Вы находитесь на странице: 1из 18

SEISMIC ANALYSIS M E T H O D S FOR IRREGULAR

BUILDINGS

By Jack P. Moehle, 1 A. M. ASCE and Luis F. Alarcon 2

ABSTRACT: A combined experimental and analytical study is made of the re-


sponse to strong base motions of reinforced concrete structures having irregular
vertical configurations. In the study, two frame-wall structures are constructed
at small scale and subjected to earthquake simulations on a shaking table. Mea-
sured responses of the structures are compared with responses computed by
several conventional analysis methods. The methods include inelastic dynamic
response history analysis, inelastic static analysis, elastic modal spectral anal-
ysis, and elastic static analysis. Based on the data presented, it is concluded
that the main advantage of the dynamic methods was that they were capable
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

of estimating maximum displacement responses, whereas the static methods


cannot be used for this purpose. In all other regards, the dynamic methods
offered no clear advantage over the corresponding static method. The inelastic
static and dynamic methods were superior to the elastic methods in interpret-
ing effects of the structural discontinuities.

INTRODUCTION

The selection of appropriate seismic analysis methods for buildings is


an important consideration in the design of safer structures. For build-
ings having "regular" configuration, current building codes (10,12) de-
scribe concise analysis methods that have been tested by research and
field experience. The recommendations for buildings having "irregular"
configuration are less clear and have not been tested extensively. A com-
m o n requirement (10,12) in seismic design of irregular buildings is that
design forces be determined considering the dynamic characteristics of
the building. In m a n y instances this is achieved in practice by using a
linear elastic modal spectral analysis of the building. It is not altogether
certain whether such a n analysis will offer advantages over the linear
elastic static analysis permitted for more regular buildings. In particular,
elastic analysis methods may not be appropriate for design of m a n y ir-
regular buildings, because the presence of an irregularity may result in
localized zones of inelastic response that cannot be represented ade-
quately by conventional elastic analysis. For such buildings, static or dy-
namic inelastic analyses may be more appropriate.
To examine the different analysis methods that might be used in the
design of irregular buildings, a combined experimental a n d analytical
study was undertaken. Two small-scale, nine-story, reinforced concrete
test structures were selected for analysis. The test structures represent
buildings having "regular" a n d "irregular" distributions of stiffness a n d
strength in the vertical plane. Measured responses to earthquake sim-
ulations are compared with responses calculated using four analysis
methods. The four methods are: (1) Inelastic dynamic analysis; (2) in-
x
Asst. Prof., Dept. of Civ. Engrg., Univ. of California, Berkeley, CA.
2
Grad. Student, Dept. of Civ. Engrg., Univ. of California, Berkeley, CA.
Note.—Discussion open until June 1, 1986. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manu-
script for this paper was submitted for review and possible publication on March
6, 1985. This paper is part of the Journal of Structural Engineering, Vol. 112, No.
1, January, 1986. ©ASCE, ISSN 0733-9445/86/0001-0035/$01.00. Paper No. 20289.

35

J. Struct. Eng. 1986.112:35-52.


elastic static analysis; (3) elastic modal spectral analysis; and (4) elastic
static analysis. Thus, the analyses permit comparison between dynamic
and static analyses, using both elastic and inelastic response assump-
tions. The objective of the study is to evaluate the capabilities of the
different analysis methods to anticipate observed behavior of relatively
uniform and relatively nonuniform structures subjected to strong earth-
quake motions.

DESCRIPTION OF TEST STRUCTURES

The responses of two small-scale, multistory, reinforced concrete test


structures are analyzed in this paper. Each structure comprised a com-
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

bination of planar frames and walls that were alined to resist unidirec-
tional lateral loads. One of the test structures, designated FFW, included
two nine-story, three-bay frames and a nine-story, prismatic wall (Fig.
1). The other structure, designated FSW, was nominally identical to
structure FFW except that the wall extended only to the first "floor"
level. Because of the continuity provided by the full height wall, struc-
ture FFW can be considered to represent a building having a reasonably
uniform distribution of stiffness and strength over the height. In con-
trast, structure FSW, with the wall cutoff at the first floor, can be con-
sidered to represent a building having a nonuniform distribution of stiff-
ness and strength over the height.

