Anda di halaman 1dari 18

Computers and Structures 72 (1999) 6380

SR5 Lake Washington Ship Canal Bridge pushover


p
analysis
Thomas A. Ballard*, Hassan Sedarat
SC Solutions Inc., Suite 202, 3211 Scott Blvd., Santa Clara, CA 95054, USA

Abstract
In this paper the results of SC Solutions' performance-based evaluation of the as-built Unit 10 frame of the SR5
Lake Washington Ship Canal Bridge are presented. This unit is a three-span frame, located between bridge piers 24
and 27 on the north approach to the bridge and extends approximately 329 ft between expansion joints. A threedimensional model of the Unit 10 frame was developed using the computer program ADINA (ADINA User
Interface Command Reference Manual, Volume 1: ADINA Model Denition, ADINA R&D, Inc., September 1997)
and ADINA's moment-curvature beam element. This paper discusses the method of evaluation of this frame, the
material and member properties used for the `pushover' analysis, properties of the as-built structure, and the results
of capacity analysis of the as-built Unit 10 assembly. Several pushover analyses were performed, consisting of mass
proportional and mode shape lateral load proles. One of the mass proportional lateral load analyses was selected
for presentation in this paper. This load pattern resulted in maximum shear forces in the frame columns. Following
the pushover analysis, a nonlinear dynamic time-history ADINA analysis was conducted to estimate maximum
displacements and to evaluate the performance of the bridge during a simulated earthquake ground motion. # 1999
Elsevier Science Ltd. All rights reserved.

1. Introduction
The SR5 Lake Washington Ship Canal Bridge is
considered to be of signicant importance such that its
collapse during a seismic event would be detrimental
to the safety and commerce of the people of Seattle,
Washington [2,3]. Therefore, the Washington State
Department of Transportation commissioned a retrot
design that would result in a predicted minimum level
of performance of `no-collapse' during the seismic
event.

Some of the original gures for this paper were generated


with a colour-producing terminal and submitted in colour.
* Corresponding author. Fax: +1-408-486-6083.
p

The SR5 Lake Washington Ship Canal Bridge is


located on Interstate 5 in Seattle, Washington. This is
a multi-level bridge with twelve lanes of trac, opened
in 1961 with an average trac load in excess of
150,000 vehicle per day. The bridge is 4430-foot long,
including an 1156-foot long south approach and a 980foot long north approach. The approach structures are
comprised of multi-span reinforced concrete frames
with box-girder deck on two levels. These approach
structures are further divided into frames or `units',
which are separated from adjacent frames by expansion joints located over split columns. Foundations
consist of spread footings and pile foundations.
This as-built performance-based evaluation was
commissioned in order to assess the eectiveness of
proposed seismic retrots. These proposed retrots

0045-7949/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 4 5 - 7 9 4 9 ( 9 9 ) 0 0 0 2 2 - X

64

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

consist of strengthening the superstructure elements


and increasing the ductility of the supporting columns,
such that all non-linearities are forced to occur at
specic locations at the top and bottom of each column, thus protecting the superstructure and foundation from damage. However, this `capacityprotection' retrot scheme did not address the extent
of damage to the columns nor did it address the relative eectiveness of each column's retrot in protecting
the superstructure. The frame which comprises Unit 10
of the north approach is a non-symmetric, non-regular
and multi-deck structure for which an evaluation using
linear elastic solution techniques was considered unreliable. The purpose of this evaluation is to determine
the order of formation of each plastic hinge in the
bent frames and to estimate the degree of damage each
column will sustain for a predicted response of the asbuilt system. The results of this `pushover' analysis
and subsequent earthquake dynamic analysis can be
extended to determine the cost associated with retrotting the bridge to a specied performance level and the
resulting safety margins and to estimate the cost of
repairs associated with a postulated seismic event.
This paper discusses the method of evaluation of the
as-built Unit 10 frame, the material and member properties used for the `pushover' analysis, properties of
the as-built structure, and results of the capacity analysis of the Unit 10 assembly. The evaluation began with
an assessment of the as-built bridge, assuming that a
strength upgrade for the superstructure will be required
to satisfy the `strong beam weak column' strategy
adopted for this evaluation. This analysis is also useful
for identifying the vulnerabilities of the system after
column retrots have been carried out, because jacketing of the columns will not increase their moment capacity signicantly, but will result in a signicant
curvature ductility increase and prevention of shear
failures. The sequence of plastic hinge formation was
predicted, and the resulting demand forces that will be
transmitted to the superstructure were estimated.

