Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

METHODOLOGY FOR FORENSIC INVESTIGATIONS OF

SEISMIC DAMAGE

By Duane Castaneda, 1 Associate Member, ASCE, and


Colin Brown, 2 Member, ASCE

ABSTRACT: A methodology is presented for conducting forensic investigations of


earthquake-damaged structures. Causes of unexpected damage, indicated by dif-
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

ferences between predicted (expected) and observed damage, are significant to the
engineering and construction professions. Due to the existence of uncertainties
associated with randomness and vagueness, methods for conducting probabilistic
and fuzzy-set damage-assessments are discussed. Measures are presented to com-
pare expected- and observed-damage assessments that are probabilistic or fuzzy.
Methods for identifying, studying, and ranking causes of unexpected damage are
presented. The use of fuzzy expert systems is proposed as a structured method for
using expert knowledge to assess the effect of observed distress on structural in-
tegrity. Desiderata are proposed to ensure reliable conclusions. A fundamental
desideratum is proposed to maintain uniform levels of precision in the required
analyses. Results are presented in layers to determine the professional significance
of the unexpected damage. This methodology is applied to the forensic investigation
of a three-story reinforced-concrete office building damaged in the 1987 Whittier
Narrows earthquake.

INTRODUCTION

Forensic investigations of e a r t h q u a k e - d a m a g e d structures have b e e n cru-


cial in the d e v e l o p m e n t of building requirements, design practice, and con-
struction methods for earthquake-resistant structures. However, studies have
encountered problems. F o r example, research teams ( S h e p a r d and Plunkett
1983; Zeris et al. 1986; Kreger and Sozen 1989) investigating the Imperial
County Services Building, which was severely d a m a g e d in the 1979 Imperial
Valley earthquake, reached very different conclusions about the causes of
the crushing of four reinforced-concrete ( R C ) columns. This was in spite
of excellent available information, including 16 accelerograms of the ground
and building accelerations during the e a r t h q u a k e ( R o j a h n and M o r k 1982).
This paper presents a m e t h o d o l o g y to improve conclusions from such in-
vestigations and is applied to the forensic investigation of a three-story
reinforced-concrete building d a m a g e d by the 1987 Whittier Narrows earth-
quake and owned by the Southern California E d i s o n C o m p a n y (SCE).

METHODOLOGY

A methodology consisting of six stages for conducting forensic investi-


gations of seismic d a m a g e is proposed. In stage one, the investigator assesses
the predicted or expected structural damage. In stage two, the o b s e r v e d
damage is assessed and c o m p a r e d with the e x p e c t e d - d a m a g e assessment to
identify unexpected damage. U n e x p e c t e d d a m a g e is significant since it m a y

~Asst. Prof., Dept. of Civ. and Envir. Engrg., Univ. of Alabama at Birmingham,
Birmingham, AL 35294.
2prof., Dept. of Civ. Engrg., Univ. of Washington, Seattle, WA 98195.
Note. Discussion open until May 1, 1995. To extend the closing date one month,
a written request must be filed with the ASCE Manager of Journals. The manuscript
for this paper was submitted for review and possible publication on September 23,
1993. This paper is part of the Journal of Structural Engineering, Vol. 120, No. 12,
December, 1994. 9 ISSN 0733-9445/94/0012-3506/$2.00 + $.25 per page.
Paper No. 7042.

3506

J. Struct. Eng. 1994.120:3506-3524.


indicate the occurrence of unexpected causes of damage not previously
recognized by the profession. If unexpected damage is identified, then stage
three is undertaken to identify possible causes of this damage. In stage four,
a damage assessment associated with each possible cause is conducted under
the premise that it was responsible for the unexpected damage. Stage five
ranks the possible causes based on how well these conditional-damage as-
sessments match the observed damage. Finally, in stage six, the results of
the investigation are presented in "layers," as shown in Fig. 1, to determine
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

the significance of the investigation to building code officials, designers,


contractors, and those responsible for the maintenance or modifications to
the structure.

STAGE 1: ASSESSING EXPECTED DAMAGE

Assessing the expected (predicted) seismic damage requires determining


expected structural demands and capacities. After reviewing past forensic
studies of earthquake-damaged structures, the writers have noted that one
aspect of the investigation was often emphasized at the expense of other
important aspects. Therefore, this methodology proposes that consistent
levels of precision should be maintained in the various analyses and gathering
of information. In particular, variables required to determine the expected
damage, i.e., variables associated with structural demands and capacities,
should be assessed with uniform levels of precision. Elms (1985) based his
principle of consistent crudeness on a similar observation that the overall
precision in engineering studies is usually governed by the least precise
information available or modeling used. This principle is proposed for use
in assessing the expected damage.