(ALL DIMENSIONS
IN mm)
rA BASE
W MOTION
STRUCTURAL
WALL

:• STRUCTURAL
FRAME

;LATERAL
DIAPHRAGMS
: 4 7 0 kg
MASSES

-FIXED BASE

SECTION A-A

FIG. 1.—-Experimental Setup; Test-Structure Dimensions

36

J. Struct. Eng. 1986.112:35-52.


Concrete in the test structures had mean compressive strength of 37
MPa (5.2 ksi), and secant modulus of 19,000 MPa (2,700 ksi). Frame and
wall longitudinal reinforcement had yield stresses of 399 and 339 MPa
(57 and 49 ksi), respectively. Longitudinal reinforcement was continuous
and uniform through the structures, having ratios of approximately 1%
for beams, columns, and walls. Details ensured that primary inelastic
action of members would be flexural. Further information on details and
component behavior is given elsewhere (1,4-7).
In the experimental setup, the two frames of a structure were situated
parallel and opposite one another with the structural wall located cen-
trally between the frames (Fig. 1). A nonstructural concrete and steel
mass was carried by the frames at the center line of each floor level to
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

increase inertial and gravity effects. A system of connections supported


these weights and acted as an effectively rigid diaphragm that linked
frames and wall together at floor levels so that lateral drifts were re-
strained to be equal, but relative rotations and vertical motions between
walls and frames were not restrained. Stiff foundation girders provided
a nearly rigid foundation that was bolted to the Univ. of Illinois shaking
table.
The planar character of the structures was maintained by subjecting
them to unidirectional horizontal base motions parallel to the plane of
the frames (Fig. 1). The base acceleration was modeled after the NS com-
ponent of the record obtained at El Centro during the 1940 Imperial Val-
ley Earthquake. To obtain realistic ratios of test structure and base mo-
tion frequencies, the time scale was compressed by a factor of 2.5. The
peak acceleration was amplified to approximately 0.4 g to achieve a de-
sired level of inelastic response during earthquake simulations. Contin-
uous records were obtained of base accelerations, floor accelerations, floor
displacements, story shears, and interaction forces between walls and
frames.
Additional details of the experiments are given elsewhere (5-7).
It is noted that the test structures are simplified representations of lat-
eral load resisting systems of reinforced concrete buildings, having con-
figurations and loadings not dissimilar to the analytical models often
used in structural design. Real buildings are generally more complex
than the simple frames analyzed in this paper.

DESCRIPTION OF ANALYTICAL MODELS

The analytical models used for the different analyses had several com-
mon characteristics. Some variations were required for the different
analysis methods. These are described in the following subsections.
General Considerations.—All analyses were based on an analytical
model having the configuration indicated in Fig. 2. The model com-
prised a frame, a wall, and lumped masses constrained to have the same
lateral deflections at floor levels. The frame had properties representing
those of the two frames in a test structure. The masses possessed lateral
inertias only that accounted for the measured nonstructural weights used
in a test structure (Fig. 1) plus tributary portions of frames and walls.
Shear and flexural deformations were considered for all members, and
axial deformations were permitted for columns. Rigid beam-column joints
37

J. Struct. Eng. 1986.112:35-52.


• • •H3"
Typical
Column
Depth = 5l -
• •• ^ 3 8 x 203
Y

• ••
Wall of
Variable
Typical Height

• ••
Beam
Depth = 38JTZ

Rigid
Joint
Cores -
HLZO 3—<r~\>—c

Axially

• ••
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

Rigid
Floor

• ••
Levels -

7777777 "7777777 ~7fr7s77

(All Dimensions In Millimeters Unless Noted)

FIG. 2 — Analytical Frame-Wall Model Used for All Analyses

were assumed (Fig. 2), although effects of beam reinforcement slip from
the joints were considered for some of the inelastic analyses. Base fixity
of frame and wall was assumed, except wall base rotation was permitted
in some of the inelastic analyses to account for reinforcement slip from
the foundation. Effects of gravity loads on column stiffness and strength
were considered, but effects of gravity loads acting through lateral dis-
placements (P-delta effect) were not considered. Experimental results re-
ported elsewhere (7) indicate that P-delta moments were not significant.
Special Modeling Considerations for the Inelastic Dynamic Anal-
yses.—Inelastic dynamic analyses were carried out using the computer
program DRAIN-2D (2). The program represents beams, columns, and
walls by an elastic line element connected to nodes by elasto-plastic springs
at the member ends. Responses to the measured base motions were cal-
culated using a time step of 0.005 sec. Viscous damping equal to 2% of
critical (using initial stiffness proportional damping) was used for all but
one of the analyses, as will be described subsequently.
Moment-curvature responses of beams, columns and walls were de-
termined assuming plane sections remain plane, and considering effects
of reinforcement strain hardening, concrete confinement, and concrete
spalling as discussed in Ref. 3. Moment-rotation behaviors of beams and
columns were then calculated using the moment-curvature relations and
assuming equal end moments that flexed the member in double cur-
vature. The resulting moment-rotation relations were subsequently
idealized as being bilinear. This approach to modeling of beams and col-
umns is reasonable for these structures because the influence of gravity
38

J. Struct. Eng. 1986.112:35-52.


STEEL STRESS

d-d' ^AVERAGE BEAM •f !


