seismic sufficiency of precast concrete I-girders in integral bridge super-structures. Such structures are not frequently used in high seismic regions due to lack of design guides and overly conservative design approaches. A half-scale, 17.8 m (58.5 ft) long test unit modeling a portion of a prototype bridge incorporating a concrete column, ten I-shaped precast concrete girders, and an inverted-tee concrete cap beam was used to experimentally verify that precast concrete members employing accelerated construction techniques can be used in integral super-structures and provide excellent seismic performance. Comparison of an as-built girder-to-cap connection detail with an improved detail shows that the as-built detail in existing bridges will satisfactorily resist positive and negative seismic moments and allow plastic hinges to develop at the column tops, though this was not the original design intent. However, the improved detail, which exhibited excellent seismic moment resistance, is recommended for new bridges to avoid potential deterioration of the girder-to-cap connection.
Content may be subject to copyright.
September–October 2015 | PCI Journal76
Accelerated bridge construction is increasingly being
pursued and promoted across the United States.
Many state transportation departments are dealing
with aging infrastructure along with increased demand due
to continuing economic and population growth.1 The rapid
construction of bridge projects to meet these needs is ben-
ecial.2 Similarly to the rest of the country, the California
Department of Transportation (Caltrans) is interested in the
benets of accelerated bridge construction techniques, pro-
vided that seismic concerns can be addressed. The desire to
improve and increase the possibilities of accelerated bridge
construction methods is highlighted in its lessons learned
report3 and the related strategic plan.4
The obvious primary benet to the incorporation of ac-
celerated bridge construction methods is the reduction of
on-site construction time, along with the associated mitiga-
tion of trafc delays. A common way to decrease time in
the eld is to employ prefabricated components as much as
possible. The use of precast concrete members instead of
cast-in-place concrete sections also results in the elimina-
tion of the need for falsework and an overall improvement
in quality control by relocating production from unpredict-
able eld conditions into a controlled shop environment.
Though accelerated bridge construction methods have
notable advantages, the incorporation of such techniques in
To promote accelerated bridge construction in seismic regions,
a large-scale experimental investigation was conducted to
examine the seismic sufciency of precast concrete I-girders in
integral bridge superstructures.
A half-scale, 17.8 m (58.5 ft) long test unit modeling a portion
of a prototype bridge was used to experimentally verify that
precast concrete members employing accelerated construction
techniques can be used in integral superstructures and provide
excellent seismic performance.
Comparison of the as-built girder-to-cap connection detail with
an improved detail shows that the as-built detail will satisfac-
torily resist positive and negative seismic moments and allow
plastic hinges to develop at the column tops in existing bridges;
however, the improved detail is recommended for new bridges
to avoid potential deterioration of the girder-to-cap connection.
A cost-effective integral
bridge system with
precast concrete I-girders
for seismic application
Justin Vander Werff, Rick Snyder, Sri Sritharan, and Jay Holombo
77PCI Journal | September–October 2015
under high seismic loading. Second, a portion of this work
focused on developing an improved girder-to-cap connec-
tion detail. This detail was shown to be able to reliably
establish an integral superstructure connection between
precast concrete cap beam and girder components with
minimal on-site construction.
The inverted-tee concept is well suited for accelerated
methods because of its incorporation of precast concrete
girders and easily constructible girder-to-cap connections
without needing temporary falsework. Also, the concept
provides the potential to use precast concrete cap beams
and make the completed structure aesthetically comparable
to cast-in-place concrete box girder bridges.
A precast concrete bridge
system for seismic regions
A frequently used precast concrete section is the California
I-girder.8 A detail that has been employed to facilitate the
use of such I-girders and accelerated bridge construction is
the inverted-tee bent cap concept (Fig. 1). It has typically
been implemented with cast-in-place concrete columns and
cast-in-place concrete inverted-tee cap beams. Once the
cap beam is built, the ledge on each side of the cap-beam
stem works well to support the dapped end of precast con-
crete girders. The girder dapped ends can subsequently be
integrated with the cap beam by the use of a cast-in-place
concrete diaphragm and by appropriate connection rein-
forcement. Finally, the cast-in-place concrete bridge deck
can be placed over the completed superstructure.
Seismic performance
and limitations
Whereas the inverted-tee bent cap concept has been em-
ployed in California, the superstructure has been designed
according to current design recommendations.8,9 Accord-
ingly, the degradation of the positive-moment connection
due to large seismic displacements and the loss of ten-
sion continuity in the girder lower ange connection are
expected. Figure 2 illustrates the moment reversal that
moderate-to-high seismic regions has been slowed because
of the poor performance of precast concrete structures in
previous earthquake events. The vulnerability of precast
concrete structures has been due to the inadequate perfor-
mance of connections and failure to ensure satisfactory
load paths. Precast concrete structures were observed to
experience connection failures (especially in buildings) in
past seismic events, including the Loma Prieta earthquake
in 19895 and the Northridge earthquake in 1994.6 Bridge
data show that cast-in-place concrete accounts for over
70% of current California bridge superstructures, while
precast concrete accounts for about 5%.7
Increased opportunities to incorporate accelerated bridge
construction techniques and the associated benets will be
realized if precast concrete connections can be developed
that are viable for quick implementation in the eld, do
not signicantly increase cost, and are able to sustain high
demands resulting from seismic loading. Capacity design
is the most common approach in designing for earthquake
loads. Using this approach, bridges are designed to exhibit
ductile behavior at the column ends, which are specically
detailed to accommodate sufcient inelastic action while
maintaining strength. These specially detailed regions are
referred to as plastic hinges. When a large seismic event
occurs, the plastic hinge regions undergo inelastic de-
formation, thereby dissipating seismic energy, while the
remainder of the structure continues to experience elastic
behavior even when subjected to high seismic demand. By
incorporating this design philosophy, structures can be eco-
nomically designed to accommodate large lateral seismic
The bridge superstructure, including the deck, is protected
from any inelastic action while allowing plastic hinge
formation in the columns. The girder-to-cap connections,
in particular, require careful attention for integral super-
structure concepts, because the girders must have sufcient
moment capacity across the cap beam. Integral designs
are advantageous in seismic regions because the moment
continuity in the superstructure above the column bents
provides a possible plastic hinge location in the column
just below the cap beam. The development of girder-to-cap
connections that facilitate rapid construction techniques in
the eld and provide sufcient shear and moment continu-
ity for integral connections in high-seismic regions will
provide greater opportunity to employ precast concrete
members and their associated benets without increasing
the cost.
Research significance
The work detailed here was undertaken to accomplish two
primary objectives. First, an existing inverted-tee cap-beam
concept had been previously used to facilitate precast con-
crete dapped-end girders, but it was not considered to have
sufcient moment capacity to be an integral connection
Figure 1. Inverted-tee bent cap concept.
September–October 2015 | PCI Journal78
design that is a competitive alternative to cast-in-place
concrete and provides the opportunity to incorporate ac-
celerated bridge construction in high-seismic regions. In
addition, total cost benets for precast concrete structures,
such as increasing quality, reducing effects on trafc, and
improving worker safety, are not integrated into the con-
struction cost of the bridge but are important advantages of
such approaches.
Prototype bridge design
To formulate the experimental plan, a prototype bridge
using the inverted-tee concept was developed (Fig. 3). The
four-span bridge incorporated reinforced concrete columns
in single-column bents, concrete inverted-tee cap beams,
and ve precast, prestressed, I-shaped concrete girders per
span. The design was based on the American Association
of State Highway and Transportation Ofcials’ third edi-
tion of the AASHTO LRFD Bridge Design Specications10
with interims and California amendments11 following the
guidelines from the Caltrans Bridge Design Aids,8 Cal-
trans Bridge Design Specications,12 and Caltrans Seismic
Design Criteria version 1.5.13 The design used Caltrans’s
deepest standard I-girder section (5.5 ft [1.7 m] depth),
with a ve-girder superstructure and single-column bents,
to develop maximum demand in the cap and connection
The overall concept behind the design of the prototype
bridge was to use a conguration that would generate max-
imum demand in the girder-to-cap connection region with
a single-column bent and California I-girders. Detailed in-
formation and design calculations for the prototype bridge
can be found in Theimann.14
Laboratory testing plan
and procedure
A large-scale experimental investigation was conducted to
determine the seismic behavior of the inverted-tee bridge
system and to carefully investigate and quantify the girder-
to-cap connection performance. The experimental work
was divided into two phases. The primary purposes of
phase 1 were the following:
occurs due to large horizontal seismic loads. Under dead
and live loads, the girder-to-cap connections are subjected
to negative moment only (Fig. 2). Under this condition,
the deck reinforcement, which runs continuously over the
top of the connection, provides tension continuity, and
therefore robust negative-moment capacity. However, when
the horizontal seismic load is added, the moment diagram
shifts (Fig. 2). Large seismic loads will produce a reversal
of moment in the connection region, resulting in tension
in the bottom of the connection region where there is no
reinforcement continuity. Therefore, the recommendations
stipulate that the cap-to-girder connection should be con-
sidered to have zero moment resistance under combined
gravity and seismic loading.