~ ADAMAGE DUE TO CODE REQ.S


__ CCEPT. PRACTICE

CODE REQUIREMENTS AND I ~ DAMAGE DUE TO


ACCEPTED PRACTICE ~SIGN
FOLLOWED

DESIGN SPECIFICATIONS DAMAGE DUE TO


FOLLOWED CONSTRUCTION

DAMAGE DUE TO
NO DETERIORATION DETERIORATION/
NOR MODIFICATIONS
J} ~ g9
ACTUAL OBSERVED
ODIFICATIONS

DAMAGE

FIG. 1. Professional Significance of Damage in Layers

3507

J. Struct. Eng. 1994.120:3506-3524.


Maintaining Uniform Levels of Precision
The expected damage to an element, or globally to the structure, may
be represented by a damage index, D e, which is a function of the input
variables xi. For example, damage due to shear forces could be represented
by the index
V
D e = -- (1)
v.
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

where Vn and V = variables representing the expected nominal shear strength


and applied shear force, respectively. However, since these variables may
be random or fuzzy, so will the expected damage index, D e. Determining
the precision of D e requires statistical information about the input variables,
xi, such as their standard deviations, ai (Benjamin and Cornell 1970). Since
this information is often unavailable, Elms (1985) expressed levels of pre-
cision on a scale from 1 to 5, with 1 representing the lowest and 5 the highest
precision. Elms suggested that verbal descriptions of precision could be
represented by fuzzy sets. The membership function of the fuzzy set, ix, can
take on values from 1.0 (indicating that the precision level fully represents
the variable's precision) to 0.0 (indicating that it does not in any way rep-
resent the variable's precision). According to the procedure proposed by
Elms, these fuzzy sets should be modified to reflect the sensitivity of D e to
the input variables. This is done using Zadeh's (1973) concentration oper-
ation of the exponent a on Ix to produce new memberships Ix~. Elms used
an expert system to assign values to a which ranged from 0.2 for "insensitive"
to 2.0 for "very sensitive" variables.
These fuzzy sets represent a sensitivity-modified assessment of the pre-
cision associated with the available information about these variables. The
intersection of these fuzzy sets provides an assessment of the precision of
D e. Since this intersection operation calculates minimums of the membership
values, the precision of D e will be equal to or less than that of any of the
input variables. Therefore, the first desideratum proposed for forensic in-
vestigations is Elms' (1985) principle of consistent crudeness (reworded):
the sensitivity-modified precision in which any variable is assessed s h o u l d not
be made significantly better than that o f any other variable.

STAGE 2: OBSERVED DAMAGE ASSESSMENT


Since information about observed damage, i.e., cracking and spalling in
RC structures, is expressed verbally and often vague, fuzzy sets which ex-
press uncertainty due to vagueness may be used to express this information.
For example, the verbal assessment of observed cracking, "The crack widths
are very large," could be expressed as follows. The phrase crack widths is
treated as a linguistic variable (Zadeh 1973) that may be assigned the values
(small, m e d i u m , large). These terms are then expressed as fuzzy sets, such
as those shown in Fig. 2. Other terms such as (not, and, or, very, quite,
highly) could be used to modify these fuzzy sets using Zadeh's (1973) cal-
culus. Therefore, in this example, crack widths would be assigned the value
very large, expressed mathematically as the fuzzy set A. Alternatively, these
fuzzy sets can be constructed graphically so that the memberships indicate
confidence that the crack widths represent those which occurred.

Fuzzy Relations for Interpreting Observed Damage Information


Since the expected damage will be expressed as a damage index, D e, it
will be necessary to determine an index of observed damage, D ~, from the
3508

J. Struct. Eng. 1994.120:3506-3524.


SMALL MEgIUH LARGE
\
ck
H
1,0j
0,8
< / \,/ /
I 0,6
0,4
,.s / \
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

W
@,E
Z
0,0
0,75 150 2,25

CRACK WIDTHS (rqm)

FIG. 2. Fuzzy Sets Representing Verbal Assessments of Crack Widths

L[3W MgDERATE SEVERE


/
/ \ \
i0
m 0,8
H
:i:
(/)
0,6
/
Y
?x/ X
p/
LnO,4
PQ
>- 0,2
L,J
Z
/
0,0
0,0 0,2 0,4 0,6 0,8 1,0
DAMAGE INTENSITY

FIG. 3. Fuzzy Set Definitions of Damage Intensity Descriptions

fuzzy set A that expresses information about the observed structural distress,
i.e., cracking or spalling. Fuzzy relations, which have been used to express
expert knowledge, may be used to express expert knowledge about the
significance of structural distress on the structural integrity, as measured by
D ~. This index is computed through the composition operation

D ~ = R oA (2)
R can be constructed using conditional statements (Zadeh 1973). For
example, the following statements represent knowledge about the relation-
ship between crack widths and D ~ for RC shear walls in the SCE building
(to be discussed later):

9 Ba: If widths are s m a l l then D ~ will be l o w


9 Bz: Else if widths are m e d i u m then D ~ will be m o d e r a t e
9 Bs: Else if widths are large then D ~ will be severe

Fuzzy sets defining (small, m e d i u m , large, l o w , m o d e r a t e , severe), as


shown in Figs. 2 and 3, can be used to express the statements Bi as fuzzy
Cartesian products. For example, if s m a l l is defined by the fuzzy set A1 and
low by El, then B1 can be determined as the Cartesian product, B~ = A1

3509

J. Struct. Eng. 1994.120:3506-3524.


TABLE 1. Fuzzy Relation between Crack Widths and D ~

Crack
width Membe~hip~ D~
(rnm) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (1 O) (11 ) (12)
0.25 1 0.7 0.3 0 0 0 0 0 0 0 0
0.50
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