BOND STRESS,
U = 2MPa
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

JOINT
—V-
FIG. 3.—Analytical Model for Beam Reinforcement Slip

loads on moments is negligible, such that inflection points are near


member midlengths. For structure FFW, the moment-rotation relation of
the wall in the first story was calculated using the same general pro-
cedure, except the moment variation was assumed to vary from a max-
imum at the base to half that value at the first floor level. Approximately
the same variation was measured at the time of peak response during
the experiments (7). The upper stories of the wall remained in the elastic
range of response. For FSW, the wall moment variation was linear to a
value of zero at the first-floor level.
For some of the analyses, a simple model was used to approximate
effects of beam reinforcement slip from beam-column joints. According
to the model (Fig. 3), slip of the reinforcement was calculated assuming
an average uniform bond stress. The associated fixed-end rotation due
to bar slip was subsequently added to the moment-rotation envelope
relations for beams (7). Similarly, effects of wall reinforcement slip from
the foundation were included in some of the analyses by adding a con-
centrated spring at the base of the wall. Given the difficulty of accurately
computing slip at this location, and its importance relative to overall re-
sponse, the slip spring stiffness was taken equal to the stiffness mea-
sured during component tests (7).
Hysteretic response of all elements followed rules of the Takeda model
(11). For columns, it is noted that the model does not consider effects
of changing axial load on stiffness and strength. This is considered rea-
sonable for the test structures because increases in stiffness and strength
in columns on one side of the structure due to overturning are com-
pensated by nearly equal decreases in stiffness and strength in columns
on the other side of the structure for the range of axial loads experi-
enced.
Special Modeling Considerations for Inelastic Static Analyses.—To
maintain consistency in modeling, inelastic static analyses were carried
out using the same computer program (DRAIN-2D) used for the inelastic
dynamic analyses. Static analysis was effected by introduction of ficti-
tious members and masses. The analytical model is depicted in Fig. 4.

39

J. Struct. Eng. 1986.112:35-52.


-SPRINGS
10 K

9K
8K

7K
6K

5K
4K

3K
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

2K
K'

/J/W//^///////J//////;A)/////////////A';///)////A';;///l'//
RIGID STRUCTURE STRUCTURE
(Mass Included) ( Mass Removed)

FIG. 4.—Model Used for Static Inelastic Analyses

As shown, all mass is removed from the original structure, and a rigid
element with large mass is added. The rigid element is attached to each
floor level by springs having stiffnesses arranged according to the de-
sired static lateral load distribution. The system is excited by a rectan-
gular acceleration pulse that slowly moves the rigid element away from
the structure. In so doing, the springs deform and develop forces that
are applied monotonically to the structure. The properties of the system
were arranged to achieve a distribution of lateral loads varying linearly
from zero at the base to a maximum at the top.
In all other respects, the analytical modeling of the test structures was
identical to that used for Analysis A of the inelastic dynamic analyses.
Effects of reinforcement slip were not considered.
Special Modeling Considerations for Elastic Dynamic Analyses.—
Linear elastic modal spectral analyses were carried out using the com-
puter program SAP 80 (13). Elastic stiffnesses were taken as the fully
cracked values, including the effect of axial load on column stiffness. It
is noted that a gross-section model is used more typically in many de-
sign offices. Results obtained using the gross-section model would be
different from those presented in this paper. However, previous studies
(6) have shown that fully-cracked stiffnesses are more appropriate for
interpreting responses of these particular test structures during the de-
sign intensity earthquake simulations. Effects of reinforcement slip were
not considered.
Modal spectral responses were determined for a smoothed linear elas-
tic acceleration response spectrum (Fig. 5). The smoothed spectrum
compares closely with spectra of the measured motions for frequencies
in the range of the first mode (Fig. 5), but exceeds the measured spectra
for frequencies in the range of the higher modes (above 5 Hz). Thus, it
can be expected that modal spectral analyses will overestimate contri-
40

J. Struct. Eng. 1986.112:35-52.


APPROXIMATE RANGE OF
FIRST-MODE PERIOD DURING
TEST FOR FFW AND FSW.

Idealized
—-—FFW
——-•FSW
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

40 30 20
10 5
0.2 0.4 0.6 0.8 1.0
FREQUENCY, Hz PERIOD, sec.

FIG. 5.—Comparison between Measured and Idealized Elastic Response Spectra

butions of the higher modes to overall response.


Viscous damping equal to 5% of critical was assumed for all modes.
The responses for the first five vibration modes were combined accord-
ing to the CQC method (14) to obtain estimates of response maxima.
Special Modeling Considerations for Elastic Static Analyses.—To
maintain consistency in modeling, elastic static analyses were carried out
using the same computer program (SAP 80) and same modeling as-
sumptions used for the elastic modal spectral analyses. Lateral floor-level
loads for the static analyses varied linearly from zero at the base to a
maximum at the top level.

