Vous êtes sur la page 1sur 12

Performance Evaluation of Damaged Steel Frame Buildings

Subjected to Seismic Loads


Kihak Lee1 and Douglas A. Foutch, M.ASCE2

Abstract: This paper presents a systematic evaluation approach that allows the design professional to estimate the performance of
damaged buildings subjected to sets of ground motions representing different hazard levels. The performance of damaged buildings for a
future seismic event prior to repair and/or rehabilitation was evaluated using an elastic model and stiffness of the structure. The linear
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

static model using the damaged state of brittle fractured connections in steel moment resisting frame buildings provides a reasonable
estimation of the interstory drift of the damaged buildings when compared to nonlinear time history analysis. Then, this result was
incorporated in a reliability-based procedure that provides a degree of confidence that a building before and after the damage will exceed
the given performance levels. Finally, examples of confidence level calculations are presented. The results showed that the linear static
model and the performance evaluation procedure are very useful and allow the design professional to determine the level of safety of
damaged buildings.
DOI: 10.1061/共ASCE兲0733-9445共2004兲130:4共588兲
CE Database subject headings: Steel frames; Earthquake engineering; Earthquake damage; Performance evaluation; Brittle
fractures; Seismic hazard.

Introduction occupancy 共Foutch 2000兲. The issue of when and where to inspect
for damage is also addressed in other guidelines 共FEMA 2000兲.
One of the important engineering observations of the Northridge This paper follows results of other studies on seismic perfor-
Earthquake of January 17, 1994, was the widespread occurrence mance of new steel moment frame buildings 共Lee and Foutch
of brittle fractures in welded beam–column connections in steel 2002a兲, existing, but undamaged pre-Northridge buildings 共Lee
buildings. Even though no structural collapse was reported after and Foutch 2002b兲 and steel frame buildings with partially re-
the Northridge Earthquake, more than 200 steel moment resisting strained moment connections. Some of the early stages of this
work were summarized by Hamburger et al. 共2000兲. The objec-
frame 共MRF兲 buildings suffered cracks or fractures in the beam–
tive of this paper is to provide simple analytical procedures for
column connections 共Bonowitz and Youssef 1995; Holguin 1998兲.
evaluating the safety of a damaged steel MRF building after an
It was expected that steel buildings would be able to provide
earthquake. These procedures may be used in conjunction with
ductile behavior with limited damage when such buildings were material provided in FEMA 352 共2000兲 to determine if a building
subjected to strong earthquake ground motion. However, the may be occupied while repairs and/or upgrades are being done.
brittle fractures at the welded joints occurred which caused a First, a number of scenarios representing back-to-back earthquake
sudden loss of strength and stiffness and resulted in significant applications will be presented to qualitatively demonstrate the
economic loss due to the damage and service disruption. The poor effects of accelerogram amplitude, year of construction, initial
performance of these building structures raised a genuine concern and postearthquake condition, and stiffness of a building. Next, it
about the level of safety of badly damaged buildings. Professional is demonstrated that the calculated stiffness of a building before
engineers had to make a decision as to whether or not to allow and after the damaging earthquake may be used in conjunction
occupancy of the damaged buildings while the inspection and with field and laboratory observations of behavior of damaged
repair process was made. There are major sociological issues re- beam–column connections to determine the relative safety of a
lated to the balance of safety versus the cost of not allowing building. A more formal procedure based on material in FEMA
352 共2000兲 is then described which allows a designer to estimate
1
Assistant Professor of Architectural Engineering, Sejong Univ., 98 the degree of confidence that the building will satisfy the defined
Gunja-Dong, Gwangjin-Gu, Seoul 143-747, Republic of Korea 共corre- performance objective. The relative safety of the building before
sponding author兲. E-mail: kihaklee@sejong.ac.kr and after the earthquake may be used in this approach. Several
2
Professor of Civil and Environmental Engineering, Univ. of Illinois examples are presented in this paper to demonstrate these proce-
at Urbana–Champaign, 205 N. Mathews Ave., Urbana, IL 61801. E-mail: dures. These and other results 共Lee and Foutch 2002a,b兲 were
d-foutch@uiuc.edu developed as input to the writers of FEMA 352.
Note. Associate Editor: Gregory A. MacRae. Discussion open until
September 1, 2004. Separate discussions must be submitted for individual
papers. To extend the closing date by one month, a written request must
be filed with the ASCE Managing Editor. The manuscript for this paper Performance of Buildings for Multiple Earthquake
was submitted for review and possible publication on August 6, 2002; Occurrences
approved on August 11, 2003. This paper is part of the Journal of Struc-
tural Engineering, Vol. 130, No. 4, April 1, 2004. ©ASCE, ISSN 0733- Several scenarios are used to demonstrate the importance of the
9445/2004/4-588 –599/$18.00. year of design, design procedure, building stiffness, and ampli-

588 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Plan and elevation view of 1994 UBC 9- and 20-story buildings