Furthermore, the Caltrans Seismic Design Criteria9
require that a static vertical load, both upward and
downward, equal to 25% of the dead load needs to be
incorporated for ordinary standard bridges where the site
peak ground acceleration is 0.6g or greater (where g is
acceleration due to gravity). When this acceleration must
be considered, longitudinal-side mild reinforcement in
the girders should be provided that is capable with shear
friction to resist 125% of the dead-load shear at the cap-
beam interface. This requirement, which exists to protect
against potential shear failures when the bottom of the
connection opens up, has been disadvantageous for two
There may not be sufcient room to place such rein-
This detail would increase the girder and assembly
costs while introducing constructability challenges.
Eliminating this requirement, which was not based on a
comprehensive investigation, would make the inverted-tee
detail more attractive in seismic regions.
Development of girder-to-cap connections that provide full
moment resistance will offer the possibility of incorporat-
ing design approaches that take advantage of fully integral
superstructure behavior while using accelerated bridge
construction. The result is a precast concrete superstructure
Figure 2. Moment diagrams due to different loading conditions.
Dead and live load Dead, live, and horizontal seismic load
79PCI Journal | September–October 2015
Figure 4 provides a schematic of the test unit. The
as-built connection was incorporated for the five gird-
ers on one side of the cap beam, while the improved
connection was incorporated for the five girders on
the other side of the cap. Detailed information on both
connection concepts is provided later in this paper.
The termination of the girders at the location rep-
resenting the prototype midspan resulted in support
locations at the approximate girder inflection points
under horizontal seismic loading. Hold-downs were
used to properly simulate the effects of gravity load in
the girder-to-cap connection region; these hold-downs
were located at the approximate girder inflection
points during the service-load-only condition, with
load application occurring in two stages as detailed in
the next section. Two pairs of horizontal actuators, one
at each end of the test unit, were used to apply qua-
sistatic horizontal seismic loads, and pairs of vertical
actuators at each end were used to provide vertical
support and to accommodate the column growth ex-
pected, due to the formation of plastic hinges, without
introducing additional load to the superstructure. This
support condition was accomplished by programming
the vertical actuator control based on the predicted
column growth at various horizontal displacement
levels, following the procedure outlined by Holombo
et al.15
validate the overall system for high-seismic regions
determine the ability of the girder-to-cap connections
to maintain elastic superstructure action while allow-
ing plastic hinges to fully develop at column ends
compare and contrast the existing Caltrans girder-to-
cap connection detail with an improved detail
The primary purpose of phase 2 was to exercise the girder-
to-cap connections to realize their full potential by apply-
ing connection demands beyond what would be permitted
by the typical overstrength capacity of the column plastic
hinge region.
Experimental configuration
The experimental configuration was developed to al-
low the investigation of the existing connection detail
for inverted-tee cap beams and dapped-end I-girders
(referred to as the as-built connection) and a modified
connection detail concept (referred to as the improved
connection) in a single test unit. The test unit was
designed at a 50% dimensional scale of the prototype
structure. It modeled the full five-girder width of the
prototype on both sides of bent 3, with the girder
length extending approximately to the midspan on
either side of the column (dashed region in Fig. 3).
Figure 3. Prototype integral bridge structure. Note: 1 ft = 0.305 m.
Longitudinal elevation
Transverse section
September–October 2015 | PCI Journal80
Staged loading to simulate
prototype gravity effects
The test conguration was designed to provide stress
simulation of the prototype girder-to-cap connection
region. To accomplish this simulation, the progression of
the prototype connection load transfer capabilities during
construction needed to be replicated as closely as possible.
In the eld construction of existing structures using the
inverted-tee system, the girders are set in place without
moment restraint prior to the diaphragm placement. The
casting and subsequent curing of the diaphragm concrete
in the girder-to-cap connection then creates a moment
connection. The initial loads between the girders and cap
beam prior to diaphragm casting are transferred as if the
girders were simply supported. However, after placement
of the diaphragm, the loads between the girders and cap are
transferred through a moment connection. The experimen-
tal test unit was not designed to model the full length of the
girders. Therefore, the hold-downs were used to simulate
additional girder dead loads, barrier loads, and wearing
surface loads. Because the girder loads would be present
prior to diaphragm placement and the barrier and wearing
surface loads would be added after deck and diaphragm
placement, the vertical load simulations were introduced in
a staged process to properly simulate the connection xity
during each stage of the loading process.
Figure 5 compares the moment prole of the prototype
and the test unit at stage 1 (dead loads prior to connec-
tion moment capacity) and stage 2 (additional dead loads,
such as barrier and wearing surface, after the connection
moment capacity is developed). Figure 5 shows how the
hold-down forces were used to accurately simulate the
moments in the connection region. Similar comparisons
were conducted to ensure proper simulation of the shear
in the connection region but, for brevity, are not included
here. The stage 1 loads (33.4 kip per girder) were applied
using a spreader beam and spacer plates on the girders to
simulate the additional girder self-weight load that would
be present prior to deck placement. The stage 2 loads
(45.2kip per girder) were applied to the spreader beam
Figure 4. Schematic of test unit conguration for phase 1. Note: 1 in. = 25.4 mm; 1 ft = 0.305 m.
Figure 5. Comparison of prototype and test-unit moment proles during stage loading (test-unit scale). Note: 1 ft = 0.305 m; 1 kip = 4.448 kN.
Stage 1 loading Stage 2 loading
81PCI Journal | September–October 2015
after deck and diaphragm placement to properly simulate
the connection moment transfer.
Seismic load protocol
A cyclic quasistatic load protocol was planned to simu-
late the effects of horizontal earthquake loads in phase 1
testing. The horizontal actuators (Fig. 4) applied the cyclic
horizontal loads. The dead load hold-downs remained en-
gaged to simulate gravity loads, and the vertical actuators
served primarily as adjustable supports. Figure 6 shows
the horizontal load sequence. Single load cycles were used
to apply loads using the horizontal actuators under force
control at peaks of ±0.25
, ±0.5
, and ±0.75
where Fy
' was the estimated rst yield strength of the
system. The remainder of the phase 1 test was conducted
using the horizontal actuators under displacement control,
using three fully reversed quasistatic displacement cycles
at system displacement ductility
levels varying from to
1.0 to 10.0. The moment demand in the connections due to
time-dependent effects was beyond the scope of the study.
Figure 7 provides the moment prole for the test-unit
superstructure when subjected to large horizontal loads
Figure 7. Comparison of test-unit and prototype moment proles. Note: 1 ft =
0.305 m; 1 kip = 4.448 kN. Figure 8. Conguration of test unit for phase 2.
Figure 6. Phase 1 load sequence. Note:
= estimated rst yield strength of bridge system;
= horizontal displacement ductility. 1 kip = 4.448 kN.
September–October 2015 | PCI Journal82
the same test unit but recongured the actuators to allow
maximum load and displacement to be applied to the gird-
er-to-cap connection regions. The dead load hold-downs
were removed, and the vertical actuators were relocated to
these locations (Fig. 8). In this conguration, the vertical
actuators were used to apply the primary load sequence,
using displacement control to cyclically exercise the gird-
ers up and down, producing large shear and moment condi-
tions in the girder-to-cap connection regions. The horizon-
tal actuators were used to maintain system stability.
Figure 9 provides sketches of the test-unit shear and mo-
ment envelopes intended during the phase 2 loading, again
assuming moment continuity in the superstructure over the
column. These shear and moment proles do not match
that simulate seismic effects, assuming integral connec-
tion behavior in both the prototype and test unit. The sign
convention used in this gure is based on positive mo-
ment, producing compression on the bottom surface of the
superstructure. The test-unit prole also shows the negative
moment peaks that result from the hold-down forces to
simulate gravity effects. Comparing the test-unit moment
prole with the prototype moment prole, also included in
this gure, conrms that the test-unit and prototype proles
match well between the tie-down locations and match
almost perfectly in the connection region near the column.
