You are on page 1of 9

Seismic Vulnerability and Mitigation during Construction

of Cable-Stayed Bridges
John C. Wilson1 and Keith Holmes2

Abstract: The implications of earthquake loading during balanced cantilever construction of a cable-stayed bridge are examined.
Finite-element models of a cable-stayed bridge were developed and multiple ground motion time history records were used to study the
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

seismic response at the base of the towers for six stages of balanced cantilever construction. Probabilistic seismic hazard relationships
were used to relate ground motions to bridge responses. The results show that there can be a high probability of having seismic responses
共forces/moments兲 in a partially completed bridge that exceed, often by a substantial margin, the 10% /50-year design level 共0.21% per
annum兲 for the full bridge. The maximum probability of exceedance per annum was found to be 20%. This occurs because during
balanced-cantilever construction the structure is in a particularly precarious and vulnerable state. The efficacy of a seismic mitigation
strategy based on the use of tie-down cables intended for aerodynamic stability during construction was investigated. This strategy was
successful in reducing some of the seismic vulnerabilities so that probabilities of exceedance during construction dropped to below 1% per
annum. Although applied to only one cable-stayed bridge, the same approach can be used for construction-stage vulnerability analysis of
other long-span bridges.
DOI: 10.1061/共ASCE兲1084-0702共2007兲12:3共364兲
CE Database subject headings: Bridges, cable-stayed; Seismic effects; Bridges, cantilever; Aerodynamics; Cables; Bridge
construction.

Introduction In this study, finite-element models were used to represent


successive stages of balanced cantilever construction of a cable-
The vulnerability of cable-stayed bridges during balanced canti- stayed bridge, ranging from a single bare tower through to the
lever construction is a well-known fact for wind engineering of completed bridge. The responses 共forces and moments兲 of the
this type of bridge. However, in seismic design, typically only the completed bridge to self-weight and seismic loads were used as
completed state of the bridge is considered; earthquake response reference values and were considered to be the conventional de-
during construction is not usually addressed. The reasoning be- sign quantities. As a measure of seismic vulnerability during con-
hind this approach is that the time of exposure during construc- struction, seismic responses at various stages of construction were
tion is short, and therefore the probability of occurrence of the compared to the response values for the completed bridge, and
earthquake ground motions used for design of the full bridge probabilities of exceedance for ground motions that would result
共e.g., corresponding to a 475-year return period兲 is much lower in the reference response values being exceeded during construc-
than it is during the service life of the bridge. tion were computed. The effectiveness of superstructure tie-down
There is a hidden weakness in this reasoning because the bal- cables 共which are typically provided for aerodynamic stability
anced cantilever method of construction puts a cable-stayed during construction兲 to reduce the vulnerability of the bridge to
bridge into a particularly precarious and vulnerable position. In earthquake loading during construction was also examined.
essence, during construction, the bridge is a very different struc-
ture than when it is in its completed state. Relatively small to
moderate earthquake ground motions 共with a much more frequent Modeling and Analysis of Construction Stages
rate of occurrence than the design-level ground motions for the
full bridge兲 may actually impose significant seismic loads on the
Bridge Details
partially completed bridge.
The study was based on the Fred Hartman Bridge, a 675-m cable-
1
Professor, Dept. of Civil Engineering, McMaster Univ., 1280 Main stayed bridge located near Houston 共Fig. 1兲. Although not located
St. W., Hamilton ON, Canada L8S 4L7 共corresponding author兲. E-mail: in a seismic zone, the bridge was selected for study because com-
jcwilson@mcmaster.ca plete structural details were available, and detailed finite-element
2
Engineer, ND LEA Consultants Ltd., Suite 600-1455 West Georgia models of the complete bridge had already been developed by one
St., Vancouver BC, Canada V6G 2T3. of the writers. The overall dynamics of the bridge during con-
Note. Discussion open until October 1, 2007. Separate discussions
struction and in the completed state are prototypical of this size of
must be submitted for individual papers. To extend the closing date by
one month, a written request must be filed with the ASCE Managing
cable-stayed bridge. This is an important overall consideration
Editor. The manuscript for this paper was submitted for review and pos- here, as the dynamics of the bridge would not be expected to be
sible publication on February 25, 2005; approved on June 22, 2006. This substantially different had the bridge been located in a seismic
paper is part of the Journal of Bridge Engineering, Vol. 12, No. 3, May zone.
1, 2007. ©ASCE, ISSN 1084-0702/2007/3-364–372/$25.00. The bridge has two double-diamond-shaped towers 共134 m in