OBSERVATIONS FROM INELASTIC DYNAMIC RESPONSE ANALYSES

Several inelastic response history analyses were conducted for each


test structure. For each analysis, different modeling assumptions were
tried in an effort to establish a "best fit" model. Results of several of the
analyses are discussed in this section.
The first analysis, designated "Analysis A," was based on computed
member moment-rotation behaviors without including effects of rein-
forcement slip. The resulting computed top-level displacement wave-
forms for the first five seconds of response are compared with measured
responses in Fig. 6(a). For structure FFW, 5% of critical damping was
assumed, which is in the range suggested by Newmark and Hall (8) for
response near the yield level. The comparison for FFW indicates that the
calculated oscillation period and amplitudes were less than measured
values. Based on the observed lower amplitudes, the damping ratio was
changed from five to two percent of critical for Analysis A of structure
FSW [Fig. 6(a)]. For both structures, the general shape of the calculated
oscillation was different from measured, indicating a shortcoming in the
analytical model.
To improve the response "predictions," a second analysis (designated
41

J. Struct. Eng. 1986.112:35-52.


STRUCTURE FFW STRUCTURE FSW MEASURED
CALCULATED

J^A^J<) (a) ANALYSIS A

$f (b) ANALYSIS B
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

(c) ANALYSIS C

1 2 3 4 5 2 3 4
T I M E , sec. TIME, sec.

FIG. 6.—Comparison between Measured and Computed Top Displacement Re-


sponse Histories from Inelastic Dynamic Analyses

"Analysis B") was conducted in which slip of reinforcement at the base


of walls and slip of beam reinforcement from beam-column joints were
included. In the analytical model, the beam slip stiffness was based on
the fixed-end rotation in Fig. 3 calculated for the reinforcement stressed
to yield. Wall slip stiffness was obtained from experimental data (7). Two
percent damping was assumed. Computed periods for both structures
are slightly longer than measured periods [Fig. 6(b)], suggesting flexi-
bilities of elements have been overestimated for this analysis. The shapes
of the calculated oscillations bear resemblance to the measured oscilla-
tions, but do not match responses correctly during times of peak re-
sponse and afterwards.
To arrive at improved response estimates, a third analysis was con-

STRUCTURE FFW STRUCTURE FSW MEASURED


- — CALCULATED

(a) BASE MOMENTS

20 20

^V^*y^ 0 ^ W * (b) BASE SHEARS


-20

2 3 4 0 1 2 3 4
TIME, sec. TIME, sec.

FIG. 7.—Comparison between Measured and Calculated Base Shear and Base
Moment Response Histories from inelastic Dynamic Analyses

42

J. Struct. Eng. 1986.112:35-52.


MEASURED •
LEVEL STRUCTURE
9T WALL
&! 0
a- ^ CALCULATED •
^ & o lo l STRUCTURE 0
/

b-

1 WALL a
1 ol
b- ¥ 0[
4- *1 o)
6- X ol
•i.- * 'i o|
1 •
. 0

i I I 1 1 1 1 1
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

0 5 10 15 kN 0 5 10 15 kN
(a) FFW (b) FSW

FIG. 8.—Comparison between Measured and Calculated Story Shear Envelopes


from Inelastic Dynamic Analyses

ducted for each test structure (designated "Analysis C"). For this anal-
ysis, beam fixed-end rotations due to slip were reduced by computing
the fixed-end rotational stiffness for bar stress levels equal to approxi-
mately half the yield stress. Although rationalizations for this modifi-
cation can be made, it is more accurate to state that the stiffness selection
is arbitrary, having been made solely to coax an incomplete analytical
model into mimicing observed behavior.
Top-level displacement responses computed by Analysis C are com-
pared with measured responses in Fig. 6(c). For both structures, the gen-
eral character of the displacement response waveforms compares well
with measured responses. Computed structure base-moment and base-
shear waveforms show similar comparison with measured responses (Fig.
7). The analyses consistently underestimate the magnitude of maximum
story and wall shears, as indicated by the comparison between mea-
sured and calculated shear envelopes in Fig. 8. However, the general
variation of shear over the height is represented adequately. Given the
close correlations between measured and calculated responses for Anal-
ysis C, results of this analysis will be considered the "best fit" results
and will be used in most subsequent comparisons.

OBSERVATIONS FROM INELASTIC STATIC ANALYSES

Partial results of the inelastic static analyses are summarized in Fig. 9,


which compares computed and measured envelope relations between
base shear and top displacement. Measured responses were obtained
from peak responses during the dynamic experiments. The close cor-
relation over the major part of the envelope suggests the static analyses
are adequate in representing the overall load-deformation response of
the structures. The discrepancies in initial stiffnesses (Fig. 9) are ex-
pected because the bilinear approximation of member moment-rotation
response does not allow consideration of uncracked stiffness at the be-
ginning of the "elastic" region.
43

J. Struct. Eng. 1986.112:35-52.


<
LLJ
I
in
UJ
< A MEASURED
CD CALCULATED

10 20 30 40 50 10 20 30 40 50
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