Phase 2 of the load protocol was planned to fully exercise
the girder-to-cap connections and provide a comparison of
the as-built and improved connection details. Phase 2 used
Figure 9. Phase 2 test-unit shear and moment envelopes.
Figure 10. Girder-to-cap connection concepts.
As-built detail Improved detail
83PCI Journal | September–October 2015
positive moment and the corresponding shear. The as-built
detail’s main limitation is the lack of positive-moment
tension continuity. To address this deciency, the improved
detail incorporated unstressed strands to provide tension
continuity between the girder bottom ange and the cap-
beam corbel (Fig. 10). Grade 270 (1860 MPa), seven-wire,
uncoated prestressing strands were used. The strands were
sized assuming that they needed to be capable of provid-
ing the needed tensile resistance of the positive moment in
the girder-to-cap connection corresponding to the ultimate
limit state of the plastic hinge region of the column. The
strands were threaded through ducts in the bottom ange
of the precast concrete girder and continued across the
girder-to-cap interface into aligning ducts in the cap beam.
After the strands were positioned, they were grouted in
place to provide anchorage in the girder and cap. This
connection detail retains the same negative moment and
vertical shear capabilities as the as-built detail.
Figure 11 shows details of the test unit. Because the length
scale factor was 0.5, the reinforcement area was deter-
mined by scaling the prototype design by 0.25 and using
bar sizes and spacing that met constructibility require-
ments. Figure 11 provides the column-to-cap connec-
tion detail, with the column longitudinal reinforcement
anchored into the cap beam. Figure 11 also shows the
cap beam and joint reinforcement details, deck reinforce-
ment, beam ledge reinforcement, and dowel bar details,
all designed to replicate the Caltrans detail at the test unit
scale. The one exception to the replication of the detail
is the unstressed posttensioning tendon identied on the
bottom-right side of the gure. On the side of the improved
connection detail, these tendons were continuous from the
girder bottom ange and through the bottom of the cap-
beam corbel and terminated at the face of the cap beam on
the side of the as-built detail. This conguration resulted in
a test unit that modeled the improved girder-to-cap connec-
tion on the left side (in the gure orientation) of the cap
and the as-built connection on the right side.
Figure 11 provides an elevation of a single girder. The
girder strand layout, designed as a scale model of a typi-
cal Caltrans I-girder CA I66 strand layout while using
constructible size and spacing details (Fig. 11). This gure
also shows the layout of the ducts for the unstressed strand
for the girders on the improved connection side. These
ducts were positioned not to interfere with the strand
layout (Fig. 11); because they were only required for the
improved connection detail, they were not included in
the girders on the as-built side of the test unit. Finally,
Fig. 11 shows the reinforcement details for the full-depth
blocked-end portion of the girders, and the details in the
dapped-end portion of the girders immediately adjacent
to the blocked end. The blocked- and dapped-end details
were developed as scaled models of the prototype details,
except for modications based on a strut-and-tie concept.
A strut-and-tie analysis was used to quantify the continu-
the prototype proles along the length of the superstruc-
ture, but the load conguration provides an excellent way
to generate maximum shear and moment demand in the
connection region. Because the primary focus of the test-
ing related to the superstructure is the performance of the
connection region, this conguration is well suited for the
desired objective. The conguration is also suitable for
simulating vertical acceleration effects in the connection
region because the load generation in the test unit is com-
ing from the vertically oriented actuators.
Test-unit details
Figure 10 shows the as-built girder-to-cap connection
that has been used for the inverted-tee system. This detail
incorporates three dowel bars that pass through ducts in
the webs of the precast concrete girders near their dapped
ends. After the girders are placed on the corbel of the
inverted tee, the dowel bars are grouted into place in the
girder webs, and a cast-in-place concrete diaphragm is
used to encase the dapped end and dowel bars, thus achiev-
ing connection continuity.
For the detail to maintain its integral performance during
seismic loading, it needs to successfully transfer vertical
shear as well as positive and negative moments. Downward
vertical shear in the as-built detail is easily transferred
from the girder dapped end to the cap-beam corbel, due to
the direct support conguration of the dapped end on the
corbel. The as-built detail also has signicant negative-
moment capacity because the deck reinforcement provides
tension continuity across the girder-to-cap joint. The dowel
bars provide some resistance to upward shear and positive-
moment loading that could occur during a large seismic
event. However, because the detail includes no tension con-
tinuity near the girder bottom ange, rapid degradation of
the girder-to-cap connection region is anticipated and will
commence under high positive-moment action. Therefore,
the Caltrans Seismic Design Criteria recommendations
currently require that it be treated as a pin connection when
subjected to seismic loading.
One of the objectives in the experimental investigation was
to determine whether Caltrans’s treatment of the existing
detail is overly simplistic. While the connection detail may
deteriorate when subjected to large seismic displacements,
the girder subjected to positive moment demand should
rst overcome the adhesion and shear friction between the
girder and diaphragm. The deterioration of the correspond-
ing mechanisms is difcult to quantify; thus, experimental
work to enhance understanding of these mechanisms and
fully quantify their behavior would be benecial to the
design and implementation of this detail.
An additional objective of this research was to develop
an improved detail that would provide a dependable load
path for the girder-to-connection interface to sustain the
September–October 2015 | PCI Journal84
tion. The footing and column were constructed rst, and
temporary shoring was erected around the column to
support the construction of the inverted-tee cap beam.
Figure12 provides a photograph of the cap connection
region for the center girder. The ve-girder conguration
results in the center girders attaching to either side of the
cap beam adjacent to the cap beam–to–column connection.
Because of this connection proximity, the strand ducts in
the cap beam for the center-girder improved connection
were curved around the column longitudinal reinforcement
cage. While the introduction of these curves was a con-
cern in terms of feeding the strand through the ducts and
successfully grouting after placement, it did not pose any
challenges during construction.
ous horizontal-and-vertical tie (Fig. 11) and to eliminate
some of the reinforcement congestion in the dapped-end
region that often occurs when using conventional analysis
approaches. Engineered wire mesh was used to provide
the transverse reinforcement in the girders. The wire mesh
was incorporated to validate its use in place of traditional
transverse reinforcement in precast concrete girders. All of
the prototype design details can be found in Theimann.14
To make the test unit as close to an actual inverted-tee
bridge as possible, typical eld construction practices and
techniques were incorporated into the test-unit construc-
Figure 11. Test unit details. Note: no. 3 = 10M; no. 4 = 13M; no. 6 = 19M; 1 in. = 25.4 mm; 1 ft = 0.305 m.
85PCI Journal | September–October 2015
strands to be positioned continuously between the gird-
ers and the cap beam. The girders were fabricated off-
site at a precast concrete production facility using typical
methods. After the girders were delivered to the labora-
Figure 12 shows the cap beam atop the column, prior to
girder placement. The ducts in the gure were trimmed
adjacent to the cap beam’s vertical face and then mated
with the ducts in the girders to allow the unstressed
Figure 12. Photographs of construction.
Cap beam reinforcement in column region Cap beam prior to girder placement
Installing as-built girders Installing strand for improved girders
Casting an abutment Temporary abutment support
September–October 2015 | PCI Journal86
Phase 1 test results
Primary purposes for phase 1 included the following:
validating the overall system for high-seismic regions
verifying the capability of the girder-to-cap connec-
tions to maintain elastic superstructure action up to
high seismic displacements (that is, the sufciency of
the girder connections to provide adequate resistance
to develop plastic hinges in the column)
comparing and contrasting the existing Caltrans gird-
er-to-cap connection detail with an improved detail
General observations
of test-unit performance
General observation of the displacement-control portion of
the testing in phase 1 indicated excellent seismic behavior.
Figure 13 shows the column during the phase 1 test-
tory, temporary shoring was used to support them in
position on the cap beam. The strands for the improved
connection were then properly positioned through the
cap-beam ducts and grouted in place. Temporary shoring
was also used to aid in the construction of the diaphragm
in the connection region. Figure 12 also shows the abut-
ment formwork and temporary shoring that was used
prior to nal conguration of the loading actuators and
support system.
To provide temporary stability to the system, the con-
crete in the lower third of the diaphragm was placed
without fully constraining the girder ends and prior to
applying the stage 1 hold-down forces. Following the
stage 1 load application, the diaphragm concrete place-
ment was completed, and the abutment and deck concrete
was placed. After the hardening of the deck concrete,
the stage 2 hold-down load was applied to each span to
simulate the additional weight of parapets and wearing
surface that would be added to the prototype structure
following deck concrete placement.