0.7 0.7 0.3 0 0. 0 0 0 0 0 0


0.75 0.3 0.3 0.3 0 0 0 0 0 0 0 0
1.00 0 0 0 0.2 0.5 0.5 0.5 0.2 0 0 0
1.25 0 0 0 0.2 0.6 1 0.6 0.2 0 0 0
1.50 0 0 0 0.2 0.6 1 0.6 0.2 0 0 0
1.75 0 0 0 0.2 0.5 0.5 0.5 0.3 0.3 0.3 0.3
2.00 0 0 0 0 0 0 0 0.3 0.5 0.7 0.7
2.25 0 0 0 0 0 0 0 0.3 0.5 0.7 1

• El. R is then constructed as the fuzzy intersections of the Cartesian


products B~
R = A B, (3)
i

The fuzzy relation to be used to assess the observed damage to the SCE
building is constructed in Table 1 using the fuzzy definitions in Figs. 2 and
3.

COMPARING EXPECTED AND OBSERVED DAMAGE ASSESSMENTS


The surprise measure by Weaver (1948) can be used to identify unexpected
damage when damage assessments are probabilistic. If the expected-damage
assessment is expressed as the probability-mass function, pe(dj), over all
possible damage-states, dj, then the surprise at the occurrence of di is
E Pe(4)
s(d~)- Jpe(d~) (4)

where the numerator is the expected probability over all possible damage-
states, dj. There may also be uncertainty in the observed-damage assessment,
which could be expressed as the probability-mass function, p~ over all
possible damage-states, dj-. In this case, the expected surprise would be
calculated as

e[s] = E po(d3 (pC(d,) J (5)


However, a different measure will be required when the expected- (pre-
dicted) and observed-damage assessments are fuzzy. Consider the expected-
damage assessment expressed as the fuzzy set D e, with membership function
pY(di) shown in Fig. 4. According to the possibility/probability consistency
principle (Zadeh 1978), the occurrence of a predicted damage-state, dj,
having a high membership is likely, while that of a state with a low mem-
bership is unlikely and, therefore, surprising. A n inverse relationship exists
between the memberships of fuzzy sets representing the expected-damage
assessment and the surprise if the damage state actually occurred. Therefore,
3510
J. Struct. Eng. 1994.120:3506-3524.
PDSSIBLE EXPECTED
SURPRISE DAHAGE

1,0 1
I
I
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

I
I
I
1,0 D
FIG. 4. Construction of Possible Surprise Fuzzy Set from Expected-Damage Fuzzy-
Set

PDSSISLE DtSSERVED SURPRISING


SURPRISE DAMAGE )AHAGE

a_ 1,0
m-
0-9
C~ \
L~
Pa
Ld
'/ \
\
//
/
>
1,0 D

FIG. 5. Construction of Surprise Fuzzy Set Representing Unexpected Damage

a fuzzy set representing the possible surprise at the occurrence of the ex-
pected damage can be calculated
~sp(dj) = 1 - ~t.e(dj) (6)
The possible surprise associated with the expected damage-assessment is
shown in Fig. 4. However, the actual surprising nature of the observed
damage must also reflect uncertainty in the fuzzy observed-damage assess-
ment with membership function ix~ Therefore, surprise at the actual
occurrence of a damage state, dj, could be represented by the membership
function I~s(dj), which could be interpreted in this context as the degree of
surprise
Ix~(dj) = min[ixse(dj), ix~ over all dj (7)
Fig. 5 illustrates calculation of the fuzzy set of surprising damage where
damage states associated with high degrees of surprise represent unexpected
damage. This surprise fuzzy set may not be normal in the fuzzy-set- sense.
The investigator may determine the significance of the unexpected dam-
age by inspecting the surprise fuzzy set or by using a measure of the overall
surprise, S, which may be calculated
3511
J. Struct. Eng. 1994.120:3506-3524.
s = w(4) (8)
J
However, since S will not be invariant to the number of damage states
considered, a surprise index, P, possessing this invariance may be calculated
S - Sm~n
I'- , O-<Ps 1.0 (9)
Smax - - Smin
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

Smin = min[l~sP(dj)]; over all dj (10)


Smax -----E ~w(d]) (11)
i
where Sm~. and Sm.x represent the least and greatest possible values of S.
Calculation of these parameters will be illustrated in the discussion of stage
5.

STAGE 3: IDENTIFYING CAUSES OF UNEXPECTED DAMAGE

If significant unexpected damage has occurred, then possible causes must


be identified. Fault trees are proposed to systematically identify these causes.