TOP DISPLACEMENT, mm TOP DISPLACEMENT, mm


(a) STRUCTURE FFW (b) STRUCTURE FSW

FIG. 9.—Comparison between Measured and Computed Envelope Relations be-


tween Base Shear and Top Displacement from Inelastic Static Analyses

TOTAL STRUCTURE
FLOOR
LEVEL WALL
9T
8- .
r- ~i
—i
r~~n
*—'
—n
7- i t

6- L i « " - 1

L
5- \ *••} U L
t '-t

4- \ S l| Li \ \
3- i [ 1 1 \ )
•I-
1- L 1 r •~i
1
1
\

1
TOP DISR TOP DISR TOP DISR
8.6 mm I4.6rrinn 28.9mm

FIG. 10.—Effect of Lateral Drift on Computed Story Shear Envelopes for Structure
FFW

For structure FFW, calculated distributions of shear between frames


and wall were observed to vary depending on the level of lateral drift
(Fig. 10). This emphasizes the importance of inelastic modeling for de-
termining interaction forces in that structure. For structure FSW, the cal-
culated interaction was not strongly dependent on the level of response.

COMPARISON BETWEEN MAXIMUM DISPLACEMENTS

Maximum top-floor displacements obtained by the experiments and


by different analysis methods are compared in Table 1. Values for the
static elastic analyses were obtained using lateral loads producing a base
shear equal to that computed for the corresponding elastic modal spec-
tral analysis. The best estimates of maximum displacement were ob-
tained by Analyses B and C of the inelastic dynamic analyses. This is
not unexpected considering that those analysis models were derived to
match measured responses. The other models produced less accurate
44

J. Struct. Eng. 1986.112:35-52.


TABLE 1.—Maximum Top-Floor Displacements (mm)

Dynamic Inelastic Analyses Elastic Analyses


Modal
Structure Measured "Analysis A" "Analysis B" "Analysis C" spectral Static
(1) (2) (3) (4) (5) (6) (7)
FFW 26.1 16.3 (0.62) 27.5 (1.05) 23.0 (0.88) 17.6 (0.67) 17.7 (0.68)
FSW 22.4 17.0 (0.76) 24.2 (1.08) 19.8 (0.88) 17.7 (0.79) 17.9 (0.80)
"Values in parentheses are ratios between calculated and measured maximum displace-
ments.

but acceptable estimates. The reasonably good correlation obtained by


Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

the elastic modal spectral analyses is consistent with the principle of en-
ergy conservation that has been observed for inelastic response of sim-
pler systems (9).

COMPARISON BETWEEN DISTRIBUTIONS OF INTERSTORY DRIFT

The magnitude of interstory drift is a sensitive indicator of structural


damage, and the variation over height is likely to be a particularly good

MEASURED
FLOOR CALCULATED o
LEVEL
9
1
8-
7-
6-
n \
\
5-
4-
3- \
2-
0
1 •
0
//// till
DYNAMIC
NELASTIC

(a) FFW
FLOOR
LEVEL
9
1
8-
7-
6-
5-
4-
K i

s.
3-
2- >
1 -

l/II III
DYNAMIC STATIC STATIC
NEL USTIC INELASTIC ELASTIC

(b) FSW

FIG. 11 .—Comparison between Measured and Computed Interstory Drifts

45

J. Struct. Eng. 1986.112:35-52.


indicator of the presence of structural discontinuities in the vertical plane
of a building. Calculated variations of interstory drift over height are
compared with measured distributions for the various analyses in Fig.
11. Values of interstory drift are normalized with respect to maximum
measured top displacement, thereby minimizing errors in magnitude at-
tributable to incorrect estimates in maximum top displacement. Results
of Analysis C are shown for the dynamic inelastic analyses. For the static
inelastic analyses, results are shown for calculated top displacement ap-
proximately equal to the maximum displacement measured during the
experiments.
For structure FFW, the measured variation of interstory drift is nearly
uniform over height [Fig. 11(a)]. This demonstrates the continuity pro-
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

vided by the continuous structural wall. Calculated variations of drift


are not so uniform, but deviations from the measured are not significant.
It is noteworthy that all the analysis methods indicate essentially the
same distribution of interstory drift.
For structure FSW, the measured variation of interstory drift is mark-
edly nonuniform [Fig. 11(b)]. As expected, maximum interstory drifts
occur in the lower intermediate stories above the wall cutoff. For all an-
alytical models, calculated drift variations follow the same general trend.
However, the elastic analyses underestimate the magnitude of maxi-
mum drift. In contrast, both of the inelastic analyses closely estimate the
maximum interstory drift.

