ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 86-S28

Modeling of Concrete Damage

by Young Soo Chung, Christian Meyer, and Masanobu Shinozuka

As concrete is subjected to loading of increasing intensity, it under- Such a model may then form the basis for rational seis-
goes different phases of damage, from microcracking up to ultimate mic risk evaluation of concrete structures.
failure. It is necessary to simulate this process mathematically to pre-
dict accurately the residual capacity of damaged structures to resist
further load. Of particular concern are members that have been sub- RESEARCH SIGNIFICANCE
jected to several damaging load cycles of a major earthquake. Conventional methods for design and analysis of
Many different damage models have been proposed in the past.
Most of these are not well suited to predict the residual stre'll(th of
concrete members subjected to load all but disregard
damaged members. This paper reviews some basic facts about con- the state of damage that these members may have suf-
crete damage and uses these to model damage systematically as a low- fered during previous load exposure. This paper re-
cycle fatigue phenomenon. The energy dissipation capacity of a views some of the known fundamental facts about
member, instead of the number of cycles to failure, is taken as the damage and proposes a new model that is suitable for
main variable, which depends on many different factors. The model
is capable of simulating reasonably well the strength and stiffness
calculating the response and predicting the residual
degradation of reinforced concrete members subjected to strong cyclic strength of damaged concrete buildings.
loads.
FUNDAMENTAL DISCUSSION OF DAMAGE AND
Keywords: cracking (fracturing); cyclic loads; damage; earthquakes; energy FAILURE OF REINFORCED CONCRETE
dissipation; failure; fatigue (materials); frames; mathematical models; rein- MEMBERS
forced concrete; strength.
Any attempt to devise mathematical models to quan-
tify damage in a rational way should begin with a clear
and precise definition of damage, because "damage" is
The concept of damage permeates many branches of
a widely used word that describes all kinds of different
concrete engineering. As concrete is subjected to loads
phenomena and is open to subjective interpretation. In
of increasing intensity, it undergoes different phases of
the context of our discussion, damage of a reinforced
damage, from microcracking up to ultimate failure. A
concrete member shall signify a specific degree of
capability to simulate this process mathematically is
physical deterioration with precisely defined conse-
desirable because it allows us to predict the residual
quences regarding the member's capacity to resist fur-
strength and serviceability of damaged or aged con-
ther load. Similarly, "failure" of a member signifies a
crete structures. Of particular concern are those mem-
specific level of damage, which corresponds to a some-
bers that have been subjected to several damaging load
what arbitrarily defined residual capacity to resist fur-
cycles, such as may be inflicted by a major earthquake.
ther load. A damage index is usually· defined as the
A considerable effort has been expended by re-
damage value normalized with respect to the (arbitrar-
searchers to develop concrete damage models. In a re-
ily defined) failure level so that a damage index value
cent study, 1 the authors critically evaluated 17 such
of unity corresponds to failure.
models, many of which are of an empirical nature or
As an illustration, Fig. I shows a typical response of
were derived originally for metal structures. Most of
a reinforced concrete cantilever beam to progressively
these models are not well suited to predict the residual
increasing load cycles. 2 The stiffness of the member de-
strength of damaged concrete members.
It is the purpose of this paper to review some basic
facts about concrete damage and to use this knowledge ACI Structural Journal, V. 86, No. 3, May-June 1989.
Rec~1ved Jan. 25, 198~. and reviewed under Institute publication policies.
to construct systematically a new model that is capable Copyng~t © 1989, Amencan Coi!cr.ete !nstitut.e. All rights reserved, including
of simulating reasonably well the strength and stiffness the makmg. of cop1es unless perm1ss10n IS obtamed from the copyright propri-
etors. Pertinent discussion will be published in the March-April 1990 ACI
degradation that accompanies the damage process. Structural Journal if received by Nov. I, 1989.