Figure 13. Test-unit photographs during and after phase 1 testing. Note: 1 in. = 25.4 mm.
Column at 7 in. lateral displacement
(displacement ductility +10.0)
Buckled bars in the column top plastic hinge region
Flexural cracking across the entire deck width
87PCI Journal | September–October 2015
bars had buckled (Fig. 13). Both the improved and as-built
connections between the cap beam and girders behaved
as xed connections and did not show signs of signicant
damage or degradation. Flexural cracking was observed
across the width of the deck, indicating that the diaphragm
action of the deck had engaged all of the girders in resist-
ing the column seismic moment. The seismic load distribu-
tion among the center, intermediate, and exterior girders
was estimated to be 0.205:0.239:0.158, and this load
distribution occurred in the girder-to-cap connection region
immediately adjacent to the cap starting at early load lev-
els. Detailed information on seismic load distribution in the
superstructure is reported in Vander Werff and Sritharan.17
Phase 1 testing was concluded when, during the displace-
ment cycle corresponding to ductility 10, a horizontal
strength loss of approximately 10% was observed due to
buckling of the longitudinal column reinforcement in the
plastic hinge region just below the cap beam (Fig. 14).
This failure mechanism was an excellent indicator that the
superstructure behaved as intended, fullling its purpose
according to the capacity design approach of maintaining
sufcient elastic strength to produce plastic deformation
and failure in the column hinge region prior to superstruc-
ture failure. In addition, a displacement ductility level of
10.0 far surpassed the minimum recommended ductility
of 4.0 required in the Caltrans Seismic Design Criteria9
and AASHTO LRFD specications10 for structures such
as the prototype with single column bents supported on
xed foundations. This performance indicated the seismic
capability of both the as-built and improved connection de-
tails because had the as-built connection deteriorated to a
pinlike condition, the excellent strength retention observed
in Fig. 14 would not have been possible.
Force-displacement response
The horizontal force-displacement response of the test
unit (Fig. 14) indicates excellent seismic performance,
ing. Plastic hinges were developed at both the base of the
column above the footing and at the top of the column just
below the cap beam, indicating successful performance of
the superstructure. The successful superstructure perfor-
mance was notable, as it contradicted Caltrans’s current
conservative treatment of the as-built connection as having
limited moment resistance at high seismic displacements.
Further, no signs of distress were observed in the girders
beyond the connection region, validating the use of wire
mesh reinforcement as a substitute for traditional trans-
verse reinforcement in precast concrete girders.
To quantify the test unit’s performance, ductility
of 1.0
was established as corresponding to idealized yield. This
was determined based on measured experimental yield
displacement and theoretical moment resistance for the
rst yield and idealized yield limit state, following the ap-
proach of Priestley et al.16 Overall, the structure achieved
a displacement ductility
of 10.0, corresponding to 7 in.
(180mm) of total horizontal displacement in each direc-
tion. At this displacement, several column longitudinal
Figure 14. Horizontal force-displacement response for system test unit. Note:
= horizontal displacement ductility. 1 in. = 25.4 mm; 1 kip = 4.448 kN.
Figure 15. Dowel-bar strains at peak displacements. Note:
= horizontal displacement ductility. 1 in. = 25.4 mm.
Positive peaks Negative peaks
September–October 2015 | PCI Journal88
strain (approximately 2000 με). The dowel-bar strains in
the as-built connections (not shown) were similar to those
in the improved connections. The relatively low strain ob-
served in all of the dowel bars indicates that the dowel bars
remained elastic throughout the phase 1 test for both the
as-built and improved connection details. In addition, the
systematic behavior of the dowel bars, with larger strains
at the top under negative moment and larger strains at the
bottom under negative moment, shows that the dowel bars
were engaged and providing moment resistance in both
directions of loading.
The improved connection added unstressed strands for
positive-moment continuity that were not included in the
as-built detail. An interesting observation related to the
dowel-bar data is that the dowel-bar strains in the im-
proved connection are not lower than the dowel-bar strains
in the as-built connection, as might be expected if some
of the positive-moment tension load is diverted from the
dowel bars to the strands. Figure 16 shows strain data
from the strands in the improved connections of one exte-
rior and two intermediate girders. All of the strands seem
to have been engaged already at low displacements, but all
exhibited a noticeable increase in engagement when the
superstructure was displaced to 1.5
(1.0 in. [25 mm]).
This sudden increase corresponds closely to the time when
a noticeable opening of the girder-cap interface of the im-
proved connection was rst observed. The opening at this
point was an indication that the concrete tensile capacity
for resisting positive moment in the girder-cap interface
was fully lost, producing signicant load transfer to the
strands. The dowel-bar strain proles from the connection
region (Fig. 15) also show the largest incremental increase
between displacement steps 1.0
and 1.5
. This behav-
ior indicates that the dowel bars and unstressed strand were
acting together, as both increased in strain proportionally
and simultaneously.
Figure 16 also shows that after the sudden engagement of
the strands between displacement steps 1.0
and 1.5
as strength retention was maintained all the way to
±8.0. Also, while longitudinal bar buckling and beginning
of connement loss occurred at
equal to ±10.0, signi-
cant strength still remained in the system. This strength
exhibits the ability of the system to continue carrying grav-
ity load even at large seismic displacements.
Figure 14 also provides the predicted force-displacement
response. This prediction was the result of an analyti-
cal investigation that incorporated a nite element model
analysis, detailed in Theimann,14 and an associated gril-
lage analysis, detailed in Snyder et al.18 The predicted
force-displacement response compared favorably with the
experimental horizontal force-displacement response of
the superstructure. While there is a slight variation in the
elastic region, the results converged more closely at higher
levels of displacement as more of the test unit began to
soften due to the development of cracks and yielding of
longitudinal reinforcement.
Connection response
The behavior of the girder–to–cap beam connections was
a primary area of interest, particularly verifying whether
the superstructure remained elastic while allowing the
column plastic hinges to fully develop. Visual observations
during phase 1 indicated that the superstructure did indeed
remain elastic, as plastic hinges were developed in the
column and no signicant spalling, bar buckling, or other
failure indicators were observed in the superstructure. Data
gathered during phase 1 testing was used to validate these
observations. Figure 15 shows dowel-bar strains measured
in the improved connections for the center girder; it also
shows dowel-bar strains at the peak displacements that
produced positive moment in the connection region, and
shows how strains are provided for the negative-moment
peaks (1.5
corresponds to 1.0 in. [25 mm] horizontal
displacement, 3
corresponds to 2.0 in. [50 mm] hori-
zontal displacement, and so on). The maximum measured
strain was approximately 1000 με, well below the yield
Figure 16. Unstressed strand strains in exterior and intermediate girders at
peak displacements producing positive moment in improved connection region.
Note: 1 in. = 25.4 mm.
Figure 17. Gap opening at bottom girder-to-diaphragm interface (phase 1).
Note: 1 in. = 25.4 mm.
89PCI Journal | September–October 2015
Load protocol
For phase 2, the relocated vertical actuators were used
as the primary control mechanism, while the horizontal
actuators were congured to remain at zero load to retain
horizontal stability in the test unit without affecting the
load condition. Figure 18 provides the load protocol used
with the vertical actuators. The actuators were initially
adjusted under load control to establish the initial condi-
tion, matching the endpoint of the phase 1 test and corre-
sponding with the left edge of the sequence. Displacement
control was then used to apply small, incremental vertical
displacements at the actuator locations down to 1.5 in. (38
mm) below the initial girder positions (producing nega-
tive moment in the connection regions) and then up to 1.0
in. (25.4 m) above the initial girder positions (producing
positive moment in the connection regions). The verti-
cal displacements were applied to both sides of the test
unit simultaneously to limit the force demand in the cap
beam. The initial displacement control sequence was
used to establish the test procedure and ensure specimen
performance without going beyond load levels produced
during phase 1. Following the initial sequence, the primary
displacement sequence used three cycles per displacement
level up to a maximum displacement of 6.0 in. (150 mm)
downward and 3.0 in. (75 mm) upward on each side of the
test unit.
General observations
The primary observation during phase 2 was the contrast
in performance between the as-built connections and the
improved connections. Throughout the test, the improved
connections exhibited no signs of damage, whereas the
as-built connections in all ve girders experienced pro-
gressive deterioration near the interfaces, with all girders
eventually pulling out of the diaphragm.