2. Method of evaluation
SC Solutions' evaluation of Unit 10 of the SR5
Lake Washington Ship Canal Bridge is designed to
estimate the performance of the frame by studying
plastic curvature of the columns, the displacements of
the decks and total base shear. The evaluation began
with construction of two-dimensional (2D) bent frame
models, which include nonlinear moment-curvature
representation of all column members. The superstructure, including the box-girder deck and cross-beams
was modeled with linear elastic elements. This assumes
that these components will be retrotted to increase

their strength, concentrating all non-linearities into the


supporting columns.
Moment-curvature properties were derived based on
material models discussed in the next section. These
material properties were incorporated into column
cross-section models which are exercised in the BIAX
program [7] which computes, using numerical integration, the cross section moment-curvature properties,
considering concrete and reinforcing stressstrain
properties. In this manner, the column's moment-curvature properties are computed for specied levels of
compressive and tensile axial force, resulting in a complete denition of the axial force-bending moment
yield surface.
This yield surface is the basis for determining column hinging during the pushover analysis. The geometry of the frames is described using linear and
nonlinear nite elements for the superstructure and
columns, elastic springs for soil stiness, and rigid
links to couple structural components.
For this evaluation, it was considered important to
provide enough renement along the length of the column to capture any plasticity that may occur. For
instance, Pier 24 columns are split at the top to provide for thermal expansion of the deck structure. This
split is carried to an elevation of top-of-footing plus 25
ft. The column will either hinge just above the split or
at the top-of-footing. In order to capture this behavior,
the entire column is modeled with moment-curvature
elements.
The analysis applied dead load followed by transverse force-controled pushes. This tested the model
and material behavior and demonstrated the frame
behavior in a direction that could be readily understood. Following this, the entire Unit 10 three-dimensional (3D) frame model was exercised, including all
four bent frames, with the push direction varied. Push
forces are computed based on mass distribution and
mode shapes of the system that have the predominant
eective modal mass participation. For the mode
shape distributions, amplitude times the lumped tributary mass of the deck and columns is the basis for the
relative push force at each deck level.
As the pushover analysis proceeds, the formation of
plastic hinges and the curvature or maximum strain at
each hinge is tracked for evaluation of damage.
3. Material properties
All material and section properties are based on the
original construction drawings [1] and recommendations provided by Priestley et al. [4].
Concrete and steel properties, for all members in the
Unit 10 frame, were specied as 3000 psi and Grade
50, respectively. Since the purpose of this evaluation is

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

65

Fig. 1. Stressstrain properties.

to estimate the most probable capacity of this frame,


the concrete strengths are factored by 1.5 and steel
strength is factored by 1.1. The concrete strength factor considers increase in strength for concrete aging
and conservative batching processes and is actually a
lower bound for typical bridges [4]. Fig. 1 shows the
concrete stressstrain properties used to develop column moment-curvatures properties.
Conned concrete properties are based on Mander's
model [5], using the 3/8 in. steel jacket to provide connement. Since the cross-section of the columns is
unique and non-standard, a computer program capable
of generating moment-curvature properties for general
sections was employed. This program, BIAX, uses a
modied KentPark concrete model for conned concrete. Therefore, the KentPark concrete model was t
to the Mander model deriving the appropriate factors.
Reinforcing steel properties are computed based on
Ref. [4] recommendations. Minimum probable yield
stress is 55 ksi with ultimate strength 1.2 times yield
stress and stress at ultimate strain equal to 60 ksi. A
strain hardening plateau is considered. Fig. 2 presents
the stressstrain diagram for the reinforcing steel.

4. Section properties
Each column is evaluated for moment-curvature
based on engineering mechanics principles and material
properties discussed in the previous section. Typical
column section geometry is shown in Fig. 3 for both
as-built and retrot congurations.
Cross-section geometry is described as an assemblage of rectangular sections with reinforcing placed
relative to the concrete rectangles. Each piece is subdivided for the purpose of numerical integration to
determine resisting force and moment for a specied
section axial strain and bending curvature. The as-built
column properties are determined considering that all
of the concrete is unconned. The proposed steeljacket retrot will not signicantly increase the column
stiness or strength at the extremities of the member,
where the plastic hinges form, and will increase the
strength and ductility only based on consideration of
the degree of connement of the enclosed concrete.
This has been demonstrated in full-scale concrete column tests at the University of San Diego [6]. This is
due to lack of composite action between the steel

66

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 2. Grade 40 reinforcing steel stressstrain properties.