364 / JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007

J. Bridge Eng. 2007.12:364-372.


Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Fred Hartman Bridge 共curvature is on approach spans, not on


cable-stayed spans兲

height兲, each with four legs, 192 cables in a fan arrangement, and
two independent continuous superstructures with composite decks
共an east and a west deck兲 with a main span of 381 m and side
spans of 147 m. The two 23.5-m-wide decks are connected to
each tower and to the anchor piers so as to transmit both longi-
tudinal and transverse shear loads into these supports. The bridge
was erected using the balanced cantilever method. Deck extension Fig. 2. Modeling of six stages of balanced cantilever construction
progressed incrementally using preassembled sections, the length
of which equaled the distance between cables 共14.6 m for the side
spans, 15.8 m for the main span兲. Temporary tie-down cables rotation of the superstructure about the global transverse y-axis of
were used to avoid potential aerodynamic instabilities during con- the connection but were fixed for motions in all other degrees of
struction; details about these cables appear later. freedom.
Note that the term “balanced cantilever construction” refers
here to the notion that, during construction, the deck is progres-
Construction Stages
sively extended in an approximately weight-balanced manner on
Six three-dimensional 共3D兲 finite-element models, shown in Fig. both sides of the towers. At a particular stage of construction, due
2, were developed to represent the following stages of construc-
tion: 共1兲 a single bare tower; 共2兲 three stages of balanced-
cantilever extension of the superstructure from a single tower
representing approximately one-eighth, one-fourth, and three-
eighths completion; 共3兲 the half bridge just prior to closure where
the midspan is free and the side span is connected to the approach
pier; and 共4兲 the full completed bridge including both towers.
These six stages were named tower, 10-15, 6-19, 4-21, half, and
full, where the numbers refer to cable locations at the ends of the
cantilevered sections shown in Fig. 3共a兲. A 3D view of the full
model is shown in Fig. 3共b兲, including the global coordinate axes
to which references will be made throughout this paper.
The concrete deck was modeled using shell elements, while
the steel girders, floor beams, and concrete towers were modeled
using 3D beam elements and the cables modeled using uniaxial
tension elements. Rigid offset links were used at the cable-girder,
cable-tower, and girder-deck connections for accurate modeling
of the geometry. To simplify the analysis, and because complete
details on the foundations were not available to the writers, the
tower bases were taken to be fixed in all degrees of freedom.
Even though cable-stayed bridges often have deep foundations,
the use of a fixed base is suitable for this study because the
analysis makes relative comparisons between responses at various
stages of construction 共all stages having a fixed base兲 and does
not evaluate the response in an absolute sense. Connections of the Fig. 3. 共a兲 Cable numbering system for cantilever construction
superstructure to the anchor piers and towers allowed for free models; 共b兲 global coordinate system for bridge

JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007 / 365

J. Bridge Eng. 2007.12:364-372.


Table 1. Modal Information for Dynamic Models Damping in cable-stayed bridges tends to be complicated and is
Effective thought to be a function of a number of parameters: cable
modal mass Shortest configuration, level of excitation, mode number and shape, com-
Number of Number of 共%兲 period in ponent coupling, and geometric and material composition. A
Construction deck modes analysis damping ratio of 2% was used for all modes for all stages of
stage segments used X Y Z 共s兲 construction. This is consistent with results reported by Stiemer et
Tower 0 / 22 14 85 86 2 0.257 al. 共1988兲, Kawashima and Unjoh 共1991兲, Wilson and Liu 共1991兲,
10–15 6 / 22 30 84 83 31 0.255 Clemente et al. 共1998兲, and Wilson and Atkins 共2000兲.
6–19 14/ 22 50 89 89 53 0.256
4–21 18/ 22 60 91 90 58 0.266 Seismic Input
Half 22/ 22 80 81 86 63 0.221
Full 45/ 45 150 85 88 69 0.258 Earthquake ground motions were selected and scaled with consid-
Note: See Figs. 2 and 3共a兲 for models. eration for the periods of the structure under study. The ratio of
peak ground acceleration 共units: g兲 to peak ground velocity 共units:
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