COMPARISONS BETWEEN ABSOLUTE MAXIMUM SHEARS

Absolute magnitudes of story and wall shear envelopes were obtained


directly from the inelastic dynamic analyses. These were compared pre-
viously with measured distributions (Fig. 8), with the conclusion that
overall trends were matched closely but absolute magnitudes were
underestimated.
Absolute magnitudes of story and wall shear envelopes obtained from
the elastic modal spectral analyses significantly exceed the measured
shears. This is not unexpected, because inelastic response occurs in the
test structures, with corresponding reductions in magnitudes of forces.
Ratios between computed and measured base shears are 1.5 and 1.6 for
structures FFW and FSW, respectively. According to one design ap-
proach (9), these ratios are likely to be of approximately the same mag-
nitude as the overall displacement ductilities. Values of approximately
1.9 and 1.7 were reported (7) based on experimental measurements.
Absolute magnitudes of shears cannot be determined from the static
analyses unless maximum drifts are known by some other analysis
method. For drift levels equal to those calculated by the corresponding
elastic or inelastic dynamic analysis, base shears for the static analyses
are very nearly equal to those anticipated by the corresponding dynamic
analysis.

COMPARISONS BETWEEN DISTRIBUTIONS OF STORY SHEARS

The capability of the different methods to anticipate distributions of


story forces can be gaged by comparison between envelopes of mea-
46

J. Struct. Eng. 1986.112:35-52.


MEASURED' Structure •
Wall
CALCULATED = Structure o
FLOOR Wall
LEVEL
9T

7
6-
5
4
3+
2
I
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

STATIC
ELASTIC

(a) FFW
FLOOR
LEVEL
9 •
8-
7-
6-
5 '
4
3
2
I

(b) FSW

FIG. 12.—Comparison between Measured and Computed Story Shear Distribu-


tions

sured and calculated story and wall shears (Fig. 12). Calculated shears
are normalized with respect to maximum measured base shear, thereby
minimizing deviations that are attributable to incorrect estimates in max-
imum base shear. Results of Analysis C are shown for the dynamic in-
elastic analyses. For the static inelastic analyses, results are shown for
calculated top displacements approximately equal to the maximum dis-
placement measured during the experiments.
For both FFW and FSW, calculated distributions of total story shears
are similar for all the analyses (Fig. 12). Although the distributions ob-
tained by the dynamic analyses are closer to measured distributions than
are corresponding distributions for static analyses, the improvements
rendered by the dynamic analyses are not significant.
In contrast to the similarity in total structure shears, calculated distri-
butions of wall shears in structure FFW are affected strongly by the anal-
ysis model [Fig. 12(a)]. The modal spectral analysis grossly overestimates
the wall shear envelopes, whereas, the static elastic analysis underes-
timates the wall shears for most of the stories. Both of the inelastic anal-
ysis methods produce close estimates of the-wall shear envelopes, except
in the top story where the analysis methods are conservative. For struc-
47

J. Struct. Eng. 1986.112:35-52.


ture FSW, all the analysis methods closely anticipate the distribution of
base shear between the frames and wall [Fig. 12(b)].
As discussed previously in reference to Fig. 10, the frame-wall inter-
action forces in structure FFW are sensitive to the level of inelastic re-
sponse. Thus, the poorer correlation obtained for the static elastic anal-
ysis [Fig. 12(a)] probably occurs because the elastic stiffnesses used in
the model do not correctly model the inelastic interaction. The even poorer
correlation obtained by the elastic modal spectral analysis [Fig. 12(a)]
may be attributable to conservatism in the procedure used to combine
modal responses and to inconsistencies between the assumed and actual
response spectra (Fig. 5). For structure FSW, wall shear was not sensi-
tive to the analysis model, despite the severe irregularity in the framing
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

system at the first floor level. Different irregular structures may be more
or less sensitive to modeling assumptions.

COMPARISON OF COMPUTED BEAM ROTATIONAL DUCTILITIES

Maximum member ductilities provide a sensitive indication of the


magnitude and distribution of inelastic action in a structure. For this
study, beam rotational ductility is examined. Experimental ductilities could
not be recorded, and can only be inferred from variations in visible dam-
age. Direct comparison between the different analytical procedures is
also not possible, because of vast differences in modeling assumptions
used by the different procedures. However, comparison of numerical
values obtained by the different methods is of interest because it reveals
the variability of different methods used in design.
Computed beam rotational ductilities from Analysis C of the dynamic
analyses are taken as the best estimate of actual values, and are used as
a basis of comparison. For that analytical model, rotational ductilities are
defined as the ratio between maximum end rotation and elastic end ro-
tation of the beams. For the inelastic static analyses, rotational ductilities
are first obtained for a top-level displacement approximately equal to the
maximum displacement calculated for the corresponding inelastic dy-
namic analysis. The numerical values of ductility are then scaled linearly
by the ratio between top level displacements (it is noted that linear scal-
ing provides only an approximation to the ductilities occurring at the
normalized displacement). Rotational ductilities for the modal spectral
analyses are estimated as the ratio between maximum calculated beam
end moment and yield moment. Values for the elastic static analyses are
computed by the same procedure, and are normalized for base shear
equal to that obtained by the corresponding modal spectral analysis.
Distributions of maximum beam rotational ductilities calculated by each
of the three simpler analysis methods (static inelastic, modal spectral,
and static elastic) are compared with distributions calculated by the in-
elastic dynamic analyses in Fig. 13. The inelastic dynamic analysis an-
ticipates a relatively uniform distribution of rotational ductility demands
for structure FFW, which is indicative of the continuity provided by the
continuous wall. For structure FSW, the computed rotational ductility
demands are concentrated in the lower intermediate floors above the
discontinuous wall. For both structures, rotational ductility demands ob-
tained by the static inelastic analysis are nearly the same as those ob-
48