Fault Trees to Identify Causes


Fault trees have traditionally been used to investigate system reliability
by determining possible causes of system failure. However, fault trees may
also be effective for determining causes of damage which have already
occurred. This effectiveness is due to the hierarchical approach used in fault-
tree analysis. Usually the top event represents system failure while inter-

CRUSHINGOFAI(._~ TOP
RCCOLUMN I E V E N T

I I
HIGHER THAN EXPECT,
AXIAL LBADS ~ INTERMEDIATE
EVENT
LOWER THAN EXPECT,
AXIAL CAPACITY
+
I
HIGHER THAN EXPECT, HIGHERTHAN EXPECT, CONCRETE STRENGTH LGWER THAN EXPECT,

o
GRAVITY LOADS
+
OVERTURNINGFORCES LOWER THAN EXPECT, REBAR STRENGTH

('-) PRIMARY/
EVENT
HIGHER THAN EXPECTS, HIGHER THAN EXPECT,1
LATERAL FQRCES| FORCES RESISTED BY
COLUMN
I
~OR GATE
i I
~GHER THAN EXPECT, GREATER THAN 1 HIGHER THAN HIGHER THAN EXPECT,I
GROUND MOTIONS EXPECT/ MASS LEXPECT,STIFFNESS FRAMESTIFFNESS
9 9 9 (2
FIG, 6. Fault Tree of Possible Causes of Unexpected Damage to Reinforced-Con-
crete Column

3512

J. Struct. Eng. 1994.120:3506-3524.


mediate events lower in the tree represent possibilities leading to the top
event's occurrence. Inferences from lower to higher events are made through
and and or gates, which represent intersections and unions of events, re-
spectively. For identifying causes of unexpected damage, the top event
would represent the unexpected damage while lower events would represent
possible causes.
As an example, the fault tree in Fig. 6 represents possible causes of the
unexpected crushing of a RC column. The identification process begins by
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

identifying major categories of causes, such as higher than expected axial


loads. Then more specific causes of this major event are identified, such as
higher than expected overturning forces. This process is continued until very
specific causes of the unexpected damage are identified.

STAGE 4: STUDIES OF POSSIBLE CAUSES


Once possible causes of the unexpected damage are identified, a limited
number of these causes must be investigated. Therefore, a second desider-
atum should be followed: investigate all equally likely causes of damage.
This is based on the premise that confidence in the investigation's conclu-
sions rests on the most likely cause left unstudied, and is useful since possible
causes should be investigated on the basis of their likelihoods rather than
the precision to which they may be assessed. Also, to improve confidence
in the conclusions, and in accordance with the consistent-crudeness principle
a third desideratum is proposed: studies of possible causes should be equally
precise.

STAGE 5: RANKING OF POSSIBLE CAUSES

Since it may not be possible to determine with absolute confidence that


a particular cause was responsible for the unexpected damage, causes should
be ranked according to their likelihoods. The likelihood of a possible cause
may be determined by comparing the assessment of the expected damage,
conditional on the occurrence of that cause, with the assessment of the
observed damage. If the assessment of the expected damage had the kth
cause occurred is expressed as the conditional probability-mass function
p~(dj) over all possible damage-states dj, then the expected surprise, E[s~],
associated with the kth possible cause may be calculated

s = ~ p~ 1. p ~ j (12)

Possible causes of the unexpected damage are ranked according to highest


likelihood. The likelihood of the kth cause is calculated
1
L~ - Else] (13)
If the assessment of the expected damage conditional on the kth cause
is expressed as a fuzzy set with membership function ix~(dj) over all damage
states dj, then the possible surprise at this expected damage may be expressed
as a fuzzy set with membership function ~P(dj)

i.{P(dj) = 1 - t*~:(g) (14)


3513

J. Struct. Eng. 1994.120:3506-3524.


The actual surprise, considering uncertainty in the o b s e r v e d - d a m a g e as-
sessment, may be calculated
Ig,(d/) = min[l~,P(dj), Ix~ (15)
The surprise measure S k associated with the kth cause m a y be found

sk = (16)
/
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

A n invariant surprise index I s associated with the kth cause m a y be cal-


culated
Sk --
k
Srnin
Is - (17)
S max -- S m,.
Also, a likelihood index associated with the kth cause m a y be d e t e r m i n e d
I~ : 1 - Is (18)

where values of IkL = 0 and IkL = 1 refer to the lowest a n d highest possible

1
0.9
0.8 , EXPECTEDDAMAGE
=. 0.7 GIVEN CAUSE 1
OCCURRED
r 0.6
'" 0.5 EXPECTED DAMAGE
GIVEN CAUSE 2
0.4
I,M OCCURRED
0.3
0.2 = OBSERVEDDAMAGE

0,1
A
0
O
o d d d o d d (5 o
D A M A G E INDEX

FIG. 7. Fuzzy Sets Representing Expected Damage Conditional on Cause 1, Ex-


pected Damage Conditional on Cause 2, a n d O b s e r v e d Damage

1
0.9
o.8
O,.
* SURPRISE GIVEN
0.7
CAUSE 1
r 0.6 OCCURRED
14J 0.5
=E 0.4 SURPRISE GIVEN
I,M
CAUSE 2
=E 0.3
OCCURRED
0.2
0.1
0
o
0 O 0 0 O 0 0 0 0

D A M A G E INDEX

FIG. 8. Fuzzy S e t s Expressing Surprise A s s o c i a t e d with C a u s e s 1 and 2

3514

J. Struct. Eng. 1994.120:3506-3524.


TABLE 2. Surprise Parameters Required to Rank Causes 1 and 2

Cause $~ S~in S~m~, Ranking


(1) (2) (3) (4) (5) (6) (7)
1 2.5 1.0 0.17 0.83
2 1.1 1.0 0.013 0.99

likelihoods, respectively. For example, if the expected-damage assessments


Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

given causes 1 and 2 are as shown in Fig. 7, then the surprise conditional
on each cause would be as shown in Fig. 8. Surprise calculations and ranking
of the causes according to (14)-(18) are summarized in Table 2.