At a displacement of +0.5 in. (13 mm), the as-built con-
nection was already subjected to a moment approximately
27% greater than the maximum positive moment achieved
during the phase 1 test. At a displacement of +0.75 in.
(19mm), the improved connection side remained un-
changed, but the as-built side began to exhibit signicant
degradation, reducing the positive-moment resistance.
Signicant spalling in the diaphragm on the as-built side
was indicative of the distress in the connection dowel bars
(Fig.19). The gap between the girder bottom anges and
the cap beam widened to approximately 0.2 in. (5 mm),
and the 1 in. (25 mm) thick grout along the bottom inter-
face between the girders and cap had begun to separate and
fall off. Cracks in the diaphragm concrete, indicative of the
girder bottom dowel bar trying to pull out, were observed
on the as-built connection side. At +1.0 in. (25 mm)
displacement, the as-built connection continued to exhibit
interface grout spalling, signicant crack opening, and
bottom ange girder pullout. A signicant crack between
the strand strains at subsequent larger peak displace-
ments decreased slightly, while still exhibiting noticeable
engagement. Because the strand strains are below the
yield strain, this behavior is not indicative of the strands
relaxing after plastic behavior. Rather, it is more likely
indicative of the load path and combined mechanism be-
tween the strands and dowel bars gradually changing due
to concrete cracking and softening once the mechanism
was fully engaged.
The relative behavior of the gap between the girder
bottom ange and the adjoining edge of the diaphragm
provides further insight into the difference between the
improved and as-built connection behaviors. Figure 17
compares the relative gap-displacement data for both the
improved and as-built connections at positive-moment-
direction peak-load conditions. The comparison reveals
similar trends for both connections, but larger gap-dis-
placement magnitudes in the as-built connection, showing
that the unstressed strand in the improved connection was
effective in reducing the gap opening under positive-
moment loading.
Strains measured in the deck reinforcement, which acted
as the primary tension reinforcement for the connections
under negative-moment loading, were also used to inves-
tigate the superstructure behavior. A primary nding was
that the deck strains exhibited the engagement of all ve
girders in resisting the column moment, from the early
load stages all the way through to the overstrength mo-
ment. More details on this topic can be found in Vander
Werff and Sritharan.17
Results from the phase 2 test
Because the superstructure, including the connection,
maintained elastic response in the phase 1 test as expected,
the phase 2 test commenced as planned to fully exercise the
girder-to-cap connections, further quantify their behavior,
and give an indication of their capability when subjected to
vertical seismic effects.
Figure 18. Test unit phase 2 load sequence. Note: 1 in. = 25.4 mm.
September–October 2015 | PCI Journal90
the underside of the deck and the top of the diaphragm on
the as-built side indicated a connection separation. The
improved connection remained unchanged.
The higher displacement cycles continued to show the
trend of increased deterioration on the as-built side, with
little change on the improved side. Figure 19 shows the
exposed dowel bars on the as-built side. (Headed bars
were used only for the exterior girders, given the limited
embedment length outside the exterior girders. How-
ever, the deterioration was similar around the interior
dowel bars as well.) Eventually, the as-built connection
deteriorated to the point of behaving as a pin connec-
tion under positive-moment loading. Figure 19 shows the
deterioration of the as-built connections at large positive
displacements. The pin behavior of the as-built connec-
tion and the plastic hinge formed at the top of the column
during phase 1 prevented any further testing to increase
the positive moment in the improved connection to the
point of ultimate failure. Based on the force-displacement
plots for the structure at a downward displacement of 6.0
in. (150 mm), both connection details seemed to have
additional negative-moment capacity. However, when the
structure was cycled to an upward displacement of 3.0 in.
(75 mm), a 42% drop in strength was observed, indicat-
ing that the as-built connection had already reached its
ultimate capacity and therefore, the test was terminated at
this point.
Figure 19. As-built connection region during latter stages of phase 2 testing. Note: 1 in. = 25.4 mm.
Spalling around dowel bars in girder connection region
Deterioration around dowel bars
near exterior girder
Partially spalled grout pad at as-built girder-to-cap interface
at +0.75 in. (19 mm) displacement
Opening of as-built connection
under large positive-moment load
91PCI Journal | September–October 2015
Force-displacement response
Figure 20 depicts the vertical load-displacement response
for the center girder on both the improved and as-built sides
during the phase 2 test. One notable difference between the
two sides was the positive-moment peak resistance at high
displacements; the improved side maintained its strength,
whereas the as-built side exhibited a decrease in resistance
as the displacement increased. Besides the positive-moment
peak behavior, initial inspection of the remainder of the data
in this gure seem to indicate similar inelastic responses on
both the improved and as-built sides of the superstructure.
However, a systematic analysis of the data was subsequently
conducted, taking into account the experimental rotations,
column load, horizontal actuator load, and vertical loads
from both sides. This analysis was used to determine the
total moment on the two sides, independent of each other.
These separated data for each girder (Fig. 21) provide a
much clearer picture of the test-unit behavior.
The response envelopes of the as-built and improved con-
nections when subjected to positive moment are different
(Fig. 21). While the as-built connection showed softened
behavior up to 0.75 in. (19 mm) and signicant strength
deterioration thereafter, the improved connection was not
subjected to relative displacements (adjusted to account for
girder rotation) higher than approximately 0.35 in. (9mm),
and the response was linear and elastic. Both connections
demonstrated reserve positive-moment strength well beyond
the maximum demand experienced under horizontal seismic
loading (dashed line on Fig. 21). However, the as-built con-
nection experienced a loss in stiffness at loading above the
maximum horizontal demand, whereas the improved con-
nection performance was elastic.
Figure 21 compares the negative-moment behavior of the
two connections. The improved connection exhibited a
slight increase in performance over the as-built connection.
The relative displacement difference between the improved
and as-built connections at the large displacements is likely
due to the loss of the grout pad in the as-built connection.
The difference in negative-moment performance between
the improved and as-built connections is less pronounced
than for the positive-moment behavior. This similarity is not
surprising, because the deck reinforcement provided the pri-
mary tension-transfer mechanism for the two connections,
Figure 20. Vertical load-displacement response from phase 2 test. Note: 1 in. = 25.4 mm; 1 ft = 0.305 m; 1 kip = 4.448 kN.
Improved side As-built side
Figure 21. Comparison of girder-to-cap connection moment behavior. Note: G = gravity load condition; H = full horizontal seismic load condition; V = seismic vertical
acceleration load condition. 1 in. = 25.4 mm; 1 ft = 0.305 m; 1 kip = 4.448 kN.
Positive moment Negative moment
September–October 2015 | PCI Journal92
seismic, and 1.0g constant vertical acceleration condi-
tion (G + H + 1.0V). The equivalent vertical acceleration
action included for the positive and negative moments has
been added in the direction that increases moment magni-
tude for each direction (that is, upward vertical accelera-
tion for positive moment, downward vertical acceleration
for negative moment).
Figure 21 shows that both the as-built and improved con-
nections exhibited moment capacity in excess of the 0.25g
vertical acceleration stipulated by Caltrans.8 In fact, the
as-built connection did not begin to lose strength until the
equivalent vertical acceleration reached 1.0g, although it
began to show reduction in stiffness for loading beyond
phase 1 testing. The improved connection exhibited elastic
positive-moment behavior throughout the test, indicating
that it had capacity in excess of 1.0g. Figure 21 shows that
both the as-built and improved connections performed suc-
cessfully under moment demands larger than what would
be expected if the prototype structure experienced the
gravity, full horizontal seismic, and 1.0g vertical accelera-
tion condition. These results show the suitability for both
connection details in resisting vertical acceleration effects
in both the upward and downward directions that are
greater than the requirement of 0.25g.8
Behavior of connection details
Figure 22 shows the dowel strain demand and the gap
opening at the lower girder-cap interface for the improved
and as-built connections in phase 2. The superior perfor-
mance of the improved connection’s data (shown in dashed
lines for both gap opening and dowel strain) is even more
pronounced in this gure. The gap opening is larger for
the as-built connection. In addition, the magnitudes of
dowel-bar strain are lower in the improved connection.
Both of these observations show the benet of the grouted,
unstressed strand.
Figure 23 shows the strain envelope at peak displacements
for the strains measured on the dowel bars for the dura-
tion of the phase 2 test. The behavior under displacements
producing negative connection moment was similar for
both the as-built and improved connections. For displace-
ments producing positive connection moment, there was
a notable difference between the as-built and improved
dowel bars. The as-built connection exhibited quickly
increasing strains at positive displacements of 1.0 in.