J. Struct. Eng. 1986.112:35-52.


DYNAMIC INELASTIC ANALYSIS
FLOOR
o ANALYSIS METHOD INDICATED
LEVEL
9-
8 t
7
6
5
4
3
2
I+
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

STATIC MODAL STATIC


INELASTIC SPECTRAL ELASTIC

(o) FFW
FLOOR
LEVEL
9-
8-
7-
6-
5-
4-
3-
2
I

STATIC MODAL STATIC


INELASTIC SPECTRAL ELASTIC

(b) FSW
FIG. 13.—Comparison between Computed Beam Rotational Ductility Demands

tained by the inelastic dynamic analysis. In contrast, both of the elastic


analysis methods tend to underestimate maximum ductility demands
relative to the inelastic analyses. This is particularly evident for structure
FSW [Fig. 13(b)], for which the elastic analyses anticipate a more uniform
distribution of inelastic action than is probable.
Distributions of residual beam crack widths measured from the ex-
periments (7) support the distributions of rotational ductilities indicated
by the inelastic analyses. For structure FFW, the residual cracks were
approximately uniform over height. For FSW, residual cracks were wid-
est in floors 1 through 4, and decreased to an immeasurably small width
at the top floor.

SUMMARY AND CONCLUSIONS

Experimentally measured inelastic responses of two multistory, rein-


forced concrete, planar, frame-wall test structures are studied using dif-
ferent analysis methods. One of the test structures had a structural wall
continuous over the structure height. The other structure was nominally
49

J. Struct. Eng. 1986.112:35-52.


identical to the first, except the structural wall was terminated at the
first floor level. The structures were constructed at small scale and sub-
jected to earthquake simulations that induced inelastic responses. The
responses are interpreted using four different analysis methods, being:
(1) Inelastic dynamic response history analysis; (2) inelastic static anal-
ysis; (3) elastic modal spectral analysis; and (4) elastic static analysis. The
objective is to evaluate the capabilities of the different analysis methods
to anticipate observed response to strong earthquake motions of struc-
tures having relatively uniform and nonuniform vertical distributions of
stiffness and strength.
Based on analysis results presented in this paper, the following con-
clusions are made:
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

1. The computed inelastic response history was sensitive to small


changes in modeling assumptions. It was necessary to model effects of
reinforcement slip from beam-column joints and wall foundations. Ef-
fects of wall base rotations were most significant. An implication is that
effects of wall foundation rotations in real buildings are likely to be sig-
nificant. It is also apparent that good response correlations could not
have been obtained reliably in this study without availability of experi-
mental data (specific to these structures) and without several random
trials. Improved analytical procedures are clearly necessary if reliable re-
sponse estimates are deemed essential to seismic design. However, it is
not clear that such methods are essential.
2. Inelastic static analyses were capable of accurately reproducing the
measured envelope relations between top displacement and base shear.
The analyses indicated that frame-wall interaction can be influenced sig-
nificantly by inelastic member behavior.
3. Maximum displacements were estimated equally well by both the
elastic and inelastic dynamic methods for both structures. For the same
base shear, the static methods produced approximately the same drift
as the corresponding dynamic methods.
4. The nearly uniform distribution of interstory drift in the structure
with continuous wall was adequately reproduced by all the analysis
methods. The nonuniform distribution in the structure with discontin-
uous wall was better reproduced by the inelastic methods than the elas-
tic methods. The dynamic methods did not show clear advantage over
the static methods.
5. Distributions of total story shears were reproduced slightly more
accurately by the dynamic methods than by the static methods. More
significantly, distributions of internal forces were better modeled by the
inelastic methods than the elastic. The dynamic methods did not show
clear advantages over the static methods with regard to distribution of
internal forces.
6. For the structure with continuous wall, all analysis methods indi-
cated approximately the same distributions and magnitudes of member
ductility demands. For the structure with discontinuous wall, the static
inelastic analysis indicated distribution and magnitudes of member duc-
tility demands comparable with the dynamic inelastic analysis. The elas-
tic methods did not adequately anticipate the concentration of inelastic
50

J. Struct. Eng. 1986.112:35-52.


action immediately above the discontinuous wall. In this regard, the elastic
dynamic analysis showed n o clear advantage over the elastic static anal-
ysis.