Fig. 3. Typical column cross-section at footing.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

jacket and concrete column at the ends. Away from


the ends the stiness and strength will be increased due
to this composite action. Therefore, no appreciable
increase in lateral load resistance of the frame will
result from this retrot detail. The pushover results
presented in this paper are for the as-built frame and,
therefore, do not reect the improvement in frame
ductility due to the retrot detail. However, the asbuilt pushover analysis can be used as a qualitative
tool to evaluate the proposed capacity protection retrot.
For this study, the as-built columns are therefore
modeled as unconned concrete and the moment-curvature properties are developed along their entire
length. The retrot columns are modeled as conned
concrete and the increase in stiness due to the jacket
and grout is not considered.
Fig. 4 is axial force-bending moment (PM ) curves,
at maximum concrete compression strains of 0.003 and
0.004, which were dened as the maximum acceptable
strain, for a typical as-built and retrot column, respectively. These PM curves were the basis for determining the axial forces for which moment-curvature
are computed. Five axial force values are selected for
this computation. These ve axial force points include

67

maximum tensile axial force, maximum compressive


axial force, and axial force associated with maximum
moment. Two additional points are selected between
the balanced axial force and maximum compressive
force and balanced axial force and maximum tensile
force. This results in a relatively good representation
of the PM curve for the section, as shown in Fig. 4.
Fig. 5 is the resulting moment-curvature curves for
this typical column as computed by BIAX.
5. Unit 10 three-dimensional evaluation
The 3D as-built pushover analysis of Unit 10 frame
is presented in this section. ADINA model details
included consideration of the deck centerline oset
from the bent cap and cross-beam. The ends of the
column connections to the upper deck are oset from
the deck by one-half the deck thickness. Lower deck
oset from cross-beam centerline is designed to capture
torsional moments in the cross-beam due to longitudinal deck forces for the 3D model. This feature was incorporated in the model to quantify the torsion in
the cross-beam member. Finite elements that form
the cross-beam-to-column joint and the cross-beam

Fig. 4. Typical concrete section PM curves.

68

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 5. Typical concrete section moment-curvature.

Fig. 6. ADINA nite element model of Unit 10 frame.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

member are modeled as linear elastic, since the capacity protection retrot strategy includes increasing
the ultimate strength of these components. The bent
cap and deck members were also modeled as linear
elastic. The nite element model is shown in Fig. 6.
This gure shows the outline of the structure as hidden
lines and the nite element elements as blue lines. All
column elements were modeled using the nonlinear
moment-curvature beam element. Soil springs were
also included based on the geotechnical information.
Two types of lateral load patterns for the pushover
analyses were selected. A mass proportional pattern
results in higher shear forces in columns, whereas
modal pattern results in a exural mechanism. Since
the results are voluminous, this paper presents only
sample results from the mass proportional pushover.
The evaluation began with an assessment of the
mode shapes and frequencies of the structure. Since
none of the important modes has an eective modal
mass ratio of more than 75%, a lateral load pattern
proportional to a single mode may not capture the importance of the higher modes. Therefore, a linear elastic response spectrum analysis was performed to
obtain the load prole in terms of combination of all
modes. Lateral load pattern for this pushover analysis

69

was proportional to displacements of the structure derived based on CQC modal and SRSS spatial combinations.
Column element properties were derived consistent
with the methodology discussed in Section 4 for both
bending axes of the sections.
5.1. Unit 10 modal properties
Mode shapes were computed by ADINA based on
the initial stiness of the moment-curvature elements.
Figs. 710 show the rst four mode shapes of the Unit
10 frame as computed by ADINA. Modes 2 and 3 are
transverse modes with the south end and north end of
the frame participating separately in their respective
modes. Mode 1 is a longitudinal mode with both
upper- and lower-deck participation. Mode 4 is a
longitudinal mode with mostly lower deck participation.
Since the behavior of the Unit 10 frame is complex
and not as easily described as the 2D pushovers, the
push force was determined in several ways. The rst
four pushover analyses were performed using mass
proportional loads. The frame and deck reactive mass
was lumped at the two deck levels and the frame was

Fig. 7. Unit 10 ADINA mode shape 11.592 s period.