ms−1兲, defined as the a / v ratio, is one suitable measure of the


to slightly different section sizes on the main and side spans, period content of a time-history record and provides a useful ap-
slightly more deck weight may be cantilevered on one side of the proach for selecting records 共Tso et al. 1991; FIB 2003兲. A low
tower than the other. So, despite the name, some unbalanced dead a / v ratio 共⬍0.8 gsm−1兲 is typically associated with longer period
load forces/moments are likely to exist at the base of the tower. ground motions from more distant earthquakes, whereas a high
ratio 共⬎1.2 gsm−1兲 is characteristic of shorter period ground mo-
Seismic Analysis tions from near source events. As this bridge is a long period
structure 共having many modes with periods longer than 1 s兲, only
Cable-stayed bridges typically have a large number of closely longer period records 共with low and intermediate a / v ratios兲 were
spaced modal frequencies within the period that is the range of considered for this study. Twelve 3-component ground motion
interest for seismic loading. This study examines time-history records were selected to obtain a diverse range of earthquake
seismic analyses, using as inputs the earthquake records described loadings: six records with low a / v ratios and six with intermedi-
in the next section. The effective modal mass for a particular ate ratios.
mode measures the fraction of the total mass of the structure that The component of each record with the largest velocity was
is excited by that mode in each global direction 共Clough and scaled to a peak ground velocity of 0.25 m / s 共a value represen-
Penzien 1993兲. The number of modes included in the analysis was tative of a 475-year event in the Vancouver, B.C. area兲, and the
selected so that the percentage of effective modal mass remained other two components were then scaled by the same factor. The
nearly the same in all the construction stage models. As the dy- use of velocity scaling 共instead of scaling by peak ground accel-
namic equilibrium equations are solved in sequence, the longest eration兲 recognizes the dependence of the response of intermedi-
period mode is solved first and the short period vertical tower ate and longer period structures to the velocity of the ground
modes are not solved until late in the analysis. motion. This dependence, demonstrated in the work of Tso et al.
For a large model 共i.e., the full model兲, the analysis can be- 共1991兲, Naeim 共1993兲, and Heidebrecht 共1995兲, is directly recog-
come preoccupied with many long-period deck modes, and the nized by the National Building Code of Canada 共1995兲 and by
vertical modes of the towers will not appear in the analysis until FIB 共2003兲. Wesolowsky and Wilson 共2003兲 have used this ap-
a much higher mode number is reached 共in this case, higher than proach in seismic analysis of other cable-stayed bridges. For
mode 250兲. In contrast, a model with many fewer degrees of time-history analysis, the 12 records were input to each model in
freedom 共i.e., tower兲 will include most vertical modes early in the one orientation with respect to the structure’s global axis system,
analysis. As all the vertical tower modes need not be included in and then the horizontal components were rotated 90° and used as
the full bridge model, those modes were excluded from the other a second set of inputs, for a total of 24 inputs.
smaller models in order to maintain consistency across all the
analyses.
As the vertical modes of the deck and tower were uncoupled,
with large differences in frequency, a mode cutoff was selected Seismic Response during Construction
for each model where the percentage of effective modal mass in
the vertical direction closely approximated the percentage of deck Seismic forces and moments were computed at the base of the
mass relative to the total mass of the structure for that particular tower for all six models in the directions of the local axes of the
stage of construction. This ensured that essentially all of the deck inside and outside legs shown in Fig. 4. While there were some
mass participated in responses in the vertical direction, and gen- differences in the responses of the inside and outside legs in the
erally about 85% or more of the structural mass participated in various construction stages, the responses of the inside leg gov-
responses in the horizontal directions. On this basis, Table 1 gives erned, almost exclusively, the seismic response of the base of the
the number of modes used for each model, the effective modal tower. These quantities were used as the basic measure of the
masses, and the shortest modal periods. The shortest modal peri- seismic response of the bridge for this study. Other parts of the
ods for each model are essentially the same, although these cor- bridge may also exhibit seismic vulnerabilities, but these were not
respond to different modes for each model. addressed within the scope of this work. Also not addressed was
Geometric nonlinearities are usually not of significant concern the issue of capacities of the structural members, because, as
for the seismic response analysis of cable-stayed bridges for stated earlier, the analysis is based on relative comparisons 共not
main-span lengths less than 500 m 共Nazmy and Abdel-Ghaffar absolute responses兲 of structural responses at various construction
1990; Endo et al. 1991; Clemente et al. 1998; Chang et al. 2001兲. stages.
Material nonlinear behavior was not considered in this study. Seismic responses in the six construction stages for each of the

366 / JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007

J. Bridge Eng. 2007.12:364-372.


Table 3. Mean Maximum Seismic Responses ⌿E = Ec / E f at Base of
inside Leg of Tower during Construction Stages Referenced to Seismic
Load Effects in Full Bridge
Construction stage
Response
quantity Tower 10–15 6–19 4–21 Half Full
Longitudinal x-force 1.58 2.64 4.10 4.64 1.21 1.00
Transverse y-force 0.78 0.95 1.30 1.12 0.87 1.00
Axial z-force 0.71 0.94 1.33 1.17 0.86 1.00
Moment about x-axis 0.82 0.99 1.35 1.17 0.88 1.00
Moment about y-axis 2.32 3.60 5.35 6.05 0.84 1.00
Torsion about z-axis 1.46 1.83 0.73 0.44 1.49 1.00
Note: Maximum value of each response quantity is underlined.
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