ACI Structural Journal I May-June 1989 259

 flaws and cracks exist in concrete even before any loads
Young Sao Chung is a research associate in the Department of Civil Engineer-
ing and Operations Research at Princeton University. He worked in the con- have been applied. 3 Sources of microcracking are seg-
struction industry for 8 years before entering Columbia University, where he regation and bleeding, particularly beneath large aggre-
earned his MS and PhD in Civil Engineering and Engineering Mechanics. His gate particles and reinforcing bars. The most frequent
research interests include nonlinear analysis of reinforced concrete structures.
microcracks originate from bond cracks at the cement-
ACI member Christian Meyer is an associate professor in the Department of aggregate interface. As concrete is subjected to stress,
Civil Engineering and Engineering Mechanics at Columbia University. He is a the stress concentrations near the microscopic flaws in-
member of AC/ Committees 442, Response of Buildings to Lateral Forces, and
118, Use of Computers. His research interests include analysis and design of variably lead to microcracks, which multiply and prop-
concrete structures, earthquake engineering, and structural dynamics. agate either along the cement-aggregate interface or
into the cement matrix itself. As the stress level exceeds
Masanobu Shinozuka is'a profeesor in the Department of Civil Engineering and
Operations Research at Princeton-Universt1y. He' has .be;m U.S. coordinator for about 0.5 f:, a more extensive and continuous macro-
a number of joint interhational r~search projects and seminars in. earthquake crack system begins to develop, which softens the ma-
engineering. His reseatch interests include seismic response and reliability of terial and gives a strongly nonlinear stress-strain rela-
building structures and bridges, -pseudodynamic testing, and lifeline systems. He
is a member of the National Academy oj Engineering. tionship.4 Beyond approximately 0.75 f:, the growth
rate of macrocracking accelerates until it becomes un-
stable, leading to failure.
creases gradually once the yield capacity of the member The strength of plain concrete and its damageability
has been exceeded. It requires a significant further in- depend upon a large number of factors. An important
crease in loading for strength to deteriorate also, i.e., factor is the water-cement ratio or, more precisely, the
when the force necessary to cause a given tip deflection material density, which is a function of the water-ce-
decreases in subsequent cycles. Fig. 1 also demon- ment ratio. Another significant influencing factor is the
strates the difficulty of defining failure. Even rubble state of stress. Under biaxial and triaxial compression,
can still carry load. Thus, any failure definition such as the strength increases appreciably and the stress-strain
"a strength reduction of 25 percent of the first yield behavior becomes more linear, indicating a delay of the
load level" is arbitrary and not sufficiently precise, softening-crack formation and damage (Fig. 2). Under
since the apparent residual strength may increase with hydrostatic pressure, it is tempting to claim that con-
further displacement increase (Fig. 1). crete cannot fail. The corresponding stress-strain curve
To understand the various factors that contribute to does indeed maintain a positive slope, presumably in-
damage, it is necessary to review the behavior of plain definitely (Fig. 3). In reality, however, the structure of
concrete at different load levels. Damage can be closely the material progressively becomes damaged with
correlated to the amount of cracking (both are irrever- higher pressure, which manifests itself in a loss of re-
sible) and expresses itself as the degree of nonlinearity sidual uniaxial compressive strength. After applying a
of the stress-strain diagram. Thus, the factors that con- hydrostatic pressure of 6f:, a drop of 25 percent has
trol concrete cracking are also responsible for the af- been reported for the uniaxial strength f: (Fig. 4). 5
flicted damage level. It is well understood that internal When studying the response of plain concrete to cyclic

1-=tbt-
Tl/d = 4.46 P +
p'/Pt 0.53
p" 0.53%

Fig. 1- Typical inelastic response of reinforced concrete member 2 (1 in. = 2.54 em;
1 kip = 4.448 kN)
260 ACI Structural Journal I May-June 1989
 ...
..c:
bO

...'="'
00
Cll l.Of~

.
Cll
.::: 0.75f~
Cll

s'"'
~

0 0.5f~
u
Strain ~
:E= 0.25f~
..
Cll
Fig. 2- Typical stress-strain curves for concrete under ~
multiaxial compression
2f~ 4f~ 6f~ Sf~
Hydrostatic Pressure