tude of ground motions considered. A number of 3-, 9-, and 20- about 1%. The top flange was assumed to fracture at a plastic
story buildings were designed in accordance with the 1973, 1985, rotation of 3% which usually occurred at a story drift of about
and 1994 Uniform Building Codes UBC-73, UBC-85, UBC-94兴. 4%. Assuming that the building did not experience a local or
Design details and other considerations are reported elsewhere global collapse, the model building in its damaged state was sub-
共Lee and Foutch 2002b兲. Fig. 1 shows floor plans and elevations jected to the same earthquake accelerogram and the additional
for the 9- and 20-story buildings. Time history and static analyses predicted damage was recorded. Results for the pre-Northridge 9-
of these buildings were run using a modified version of DRAIN- and 20-story buildings designed in accordance with the 1994
2DX 共Prakash et al. 1993; Shi and Foutch 1997兲 program which UBC provisions are shown in detail below.
contained a new connection element that was able to simulate the Fig. 2 shows the damage pattern that the 1994 nine-story
effects of a fracturing connection. The earthquake accelerograms building experienced during the seismic events representing dif-
used for this study were developed by Somerville et al. 共1997兲. ferent hazard levels. At each beam–column joint, a circle having
Three hazard levels were considered, 2, 10, and 50% 共2/50, 10/50, four sections is provided to indicate the location of fractured
and 50/50兲 probability of exceedance in 50 years. Twenty accel- flanges. A darkened segment indicates fracture of a top or bottom
erograms representing each hazard level were derived by Somer- flange on the left- or right-hand side of the joint. Fig. 2 at the top
ville et al. 共1997兲. left-hand side 关Fig. 2共A-1兲兴 shows the damage that the nine-story
After an earthquake, inspection of a steel frame building might building suffered after the first application of the 50/50 ground
reveal that some of the beam columns have been damaged. One motion while the one at the bottom left-hand side 关Fig. 2共A-2兲兴
question that has been asked by several building owners and shows the damage pattern after the second application of the same
structural engineers is, ‘‘will the damaged building be able to earthquake. The same information is illustrated in the middle
withstand an earthquake of the same intensity without collapse?’’ 关Figs. 2共B-1 and B-2兲兴 and right-hand side column figures 关Figs.
In order to help answer this question, a series analyses were com- 2共C-1 and C-2兲兴 for the 1994 nine-story building subjected to the
pleted. For each scenario, one of the model buildings was sub- accelerograms representing the 10/50 共LA 14, peak acceleration
jected to one of the accelerograms described above using a time ⫽0.59 g兲 and 2/50 共LA28, peak acceleration⫽1.1 g兲 hazard lev-
history analysis and the modified DRAIN-2DX program. For each els, respectively.
case, the accelerogram that had a spectral ordinate closest to the After the first and second applications of the 50/50 accelero-
average response spectrum for the particular hazard level was gram, only a small number of flanges experienced fractures. The
used. The maximum interstory drifts were recorded along with the small note above each figure gives the total number and percent-
location of fractured connections. A bottom flange fracture was age of flanges that fractured. A total of 17/180 共9%兲 flanges was
assumed to occur when the moment reached the plastic moment fractured during the first application of the 50/50 motion. After
capacity of the beam which usually occurred at a story drift of the second earthquake with the same hazard level, a total of 23/

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 589

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Observed damage from the first and second ground motion excitation for 1994 Uniform Building Code 9-story building 共back-to-back
earthquake with same hazard level兲

180 共13%兲 was fractured. It should be noted that no global or 3共B-1 and B-2兲 show the damage from the first application of the
local collapses have occurred. For this study, a local collapse is 2/50 accelerogram followed by the second application of the
assumed to have occurred if top and bottom flanges have frac- 10/50 accelerogram. The first 2/50 accelerogram produced 81/180
tured at each end of a beam. 共45%兲 bottom flange fractures and the second 10/50 accelerogram
The middle two figures 关Figs. 2共B-1 and B-2兲兴 show the results caused five top flange fractures in the upper stories. Figs. 3共C-1
for the 10/50 accelerogram. After the first earthquake, a total of and C-2兲 showed the damage from the first 2/50 and the second
63/180 共35%兲 flanges fractured during the first 10/50 accelero- 50/50 or 50/30 accelerogram which is less severe but a more
gram, while a total of 73/180 共41%兲 flanges experienced fractures likely event. The application of the second 50/50 or 50/30 ground
after the second 10/50 accelerogram. No local collapse was ob- motion produced no additional damage. The ratio of the fractured
served during the two applications. flanges is 45% for the first and second applications. It should be
The first application of the 2/50 accelerogram produced frac- noted that only the scenario of back-to-back 2/50 events resulted
tures in all bottom flanges except for the first floor. The second in local or global collapse.
application fractured all top and bottom flanges in the top four Fig. 4 shows the damage from the first 10/50 accelerogram and
floors, and resulted in 16 local collapses and four virtual story the second 50/50 and 50/30 earthquakes. The ratio of the frac-
mechanisms. Occupancy of this building should not be allowed tured flanges increased from 35 to 41% after the back-to-back
after the first 2/50 earthquake. Nevertheless, the performance earthquakes. On the other hand, the ratio is 37% after the second
against the first 2/50 event can be considered acceptable behavior earthquake with 50/50 and 50/30 hazard levels. None of these
since it survived the 2/50 motion. The ultimate goal for the 2/50 scenarios led to local or global collapse.
earthquake is collapse prevention 共CP兲 to protect human life. Fig. 5 shows the results for the 1994 UBC 20-story buildings.
Thus, it really should not be expected to be useable after the Essentially no damage occurred for either application of the 50/50
strong event which has a return period of 2,475 years. accelerogram. The first application of the 10/50 accelerogram
It is also highly unlikely that back-to-back earthquakes of this produced 99/440 共23%兲 bottom flange fractures with no local col-
magnitude will occur within a short period of time. A more real- lapse and the second application fractured only nine more bottom
istic test would be to subject the damaged building to a 50/50 or flanges. The two applications of the 2/50 earthquake fractured
50/30 accelerogram which corresponds to a return period of 72 152/440 共35%兲 bottom flanges but no top flanges. This is very
and 44 years, respectively. In this study, the 50/30 accelerograms good performance since no local or global collapse occurred.
were derived since these ground motions were not available for A summary of the results from all of the back-to-back seismic
the SAC research. These ground motions were developed by mul- analyses is given in Table 1. The pattern discussed in the previous
tiplying a factor to the 50/50 ground motions to match the average part is repeated here. The older the building is, the more likely it
response spectra for the 50/30 hazard level. Fig. 3 shows the is that local or global collapse will occur. No local or global
damage caused by a less severe accelerogram for the second ap- collapse occurred during the first or second application of the
plication. Figs. 3共A-1 and A-2兲 illustrate the damage after back- 50/50 accelerogram. Two three-story buildings designed in accor-
to-back earthquakes representing the 2/50 hazard level. Figs. dance with the seismic design requirements of the 1973 UBC