STAGE 6: ASSESSING THE PROFESSIONAL SIGNIFICANCE


The professional significance of the investigation's conclusions may be
determined using the layer system shown in Fig. 1. This system represents
the sequence of events leading to the structure's condition prior to the
earthquake and is similar to a scheme presented by Thiel (1984). The first
layer represents the expected damage had the structure satisfied all building
requirements and accepted practice in effect when the structure was con-
structed. Given that the structure was built, constructed, maintained, and
modified according to building requirements in effect during construction,
excessive damage in the first layer may indicate requirements which need
to be modified. The second layer represents the expected damage had the
structure been built exactly according to the design specifications. Since the
structural engineer is responsible for designing a structure that is consistent
with the code and accepted design procedures, a decrease or increase in the
expected damage from the previous layer must be attributed to the quality
of the engineered design.
The third layer represents the expected damage had the earthquake oc-
curred just after the structure was completed; that is, had no deterioration
or modifications occurred to the original as-built structure. Since the con-
tractor is required to erect the structure according to the structural designer's
specifications and accepted construction practices, changes either beneficial
or detrimental in the expected damage from the previous layer must be
attributed to the construction quality. The fourth layer represents the ob-
served damage. Since the owner is responsible for maintaining a safe struc-
ture, excessive damage as compared with the previous layer may indicate
that accepted maintenance or modification practices need to be revised.

INVESTIGATION OF THREE STORY RC BUILDING


This methodology is applied to investigate damage to an office building
owned by the Southern California Edison Company (SCE) resulting from
the Whittier Narrows earthquake, which occurred on October 1, 1987 less
than 1,500 m from the site.

Building Description and Observed Damage


This is a three-story RC building located in Los Angeles County. As
shown in Fig. 9, the building is 68.3 m by 97.5 m in plan with 5.18-m-story
heights. Shear walls are numbered for later discussion. The foundation
consists of vertical piles which resist vertical and lateral forces. These forces
are distributed to the piles by a horizontal RC truss system. Floors consist
3515
J. Struct. Eng. 1994.120:3506-3524.
i
I
@- I
I #5
@-
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

. . . . . . i . . . . .

@- E
CO
00
d3

i
@-_ I
i L
@-- I
I
|
M~
97,5 m ~"

FiG. 9. First-Floor Plan of Three-Story SCE Reinforced-Concrete Building

of 254-mm-deep hollow-core prestressed planks with a concrete topping that


are supported on precast girders and columns. Shear walls are 305-mm thick
and resist longitudinal (N-S) or transverse (E-W) lateral forces. Fig. 10
shows a typical shear wall where precast columns act as boundary elements.
Nonstructural RC precast panels adjacent to first-story shear walls also exist.
All shear walls were damaged with diagonal cracking occurring at about
45 ~ to the horizontal. Average crack-widths were about 0.76 mm and max-
imum widths were about 1.5 mm. First-floor wall #10 was most severely
damaged, followed by first-floor walls #,5, 6, and 13. Cracks in the second-
floor topping slab were observed at column lines 2 and 10. A #3 bar was
sheared at column line 2 between lines G and H.

Assessment of Expected Damage


Before conducting a consistent precision analysis required by the first
desideratum, three-dimensional linear elastic analyses using the SAP90 pro-
gram (Wilson and Habibullah 1988) were conducted using the mean values
of the required parameters. This was done to determine if the expected
response of the first-floor shear wall was inelastic. Damping of 6.6% of
critical was used in all elastic modes and was based on Haviland's (1976)
data as discussed in Castaneda (1992). All three channels of a SMA Ki-
nemetrics type 2 accelerograph, located on the ground 180 m from the
building, triggered during the earthquake. The resulting accelerograms were
assumed to represent the ground accelerations at this building and were
corrected as discussed in Castaneda (1992).
Results from the analysis indicated periods in the first three modes of
0.351 s, 0.282 s, and 0.186 s, respectively. The response in the first three
3516

J. Struct. Eng. 1994.120:3506-3524.


9.75 m
ROOF

#6 @ 0.305 m PRECAST
[ E.W.E.F. COLUMN
> I (TYP.)
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

3RD FLOOR

#6 @ 0.305 m
E.W.E.F.

2ND FLOOR

1E.F.
I 5- #11E.F. ~ i

,,4 #6 @ 0.305 m
E.W.E.F.
1ST FLOOR

CAP
I I I
FIG. 10. Typical Shear Wall in SCE Building

modes was primarily N-S, primarily E-W, and primarily torsional, respec-
tively. A response spectrum analysis was conducted using the first five modes
and ground accelerations along the longitudinal and transverse building
axes. Applied moments determined from this analysis exceeded the expected
flexural strengths of 17 of 21 first-floor shear walls, in some cases by factors
of 2-2.5. This indicated that inelastic dynamic analyses were required.
Nonlinear, inelastic dynamic analyses were conducted to assess the ex-
pected behavior before conducting consistent precision analyses, which would
be required before making refinements to the modeling. These analysis were
conducted using the plane-frame program Drain 2-D (Kanaan and Powell
1975). Since this is a plane-frame program, response only due to the lon-
gitudinal ground motions was determined. Beam-column bilinear elements
with 5% strain-hardening branches were used to model the shear walls.
Mass- and stiffness-dependent damping was used such that 6% of critical
damping existed in the first and fifth linear elastic modes. Inelastic response
due to the first 10 sec of the record only was calculated since this was the
period of most severe elastic response. Eight of nine first-floor walls and
four of eight second-floor walls reached their yield strengths.
Using results of the inelastic analysis, expected damage to the longitudinal
first-floor walls was assessed. A damage index, D e, in the range [0,1] was
used to express severity of the expected-damage with 0 representing no
3517