response to load in each direction. For example, the greatest effect


of seismic loading 共compared to self-weight兲 is clearly the mo-
ment about the x-axis where the response quantities have the larg-
est values of ⌿D.
The seismic force/moment responses in the various construc-
Fig. 4. Modeling and axes details at base of tower tion stages in Table 2 were compared to the seismic response of
the full bridge using
six force/moment quantities at the base of the inside leg are pre-
⌿E = Ec/E f 共2兲
sented in Table 2 in the normalized form of
where E f is the corresponding mean of the 24 maximum re-
⌿D = Ec/D f 共1兲
sponses of the full bridge under seismic load alone. The ⌿E val-
where Ec is the mean of the 24 maximum responses 共e.g., for ues are reported in Table 3; ⌿E is helpful in identifying the force/
longitudinal force兲 from the seismic time history analyses in a moment quantities where there is significant vulnerability and in
particular construction stage, and D f is the fixed value of the same understanding the seismic behavior of the structure at various
response quantity due to dead load 共self-weight兲 acting on the full stages of construction. Self-weight effects are incorporated into
bridge 共hence the subscript D on ⌿兲. Thus, ⌿D relates the earth- this equation later. A value of ⌿E ⬎ 1 implies that the seismic
quake forces/moments at the various stages of construction to the response in the construction stage exceeds the seismic effect ex-
corresponding self-weight forces/moments in the full bridge 共note perienced by the full bridge.
that ⌿D is not a “base shear coefficient,” as D f is not the total From Table 3 it is evident that for every response quantity, the
weight of the structure兲 and provides an indication of the suscep- seismically generated forces/moments during one or more con-
tibility of the various directions and construction stages to seismic struction stages exceed the corresponding seismic response of the
loading. For example, the mean maximum seismic force in the full bridge. For each response quantity, the maximum value oc-
longitudinal direction for the inside leg of the tower stage is 0.52 curs in one of the three balanced cantilever stages; rarely is the
of the corresponding dead-load effects in the longitudinal direc- half stage of concern. As shown in Table 3, the two quantities
tion at this location in the full bridge. with the greatest response in each of the first four stages of con-
As seen in a number of cases in Table 2, the seismic forces struction are the seismic force in the longitudinal direction and
during the first four construction stages exceed the corresponding the moment about the transverse axis. These quantities are at a
dead-load effects on the full bridge 共⌿D ⬎ 1兲. However, for the maximum in the 4-21 stage when there is maximum cantilever
half and full bridge stages, Ec exceeds the corresponding D f only extension of the deck. An understanding of the seismic behavior
for transverse force and moment about the longitudinal axis. Only in these two critical directions is necessary for subsequent analy-
for the vertical direction and for torsion are the seismic effects sis of the seismic mitigation strategies. With only one exception
less than the dead-load effects in all stages. The behavior ob- 共torsion in the 10-15 stage兲 is the maximum response for the other
served in Table 2 reflects the mechanisms involved in the bridge’s four quantities less than 1.5; in the half and tower stages, except
for moment about the y-axis, the responses are less than 1.6.
Table 4 gives the periods and effective modal masses for the
Table 2. Mean Maximum Seismic Response ⌿D = Ec / D f at Base of first two longitudinal modes of the first four construction stages.
inside Leg of Tower during Construction Stages Referenced to Dead
These two modes captured approximately 75 to 85% of the par-
Load Effects in Full Bridge
ticipation in the longitudinal direction. The first longitudinal
Response
Construction stage mode for the first four models 关Fig. 5共a兲兴 was distinguished by
quantity Tower 10–15 6–19 4–21 Half Full single curvature cantilever bending of the tower and no bending
in the deck. The second longitudinal mode 关Fig. 5共b兲兴 was distin-
Longitudinal x-force 0.52 0.87 1.35 1.53 0.40 0.33
guished by double curvature in the tower and slight antisymmetric
Transverse y-force 1.64 2.00 2.73 2.36 1.82 2.10 bending of the deck 共one end up, one end down兲. The behavior of
Axial z-force 0.49 0.65 0.92 0.81 0.59 0.69 the half and full models did not require a detailed analysis as the
Moment about x-axis 13.27 15.96 21.73 18.85 14.23 16.15 side-/main-span closures defined their longitudinal behavior.
Moment about y-axis 1.00 1.55 2.30 2.60 0.36 0.43 In the first four construction stages, the longitudinal seismic
Torsion about z-axis 0.60 0.75 0.30 0.18 0.64 0.41 force is carried by only the base of the tower. As construction
Note: Maximum value of each response quantity is underlined. progresses, this force increases steadily until the 4-21 stage 共Table

JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007 / 367

J. Bridge Eng. 2007.12:364-372.


Table 4. Characteristic of Significant Longitudinal Modes Table 5. Response Values of ⌿D+E = 共Dc + Ec兲 / 共D f + E f 兲 at Base of inside
Leg of Tower for Two Response Quantities with Greatest Vulnerability
Total
1st longitudinal mode 2nd longitudinal mode effective Construction stage
model
Response quantity Tower 10–15 6–19 4–21 Half Full
Effective Effective mass of
modal modal first 2 Longitudinal x-force 0.40 0.67 1.07 1.23 0.88 1.00
Construction Period mass Period mass modes Moment about y-axis 0.70 1.17 2.14 2.68 0.85 1.00
stage 共s兲 共%兲 共s兲 共%兲 共%兲 Note: Maximum value of each response quantity is underlined.
Tower 3.51 53 0.68 24 77
10–15 4.24 43 1.05 33 76
the seismic response, the dominant contribution to the effective
6–19 7.21 15 1.73 70 85
modal mass shifts from the first to the second mode. This shift is
4–21 9.73 9 1.96 78 87 evident in the values in Table 4. The period of the second mode
remains within the range of significant seismic excitation.
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

The moment about the transverse axis is a direct function of


2兲. The decrease in force at the half stage occurs because the the longitudinal forces induced by the mass of the deck, with the
connection at the end of the side span causes the longitudinal tower acting as a moment arm. Therefore the increase in seismic
force to be shared between the tower and the anchor pier. The moments about the transverse axis through the construction stages
increase in longitudinal seismic force as the deck is extended shown in Table 2 follows a pattern similar to the variation in
occurs because, as the period of the first mode lengthens to the longitudinal forces.
point where 共in the 6-19 and 4-21 stages兲 it no longer influences Eq. 共2兲 was modified to include the combination of dead load
共self-weight兲 plus seismic loads 共each with a load factor of 1.0兲.
This is expressed as