590 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Observed damage from the first 共2/50兲 and second ground motion excitation for 1994 Uniform Building Code 9-story building 共with less
hazard level for the second application兲

provisions 共1973兲 suffered local collapse during the first applica- building. For the 1985 buildings, the three-story building suffered
tion of the 10/50 earthquake. Global collapse occurred during the collapse for the second application of the 10/50 earthquake and
second earthquake. The 1973 nine-story building designed with- the first occurrence of the 2/50 accelerogram. The other general
out a drift limit suffered local collapse during the second appli- observation that can be made is that the shorter buildings should
cation of the 10/50 record and the first occurrence of the 2/50 be expected to perform worse than the taller buildings. Another
earthquake. Similar results were found for the 1973 20-story important observation is that in only a single case 共1973 20-story

Fig. 4. Observed damage from the first 共10/50兲 and second ground motion excitation for 1994 Uniform Building Code 9-story building 共with less
hazard level for the second application兲

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 591

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 5. Observed damage from the first and second ground motion excitation for 1994 Uniform Building Code 20-story building

for 10/50 accelerogram兲 did a local or global mechanism form bottom flanges fractured at drifts of 1% and top flanges fractured
during the second application of the same accelerogram for a at drifts of 4%. All of the static analyses were done using the
building that had only bottom flange fractures during the first Equivalent Lateral Force 共ELF兲 procedure as described in the
application. This suggests that a damaged building is probably 1997 NEHRP provisions 共FEMA 302 1998兲.
safe during a repair period if only bottom flange fractures are Very few design offices have access to a program that has
observed even if all bottom flanges have fractured. sophisticated connection elements that can model the hysteretic
behavior of the pre-Northridge connection. However, meaningful
analyses can still be undertaken using a linear static model. If the
Use of Stiffness as a Measure of Performance bottom flanges at each end of a beam have fractured, then positive
moment will open the crack resulting in essentially no stiffness,
The next part of the study was conducted to determine if stiffness or pinlike behavior. Negative moment will cause the crack to
is a good indicator of expected performance. A static analysis of close resulting in no loss of flexural stiffness. Since sway defor-
each building was completed using a finite element model of the mations in a building result in one end of a beam being in positive
building and the elastic response spectrum of the given earth- moment and the other end in negative moment, then one end or
quake. The damage pattern for the original building model was the other will always be in negative moment and the other in
observed. Next, a static analysis of the damaged building was positive moment. The effect of this on the elastic drift response
conducted for the same earthquake spectrum. The original model can be modeled by putting a hinge at one end of the beam and a
was changed to reflect the damage indicated by the nonlinear time moment connection at the other end of the beam in the elastic
history 共NLTH兲 analysis. The building model of the damaged model. This was the assumption used for this study. The moments
building had a longer period than the undamaged building so this calculated in the beams will not be correct, but the story drifts
resulted in smaller seismic forces for the second analysis. The will be calculated accurately, and all of the SAC recommenda-
fundamental period of vibration of 9- and 20-story buildings be- tions are keyed to story drift 共Foutch 2000兲.
fore and after each application of the earthquakes is given in Results for the nine-story building are shown in Fig. 6. Fig. 6
Table 2. This change in period also reflects the loss of stiffness of in the upper left-hand side portion 关Fig. 6共A-1兲兴 shows the maxi-
each building after each event. For each case, it was assumed that mum story drifts that resulted from the first and second applica-

592 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Table 1. Results of Sequential Applications of Ground Motions to Pre-Northridge Buildings
50/50 acceleration 10/50 acceleration 2/50 acceleration
Year/building First earthquake Second earthquake First earthquake Second earthquake First earthquake Second earthquake
1973/3 story, 18/36 共50%兲 a
18/36 共50%兲 25/36 共69%兲 32/36 共89%兲 36/36 共100%兲 36/36 共100%兲
without drift limit local collapse story mechanism story mechanism story mechanism
1973/9 story, 34/180 共19%兲 36/180 共20%兲 76/180 共42%兲 91/180 共51%兲 108/180 共60%兲 Global collapse
without drift limit local collapse local collapse
1973/3 story, 18/36 共50%兲 20/36 共56%兲 27/36 共75%兲 32/36 共89%兲 36/36 共100%兲 36/36 共100%兲
with drift limitb local collapse story mechanism story mechanism story mechanism
1973/9 story, 15/180 共8%兲 27/180 共15%兲 60/180 共33%兲 78/180 共43%兲 82/180 共46%兲 122/180 共68%兲
with drift limit story mechanism
1973/20 story 50/440 共11%兲 55/440 共13%兲 121/440 共28%兲 180/440 共41%兲 204/440 共46%兲 240/440 共55%兲
local collapse local collapse story mechanism
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

1985/3 story 18/36 共50%兲 19/36 共53%兲 21/36 共58%兲 32/36 共89%兲 36/36 共100%兲 36/36 共100%兲
story mechanism story mechanism story mechanism
1985/9 story 26/180 共14%兲 31/180 共17%兲 76/180 共42%兲 93/180 共52%兲 96/180 共53%兲 124/180 共69%兲
story mechanism
1985/20 story 43/440 共10%兲 57/440 共13%兲 156/440 共36%兲 174/440 共40%兲 190/440 共43%兲 202/440 共46%兲
1994/3 story 16/36 共44%兲 18/36 共50%兲 24/36 共67%兲 27/36 共75%兲 36/36 共100%兲 36/36 共100%兲
local collapse story mechanism story mechanism
1994/9 story 17/180 共9%兲 23/180 共13%兲 63/180 共35%兲 73/180 共41%兲 81/180 共45%兲 125/180 共69%兲
story mechanism
1994/20 story 3/440 共1%兲 4/440 共1%兲 99/440 共23%兲 108/440 共25%兲 138/440 共41%兲 152/440 共35%兲
a
Number of fractured connection/total number of connections 共ratio兲.
b
Elastic drift limit of 0.0025 was used for seismic design.