(25mm) and continued to experience an increase in strain
until it yielded at a girder displacement of 1.5 to 2.0 in.
(38 to 50 mm). However, the improved connection strain
plateaued at a positive displacement of 1.0 in. (25 mm)
and exhibited no increase during the remainder of the test.
This plateau behavior was attributed to the deterioration in
the as-built connection preventing signicant increase in
moment in the improved connection at the high displace-
ments. Because the dowel strain continued to increase until
and both connections used the same deck-reinforcement
detail. Both connections exhibited excellent negative-
moment performance, resisting moments that were 2.5 to
3.0 times higher than the maximum demand realized under
horizontal seismic loading.
Comparison of phase 2 load
with vertical acceleration effects
As previously mentioned, Caltrans recommendations call
for the incorporation of a static vertical load, both upward
and downward, equal to 25% of the dead load for ordinary
standard bridges where the site peak ground acceleration is
0.6g or greater. In addition, these bridges need to include
longitudinal-side mild reinforcement in the girders that is
capable by means of shear friction to resist 125% of the
dead-load shear at the cap beam interface. Neither connec-
tion detail in this test incorporated the additional longitudi-
nal steel, yet both details showed successful performance
when subjected to vertical load.
To correlate the phase 2 performance to the current recom-
mendations, the loads reached during phase 2 were used to
determine the equivalent vertical acceleration magnitudes
that were reached during the test. Figure 21 includes lines
that indicate the connection moment that would be expect-
ed during the gravity G, full horizontal seismic H, and no
vertical acceleration V condition (G + H); the gravity, full
horizontal seismic, and 0.25g constant vertical acceleration
condition (G + H + 0.25V); and the gravity, full horizontal
Figure 22. Gap opening and dowel strain at girder-cap interface (phase 2).
Note: 1 in. = 25.4 mm.
Figure 23. Dowel-bar strains in center girder midlevel bars at peak displace-
ments. Note: 1 in. = 25.4 mm.
93PCI Journal | September–October 2015
The as-built cap-to-girder connection successfully re-
sisted positive moment demand equivalent to the grav-
ity, full horizontal seismic, and 1.0g constant vertical
acceleration condition but exhibited nonlinear response
under positive and negative moments. This capacity
surpasses current recommendations related to vertical
acceleration and suggests that the longitudinal-side
mild-reinforcement requirement in the Caltrans Seismic
Design Criteria can be eliminated for the I-girders.
The improved connection detail, which incorporated
grouted, unstressed strands for positive moment conti-
nuity, provided excellent performance. The improved
connection remained elastic under positive moments
throughout phase 1 testing and exhibited lower relative
displacement at the girder–to–cap beam interface than
an as-built connection. The improved detail also suc-
cessfully transferred shear forces and did not allow verti-
cal unseating or collapse of the superstructure. However,
the test setup did not allow full quantication of the
improved detail. Further investigation into this detail is
under way and can be found in Sritharan et al.19,20
The superior performance of the improved connection
was more apparent during phase 2 loading, which pro-
duced maximum shear and moment conditions in the
connection region that were approximately double the
expected maximum demand from the gravity and full
horizontal seismic condition. The improved connection
remained elastic and produced a gap opening at the
lower girder-cap interface under positive-moment load-
ing that was only 6% of the corresponding opening in
the as-built connection. The elastic performance of the
improved connection in the positive-moment connec-
tion was veried at moment demands well in excess of
the gravity, full horizontal seismic, and 1.0g constant
vertical acceleration simulated load condition.
The as-built connection deteriorated when subjected to
the phase 2 load sequence that exercised the connection
beyond the gravity and full horizontal seismic condi-
tion. While the as-built connection produced a positive
moment resistance that was over 50% higher than the
expected gravity and full seismic condition, it did not
maintain this resistance at large vertical girder displace-
ments, and the girders eventually pulled out of the cap.
While the as-built connection is sufcient for existing
structures, the improved connection is recommended
for new structures because of its superior performance.
The unstressed strand in the improved connection reduced
the girder–to–cap beam gap opening and maintained
elastic behavior of the connection. However, it did not
drastically reduce the dowel-bar strains except at high dis-
placements. The results demonstrated that the unstressed
strand and dowel bars worked together to form a viable
positive-moment-tension load-transfer mechanism.
the moment plateaued, the dowel bars were engaged in the
improved connection, though the addition of the strand
provided signicant positive-moment tension continuity.
Although the unstressed strand improved the connection
performance, the combined mechanism of the dowel bars
and the unstressed strand was responsible for the positive-
moment tension resistance.
The strain in the improved connection’s unstressed strand
during the phase 2 test (data not included for brevity)
exhibited similar behavior in the exterior, intermediate,
and center girders up to a girder vertical displacement of
0.75in. (19 mm). In each of the girders, the strain gradu-
ally increased as the girder displacement increased. At
higher displacements, the strand strain behaviors among
the three girders were less consistent, likely due to shifting
loading mechanisms as the as-built side of the superstruc-
ture began experiencing distress. However, even at the high
displacements, the maximum measured strain values were
around 2000 με. This magnitude is less than 40% of the
strand yield strain, showing that the strand remained intact
throughout the duration of testing.
Details that facilitate the use of accelerated bridge con-
struction methods in high-seismic regions are desirable.
The inverted-tee cap beam offers an excellent possibility
for using precast concrete elements in bridge superstruc-
tures susceptible to high seismic loads, provided that the
girder-to-cap connection region is properly addressed so
that the bridges can be designed cost effectively. This study
has shown that connections that are sufcient for high
seismic load are feasible. In particular, the combination of
inverted-tee cap beams and I-girders was demonstrated to
be a viable accelerated bridge construction system. Con-
structibility of the detail was straightforward, and sufcient
seismic performance was demonstrated. Based on this
study, the following conclusions can be drawn:
The as-built girder-to-cap connection as prescribed in
the Caltrans Seismic Design Criteria was successful
in behaving as a rigid connection during the phase 1
testing, simulating gravity and full horizontal seismic
conditions. It remained elastic throughout the phase
1 test while allowing plastic hinges to fully develop
at the column ends, though its relative displacements
at the girder-to-cap interface were larger than for the
improved connection. The as-built detail also success-
fully transferred shear forces and did not allow vertical
unseating or collapse of the superstructure. The as-
built connection’s successful performance indicates
that existing inverted-tee dapped-end bridge systems
are sufcient without retrotting to meet current
seismic recommendations. Because the column top
end was not expected to form a plastic hinge in past
design, a retrot of this region may be necessary.
September–October 2015 | PCI Journal94
the Fourth Kwang-Hua Forum, Opening Symposium
of the Tongji Shaking Table Array, Shanghai, China,
December 2012.
8. Caltrans. 2005. Bridge Design Aids. Sacramento, CA:
9. Caltrans. 2013. Seismic Design Criteria. Version 1.7.
Sacramento, CA: Caltrans.
10. AASHTO (American Association of Safety and High-
way Transportation Ofcials). 2003. AASHTO LRFD
Bridge Design Specications. 3rd ed. Washington,
11. Caltrans. 2006. Interims and California Amendments
to AASHTO LRFD Bridge Design Specications. Sac-
ramento, CA: Caltrans.
12. Caltrans. 2008. Bridge Design Specications. Sacra-
mento, CA: Caltrans.
13. Caltrans. 2009. Seismic Design Criteria. Version 1.5.
Sacramento, CA: Caltrans
14. Theimann, Z. J. W. 2009. “Pretest 3-D Finite Element
Analysis of the Girder-to-Cap-Beam Connection of an
Inverted-Tee Cap Beam Designed for Seismic Load-
ings.” MS thesis. Iowa State University.
15. Holombo, J. M., J. N. Priestley, and F. Seible. 2000.
“Continuity of Precast Prestressed Spliced-Girder Bridg-
es under Seismic Loads.” PCI Journal 45 (2): 40–63.
16. Priestley, M. J. N., F. Seible, and G. M. Calvi. 1996.
Seismic Design and Retrot of Bridges. New York,
NY: John Wiley and Sons.
17. Vander Werff, J., and S. Sritharan. 2015. “Girder Load Dis-
tribution for Seismic Design of Integral Bridges.Journal
of Bridge Engineering 20 (1): 04014055-1–04014055-11.
18. Snyder, Rick, Justin Vander Werff, Zach Thiemann,
Sri Sritharan, and Jay Holombo. 2011. Seismic Perfor-
mance of an I-Girder to Inverted-T Bent Cap Connec-
tion. Sacramento, CA: Caltrans. 2011.