J. Struct. Eng. 1994.120:3506-3524.


damage and 1 total failure. A damage function discussed in Castaneda (1992)
was used to calculate D e based on the compression field theory (Hsu 1993)
and tests conducted by Stevens et al. (1991) on RC panels subjected to
cyclic shear loadings. Highest values of the expected damage index, D e,
were determined to be 0.68, 0.67, and 0.65 at wails #2, 13, and 16, re-
spectively. Note that this damage index assesses damage only to individual
walls and not to the overall lateral-force-resisting system of the structure.
Failure of a wall may not cause failure to the entire structure, due to re-
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

dundancy in the system.


Consistent Precision Analyses
To determine how the expected-damage assessment could be improved,
a consistent precision analysis was conducted as required by the first desid-
eratum. Variables used to determine D e are shown in the hierarchical tree
shown in Fig. 11. In this tree, variables at one level are functions of those
at the level below. For example, the shear resisted by the compression struts
in the wall, Vs, was determined from V, the applied shear force to the wall,
minus Vc, the strength associated with shear-resisting mechanisms other than
the struts. Input variables include fy, 7",, r, Vc, and Vsc. Other variables are

(Mu"~ WALL FLEXURAL "(rSRESULTANT

STRESS
FIG. 11. Tree of Variables to Calculate Index of Expected Damage to Reinforced-
Concrete Shear Wall

TABLE 3. Precision and Sensitivity Information for Expected-Damage Index Var-


iables
Precision
Variable Wi level Sensitivity (3/- t

(1) (2) (3) (4) (5) (6)


Tzr 0.144 average 0.521 sensitive 2.0
L 0.0832 good 1.06 very sensitive 4.0
r 0.210 poor 1.32 very sensitive 4.0
v~ 0.260 very poor 0.318 sensitive 2.0
0.166 average 1.00 very sensitive 4.0

3518

J. Struct. Eng. 1994.120:3506-3524.


intermediary and are not associated with input information. Table 3 lists
the coefficients of variation, V/, of the input variables x~, which were assigned
levels of precision according to the following rules:

good: 0 < Vi -< 0.08


average: 0.08 < V~ _< 0.15
poor: 0.15 < Vi -< 0.25
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

very poor: 0.25 < V~

The sensitivities of De to the input variables are also shown in Table 3, as


defined
'Yi = (mi/mDe)(ODe/Oxi) (19)

where mi and moe = mean values of x i and D e, respectively. Fuzzy sets

1 ANCHOR
0.9 STRENGTH
0.8
REBAR YIELD
r 0.7 STRESS
iv. 0.6
Ill 0.5 RESULTANT
m
HEIGHT
ULI
0.4
:E 0.3 SHEAR RESISTED
0.2 BY OTHER
0.1
SHEAR CAPACITY
0
OF STRUTS
2 3 4 5
PRECISION LEVEL DAMAGE

FIG. 12. Fuzzy Sets Representing Sensitivity-Modified Precision Assessments of


Input Variables Required to Calculate Expected Damage to Shear Wall # 2

1
0.9
0.8
0. 0.7
Z
0.6
E
uu 0.5
0.4
g,,,I
0.3
0.2
0.1
I I I

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


DAMAGEINDEX
FIG. 13. Fuzzy-Set Assessment of Expected (Predicted) Damage to Shear Wall
#2

3519

J. Struct. Eng. 1994.120:3506-3524.


1
0.9
0.8
9 POSSIBLE
0,. 0.7
.,r SURPRISE
n,. 0.6
uJ 0.5 OBSERVED
m
0.4 DAMAGE
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

0.3
0.2
0,1
0 []
o
o o d o o d d d o
DAMAGE INDEX

FIG. 14. F u z z y - S e t A s s e s s m e n t o f Possibly Surprising and Observed Damage t o


Shear Wall # 2

0,5
0.45
0.4
" 0.35
u~ 0.3
E
m 0.25
0.2
g,,I
0,15
0.1
0.05

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1


DAMAGE INDEX
FIG. 15. Fuzzy-Set Assessment of Surprise Associated with Observed Damage
t o S h e a r Wall # 2 Given Expected-Damage Assessment

representing the sensitivity-modified precision of the input variables as mod-


ified by the exponent oLare shown in Fig. 12. As can be seen in this figure,
uncertainty in assessing Vc, the shear force resisted by mechanisms other
than the concrete compression struts, governs the precision in which the
index of expected damage, D e, can be assessed. Since additional information
about this variable was unavailable, additional analyses to increase the pre-
cision of D e were not conducted.