tions of the 50/50 earthquake accelerogram using NLTH analyses. lected number of floors. The reason for this also relates to the fact
The damage patterns for each case are shown in Fig. 2. In this that damage occurred over most of the height of the structure
case, no significant damage occurred for the second application of during the first event.
the 50/50 event. Only 9% of the connections experienced bottom With the application of the 2/50 accelerogram, 90% of the
flange fractures for the first analysis. The damage may not even bottom flanges fractured as opposed to about 70% for the 10/50
be found during an inspection which is acceptable for this case. A earthquake. As more bottom flanges occur during an earthquake
small amount of additional damage occurred for the second analy- the greater is the likelihood that several stories approached 4%
sis. The results determined using the ELF analysis and the NLTH drift. Again, this provides a basis for estimating the actual mag-
analysis are in good agreement that a few upper stories had drift nitude of the seismic forces in the first event. Applying these
demands so some bottom flanges fractured. No story drifts ex- forces to the damaged building results in story drifts of 7% indi-
ceeded 4% so no top flange fractures were likely to have oc- cating that global collapse is likely to occur. It is due to the fact
curred. that the top four stories in the elastic model of the damaged build-
The first application of the 10/50 accelerogram resulted in bot- ings experienced a story mechanism causing frame instability.
tom flange fractures in about two thirds of all of the beams, but no Results for the 1994 20-story building are shown in Fig. 7.
top flange fractures. As a result, the story drifts exceeded 1% in Similar observations could be made for this case. At all levels of
almost all of the stories, but not 4%. The static analysis of the earthquake no local or global collapses occurred and none would
undamaged frame resulted in about 2% drift over the entire height have been predicted by the static analyses.
of the building. This is consistent with the time history results. In the actual real-life earthquake scenario, the situations will
The results of the second analysis reveal that if the earthquake not be as nicely laid out as the ones described here. One compli-
occurs again, drifts of two or three stories may exceed 4% by a cating feature will be that some top flange fractures will occur at
small amount. This indicates that local collapses may occur in a story drifts of less than 4% and some bottom flanges will fracture
few beams. One should not expect that the analysis is able to at drifts of less than 1%. Even for tests of laboratory specimens of
identify exactly which floors might experience this local collapse pre-Northridge connections, there was scatter in the results. This
which could be used, for instance, to deny occupancy on a se- will make estimation of the actual drift levels that the building

Table 2. Fundamental Period of Vibration of 1994 Uniform Building Code Nine-Story Building before and after Ground Excitation for 50/50,
10/50, and 2/50 Hazard Levels
From 50/50 Eq. From 10/50 Eq. From 2/50 Eq.
T 共undamaged兲 T 共damaged兲 T 共damaged兲 T 共damaged兲
Building 共s兲 共s兲 共s兲 共s兲
9 story 2.51 2.68 3.58 4.01
20 story 4.24 4.25 5.36 6.24

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 593

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Drift demands from both dynamic and static analyses of damaged 1994 Uniform Building Code 9-story building for 50/50, 10/50, and
2/50 hazard levels

experienced and scaling of the static forces more difficult. As a consider. In other regions, it might be prudent to retain an expe-
result, a range of different scenarios should be investigated. In rienced consultant to provide this information.
real life, it is very unlikely that two earthquakes of the same
magnitude will occur over the amount of time required to repair Performance Prediction Using Confidence Level
damaged buildings. As a result, it might be prudent to examine a Calculation
number of possible occurrences of aftershocks of smaller magni-
tude. Of course, the possibility exists that the first occurrence is a The performance evaluation procedures developed as part of the
preshock of a larger event to come. SAC Phase 2 program permit an estimation of a level of confi-
Analysis of the undamaged building can provide important dence that a structure will be able to satisfy a desired performance
information. If estimates of the ground motion are available, this objective. The randomness and uncertainty associated with pre-
analysis can indicate whether the amount of damage is consistent dicting the capacity and demand are explicitly accounted for in
with the damage. If only a few damaged connections are found the reliability-based probabilistic approach. Two performance ob-
then the maximum drift was probably less than 0.01. If a lot of jectives are defined. The CP performance level is defined as the
bottom flange connections have failed, the maximum drifts were postearthquake damage state in which a building is on the verge
probably between 0.01 and 0.04. If some top flanges have frac- of experiencing local or global collapse. This is paired with the
tured, then drifts may have exceeded 0.04. If the damage does not seismic hazard level of 2% probability of being exceeded in 50
match the calculated results, then the assumed ground motion years 共2/50兲. The immediate occupancy 共IO兲 performance level is
level assumed for the site may be incorrect. a state that a building will experience only limited structural dam-
This paper presumes that the building under consideration has age. Damage is expected to be so light that if not found during
been inspected and damage has been found. The FEMA 352 pre- inspection there would be no cause for alarm. This is paired with
sents methods of determining which joints to inspect. We recom- the seismic hazard level of 50% probability of being exceeded in
mend that the bottom flange of each beam–column joint be in- 50 years 共50/50兲.
spected. If no bottom flange damage is found, no additional A factored-demand-to-capacity ratio, ␭, is defined as
inspection of the joint should be necessary. This will take some ␥•␥ a •D
time and the largest aftershock will likely have occurred by then. ␭⫽ (1)
␾•C
In a real situation, the choice of what earthquake spectrum to use
may be difficult. In regions of the country where many strong where C⫽median capacity of the building for local or global
motion accelerographs have been installed by U.S. Geological collapse; D⫽median demand calculated from the structural analy-
Survey or other agencies, response spectrum values for the vicin- sis for the design earthquake; ␾⫽resistance factor; ␥⫽demand
ity of the building site will probably be available. Also, FEMA factor; ␥ a ⫽analysis uncertainty factor; and ␭⫽factored-demand-
352 contains guidelines on the choice of input ground motions to to-capacity ratio.