19. Sritharan, S., R. Bromenschenkel, J. Vander Werff,
and R. Peggar. 2013. “Two Alternate Connections for
Integral Precast Concrete Girder Bridges in Seismic
Regions.” Report presented at the Seventh National
Seismic Conference on Bridges and Highways, Oak-
land, CA, May 2013.
20. Sritharan, Sri, Justin Vander Werff, Robert Peggar, and
R. Bromenschenkel. 2014. “Seismic Performance of
Precast Girder-to-Cap Beam Connections Designed
Wire-mesh reinforcement was demonstrated to be an
acceptable detail in lieu of traditional transverse rein-
forcement in precast concrete girders.
The authors thank the following individuals for their sup-
port and assistance in the completion of the research pre-
sented in this paper: Caltrans for sponsoring this research
project and Charlie Sikorsky for serving as the project
manager; Michael Keever, Jim Ma, Paul Chung, and other
members of the Caltrans Project Advisory Panel for their
advice and assistance; Sami Megally and PBS&J for their
expertise and guidance in the design and construction of
the test unit; and Professor José Restrepo and the staff of
the Charles Lee Powell Laboratories of the University of
California at San Diego, including Andrew Gunthardt, Paul
Greco, Noah Aldrich, Matt Stone, Christopher Latham,
Bob Parks, John Ward, Bob Peters, Dan McAuliffe, Tim
McAuliffe, Josh Nickerson, Michael Germeraad, Chris
Horiuchi, Habib Charbel, and Taylor Gugino, for all their
hard work, assistance, and expertise in the construction of
the test unit. The hospitality was greatly appreciated. With-
out their help, this research would not have been possible.
1. ASCE (American Society of Civil Engineers). 2013. 2013
Report Card for America’s Infrastructure. Reston, VA: ASCE.
2. Chen., W. F., and L. Duan. 2014. Bridge Engineering
Handbook, Second Edition: Construction and Mainte-
nance. Boca Raton, FL: CRC Press.
3. Chung, Paul C., R. W. Wolfe, J. Q. Fang, ABC Ad-
visory Council, and California Department of Trans-
portation. 2008. Accelerated Bridge Construction
Applications in California: A Lessons Learned Report.
Sacramento, CA: California Department of Transpor-
tation, Division of Engineering Services.
4. Caltrans Accelerated Bridge Construction Council.
2008. “Caltrans ABC Strategic Plan—Development of
Practice and Policy for Future Bridge Projects.” Tech-
nical report, version 1.1. Sacramento, CA: Caltrans.
5. Housner, George W., and Charles C. Thiel Jr. 1990.
“Competing against Time: Report of the Governor’s
Board of Inquiry on the 1989 Loma Prieta Earth-
quake.Earthquake Spectra 6 (4): 681–711.
6. SEAOC (Structural Engineers Association of California)
Seismology Committee. 2010. “Seismic Design of Con-
crete Parking Structure Ramps.Structure (July): 8–10.
7. Hida, Sue. 2012. “California Department of Transpor-
tation’s Next Generation Bridges.” Report presented at
95PCI Journal | September–October 2015
G = gravity load condition
H = full horizontal seismic load condition
V = seismic vertical acceleration load
= horizontal displacement ductility
for ABC.” In Proceedings of the 2014 National Accel-
erated Bridge Construction Conference, Miami, FL:
December 2014.
' = estimated rst yield strength of bridge system
g = acceleration due to gravity
About the authors
Justin Vander Werff, PhD, PE, is
assistant professor and department
chair of the Engineering Depart-
ment at Dordt College in Sioux
Center, Iowa.
Rick Snyder, PE, is senior engineer
at SRF Consulting Group in
Minneapolis, Minn.
Sri Sritharan, PhD, is the Wilson
Engineering Professor in the Civil,
Construction, and Environmental
Engineering Department at Iowa
State University in Ames, Iowa.
Jay Holombo, PhD, PE, is senior
bridge engineer at T. Y. Lin
International in San Diego, Calif.
To promote accelerated bridge construction in seismic
regions, a large-scale experimental investigation was
conducted to examine the seismic sufciency of
precast concrete I-girders in integral bridge super-
structures. Such structures are not frequently used in
high seismic regions due to lack of design guides and
overly conservative design approaches. A half-scale,
17.8 m (58.5 ft) long test unit modeling a portion of a
prototype bridge incorporating a concrete column, ten
I-shaped precast concrete girders, and an inverted-tee
concrete cap beam was used to experimentally verify
that precast concrete members employing accelerated
construction techniques can be used in integral super-
structures and provide excellent seismic performance.
Comparison of an as-built girder-to-cap connection
detail with an improved detail shows that the as-built
detail in existing bridges will satisfactorily resist
positive and negative seismic moments and allow
plastic hinges to develop at the column tops, though
this was not the original design intent. However, the
improved detail, which exhibited excellent seismic
moment resistance, is recommended for new bridges
to avoid potential deterioration of the girder-to-cap
Accelerated bridge construction, connection, experimen-
tal, girder, large scale, seismic.
Review policy
This paper was reviewed in accordance with the Precast/
Prestressed Concrete Institute’s peer-review process.
Reader comments
Please address reader comments to or
Precast/Prestressed Concrete Institute, c/o PCI Journal,
200 W. Adams St., Suite 2100, Chicago, IL 60606. J
... This lack of code provisions stems from the infrequent use of precast concrete historically in seismic design of bridges. Specifically, the use of precast girders in bridges has been hampered by two design criteria: (1) assuming that a positive moment connection between precast girders and the bent cap beam cannot be reliably developed, formation of a plastic hinge at the column (or pier) top is not permitted, making this construction option less cost effective compared with the cast-in-place alternative (Caltrans 2013;Vander Werff et al. 2015); and (2) a connection involving mild steel reinforcement is required between the cap beam and girders to ensure satisfactory shear transfer across the cap beam-to-girder joint when vertical acceleration exceeds 0.25 g (Caltrans 2013). As such, the use of precast girders in bridge structures, when it occurs, is often motivated by overcoming construction challenges associated with cast-in-place concrete. ...
... Sritharan et al. (2001) investigated the use of precast cap beams in multicolumn bents, and Holombo et al. (2000) investigated the seismic response of spliced precast I-girders and bathtub girders. Restrepo et al. (2011) andVander Werff et al. (2015) reported the response of two different precast I-girder-to-cap-beam connections. In box girder bridges, large precast segments connected with bonded and/or external (unbonded) PT can be used along the bridge length [ Fig. 12(b)]. ...
Precast concrete facilitates a construction method using durable and rapidly erectable prefabricated members to create cost-effective and high-quality structures. In this method, the connections between the precast members as well as between the members and the foundation require special attention to ensure good seismic performance. Extensive research conducted since the 1980s has led to new precast concrete structural systems, designs, details, and techniques that are particularly suited for use in regions of high seismic hazard. This paper reviews the state of the art of these advances, including code developments and practical applications, related to four different systems: (1) moment frames; (2) structural walls; (3) floor diaphragms; and (4) bridges. It is concluded from this review that the widespread use of precast concrete in seismic regions is feasible today and that the jointed connection innovation introduced through precast research leads to improved seismic performance of building and bridge structures.
... An accelerated method has been applied on a bridge construction in high earthquake area [8]. The experiment found that an inverted tee-cap beams can be superior for using precast concrete elements in the bridge superstructure. ...
... The current California Department of Transportation (Caltrans) Seismic Design Criteria 6 assumes this connection degrades to a pinned connection during a seismic event, which means that precast concrete girder bridges are not cost-effective in seismic regions. 7 Consequently, Vander Werff et al. 8 studied the seismic response and overall moment resistance of this inverted-tee connection concept and found that when subjected to seismic loading, it developed sufficient moment resistance to ensure formation of plastic hinges at the column top. Figure 1 shows two proposed design concepts for cost-effective and easily implementable connections in seismic regions. ...