70

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 8. Unit 10 ADINA mode shape 21.208 s period.

Fig. 9. Unit 10 ADINA mode shape 30.9295 s period.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

71

Fig. 10. Unit 10 ADINA mode shape 40.424 s period.

pushed to collapse. The two horizontal directions were


pushed simultaneously with the signs of the forces varied to derive four force vectors. These cases were for
positive longitudinal and positive transverse forces;
positive longitudinal and negative transverse forces;
negative longitudinal and negative transverse forces;
and negative longitudinal and positive transverse
forces. The results of four push force proles showed
similar forcedisplacement relations. Therefore, detail
results are only presented for the positive longitudinal
and positive transverse force case.
The second method of pushing the frame was based
on an CQC modal and SRSS spatial combinations of
the rst forty modes of the system with static corrections made for the missing higher modes. A linear elastic response spectra analysis was carried out using the
design spectrum.
5.2. Unit 10 mass proportional pushover results
This section presents a small fraction of the results
of the mass proportional pushovers. The sequence of
plastic hinge formation was determined and column
shear demand forces were computed for comparison to
shear capacities. Column axial forces were also computed. Column bending moments were computed for
information only, since curvature ductility is the basis
for capacity calculations. Deck axial forces, vertical

and transverse shear forces and transverse bending


moments are also computed. Bent-cap and cross-beam
moments and forces are also recovered and presented.
5.2.1. Unit 10 pushover: mass proportional
The mass proportional pushover results for the Unit
10 frame are presented in this section. Fig. 11 shows
the pushover curve presented in the form of percent of
total Unit 10 frame mass versus resultant deck displacements. This curve is used to determine the total base
shear versus equivalent displacement for a target
demand on the structure.
The collapse scenario is presented in Fig. 12. In this
gure, each plastic hinge is shown as a red or blue circle at the approximate location of hinge formation.
Red hinges represent yielding about the transverse axis
of the bridge and blue hinges represent yielding about
the longitudinal axis of the bridge. Hinge sequences
are numbered in the order of their initial formation,
i.e., the point at which elastic curvatures are exceeded
in the section. The point at which a hinge exceeds its
ultimate curvature capacity is indicated by placing the
sequence number for that event in parenthesis.
This yielding sequence indicates that the ramp is
rst to experience plastic hinging at 17% g equivalent
lateral load and collapses at about 25% g equivalent
lateral load or 0.75 ft of resultant upper deck displacement. Pier 26 lower columns form hinges at about

72

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 11. Mass proportional pushover curve.

Fig. 12. Pushover: plastic hinge sequence.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

0.19 g equivalent lateral load. At 0.31 g equivalent lateral load and 1.00 ft resultant upper deck displacement,
the two east columns of Pier 26 exceed curvature capacity at their lower ends. The remaining gures in this
section are samples of the resulting forces and moments
in the system as the unit is pushed. These graphs are
used to determine the demands in the elastic members of
the structure at each point in the pushover history and
to check if there is any shear failure in the columns.
Figs. 1320 present typical elastic and inelastic forces
and moments.
6. Dynamic analysis of Unit 10
In a pushover analysis the response is controlled by
a single `pre-dened deformed shape' or `mode' and its
shape is assumed to remain unchanged throughout the
push. Therefore, eects of higher modes may not be
addressed properly. Establishing the direction of push
in a complex 3D structure so that meaningful results
can be interpreted is not straightforward. In the foregoing pushover analyses, we used response spectrum
analysis technique as well as mass proportional loadings to estimate a most probable collapse mode for the
push and to include the eects of higher modes.
However, when the structure undergoes nonlinear
behavior the mode shapes and frequencies changes and
the direction of the push that was obtained with these
push load proles will not hold in general. Due to the