594 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

Fig. 7. Drift demands from both dynamic and static analyses of damaged 1994 Uniform Building Code 20-story building for 50/50, 10/50, and
2/50 hazard levels

Both the capacity, C, and demand, D, are defined in terms of unreinforced connections employing weld metal with low notch
interstory drift and column compressive and tensile splice forces. toughness should be considered type 2 connections 共FEMA 352
Theoretical aspects were first developed by Jalayer and Cornell 2000兲. Ductile connections, on the other hand, that are able to
共1998兲 and reported in detail by Cornell et al. 共2002兲. Complete resist 0.04 radians without fracturing or significant strength deg-
details of the development and implementation of the procedure radation, should be considered as type 1 connections.
are reported by FEMA 352 共2000兲, Hamburger et al. 共2000兲, Yun The resistance factor, ␾, accounts for the fact that the estimate
and Foutch 共2000兲, and Yun et al. 共2002兲. of C is affected by randomness and uncertainty in the estimation
The demand factor, ␥, is associated with the variability inher- process. The capacity of the building against global collapse is a
ent in the prediction of demand related to assumptions made in function of the earthquake accelerograms used in the Incremental
structural modeling and prediction of the character of earthquake Dynamic Analysis procedure. This is the ultimate building capac-
accelerograms. The analysis uncertainty factor, ␥ a , accounts for ity against total collapse. On the other hand, the capacity of the
bias and uncertainty inherent in the specific analytical procedure building against local collapse is limited by the local behavior of
used to estimate demand as a function of ground shaking inten- beam–column connections. The local drift capacity is the maxi-
sity. Both ␥ and ␥ a are reproduced from the FEMA 352 report mum drift angle that the connection can sustain without losing its
and given in Tables 3 and 4 for CP and IO performance levels, gravity load carrying ability. Table 5 provides values for global
respectively. All values shown in Tables 3 and 4 are for type 2 drift capacity for regular well-configured structures as well as the
connections. Using nomenclature defined in the FEMA 352 re- resistance factor. Table 6 shows values for local drift capacity for
port, type 2 connections are capable of resisting median total type 2 connections.
interstory drift angle demands of only about 0.01 radians without
fracturing or strength degradation. Typical connections in older
welded steel moment frame buildings constructed using welded
Table 4. Interstory Drift Angle Analysis Uncertainty Factors, ␥ a , for
Type 2 Connections Against Collapse Prevention
Table 3. Interstory Drift Angle Demand Variability Factors, ␥, for Structure type LSP LDP NSP NDP
Type 2 Connections Against Collapse Prevention
Low rise 共⬍4 stories兲 1.03 1.40 1.35 1.06
Building height ␥
Mid rise 共4 –12 stories兲 1.25 1.70 1.46 1.11
Low rise 共⬍4 stories兲 1.7 High rise 共⬎12 stories兲 0.96 1.51 1.71 1.17
Mid rise 共4 –12 stories兲 2.0 Note: See FEMA 共2000兲; LSP⫽linear static procedure; LDP⫽linear dy-
High rise 共⬎12 stories兲 2.6 namic procedure; NSP⫽nonlinear static procedure; and NDP⫽nonlinear
Note: See FEMA 共2000兲. dynamic procedure.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 595

J. Struct. Eng. 2004.130:588-599.


Table 5. Global Interstory Drift Angle Capacity, C, and Resistance Table 7. Values of Uncertainty Coefficient ␤ UT for Global Interstory
Factor, ␾, for Type 2 Connections Against Collapse Prevention Drift Evaluation Against Collapse Prevention
Interstory drift Resistance Building height Collapse prevention
Structure type capacity, C factor, ␾
Low rise 共⬍4 stories兲 0.35
Low rise 共⬍4 stories兲 0.10 0.75 Mid rise 共4 –12 stories兲 0.45
Mid rise 共4 –12 stories兲 0.079 0.60 High rise 共⬎12 stories兲 0.55
High rise 共⬎12 stories兲 0.057 0.60 Note: Value of ␤ UT should be increased by 0.05 for the linear static
Note: See FEMA 共2000兲. procedure analysis and may be reduced by 0.05 for nonlinear dynamic
procedure analysis. FEMA 共2000兲.

The values of the parameter, ␤ UT , are used to account for the


uncertainties inherent in the estimation of demands and capaci-
fidence level that a building before and after the damage will
ties. This factor is included by considering a variety of assump-
satisfy the CP performance level for future events could be cal-
tions, such as assumed properties due to the analysis procedure,
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

culated. The seismic demands would be estimated using the static


values of damping, structural period, and strength of materials.
elastic analyses. The observed damage to the building along with
␤ UT is calculated as
the static analysis results could provide estimates of the magni-
␤ UT⫽ 冑␤ UC
2
⫹␤ UD
2
(2) tude of the first earthquake in the same fashion as described
above. Estimates of the magnitude of shaking at the building site
where ␤ UC⫽standard deviation of the natural log of the story might also be available for this purpose. The estimate of confi-
drifts associated with the capacity; and ␤ UD⫽standard deviation dence could be used along with other data to decide the fate of the
of the natural log of the story drifts associated with the demand. building.
Table 7 provides values of uncertainty coefficient ␤ UT for global The following example for the 1994 UBC nine-story building
and local interstory drift evaluation. subjected to the 50/50 accelerogram demonstrates the procedure
Then, the confidence that the structure will be able to meet the of the confidence level calculations before and after the damage.
given performance objective is backcalculated from the equation With the first application of the 50/50 accelerogram, the damage
␭⫽e ⫺␤ UT共 K x ⫺k␤ UT/2b 兲 (3) pattern was investigated and is shown in Fig. 2. The bottom
flanges in two of the upper stories experienced fractures resulting
where k⫽logarithmic slope of the hazard curve; b⫽a coefficient in 9% of total flanges being fractured. The second application of
that represents the change in demand; and K x ⫽standard Gaussian the 50/50 accelerogram produced a small amount of additional
variate associated with probability x of not being exceeded. Table damage as described above. This is supported by the NLTH re-
8 provides the level of confidence associated with the factored- sults shown in the upper left portion in Fig. 6. Then, elastic mod-
demand-to-capacity ratio ␭ calculated using Eq. 共3兲. Table 8 was els of damaged buildings were developed and corresponding pe-
developed based on k⫽5.0 which is representative of the as- riods were determined. The results of the elastic static analyses
sumed regional seismicity during the year following a major are shown in Fig. 6. Next, confidence levels are calculated.
earthquake 共FEMA 352 2000兲.