Full-text available
This paper investigates the use of unstressed strands as connection reinforcement for precast concrete members through both experimental and analytical programs. The primary focus of the study was to establish positive moment connections using un-stressed strands between precast concrete girders and a cap beam for seismic applications. ■ An experimental program was designed to investigate the fundamental load-transfer characteristics of unstressed strands. A total of 20 pullout tests were conducted with varying parameters, including strand diameter, anchorage detail, embedment length, and loading type. ■ An analytical program examined the bond stress of unstressed strands, and design guidelines for using unstressed strands to connect precast concrete members are presented. P refabricated construction techniques using precast concrete members are known to provide better quality control and a more efficient construction process than traditional construction techniques involving cast-in-place concrete. The precast concrete members are generally fabricated off-site and connected on-site to form an integral structural system. Therefore, effective and feasible field connections are essential to ensure satisfactory performance of structures designed with precast concrete components. In current industry practice, dry and wet connections are two commonly used techniques to connect precast concrete members and develop adequate capacity within the connection region. A wet connection is defined as reinforcement in the connection region that is spliced with mechanical couplers, welds, or lap splices, and filled with grout or cast-in-place concrete. A dry connection is any other type of connection. 1 Although not used frequently, extending prestressing strands from the ends of precast concrete members into the connection region provides a cost-effective and easily imple-mentable wet connection for several reasons. First, strands are flexible, and thus they can be easily routed and positioned within the connection. Second, strands present high tensile strength, therefore requiring significantly fewer of them for the same connection capacity compared with mild reinforcing bars. Third, strands used for pretensioning the precast concrete members do not have to be cut off after full development of concrete strength; they can be extended from PCI Journal (ISSN 0887-9672) V. 66, No. 3, May-June 2021.
Shake-table studies were conducted on a large-scale, two-span accelerated bridge construction (ABC) bridge model that incorporated 38 precast concrete elements and six emulative ABC connection types. The studies were aimed to assess the seismic performance of the bridge model and the six ABC connections under biaxial horizontal ground motions simulated at the Earthquake Engineering Laboratory at the University of Nevada, Reno. The bridge model was subjected to successive earthquake motions ranging from 20% to 200% of the design earthquake motion. Test results showed that the ABC bridge model matched the target performance of cast-in-place (CIP) bridges by limiting plastic hinging to columns and maintaining other components as damage-free. The ABC connections maintained the structural integrity of the system even under drift ratio demands on the columns higher than 7.5% combined with substantial in-plane rotation (e.g., larger than 0.008 rad) of the superstructure.
Technical Report
Full-text available
This report presents the research conducted as part of an investigation for the California Department of Transportation (Caltrans) regarding the seismic response and overall moment capacity of precast I-girder to inverted-T bent cap bridge connections for seismic applications. The current design practice, as outlined by Caltrans’ Seismic Design Criteria, assumes that the connection between the precast I-girders and the inverted-T bent cap will degrade in a seismic event and shall therefore be designed as a pinned connection, making the precast girder option for seismic bridges inefficient. A prototype I-girder to inverted-T bent cap bridge and a 50% scale test unit was designed in order to investigate the behavior of the as-built girder-to-cap connection region. Additionally, per the request of Caltrans, an improved girder-to-cap connection detail was developed in order to ensure a fully continuous moment connection between the I-girders and inverted-T bent cap. A finite element grillage model was developed using ABAQUS and SAP2000 and was used to predict the global and local responses of various aspects of the test unit. The test unit was constructed and tested in two phases of quasi-static cyclic testing. The first phase was a horizontal load test phase, which simulated the effects of gravity and seismic loads on the entire test unit. The second phase was a vertical load test phase, which specifically focused on the positive and negative moment capacity of the connections. Both the results of the finite element grillage model and the testing were used to make conclusions regarding the performance of I- girder to inverted-T bent cap bridges. It was concluded that the current I-girder to inverted-T bent cap bridge connection is capable of acting as a fully continuous connection for both positive and negative moments during both gravity and seismic loading, contrary to the design assumptions stated in Caltrans’ Seismic Design Criteria. The improved connection detail demonstrated the ability to ensure a fully continuous moment connection between the I-girders and inverted-T bent cap. Both connection details also exhibited a significant moment resistance beyond what was expected during the vertical load test although the as-built connection eventually failed under positive moments at moderate to large displacements.
Full-text available
Current seismic design practice related to integral bridge girder-to-cap beam connections allows little or no lateral seismic load to be distributed beyond the girders immediately adjacent to the column. However, distribution results from several large-scale tests have shown that the distribution of column seismic moment typically engages all the girders. An approach utilizing simple stiffness models to predict distribu-tion in integral bridge structures is presented in detail; distribution predictions based on grillage analyses also are compared. The experimental results and the analytical results from the stiffness and grillage models show that current design methods related to vertical load distribution are sufficiently accurate. However, when applied to the distribution of lateral load, similarly obtained results reveal that current design practice does not appropriately account for the amount of load that is distributed beyond the girders adjacent to the column to the nonadjacent girders. The current practice leads to excessive girder-to-cap connection reinforcement, increased girder depth, unnecessarily high seismic mass, and increased construction cost. Finally, this paper makes recommendations for more appropriate distribution of seismic lateral load in integral bridge superstructures.
The Principles and Application in Engineering Series is a series of convenient, economical references sharply focused on particular engineering topics and subspecialties. Each volume in this series comprises chapters carefully selected from CRC's bestselling handbooks, logically organized for optimum convenience, and thoughtfully priced to fit every budget. From the award-winning Bridge Engineering Handbook, Bridge Engineering: Construction and Maintenance takes an in-depth look at the construction engineering and maintenance aspects of steel and concrete bridges. Topics include effective project management, construction procedures and practices, construction and maintenance inspections and ratings, and bridge strengthening and rehabilitation.
A design methodology is developed for precast, prestressed concrete spliced-girder bridges with superstructure-column continuity under seismic loads. Verification of the method was provided by constructing two 40 percent scale model bridge structures and testing them under fully reversed simulated seismic forces and displacements in the longitudinal direction. These model bridges included a column and the contributory bentcap with the superstructure extending from midspan to midspan of the prototype. This allowed model construction and testing to occur under realistic field conditions and seismic forces. One model bridge featured precast bulb-tee girders and the other bathtub girders. The performance of the model bridge structures exceeded the design requirements during the tests. Particular emphasis is placed on the integral column-superstructure bentcap design and its performance.
Introduction Constraints Seismic Design Alternatives Deep-Valley-Crossing Considerations Long-Span-Bridge Considerations
This thesis provides a pretest analysis of a commonly used and increasingly popular superstructure-to-cap-beam connection used in the state of California subjected to seismic loading. The current connection investigated is used for precast girders placed on inverted-tee cap beams. The development of a prototype bridge and a 50% scale test unit was discussed. Then multiple models were analyzed ranging from a single girder with a cap beam to a assembly with a column, cap beam, diaphragms, deck slab and ten girders based on the prototype and test unit. The analysis models were analyzed in the general purpose finite element program ABAQUS. A new connection detail utilizing an unstressed, prestressing strand placed through the girders was then modeled and proposed to provide a moment resistant connection and allow a plastic hinge to form in the top of the column. By allowing a plastic hinge to form in the top of the column the system is more economical then the current design detail.
A Board of Inquiry was appointed by the Governor of California to investigate the 1989 Loma Prieta Earthquake. The formation of the Board was prompted by earthquake damage to bridges and freeway structures and the desire to know not only what happened, but how to prevent such destruction in future earthquakes. The Board made fifty-two specific findings and eight recommendations. They identified three essential challenges that must be addressed by the citizens of California, if they expect a future adequately safe from earthquakes: 1) ensure that earthquake risks posed by new construction are acceptable; 2) identify and correct unacceptable seismic safety conditions in existing structures, and; 3) develop and implement actions that foster the rapid, effective, and economic response to and recovery from damaging earthquakes The Loma Prieta earthquake should be considered a clear and powerful warning to the people of California. Although progress has been made during the past two decades in reducing earthquake risks, much more could have been done, and awaits doing. More aggressive efforts to mitigate the consequences of future, certain earthquakes are needed if their disastrous potential is to be minimized and one of the most fundamental of responsibilities of government is to be fulfilled—to provide for the public safety. The Governor signed an Executive Order implementing the principal recommendations of the Board that may prove to be the most significant step to improve seismic safety within the State taken in the last several decades. It establishes the policy that all state owned and operated structures are to be seismically safe and that important structures are to maintain their function after earthquakes.
Two Alternate Connections for Integral Precast Concrete Girder Bridges in Seismic Regions
  • S Sritharan
  • R Bromenschenkel
  • J Vander
  • R Werff
  • Peggar
Sritharan, S., R. Bromenschenkel, J. Vander Werff, and R. Peggar. 2013. "Two Alternate Connections for Integral Precast Concrete Girder Bridges in Seismic Regions." Report presented at the Seventh National Seismic Conference on Bridges and Highways, Oakland, CA, May 2013.