73

complexity of the bridge, both in plan and elevation,


and in order to make sure that the eects of higher
modes are taken into account properly, it was necessary to validate the pushover study with at least one
inelastic dynamic time-history analysis. With this type
of analysis the response of structure is not limited by
the `assumed deformed shape' that might be used for
the pushover analysis, because nonlinearity in the
structure will modify the response in time.
6.1. Input motions
The SR5 bridge was subjected to the S90W and vertical components of the 1941 El Centro ground
motions. The Peak Ground Acceleration (PGA) of the
input ground motions in longitudinal, transverse, and
vertical directions were scaled to 0.29 g. Fig. 21 shows
the input ground motion time-histories.
6.2. Summary of the results
In this section, we are summarizing the results
obtained from the time-history analysis of the bridge.
Fig. 22 shows the time-history of the displacement at
the upper level of the bridge in the transverse and
longitudinal directions. The response displacement
time-histories were obtained by averaging the displacements of each node in the upper level deck.
The forcedisplacement relations shown in Figs. 23
and 24 are total base shear versus the averaged

Fig. 13. Pushover: typical column longitudinal shear forces.

74

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 14. Pushover: typical column transverse shear forces.

Fig. 15. Pushover: typical column longitudinal bending moments.

Fig. 16. Pushover: typical column transverse bending moments.

Fig. 17. Pushover: typical column axial forces.

76

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 18. Pushover: typical lower cross-beam and bent-cap vertical shear.

Fig. 19. Pushover: typical lower cross-beam and bent-cap longitudinal shear.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 20. Pushover: typical lower cross-beam and torsional moments.

Fig. 21. 1941 El centro ground motion: input displacement time-histories.

77

78

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

Fig. 22. Response displacement time-histories at the upper level.

Fig. 23. Forcedisplacement relations in the longitudinal direction.

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

79

Fig. 24. Forcedisplacement relations in the transverse direction.

displacement of the upper level of the bridge. The


pushover curve obtained from the modal push is also
overlayed on these graphs. These graphs clearly show
that the pushover curve bound the response and that
the maximum displacement is not more than 1.5 ft.
The sequence of plastic hinges were similar to the
pushover analyses, conrming the results obtained
from these analyses.

7. Conclusions
Unit 10 of SR5 Lake Washington Ship Canal Bridge
was modeled with the as-built properties. Two-dimensional models of bents 24 through 27 and 3D model of
the entire Unit were developed. The results of the
pushover analyses were briey summarized in this
report. It was shown based on Unit 10 pushover analyses that the frame's ultimate displacement capacity is
between 2.0 and 3.0 ft provided column retrots were
implemented to increase curvature ductility. Column
hinge sequence and locations were predicted. Axial
forces, shear forces and bending moments in columns
and decks were summarized. Shear forces, torsional
moments, and bending moments of cross beam and
bent cap were also summarized. These results can be
readily employed in a performance based design

method using Acceleration-Displacement-ResponseSpectrum (ADRS). Performance of the `force controlled action' components and `deformation controlled action' components can be easily evaluated
using these results.
Three-dimensional pushover analysis of structures is
a complex but useful procedure to estimate the capacity of structures. The same complexity in this analysis demands more careful attention in determination of
the target displacement. Once the target displacement
is established, the extend of columns retrots need to
be determined based on the types and locations of failure.
A minimum of one 3D nonlinear dynamic time-history analysis is recommended to validate the target displacement and retrot design.

References
[1] Primary State Highway No. 1, Seattle Freeway, Lake
Washington Ship Canal Bridge, Washington State
Highway Commission, Department of Highways,
Olympia, Washington, 1958, pp. 1013, 1720, 2356.
[2] SR5 Lake Washington Ship Canal Bridge, Seismic
Vulnerability Assessment, Part 1, Seismic Assessment
Prole, Andersen Bjornstad Kane Jacobs, 1993.

80

T.A. Ballard, H. Sedarat / Computers and Structures 72 (1999) 6380

[3] SR5 Lake Washington Ship Canal Bridge, Seismic


Vulnerability Assessment, Part 2, Detailed Seismic
Assessment, Andersen Bjornstad Kane Jacobs, 1993.
[4] Priestley MJN, Seible F, Calvi GM. Seismic Design and
Retrot of Bridges. New York: Wiley, 1996.
[5] Mander JB, Priestley MJN, Park R. Theoretical stress

strain model for conned concrete. Journal of Structural


Division, ASCE 1988;114(8).
[6] Correspondence with F. Seible, April 29, 1998.
[7] Wallace JW, Moehle JP. BIAX: a computer program for
the analysis of reinforced concrete sections, Report No.
UCB/SEMM-90/12, University of California at Berkeley,
Berkeley, CA, July 1989.

Anda mungkin juga menyukai