Example: Collapse Prevention for the 50Õ50 Hazard


Example of Confidence Level Calculations Level
before and after the Damage • Fundamental period of building (T)⫽2.51 s 共undamaged兲,
• Seismic base shear (V)⫽1,539 kips 共ELF procedure; 1997
Confidence level calculations were used to predict performance of NEHRP provisions兲,
a nine-story building subjected to the 50/50 ground motion. The • Maximum elastic drift angle (␦ xe )⫽0.013,
nine-story building considered here was designed in accordance • Median drift demand (D)⫽␦ xe ⫻C d ⫽0.013⫻1.0⫽0.013 (C d
with the 1994 UBC seismic provisions and assumed to have ⫽1.0 for 50/50 hazard level; 1997 NEHRP provisions兲,
brittle connections. The first step would be to examine the build- • Fundamental period of building (T)⫽2.68 s 共damaged; after
ing for damage status and gather information with respect to de- the first earthquake兲,
sign and construction properties. Then, a computer model of the • Seismic base shear (V)⫽1,403 kips,
building could be developed for the undamaged status. The con- • Maximum elastic drift angle (␦ xe )⫽0.019, and
• Median drift demand (D)⫽␦ xe ⫻C d ⫽0.019⫻1.0⫽0.019.
1. Global collapse
Table 6. Local Interstory Drift Angle Capacity and Resistance • Demand variability factor 共␥兲⫽2.0: given in Table 3,
Factors • Analysis uncertainty factor (␥ a )⫽1.25: Given in Table 4
Interstory drift Resistance 关linear static procedure 共LSP兲兴,
Connection type capacity, C factor, ␾ • Median drift capacity (C)⫽0.079: given in Table 5 共type 2
connections兲,
Pre-Northridge connection with low 0.053– 0.0006d b 0.7 • Resistance factor 共␾兲⫽0.60: given in Table 5 共type 2 con-
notch toughness weld metal nections兲,
Pre-Northridge connection with notch 0.060– 0.0006d b 0.85 • Confidence factor, ␭⫽␥␥ a D/␾C: Eq. 共1兲
tough weld metal
Shear tab connections 0.16– 0.0036d b 0.7 2.0⫻1.25⫻0.019
Post-Northridge connection intended 0.04 0.85
␭⫽ ⫽1.00
0.60⫻0.079
forsteel moment-frame service • Slope of hazard curve (k)⫽5.0,
Note: See FEMA 共2000兲. • Uncertainties (␤ UT)⫽0.45⫹0.05⫽0.50: given in Table 7

596 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Table 8. Confidence Levels for Various Values of ␭, Given ␤ UT
Uncertainty
parameter
␤ UT Factored-demand-to-capacity ratios, ␭
0.2 1.43 1.31 1.23 1.16 1.11 1.05 0.99 0.93 0.86 0.79 0.70
0.3 1.84 1.62 1.47 1.35 1.25 1.16 1.07 0.97 0.85 0.76 0.63
0.4 2.49 2.10 1.84 1.65 1.49 1.35 1.21 1.06 0.89 0.77 0.59
0.5 3.54 2.86 2.44 2.12 1.87 1.65 1.43 1.22 0.99 0.82 0.59
0.6 5.30 4.10 3.38 2.86 2.46 2.12 1.79 1.48 1.14 0.91 0.62
Confidence 10 20 30 40 50 60 70 80 90 95 99
level 共%兲
Note: See FEMA 共2000兲.
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

共LSP兲, and produced the maximum drift demand of 0.074 and confidence
• Confidence level⫽90%: Determined using Table 8 共after level of 7%. These hazard levels correspond to a return period of
the 50/50 back-to-back earthquake兲. 72 and 44 years, respectively. For the 1994 UBC 3-, 9-, and
2. Local collapse 20-story buildings, the confidence levels are calculated to be 50,
• Demand factor 共␥兲⫽2.0: Given in Table 3, 46, and 66%, respectively. After the second applications with the
• Analysis demand factor (␥ a )⫽1.25: Given in Table 4, 50/50 and 50/30 hazard levels, they are almost identical. Results
• Median drift capacity (C)⫽0.042: given in Table 6 共pre- for the first application of the 10/50 earthquake and the second
Northridge connections with notch tough weld material, with less severe hazard level are shown in Table 10. The lowest
assume d b ⫽30), confidence levels were calculated for the 1973 and 1985 UBC
• Resistance factor 共␾兲⫽0.85: Given in Table 6 共pre- 20-story buildings.
Northridge connections兲, Table 11 gives the confidence level for the 50/50 hazard level
• Confidence factor, ␭⫽␥␥ a D/␾C: Eq. 共1兲 for the first earthquake. All 3- and 9-story buildings provide high
confidence levels after the second earthquake for achieving the
2.0⫻1.25⫻0.019
␭⫽ ⫽1.33 collapse prevention objective. The 1973 20-story building pro-
0.85⫻0.042 vides the lowest confidence levels of 19 and 11%. All buildings
• Slope of hazard curve (k)⫽5.0, which were designed based on the 1994 UBC provisions provide
• Uncertainties (␤ UT)⫽0.45⫹0.05⫽0.50: Given in Table 7, high confidence levels ranging from 90 to 99%.
and Recommended minimum confidence level for a new building
• Confidence level⫽74%: Determined using Table 8 共after to satisfy the collapse prevention performance level should be
the 50/50 back-to-back earthquake兲. about 90% for the 2/50 hazard level which has a return period of
After the first application of the 50/50 accelerogram, a total of 2,475 years. However, it is a very difficult task if an engineer tries
17/180 flanges were fractured during the first 50/50 ground mo- to bring an existing building, which has a finite 共perhaps small兲
tion. The maximum story drift from the ELF procedure defined in life expectancy, up to this performance level. This can be seen in
the 1997 NEHRP provisions was observed to be 0.013. After the
second 50/50 ground motion, the maximum story drift was esti-
mated to be 0.019. To address the damaged state, the elastic build-
Table 9. Confidence Level Calculations Against Global Collapse
ing model that employs its damaged state and the reduced seismic
Level 共2/50 Earthquake for the First Application兲
base shear were considered. The loss of stiffness after the first
earthquake resulted in a longer period from 2.51 to 2.68 s. After First earthquake Second earthquake
the second 50/50 seismic event, the confidence levels that the Year/building H. L. Drift C. L. H. L. Drift C. L.
damaged nine-story building will meet the global and local col- a
1973/3 story, 2/50 0.072 9% 50/50共50/30兲 0.074 7%
lapse performance levels are 90 and 74%, respectively. The lower
without drift limit
confidence level for local collapse is mainly due to the fact that
1973/9 story, 2/50 0.045 10% 50/50共50/30兲 0.047 9%
the median drift capacity was set lower to account for brittle
without drift limit 0.047 9%
connection behavior.
Tables 9–11 show the confidence level calculations against 1973/3 story, 2/50 0.060 23% 50/50共50/30兲 0.061 20%
global collapse for all buildings designed in accordance with with drift limit
UBC provisions from different years. As shown in Table 5, the 1973/9 story, 2/50 0.027 49% 50/50共50/30兲 0.029 45%
interstory drift capacity, C, against CP performance level was set with drift limit
to be 0.10, 0.079, and 0.057 for the 3-, 9-, and 20-story buildings, 1973/20 story 2/50 0.024 22% 50/50共50/30兲 0.024 21%
respectively. The drift demands for all buildings shown here were 1985/3 story 2/50 0.063 17% 50/50共50/30兲 0.063 17%
obtained from the NLTH analyses using brittle connection mod- 1985/9 story 2/50 0.035 27% 50/50共50/30兲 0.035 26%
els. 1985/20 story 2/50 0.022 29% 50/50共50/30兲 0.022 28%
Table 9 shows that the 1973 three-story building 共without drift 1994/3 story 2/50 0.048 50% 50/50共50/30兲 0.048 50%
limit兲 experienced the maximum drift demand of 0.072 during the 1994/9 story 2/50 0.028 46% 50/50共50/30兲 0.028 46%
first application of the 2/50 earthquake. The confidence level that 1994/20 story 2/50 0.013 66% 50/50共50/30兲 0.014 60%
the building will not collapse from this damage state was calcu- Note: H. L.⫽hazard level; Drift⫽maximum interstory drift; C. L.
lated to be 9%. For the second applications, both 50/50 and 50/30 ⫽confidence level against global collapse.
a
ground motions, which are less severe but a more likely event, Both 50/50 and 50/30 ground motions produced the same drift demands.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 597