⌿D+E = 共Dc + Ec兲/共D f + E f 兲 共3兲


where 共Dc + Ec兲⫽mean maximum response in a construction stage
due to the combined effects of dead load at the construction stage
共Dc兲 and mean maximum seismic load 共Ec兲, and 共D f + E f 兲⫽corres-
ponding response of the full bridge under combined dead and
mean maximum seismic loads.
Table 5 shows ⌿D+E values for the two orientations of greatest
vulnerability. For longitudinal force, the maximum ⌿D+E value of
1.23 is in stage 4-21. With the inclusion of the dead-load effects,
the previously identified vulnerability of the 4-21 stage under
seismic loading alone was greatly reduced, due to the combined
effect of small Dc and large D f values in Eq. 共3兲. For moments
about the transverse axis, the maximum ⌿D+E value of 2.68 is
also in the 4-21 stage. This is greater than the corresponding value
for longitudinal force because of the unbalanced dead-load mo-
ments about the y-axis. At the 4-21 stage, the dead-load moments
about the y-axis were about 20% larger than they were for the full
bridge.

Probabilistic Assessment

The probability of exceedance P共t兲 of a specified peak ground


motion parameter during t years is given by

P共t兲 = 1 − 共1 − p兲t 共4兲


where p is the probability of exceedance per annum. The seismic
inputs used in these analyses were based on a peak ground veloc-
ity of 0.25 m / s having a 10% probability of exceedance in
50 years 共10% /50-year event兲. For this event the probability of
exceedance per annum is p = 0.0021; the return period is
p−1 = 475 years. For a bridge with an intended design life of
100 years 共typical for many major bridges兲, the probability of
exceedance of the 10% /50-year event over 100 years is 19%, as
calculated from Eq. 共4兲. If the entire duration of bridge construc-
tion is, say, 2 years, the probability of exceedance is 0.42%.
If one considers a critical stage of construction that might exist
Fig. 5. Longitudinal modes of balanced cantilever stages: 共a兲 first for, say, 4 months, the probability of exceedance during this short
mode; 共b兲 second mode time frame is 0.07%. On this basis, one might consider that the

368 / JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007

J. Bridge Eng. 2007.12:364-372.


chance of exceeding the design-level earthquake during the time
of construction is so small that it could be neglected, and even
that the change of the design earthquake occurring during this
period is, in effect, negligible. However, another scenario that
needs to be considered is the smaller but more frequently occur-
ring earthquake that acts on a rather precariously positioned
partial bridge structure.
The following question is now addressed: What is the prob-
ability of occurrence of ground motion levels smaller than the
475-year design earthquake that would cause a response 共forces/
moments at the base of the tower兲 in the partially completed
bridge equal to the response in the completed bridge to the full
design event? It is important to emphasize that this question looks
at the ground motions that will achieve responses in the partially
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

completed bridge that are the same as in the completed bridge. It


does not address the issue of whether the partially completed
bridge will be damaged by this level of ground motion; that will
depend upon the performance criteria set for the 475-year design
earthquake. The performance criteria will, in turn, establish the Fig. 6. Seismic hazard curve for peak ground acceleration used in
capacities required for the various structural components. For this study, based on seismicity of Vancouver area and scaled to 0.25g
many major cable-stayed bridges, the performance level set for for a 10% /50 year event
the completed bridge under a 475-year design earthquake is es-
sentially elastic behavior, or, at most, minimal localized inelastic
behavior. If this is the case, then the performance of the partially moments at the base of the tower will be exceeded during con-
completed bridge would be the same. struction than when the bridge is in its completed state.
To address the above question, Eqs. 共2兲 and 共3兲 were used to The most extreme case occurs for moment about the y-axis
calculate a reduced peak ground acceleration aR 共units: g兲 that due to combined dead load plus seismic load where there is a
would result in force/moment responses at the base of the tower 20% probability of exceedance per annum of the full bridge de-
during the most critical stage of construction to be equal to the sign moment. The physical reason for this high vulnerability, as
responses of the full bridge to the 475-year 0.25 m / s design explained earlier, is that the static and dynamic characteristics of
earthquake the partially completed bridge are dramatically different than for
the full bridge. In this 4-21 stage, the unbalanced dead load in-
For seismic load only: aR = 0.25共E f /Ec兲g 共5兲 creases the vulnerability so that the full bridge design level re-
sponse is reached at a peak ground acceleration of only one-half
of that required to raise the seismic force alone to this level 共0.021
For dead load plus seismic load: aR = 0.25共E f + D f − Dc兲/Ecg
versus 0.041g兲.
共6兲 On the other hand, for forces in the longitudinal direction, the
The analysis in this part uses peak ground acceleration as a dead load acts to reduce the vulnerability of the 4-21 stage 共peak
quantity required for evaluating the probability of annual exceed- ground acceleration required to reach the full bridge design force
ance. The original time-history records were scaled to a peak in the longitudinal direction increases from 0.054 to 0.20g兲. In
ground velocity of 0.25 m / s. For Vancouver, the peak ground this case, the probability of exceedance per annum decreases from
acceleration 共in units of g兲 is numerically equal to the peak 3.5 to 0.32%, closer to but still somewhat above the 0.21% full
ground velocity 共in units of m/s兲, and therefore peak ground ac- bridge design level. Therefore, the probability of exceeding the
celeration is equal to 0.25g.
Probabilities of exceedance per annum for the reduced accel-
erations were determined using the seismic hazard curve shown in Table 6. Vulnerability of 4-21 Construction Stage
Fig. 6. This curve is based on data in the NBCC commentary for Reduced peak
Vancouver for probabilities of exceedance of various mean peak ground Probability of Return
ground accelerations, and scaled for a 10% /50-year event peak accelerationa exceedanceb period
ground acceleration of 0.25g. Adams and Atkinson 共2003兲 and Response quantity Loads 共g兲 共% per annum兲 共years兲
Heidebrecht 共2003兲 provide more detailed discussion of hazard
Longitudinal force E 0.054 3.5 30
curves. These authors note that the slopes of the hazard curves
D+E 0.20 0.32 315
can vary considerably from one seismic region to another. In this
work, a single slope representative of the seismic hazard of the Transverse force E 0.19 0.35 285
Vancouver area was used. Axial force E 0.19 0.35 285
Using Fig. 6 and the aR values, seismic vulnerability results Moment at E 0.19 0.35 285
are reported in Table 6 for all orientations under seismic load 共E兲 longitudinal axis
and for the two most vulnerable orientations, longitudinal forces Moment at E 0.041 6.0 15
and moments about the transverse axis, under combined dead and transverse axis D+E 0.021 20 5
seismic loads 共D + E兲. In all cases, the probabilities of exceedance Torsion E 0.14 0.61 165
a
per annum for seismic loading alone are greater than 0.21%, with Computed from Eqs. 共5兲 and 共6兲 and values in Tables 2 and 5.
a value as high as 5.9%. In other words, there is a much greater b
From Fig. 6, accelerations and probabilities of exceedance reported to
probability per annum that the full design earthquake forces/ two significant figures; return periods rounded to nearest 5-year value.

JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007 / 369

J. Bridge Eng. 2007.12:364-372.


level of the full bridge design forces/moments during construction
depends upon how the dead load affects the forces/moments in
the various construction stages.
An assessment of acceptable risk will depend on several fac-
tors 共e.g., financial, political兲 specific to a given project. This
assessment may include a decision to have a separate set of per-
formance criteria for construction that may accept a higher risk
and consequently the possibility of occurrence of earthquake-
induced damage in the partially completed structure that might
not be acceptable in the completed bridge. Notwithstanding these
important decisions, however, the results for probabilities of ex-
ceedance presented here show that there can be a marked differ-
ence in the levels of vulnerability of the bridge at various stages
of construction. Knowledge of these levels can lead to more in-
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

formed decisions on how to handle the seismic risk involved dur-


ing construction. Furthermore, they suggest that a strategy to
reduce the seismic vulnerability during construction is well worth
consideration.

Seismic Mitigation

Mitigation Strategy and Analysis


Tie-down cables are commonly used to reduce the potential aero-
dynamic instability of a cable-stayed bridge during cantilever
construction 共Green 1990兲. This section examines potential ben-
efits of making use of these cables to reduce the seismic vulner-
ability of the bridge. The critical response quantities, longitudinal
force and moment about the transverse axis, both of which reach
a maximum just prior to side-/main-span closure at the 4-21 stage,
are considered here. Responses are based on the same seismic
input used in the original analysis and are presented here only for
this critical 4-21 stage.
Two spatial configurations of tie-down cables were considered
for the 4-21 stage: 共1兲 TIE 1 关Fig. 7共a兲兴 was based on the con-
figuration used for aerodynamic stability during the construction
of the Hartman Bridge, and 共2兲 TIE 2 关Fig. 7共b兲兴 is a modification
of TIE 1, where the cables are anchored at the deck farther from
the tower. In both configurations, a total of 12 tie-down cables Fig. 7. Seismic mitigation schemes using tie-down cables: 共a兲 TIE 1;
were used in three sets anchored at the edges of the deck: 8 cables 共b兲 TIE 2
on the main span and 4 on the side span.
At ground level, the main-span tie-downs were anchored at the
base of the tower and the side-span tie-downs were anchored to vulnerability of the 4-21 stage. TIE 1B was essentially ineffective
two specially constructed concrete blocks 45 m from the tower. in being able to reduce the response. For the longitudinal force,
The geometric and material properties of the tie-down cables these three tie-down arrangements reduced ⌿D+E to values below
were based on the specific detailing of those used for the Hartman what would be experienced by the bridge in its fully completed
Bridge. For each TIE configuration, two sets of cable prestress state 关recall from Eq. 共3兲 that ⌿D+E ⬍ 1 means that the response in
values were examined: the first 共Set A兲, was based on the mini- the construction stage is less than in the completed bridge兴. These
mum prestress needed to avoid loss of tension in the cables under same three tie-down arrangements also reduced the ⌿D+E values
seismic loading, and the second 共Set B兲 was based on aerody- for moments about the transverse axis, but in this case they re-
namic tie-down specifications for TIE 1 and for improved seismic mained higher than would be experienced by the bridge in its
performance for TIE 2. completed state. TIE 2B was the single most effective arrange-
The improvements provided by the tie-down strategy for 4-21 ment for reducing the vulnerability of the 4-21 stage.
are summarized in Table 7 with updated values of ⌿E and ⌿D+E. The responses with the tie-down cables varied significantly
While the spatial configuration of TIE 1 and TIE 2 affects values and depended on the prestress forces. As discussed earlier for the
of these parameters, the prestressing forces are considered as dead original configuration, under the combined loading the longitudi-
loads and therefore their magnitude affects only the ⌿D+E values. nal force vulnerability was significantly reduced, but moments
Hence, there are double entries for ⌿D+E in Table 7. For the about the transverse axis were larger as the unequal lengths in
tie-downs, ⌿E ranges from 2.70 to 4.82, and ⌿D+E ranges from side and main spans at the 4-21 stage generated unbalanced dead-
0.78 to 2.54. In all cases, the 4-21 stage is more vulnerable to the load moments at the base of the tower. Appropriate prestress in
transverse moment response than to longitudinal force. TIE 1A, the tie-down cables, however, can be used to counter these unbal-
TIE 2A, and TIE 2B were all effective in reducing the seismic anced dead-load moments. This was clearly illustrated for the TIE