Based on the preceding analysis, it is apparent that the main advan-


tage of the dynamic analysis m e t h o d s was that they provided an indi-
cation of the maximum displacement response, whereas the static m e t h -
ods alone are generally incapable of indicating displacement amplitudes
for a given seismic event. Otherwise, the dynamic m e t h o d s s h o w e d n o
clear advantage over the corresponding static method. The inelastic
analysis methods showed various advantages over the elastic m e t h o d s ,
in particular, in recognizing the severity of the discontinuity in the struc-
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

ture with discontinuous wall. In this regard, the static inelastic analysis
was comparable to the dynamic inelastic analysis.
It is noted that the conclusions are based solely on the study reported
in this paper. Although it has been s h o w n that the inelastic m e t h o d s
offer advantages over the elastic methods, it is not correct in such a
limited study to conclude in general that dynamic methods do not offer
advantages over static methods. Further study of this topic is recom-
mended.

ACKNOWLEDGMENT

The experimental work was conducted in the N e w m a r k Civil Engi-


neering Laboratory at the Univ. of Illinois at Urbana, IL u n d e r the guid-
ance of M. A. Sozen, and was supported by a grant from the National
Science Foundation. The analytical study was conducted at the Univ. of
California, Berkeley, CA in partial satisfaction of the requirements for
the degree of Master of Engineering of the second writer u n d e r the su-
pervision of the first writer. F. Charney is thanked for assistance in op-
erating the inelastic computer programs and is attributed with devel-
oping the concept for the inelastic static analysis.

APPENDIX.—REFERENCES

1. Gilbertsen, N. D., and Moehle, J. P., "Experimental Study of Small-Scale R/


C Columns Subjected to Axial and Shear Force Reversals," Civil Engineering
Studies, Structural Research Series No. 481, Univ. of Illinois, Urbana, IL, luly,
1980.
2. Kanaan, A. E., and Powell, G. H., "DRAIN-2D, a General Purpose Com-
puter Program for Dynamic Analysis of Inelastic Plane Structures," (With
User's Guide and Supplement), Reports No. 73-6 and No. 73-22, Earthquake
Engineering Research Center, Univ. of California, Berkeley, CA, Sept., 1973.
3. Kent, D. C , and Park, R., "Flexural Members with Confined Concrete," Journal
of the Structural Division, ASCE, Vol. 97, No. ST7, luly, 1971, pp. 1969-1990.
4. Kreger, M. E., and Abrams, D. P., "Measured Hysteresis Relationships for
Small-Scale Beam-Column Joints," Civil Engineering Studies, Structural Re-
search Series No. 453, Univ. of Illinois, Urbana, IL, Aug., 1978.
5. Moehle, J. P., "Seismic Response of Vertically Irregular Structures," Journal
of Structural Engineering, ASCE, Vol. 110, No. 9, Sept., 1984, pp. 2002-2014.
6. Moehle, J. P., "Seismic Analysis of R/C Frame-Wall Structures," Journal of
Structural Engineering, ASCE, Vol. 110, No. 11, Nov., 1984, pp. 2619-2634.
51

J. Struct. Eng. 1986.112:35-52.


7. Moehle, J. P., and Sozen, M. A., "Experiments to Study Earthquake Re-
sponse of R/C Structures with Stiffness Interruptions," Civil Engineering Studies,
Structural Research Series No. 482, Univ. of Illinois, Urbana, IL, Aug., 1980.
8. Newmark, N. M., and Hall, W. J., Earthquake Spectra and Design, Earthquake
Engineering Research Institute, Berkeley, CA, 1982.
9. Newmark, N. M., and Rosenblueth, E., Fundamentals of Earthquake Engineer-
ing, Prentice-Hall, Englewood Cliffs, NJ, 1971.
10. Recommended Lateral Force Requirements and Commentary, Seismology Commit-
tee, Structural Engineers Association of California, San Francisco, CA, 1980.
11. Takeda, T., Sozen, M. A., and Nielsen, N., "Reinforced Concrete Response
to Simulated Earthquakes," Journal of the Structural Division, ASCE, Vol. 96,
No. ST12, Dec, 1970.
12. Uniform Building Code, International Conference of Building Officials, Whit-
tier, CA, 1982.
Downloaded from ascelibrary.org by University of Manitoba on 08/26/15. For personal use only.

13. Wilson, E. L., "The SAP-80 Series of Structural Analysis Programs," Version
84.00, Jan., 1983.
14. Wilson, E. L., Der Kiereghian, A., and Bayo, E. P., "A Replacement for the
SRSS Method in Seismic Analysis," Earthquake Engineering and Structural Dy-
namics, Vol. 9, No. 2, Mar.-Apr., 1981, pp. 187-194.

52

J. Struct. Eng. 1986.112:35-52.

Вам также может понравиться