J. Struct. Eng. 2004.130:588-599.


Table 10. Confidence Level Calculations Against Global Collapse Table 11. Confidence Level Calculations Against Global Collapse
Level 共10/50 Earthquake for the First Application兲 Level 共50/50 Earthquake for the First Application兲
First earthquake Second earthquake First earthquake Second earthquake
a b c a b c
Year/building H. L. Drift C. L. H. L. Drift C. L. Year/building H. L. Drift C. L. H. L. Drift C. L.
d
1973/3 story, 10/50 0.036 83% 50/50共50/30兲 0.036 83% 1973/3 story, 50/50 0.023 99% 50/30 0.025 99%
without drift limit without drift limit 50/50 0.028 96%
1973/9 story, 10/50 0.028 47% 50/50共50/30兲 0.030 42% 1973/9 story, 50/50 0.018 85% 50/30 0.018 84%
without drift limit without drift limit 50/50 0.020 80%
1973/3 story, 10/50 0.044 62% 50/50共50/30兲 0.044 62% 1973/3 story, 50/50 0.025 98% 50/30 0.027 97%
with drift limit with drift limit 50/50 0.031 92%
1973/9 story, 10/50 0.018 83% 50/50共50/30兲 0.019 83% 1973/9 story, 50/50 0.014 95% 50/30 0.015 94%
with drift limit with drift limit 50/50 0.017 89%
1973/20 story 10/50 0.035 6% 50/50共50/30兲 0.036 6% 1973/20 story 50/50 0.023 24% 50/30 0.025 19%
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