Fig. 4-Residual compressive strength after application

and removal of hydrostatic pressure 5

.,...'"'"
Volumetric Strain 00

Fig. 3-Hydrostatic stress-volumetric strain behavior 5

Strain
loads, the most striking aspect of the material behavior
is how the material "remembers" the amount of dam-
age sustained in all preceding load cycles. As a conse- Fig. 5-Cyclic stress-strain curve 6
quence, the cyclic stress-strain curve appears to be
bounded by the monotonic stress-strain curve (Fig. 5). 6 more likely to fail in bond or shear than in flexure,
After one load cycle has been completed and a certain even if properly designed for monotonically applied
amount of damage inflicted, the material responds to a loads.
subsequent load cycle as a simple continuation of the The progressive accumulation of damage in a mate-
previous cycle beyond the maximum previous load rial up to the point of failure under repeated load ap-
level. Note the considerable residual strain, which is an plication is generally known as "fatigue." Each load
indication of the inflicted damage. cycle inflicts a certain amount of irreversible damage
The response of reinforced concrete to load is com- and can be compared to the passage of some time unit
plicated by the complex interaction between steel and of the life span of the material. The relationship be-
concrete. This is reflected by the numerous possible tween stress levelS; and the number of cycles to failure
failure modes in flexure, shear, or bond. Conventional N; is plotted in a so-called S-N curve, typically on a
reinforced concrete design philosophy calls for member logarithmic scale. If the material is subjected to a his-
detailing such that all but a few ductile failure modes tory of varying stress levels, predicting the fatigue life
are precluded. For dynamically applied cyclic loads, it is much more difficult. In this case, it is common prac-
is difficult to predict the failure mode for even "prop- tice to utilize Miner's hypothesis
erly" detailed members. There are two reasons for this.
First, the so-called strain-rate effect influences the var-
ious failure modes to different degrees. (Thus, even a (1)
ductile steel member can experience a brittle fracture,
because the resistance against sliding increases more
rapidly with increasing loading rate than does the resis- where N, is the number of cycles with stress level S;
tance to separation.) The second reason is that load re- leading to failure, and n; is the number of cycles with
versals inflict different levels of damage in the various stress levelS; actually applied.
modes. Bond deterioration and shear cracking typically Eq. (1) assumes that the accumulation of damage is
progress more rapidly under cyclic loading than flex- linear and independent of the load sequence. Thus, the
ural-strength degradation. As a result, reinforced con- three load histories of Fig. 6 are assumed to result in
crete members subjected to earthquake-type loads are exactly the same amount of damage in the end.
ACI Structural Journal I May-June 1989 261
 "'0
cd
0
rJ
~ >;
b.O
'"'
Q)

Time
~
=
"'0
a
-s
Q)
.!:!!
cd

'0"'
z
b

Normalized Deformation, Di

Fig. 8-Ej-Dj curve for reinforced concrete member

c
pendent variable that can be determined experimentally
Fig. 6- Three different load histories for a given stress level S;. The energy dissipated in a
single load cycle, on the other hand, is dependent on
the deformation level. In fact, if the deformation level
were to be chosen so arbitrarily low that the load-de-
formation relationship would remain linear, no energy
would be dissipated at all. Thus, based on this argu-
ment and limited experimental evidence, a D;-E; curve
must have a qualitative shape as shown in Fig. 8, with
a load-amplitude range for which a member will dissi-
pate a maximum amount of energy.
The energy-dissipation capacity of a reinforced con-
E crete member is dependent on a number of variables,
Eo
including the amount of confinement reinforcement;
concrete strength; amount of longitudinal reinforce-
Deformation ment; member detailing; amount and arrangement of
shear reinforcement; shear span; axial force; and the
Fig. 7- Typical definition of normalized dissipated en- load sequence.
ergy Ei In particular, it can be observed from some labora-
tory experiments 8 that the failure mode is closely re-
For metals, the straight application of Miner's rule is lated to the formation of initial cracks that eventually
not well supported by experimental evidence, so that may become critical. Consider the first two load histo-
so-called modified Miner's rules have been proposed. 7 ries shown schematically in Fig. 6. Assume that the
There is even less justification to apply Miner's rule in four low-level load cycles of Fig. 6(a) cause some dam-
the form of Eq. (1) to reinforced concrete. age associated with a specific kind of cracking, and that
In studies of low-cycle fatigue of reinforced con- the energy absorption capacity will be almost intact be-
crete, the number of load cycles to failure is typically fore the member is subjected to the critical fifth load
replaced by the cumulative dissipated energy, which is cycle. The damage caused by this final load cycle is not
often normalized with regard to the energy stored when only more severe, but the associated cracking also may
the member is stressed to the yield level (Fig. 7). If the predispose the member to a different failure mode. If
member were not to degrade under inelastic load cycles, the member were subjected to the load history of Fig.
such a dissipated energy index E; would be exactly pro- 6(b), it is likely that the response and energy absorp-
portional to the number of load cycles. In reality, how- tion capacity would be completely different. For in-
ever, the amount of energy dissipated in each cycle de- stance, the initial critical load cycle may severely dam-
creases with progressive damage until failure. Even age the member and reduce its residual strength and
then, given the necessary experimental data, it would be energy absorption capacity such that it may not even be
possible to present a relationship between constant de- able to sustain the remaining four low-level load cycles.
formation level D; and total energy dissipation capacity Not only the total energy-dissipation capacity of the
E; in analogy to an S-N curve. member is dependent on the load history, but its fail-
As was pointed out, the number of cycles to failure ure mode might be as well.
and total energy dissipation capacity do not correspond A model to predict the remaining energy-dissipation
one-to-one. In an S-N relationship, N; is a truly inde- capacity or life of a member must recognize these facts.
262 ACI Structural Journal I May-June 1989
 &300
~250
J-4
(I)
s:= 200
r.l
"g 150
1001
Deformation, 8
~ 100
8J-4
0 50
Da = ~ z
Fig. 9-Definition of damage ratio 18