Fuzzy Assessment of Expected Damage to Wall # 2


Since wall #2 was expected to have the greatest damage, a fuzzy set
representing uncertainties in the expected-damage assessment for this wall
was constructed. Uncertainties in the information associated with the input
variables shown in Fig. 11 were assessed using information-entropy theory
(Castaneda 1992). Entropies associated with input variables were used to
determine the entropy in expected-damage index D e. This entropy was then
used to construct a probability-mass function representing the overall un-
certainty in the expected-damage assessment. This function was normalized
3520

J. Struct. Eng. 1994.120:3506-3524.


by its maximum value to construct the fuzzy-set assessment of the expected
damage, D e , shown in Fig. 13. The wide range of possible-damage indices
is due to the imprecise information associated with the input variables. The
fuzzy set representing the possibly surprising damage was calculated ac-
cording to (6) and is shown in Fig. 14.

Assessment of the Observed Damage for Wall # 2


The actual damage to wall #2 was assessed from visual observations of
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

the diagonal-cracking severity and was expressed as the fuzzy set, A, of


crack widths
Z = 0.5[0.50 + 1.010.75 + 1.011.00 + 1.011.25 + 0.511.50 mm (20)
The fuzzy relation shown in Table 1 was used to convert this fuzzy crack-
width assessment to one expressing the severity of observed damage to wall
#2, which is shown in Fig. 14.

Comparison of Expected and Observed Damage Assessment for


Wall #2
The fuzzy set representing the unexpected (surprising) damage to wall
#2 was constructed according to (7) as shown in Fig. 15. Note that the right-
most area of intersection shown in Fig. 14 is not plotted in Fig. 15 since the
intersection was taken at discrete values of damage index. Although sur-
prising damage in the range 0 --- D < 0.3 occurred, the surprise index I s,
was only 0.26. Since values of I s of 0 and 1 indicate no surprise and total
surprise, respectively, 0.26 indicates that this damage is not surprising.
Therefore, unexpected causes of damage were not indicted, and stages 3 -
6 were not conducted.

CONCLUSIONS
A methodology for conducting forensic investigations of seismic damage
has been proposed to identify and study causes of unexpected damage that
are important to designers and builders. Unexpected damage is identified
in stages 1 and 2 by comparing the assessment of predicted (expected)
damage with the observed damage. These assessments may be expressed as
probability functions or fuzzy sets defined over all possible damage states,
dj. A method is presented to use observations of structural distress expressed
as fuzzy sets to determine a fuzzy-set damage index. Unexpected damage,
indicated by differences between the expected- and observed-damage as-
sessments, is identified using the surprise measure when these assessments
are probabilistic. A modified surprise measure for fuzzy-set damage as-
sessments has been developed. If professionally significant unexpected dam-
age has occurred, then further studies are conducted to identify and assess
possible causes. In stage 3, fault trees are used to structure the process of
identifying possible causes. The expected damage had each cause occurred
is assessed in stage 4. Stage 5 compares these conditional expected-damage
assessments with the observed damage. The likelihood of each possible cause
is determined and causes are ranked accordingly. The professional signifi-
cance of the investigation in layers is determined in stage 6 for people
charged with designing, constructing, and maintaining structures, as shown
in Fig. 1.
Desiderata have been proposed to mitigate problems that have occurred
in previous investigations. The first desideratum requires maintaining uni-
3521

J. Struct. Eng. 1994.120:3506-3524.


form levels of precision in the expected-damage analyses. The second de-
sideratum calls for investigating all equally likely causes of damage, since
confidence in the conclusions rests on the most likely cause not studied.
The third desideratum requires maintaining similar levels of precision in the
expected-damage assessments conditional on the possible causes of unex-
pected damage.
This methodology has been applied to a three-story reinforced-concrete
building damaged in the 1987 Whittier Narrows earthquake. Damage to the
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

first-floor shear walls was investigated using elastic and inelastic dynamic
analyses and. records of nearby ground accelerations. A consistent precision-
analysis indicated that significant improvements in the expected-damage
assessments were not possible. Fuzzy sets representing the expected and
observed damage to a first-floor shear wall predicted to be most highly
damaged were constructed. Comparisons of these fuzzy sets indicated no
unexpected damage, and further studies were not pursued.

ACKNOWLEDGMENTS
The writers wish to thank the Southern California Edison Company for
its invaluable cooperation, especially Dr. Dennis Ostrom. Thanks also to
Dr. Gerald Brady at the U.S. Geological Survey (USGS), Menlo Park;
Peter Yanev, Dr. Charles Scawthorn, and Ronald O. Hamburger at E Q E
Inc.; and Dr. Gary Hart, University of California, Los Angeles, who were
of great assistance. This research was sponsored by the National Science
Foundation.