1985/3 story 10/50 0.031 92% 50/50共50/30兲 0.031 92% 50/50 0.030 11%
1985/9 story 10/50 0.024 63% 50/50共50/30兲 0.024 62% 1985/3 story 50/50 0.025 98% 50/30 0.027 97%
1985/20 story 10/50 0.020 35% 50/50共50/30兲 0.020 34% 50/50 0.032 91%
1994/3 story 10/50 0.038 78% 50/50共50/30兲 0.038 77% 1985/9 story 50/50 0.016 92% 50/30 0.016 90%
1994/9 story 10/50 0.019 81% 50/50共50/30兲 0.019 80% 50/50 0.021 75%
1994/20 story 10/50 0.014 62% 50/50共50/30兲 0.017 45% 1985/20 story 50/50 0.016 54% 50/30 0.016 50%
a
Hazard level. 50/50 0.017 45%
b 1994/3 story 50/50 0.014 99% 50/30 0.017 99%
Maximum interstory drift.
c 50/50 0.027 97%
Confidence level against global collapse.
d
Both 50/50 and 50/30 ground motions produced the same drift demands. 1994/9 story 50/50 0.014 96% 50/30 0.014 95%
50/50 0.016 90%
1994/20 story 50/50 0.006 98% 50/30 0.006 98%
50/50 0.006 98%
Table 9 which indicates that no undamaged building designed in a
1994 or earlier satisfies this goal. The confidence level for these Hazard level.
b
undamaged buildings ranges from 9 to 66%. It is recommended in Maximum interstory drift.
c
this study that the minimum acceptable confidence level for a Confidence level against global collapse.
damaged building should be about 90% for global collapse
against an earthquake with a 44 year return period 共50/30 hazard
remain occupied. Almost all of the damage scenarios investigated
level兲. For local collapse of the beams, a 50% confidence level
in this study satisfy this condition.
is recommended to determine the relative safety of a damaged
A note of caution should be given. The purpose of the paper is
building and whether a building should remain occupied while it
to describe a tool that can be used by design professionals to
is repaired and/or upgraded.
evaluate the safety of a damaged building. General conclusions
about the safety of buildings, in general, should not be made
based on the results presented herein. Different accelerograms
Conclusions and/or different building configurations might produce different
results. The evaluation procedure, however, would be the same.
This study focused on the problem of evaluating the safety of Also, it would be prudent to use several accelerograms when
steel MRF buildings that have experienced connection fractures evaluating an actual building where safety of the occupants is the
in an earthquake. The main purpose of the study was to provide chief concern.
design professionals with analytical tools that can be used in de-
termining whether or not a damaged building may remain occu-
pied prior to repair. Acknowledgments
Several scenarios were presented which studied the effects of
back-to-back earthquakes on steel MRF buildings designed under This research was sponsored by the SAC Joint Venture with funds
past seismic design codes. The results indicated that, in most provided by the Federal Emergency Management Agency. This
cases, the second application of an earthquake of equal or lower support is greatly acknowledged. Opinions and advice provided
intensity caused only a small number of additional flange frac- by the Technical Advisory Panel were also very helpful. Any
tures. Only the oldest buildings subjected to the earthquake rep- findings, results or conclusions are solely those of the writers and
resenting the 2/50 hazard level were found to be unsafe for occu- do not represent those of the sponsors.
pation. A linear static analysis approach was investigated and
shown to be an effective tool for evaluating the safety of a dam-
aged building. A performance-based approach developed for References
FEMA 352 was also presented. The confidence level that a build-
ing will satisfy a given performance objective is used to deter- Bonowitz, D., and Youssef, N. 共1995兲. ‘‘SAC survey of steel moment-
mine the safety of a damaged building. It is recommended that a resisting frame buildings affected by the 1994 Northridge earth-
confidence level of 90% the building will satisfy the local and quake.’’ SAC 95-06, SAC Joint Venture, Richmond, Calif.
global collapse conditions for the 50/30 共44 year return period兲 Cornell, C. A., Jalayer, F., Hamburger, R. O., and Foutch, D. A. 共2002兲.
hazard level be used for determining if a damaged building may ‘‘Probabilistic basis for 2000 SAC Federal Emergency Management

598 / JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004

J. Struct. Eng. 2004.130:588-599.


Agency steel moment frame guidelines.’’ J. Struct. Eng., 128共4兲, 526 – ground Document, SAC Joint Venture, Richmond, Calif.
533. Lee, K., and Foutch, D. A. 共2002a兲. ‘‘Performance evaluation of new steel
Federal Emergency Management Agency 共FEMA兲. 共1998兲. ‘‘NEHRP rec- frame buildings for seismic loads.’’ Earthquake Eng. Struct. Dyn.,
ommended provisions for seismic regulations for new buildings and 31共3兲, 653– 670.
other structures.’’ FEMA 302, 1997 Ed., FEMA, Washington, D.C. Lee, K., and Foutch, D. A. 共2002b兲. ‘‘Seismic performance evaluation of
Federal Emergency Management Agency 共FEMA兲. 共2000兲. ‘‘Recom- pre-Northridge steel frame buildings with brittle connections.’’ J.
mended postearthquake evaluation and repair criteria for welded steel Struct. Eng., 128共4兲, 546 –555.
moment-frame buildings.’’ FEMA 352, FEMA, Washington, DC. Prakash, V., Powell, G. H., and Campbell, S. 共1993兲. DRAIN-2DX, Ele-
Foutch, D. A. 共2000兲. ‘‘State of art report on performance prediction and ment description and user guide, University of California, Berkeley,
evaluation of moment-resisting steel frame structures.’’ SAC Rep. No.
Calif.
FEMA 355f, FEMA, Washington, D.C.
Shi, S., and Foutch, D. A. 共1997兲. ‘‘Connection element 共type 10兲 for
Hamburger, R. O., Foutch, D. A., and Cornell, C. A. 共2000兲. ‘‘Perfor-
DRAIN-2DX.’’ Civil Eng. Rep., Univ. of Illinois at Urbana–
mance basis of guidelines for evaluation, upgrade, and design of
Champaign, Urbana, Ill.
moment-resisting steel frames.’’ Proc., 12th World Conf. on Earth-
quake Engineering, New Zealand Society for Earthquake Engineer- Somerville, P., Smith, N, Puntamurthula, S., and Sun, J. 共1997兲. ‘‘Devel-
opment of ground motion time histories for phase 2 of the FEMA/
Downloaded from ascelibrary.org by Ryerson University on 07/29/13. Copyright ASCE. For personal use only; all rights reserved.

ing, Upper Hutt, New Zealand. Paper Ref. No. 2543.


Holguin, R. 共1998兲. ‘‘Building department’s response to earthquake dam- SAC steel project.’’ SAC Background Document No. SAC/BD-97/04,
age to steel frame buildings.’’ J. Perform. Constr. Facil., 12共4兲, 199– SAC Joint Venture, Richmond, Calif.
201. Yun, S.-Y., and Foutch, D. A. 共2000兲. ‘‘Performance prediction and evalu-
International Conference of Building Officials 共ICBO兲. 共1973, 1985, ation of low and intermediate moment frame buildings for seismic
1994兲. Uniform building code, structural engineering design provi- loads.’’ SAC Background Document No. SAC/BD-00/26, SAC Joint
sions, Vol. 2, International Conference of Building Officials, Whittier, Venture, Richmond, Calif.
Calif. Yun, S.-Y., Hamburger, R. O., Cornell, A. C., and Foutch, D. A. 共2002兲.
Jalayer, F., and Cornell, C. A. 共1998兲. ‘‘Development of a probability- ‘‘Seismic performance evaluation for steel moment frames.’’ J. Struct.
based demand and capacity factor design seismic format.’’ SAC Back- Eng., 128共4兲, 534 –545.

JOURNAL OF STRUCTURAL ENGINEERING © ASCE / APRIL 2004 / 599

J. Struct. Eng. 2004.130:588-599.

Vous aimerez peut-être aussi