PREVIOUS DAMAGE MODELS

Fig. JO(a)-Experimental damage trajectories up to
Numerous models have been proposed in the past to fai/ure 19
represent damage of structural members or entire
structures. Some of these were derived for metal struc-
tures.9.11 Because of fundamental differences between 09
reinforced concrete and homogeneous materials such as
steel, these models are not directly applicable to rein- GO
II)
forced concrete. Other models are based on empirical cic-.
damage definitions!w These all but disregard the me- ~
chanics of the materials involved and therefore do not ~

lend themselves to rational predictions of the str~~gth­ ~

reserve and response characteristics of a structure with II

a specified degree of damage. *c-.
Several investigators have introduced energy indexes ~
that are a function of a few selected parameters. 8 • 16 ·17
For example, Darwin and Nmai 16 defined a normalized
dissipated energy index as

(2)
Fig. JO(b)-Contours of equal probability of failure
and experimental failure points in Df-D1 p/ane19

where E is the total dissipated energy, and A; and As

where 0, = yield rotation, 1:\0o\ = the absolute sum of
are the areas of compression and tension steel, respec-
all plastic rotations associated with loading, and the
tively. Using linear regression analysis, they found a
normalized dissipated energy E.(t)
linear relationship between the energy-dissipation index
D; = E/E and the quantity (vsfc') · /v~ 5 , where vm =
0 5

maximum applied shear stress and vs = shear rein-

forcement capacity. The damage ratio of Lybas and
t M(T)(J(dr)

Sozen is a well-known modeP 8

E.(t) = ---- (5)
M~
y 2

(3)
where !~M(T)(J(dr) is the energy dissipated up to time t
and M,0/2 is the maximum energy that can be stored
where ko = initial stiffness and k, = reduced secant
elastically. Both the flexural damage ratio FDR =
stiffness associated with maximum displacement (Fig.
k1 1Kr and E.(t) can be considered as damage state var-
9). Such a simple damage definition ignores many im-
iables D 1 and D 2 , respectively. By tracing the history of
portant factors.
a member in the D 1-D2 plane [Fig. lO(a)], Banon and
Banon and Veneziano 19 defined a normalized cumu-
Veneziano 19 were able to derive contours of equal prob-
lative rotation NCR
ability of failure [Fig. lO(b)]. This approach obviously
focuses on the important aspects of residual strength
~\Oo\ and low-cycle fatigue life, even though some important
NCR=-- (4)
Oy influence factors are ignored.
ACI Structural Journal I May-June 1989 263
 Moment, M

I
I
I
I
I
I
I
I

(E~)3
I
I
I
'....-:::::r-
--;?'.1- (EI)2
/

.... ---
Curvature, ¢;