370 / JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007

J. Bridge Eng. 2007.12:364-372.


Table 7. Influence of Tie-Down Cables on Seismic Response of 4-21 Stage
Tie-down Tie-down
response/original response/original
Response Spatial response response
quantity configuration ⌿E 共%兲 Prestress ⌿D+E 共%兲
Longitudinal force Original 4.64 — — 1.23 —
4-21 stagea
TIE 1 3.52 76 TIE 1A 0.90 73
TIE 1B 1.14 93
TIE 2 2.70 58 TIE 2A 0.78 63
TIE 2B 0.80 65
Moment about Original 6.05 — — 2.68 —
transverse axis 4-21 stagea
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

TIE 1 4.82 80 TIE 1A 1.82 68


TIE 1B 2.54 95
TIE 2 3.81 63 TIE 2A 1.54 57
TIE 2B 1.26 47
a
No tie-downs; values are from Tables 2, 3, and 5.

2A and TIE 2B configurations, where the first set of prestress less than the full bridge design criterion of 0.21% per annum
values was based on the minimum prestress necessary to prevent 共hence these values are not shown in Table 8兲. So for this re-
loss of tension in the cables, while the second set was selected to sponse quantity, all of the tie-down configurations are very effec-
minimize the unbalanced dead-load moments. For moments about tive at reducing the seismic vulnerability. For transverse moments
the transverse axis, this refinement in prestress values from TIE under combined loading, the probability of exceedance of the full
2A to TIE 2B resulted in a 15% reduction in ⌿D+E. More system- bridge design level was 1% per annum or less for all configura-
atic optimization of the configuration and prestressing of the tie- tions except TIE 1B. The lowest value, 0.41% per annum, while
down cables could be done, but this was beyond the scope of this still twice the target design value for the full bridge, is still a vast
study. improvement over the 20% per annum for the original 4-21 stage
with no tie-downs. TIE 1B provided effectively no reduction from
Probabilistic Assessment the 20% value for moments from the original configuration.
The probabilistic approach used to assess the tie-down structures
was identical to that described previously. A summary of prob-
abilities associated with the mitigation strategy is presented in Conclusions
Table 8. The probability of exceedance per annum of the full
bridge design-level moments for the critical 4-21 construction The seismic response of a cable-stayed bridge during various
stage was as high as 4.2% for TIE 1 and 2.6% for TIE 2. The stages of balanced cantilever construction has been examined.
reduction in probability of exceedance per annum for both longi- The approach is also applicable to other cable-stayed and long-
tudinal force 共from 3.5 to 1.4%兲 and transverse moment 共from 5.9 span bridges. The results presented here challenge a common mis-
to 2.6%兲 for TIE 2 reflects the substantial impact of this tie-down conception that seismic loading during construction need not be
configuration in reducing the seismic response. considered because of the relatively short duration of construc-
All tie-down configurations reduced the probability of exceed- tion. The analysis presented here shows that the seismic vulner-
ance of dead plus seismic forces in the longitudinal direction to ability during construction may be substantially greater than when

Table 8. Seismic Mitigation in 4-21 Construction Stage with Tie-Down Cables


Seismic load Dead load+ seismic load +prestress load in tie-downs
Probability of Return Probability of Return
Spatial exceedance period exceedance period
Response quantity configuration 共% per annum兲 共years兲 Prestress 共% per annum兲 共years兲
a
Longitudinal force Original 4-21 stage 3.5 30 Original 4-21 stagea 0.32 315
TIE 1 2.3 45 See note below
TIE 2 1.4 70
Moment about Original 4-21 stagea 6.0 15 Original 4-21 stagea 20 5
transverse axis TIE 1 4.2 2.5 TIE 1A 1.0 100
TIE 2A 0.67 150
TIE 2 2.6 40 TIE 2B 0.41 245
Note: Blank cells are where response during construction was less than response of full bridge and hence probabilities were not calculated; TIE 1–2 is not
shown for moment response as there was no significant improvement from the original 4-21 configuration.
a
No tie-downs; values are from Table 6; probabilities of exceedance reported to two significant figures; return periods rounded to nearest 5-year value.

JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007 / 371

J. Bridge Eng. 2007.12:364-372.


the bridge is in its completed state. In particular, the bridge is in challenge of a long cable-stayed bridge—Tatara Bridge.” Proc., Semi-
its most vulnerable state during the final stages of superstructure nar on Cable-Stayed Bridges—Recent Developments and Their Fu-
extension, prior to closure of the side and main spans. Using a ture, M. Ito, Y. Fujino, T. Miyata, and N. Narita, eds., Elsevier Sci-
probability of exceedance of 10% in 50 years 共probability of ex- ence, New York, 417–438.
ceedance of 0.21% per annum兲 as the assumed design criterion of Heidebrecht, A. C. 共1995兲. “Insights and challenges associated with de-
the full bridge reveals that the seismic loads in the partially com- termining seismic design forces in a loading code.” New Zealand Nat.
pleted structure can have up to 6% probability of exceedance per Soc. Earthquake Eng. Bull, 28共3兲, 224–246.
annum. Considering both dead and seismic loads, the probability Heidebrecht, A. C. 共2003兲. “Overview of seismic provisions of the pro-
posed 2005 edition of the National Building Code of Canada.” Can. J.
of exceedance per annum can be as great as 20%.
Civ. Eng., 30, 241–254.
A strategy was proposed that uses tie-down cables to effec-
International Federation for Structural Concrete 共FIB兲. 共2003兲.
tively mitigate seismic vulnerability during construction. It was “Displacement-based seismic design of reinforced concrete build-
shown that the probability of exceedance of the full bridge design ings.” State-of-Art Rep., Lausanne, Switzerland.
loads during construction could be reduced to 1% per annum or Kawashima, K., and Unjoh, S. 共1991兲. “Seismic behavior of cable-stayed
less. The reduction in vulnerability depended upon both the spa-
Downloaded from ascelibrary.org by New York University on 05/14/15. Copyright ASCE. For personal use only; all rights reserved.

bridges.” Proc., Seminar on Cable-Stayed Bridges—Recent Develop-


tial configuration of the tie-down cables and their prestress. ments and Their Future, M. Ito, Y. Fujino, T. Miyata, and N. Narita,
eds., Elsevier Science, New York, 193–212.
Naeim, F. 共1993兲. “Selection of earthquake records for seismic design of
Acknowledgments tall buildings.” Struct. Des. Tall Build., 2, 255–293.
National Research Council of Canada 共NBCC兲. 共1995兲. National building
The writers thank the Natural Sciences and Engineering Research code of Canada, Associate Committee on the National Building Code,
Council of Canada for its financial support. Thanks are also due to Ottawa.
the anonymous reviewers who provided thoughtful comments that Nazmy, A. S., and Abdel-Ghaffar, A. M. 共1990兲. “Non-linear earthquake-
helped to clarify several aspects of the paper. response analysis of long-span cable-stayed bridges: Theory.” Earth-
quake Eng. Struct. Dyn., 19, 45–62.
Stiemer, S. F., Taylor, P., and Vincent, D. H. C. 共1988兲. “Full scale dy-
References namic testing of the Annacis Bridge.” IABSE Proc., P-122/88, IABSE,
Zurich, 1–16.
Adams, J., and Atkinson, G. M. 共2003兲. “Development of seismic hazard Tso, W. K., Zhu, T. J., and Heidebrecht, A. C. 共1991兲. “Engineering
maps for the proposed 2005 edition of the National Building Code of implication of ground motion A/V ratio.” Soil Dyn. Earthquake Eng.,
Canada.” Can. J. Civ. Eng., 30, 255–271. 11, 133–144.
Chang, C. C., Chang, T. Y. P., and Zhang, Q. W. 共2001兲. “Ambient vibra- Wesolowsky, M. J., and Wilson, J. C. 共2003兲. “Seismic isolation of cable-
tion of long-span cable-stayed bridge.” J. Bridge Eng., 6共1兲, 46–53. stayed bridges for near-field ground motion.” Earthquake Eng. Struct.
Clemente, P., Marulo, F., Lecce, L., and Bifulco, A. 共1998兲. “Experimen- Dyn., 32共13兲, 2107–2126.
tal modal analysis of the Garigliano cable-stayed bridge.” Soil Dyn. Wilson, J. C., and Atkins, J. C. 共2000兲. “Analysis of damping in the
Earthquake Eng., 17, 485–493. earthquake response of cable-stayed bridges.” Proc., 12th World Conf.
Clough, R. W., and Penzien, J. 共1993兲. Dynamics of structures, McGraw- on Earthquake Engineering.
Hill, New York. Wilson, J. C., and Liu, T. 共1991兲. “Ambient vibration measurements on a
Endo, T., Iijima, T., Okukawa, A., and Ito, M. 共1991兲. “The technical cable-stayed bridge.” Earthquake Eng. Struct. Dyn., 20共8兲, 723–747.

372 / JOURNAL OF BRIDGE ENGINEERING © ASCE / MAY/JUNE 2007

J. Bridge Eng. 2007.12:364-372.

You might also like