APPENDIX I. REFERENCES
Benjamin, J. R., and Cornell, C. A. (1970). Probability, statistics, and decision for
civil engineers. McGraw-Hill, Inc., New York, N~Y.
Castaneda, D. E. (1992). "A methodology for conducting forensic investigations of
earthquake damaged buildings," PhD dissertation, Univ. of Washington, Seattle,
Wash.
Collins, M. P. (1978). "Towards a rational theory for RC members in shear." J.
Struct. Div., ASCE, 104(4), 649-666.
Elms, D. G. (1985). "The principle of consistent crudeness." Proc., National Science
Foundation (NSF) Workshop on Civ. Engrg. Applications of Fuzzy Sets, C. B.
Brown, J. L. Chameau, R. Palmer, and J. T. P. Yao, eds., Purdue Univ., School
of Civ. Engrg., West Lafayette, Ind., 35-44.
Haviland, R. (1976). "Study of the uncertainties in the fundamental translational
periods and damping values for real buildings." Rep. No. 5, Pub. No. R76-12:
Evaluation of Seismic Safety of Buildings, Mass. Inst. of Technol., Cambridge,
Mass.
Hsu, T. T. C. (1993). Unified theory of reinforced concrete. CRC Press, Inc., Boca
Raton, Fla,
Kanaan, A. E., and Powell, G. H. (1975). "Drain-2D. A general purpose computer
program for dynamic analysis of inelastic plane structures with users' guide." Reps.
No. EERC 73-6 and 73-22, Earthquake Engineering Research Center, Berkeley,
Calif.
Kreger, M. E., and Sozen, M. A. (1989). "Seismic response of Imperial County
Services Building in 1979." J. Struct. Engrg., ASCE, 115(12), 3095-3111.
Rojahn, C., and Mork, P. N. (1982). "An analysis of strong-motion data from a
severely damaged structure--the Imperial County Services Building, El Centro,
California." The Imperial Valley, California, Earthquake of October 15, 1979;
Profl. Paper 1254, U.S. Geological Survey, 357-375.
Shepard, R., and Plunkett, A. W. (1983). "Damage analysis of the Imperial County
Services Building." J. Struct. Engrg., ASCE, 109(7), 1711-1726.
3522
J. Struct. Eng. 1994.120:3506-3524.
Stevens, N. J., Uzumeri, S. M., and Collins, M. P. (1991). "'Reinforced concrete
subjected to reversed cycfic shear-experiments and constitutive model." ACI Struct.
J., 88(2).
Thiel, C. C. Jr. (1984). "Divergence between estimated building vulnerability and
observed damage: a fuzzy set theory reconciliation." Appl. Technol. Rep. ATC-
tO-l: Critical Aspects of Earthquake Ground Motion and Building Damage Poten-
tial, Applied Technology Council, Berkeley, Calif., 115-129.
Weaver, W. (1948). "Probability, rarity, interest, and surprise." The Scientific Monthly,
67(6), 390.
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

Wilson, E. L., and Habibullah, A. (1988). SAP90: A series of computer programs


for the static and dynamic analysis of finite element analysis of structures. Computers
and Structures, Inc., Berkeley, Calif.
Zadeh, L. A. (1978). "Fuzzy sets as a basis for a theory of possibility." Fuzzy Sets
and Systems, 1(1), 3-28.
Zadeh, L. A. (1973). "Outline of a new approach to the analysis of complex systems
and decision processes." IEEE Trans. on Systems, Man, and Cybernetics, 3(1),
28-44.
Zeris, C. A., Mahin, S. A., and Bertero V. V. (1986). "Analysis of the seismic
performance of the Imperial County Services Building." Proc., 3rd U.S. Nat. Conf.
on Earthquake Engrg., Aug. 24-28, Charleston, S.C., Vol. H, Earthquake En-
gineering Research Institute (EERI), Oakland, Calif., 847-858.
Zsutty, T. C. (1968). "Beam shear strength prediction by analysis of existing data."
ACI J., 65(11).

A P P E N D I X II. NOTATION

The following symbols are used in this paper:

A = fuzzy set of observed-damage information;


Bi --- ith statement describing relationship between crack widths and
Do;
D e --- index of expected damage;
D ~ -- index of observed damage;
dj = jth possible value of D;
E[s] = expected surprise;
E[sk] = expected surprise conditional on kth cause;
I s -~ index of surprise;
I~ = likelihood index of kth cause;
I s -~ surprise index of kth cause;
Lk = likelihood of kth cause;
M, = flexural strength of shear wall at yield;
pe = probability-mass function of De;
po = probability-mass function of D~
p~ = probability-mass function of expected damage conditional on kth
cause;
R = fuzzy relation;
s = surprise function;
S = measure of overall surprise;
Sk = measure of overall surprise conditional on kth cause;
Sm,x = maximum value of S;
S ~ , = minimum value of S;
S~max = maximum value of Sk;
S~i, = minimum value of S~;
T, = strength of anchor from shear wall to foundation;

3523

J. Struct. Eng. 1994.120:3506-3524.


V expected applied shear force;
~= coefficient of variation of x;;
~= expected nominal shear strength;
Xi = input variables of De;
O[ : exponent used in concentration operation on I~;
sensitivity of D e to xi;
i~ = membership function;
~= membership function of De;
Downloaded from ascelibrary.org by Remote User on 03/12/14. Copyright ASCE. For personal use only; all rights reserved.

Ixek= membership function of expected damage conditional on kth cause;


~o =
membership function of D ~
#= membership function of surprising damage;
~= membership function of surprising damage conditional on kth cause;
~= membership function of possibly surprising damage; and
~: membership function of possibly surprising damage conditional
on kth cause.

3524

J. Struct. Eng. 1994.120:3506-3524.

You might also like