Fig. 11- Typical hysteretic moment-curvature relationship

Of the more recent damage models, the widely cited ious branches of hysteretic behavior: 1) elastic loading
model of Park and Ang 20 should be noted and unloading; 2) inelastic loading; 3) inelastic unload-
ing; 4) inelastic reloading during closing of cracks; and
5) inelastic reloading after closing of cracks.
(6)
In a reversed load cycle with high shear, previously
opened shear cracks tend to close, leading to an in-
where Omax = maximum deformation experienced so crease in stiffness and a characteristic pinched shape of
far, o" = ultimate deformation under monotonic load- the moment-curvature curve. Roufaiel and Meyer 22
ing, QY = calculated yield strength, dE = dissipated have modeled this effect by introducing the crack-clos-
energy increment, and ing moment M; associated with curvature¢; (Fig. 12).
If shear stresses are negligible and the hysteresis loops
are stable during cyclic loading, no pinching is likely to
{3 = ( -Q.447+0.73~+0.24n 0 +0.314P,) 0.1Pw (7) occur and Branches 4 and 5 will form a single straight
line. In this case, the point (M; ,¢;) will degenerate to
where II d = shear span ratio, no = normalized axial a point of no pinching (M; ,¢;). A pinching factor aP
force, Pw = confinement ratio, and P, = longitudinal is introduced such that ap = 1 if the shear effect is neg-
steel ratio. ligible, and ap = 0 if the shear effect completely con-
Both the assumptions that damage is a linear combi- trols the load-deformation behavior.
nation of a deformation index and a dissipated energy At its present stage of development the model does
index [Eq. (6)] and the form in which the {3-factor con- not reproduce the strength and stiffness degradation
siders important structural parameters betray a certain associated with bond slip and bond failure.
arbitrariness and are not supported by experimental
evidence. Recently, Nishigaki and Mizuhata 21 expanded Strength deterioration
Park's model by employing a modified Miner's rule to In addition to stiffness degradation, reinforced con-
represent the accumulation of low-cycle fatigue dam- crete members experience strength deterioration under
age. cyclic loading beyond the yield level. Atalay and
Penzien 23 had noticed some correlation between com-
NEW DAMAGE MODEL mencement of strength deterioration and the spalling of
Under load reversals, a reinforced concrete member the concrete cover. But Hwang's experiments 8 showed
experiences a progressive stiffness reduction due to that strength deterioration can start at considerably
concrete cracking and bond deterioration of the steel- lower load levels. Even for loads slightly above the
concrete interface. The hysteretic behavior has been yield level, damage and strength deterioration can be
modeled by various authors. Used herein is Roufaiel observed, provided a sufficiently large number of load
and Meyer's model, 22 which takes into account the fi- cycles is applied. Roufaiel and Meyer 22 found a strong
nite size of plastic regions. Fig. 11 illustrates the var- correlation between the onset of strength deterioration
264 ACI Structural Journal I May-June 1989
 Curvature, ¢

Fig. 12-Typica/ crack closing moment and strength deterioration

and a "critical" displacement level at which the con- Moment,M

crete in the extreme compression fiber is strained to
some limit value. But their investigation relied on test
data with relatively small numbers of cycles for each
load level, such as the test series by Ma, Bertero-, and
Popov. 2 More significantly, it is unreasonable to stipu-
late such a precisely defined point of failure initiation,
i.e., if this point is exceeded by a small amount,
strength deterioration is initiated, but if it is missed by
a small amount, no such strength deterioration takes
place. It is therefore suggested that strength deteriora-
tion is initiated as soon as the yield load level is ex-
Curvature,¢>
ceeded, and the strength deterioration accelerates as the
critical load level is reached. For this purpose, a
Fig. 13-Strength-deterioration curve
strength drop index Sd (Fig. 13) is proposed
formation curve is aimed during reloading (Fig. 12),
(8) such that

where tJ.M = moment capacity (strength) reduction in (10)

a single load cycle up to curvature ¢, tJ.M1 = fictitious
moment capacity (strength) reduction in a single load
and the effective stiffness during reloading becomes
cycle up to failure curvature ¢ 1 For analysis purposes,
the strength drop is measured from the second branch
of the primary moment-curvature curve. The actual
strength reductions are indicated by the shaded area in
Fig. 13. For the parameter w, calibration studies 1 sug-
gest a value of 1.5. where (¢x,MJ is the point of maximum previous load-
With tJ. M denoting the strength drop in one load ing. The actual strength M; reached at curvature ¢x af-
cycle to some curvature ¢, the residual strength after ter one cycle is defined as
this one load cycle (Fig. 13) is given by
(12)

Then, using Eq. (10), (11), and (12), the coordinates of

(9) the imaginary point become

To incorporate this concept of strength deterioration 1

cPx= - - - - -
into the hysteresis model, an imaginary point with co- (£/)- p(EI). (13)
ordinates (¢x ,M x) is introduced, at which the load-de- [My - M~ - c/>y p(EI). + ¢~ (£/)]
ACI Structural Journal I May-June 1989 265
 Moment, M
Strength Drop up to
Failure ~oment
Primary Moment-
Curvature Curve

I
I I
1Failure Moment Curve
I I I
I I
I I

Curvature Ratio, ~ = <P/<Pr

Fig. 14-Definition of failure

(EI) To relate the failure moment at some arbitrary cur-

M = X M, + --===-':___:____ vature c/>; to the actual load history, it is necessary tore-
(EJ) - p(EI). (14)
fer to the strength drop due to one load cycle tJ.M [Eq.
[My - M, + (¢~ - c/>y)p(El),}
(9)]. If the total strength drop down to the failure mo-
ment M 1; at some curvature ¢; is known, the number of
Definition of failure cycles to this curvature level needed to cause failure can
For reinforced concrete members undergoing cyclic be determined as N; = (M;- Mfl)/ ilM;. The failure mo-
loading, several investigators 8• 16•20•23 have defined failure ments for different curvature levels are plotted in the
as the point where the member strength (moment) at failure curve of Fig. 14
maximum displacement (curvature) has dropped below
75 percent of the initial yield strength (moment). But if
(15)
the member is subsequently loaded beyond this maxi-
mum displacement or curvature, its moment can be ob-
served to increase well above the 7 5 percent level, 8 even where
though it has already been assumed to have "failed"
(Fig. 1). For this reason it is necessary to relate the fail- failure moment for given curvature level c/>;
ure definition to the actual strength reserve or residual failure moment for monotonic loading
strength, which is a function of the experienced loading cl>; • )
= - (curvature ratiO
history. ¢!
First, the failure moment M 1 and the corresponding = failure curvature for monotonic loading
curvature ¢ 1 need to be defined for monotonic loading.
These depend on a number of different variables such According to Fig. 14, the failure moment Mfl decreases
as concrete strength, confinement ratio, and axial force, with smaller curvature levels c/>;, i.e., larger strength
which in turn determine the failure mode. A section drops from the monotonic loading curve are needed to
may fail in flexure due to concrete crushing or fracture lead to failure. As can be seen, a simple definition of
of the tensile steel, or it may fail because the compres- Mfl = 0.75 My is not very meaningful.
sion steel buckles. Other failure modes are related to
shear or bond failures, failure modes which are less New damage index
common under monotonic loadings. The failure curva- Based on this definition of failure, we shall now pro-
ture ¢ 1 is defined to be the curvature at which one of pose a damage index as a rational measure of damage
the following three critical strains is reached first: 1) the sustained by reinforced concrete members undergoing
concrete's crushing strain, 2) the tensile steel's rupture strong cyclic loading. It is emphasized again that dam-
strain, or 3) the compression steel's buckling strain, age increments are functions of both the amount of
following spalling of the concrete cover. Given the dissipated energy and displacement or curvature levels.
complete stress-strain curves for steel and concrete, and The proposed damage index D, combines a modified
the cross-sectional dimensions, it is relatively straight- Miner's hypothesis with damage modifiers that reflect
forward to compute the monotonic moment-curvature the effect of the loading history, and considers the fact
curve by determining the moment M; associated with an that reinforced concrete members typically respond
arbitrary curvature ¢;. 1 differently to positive and negative moments
266 ACI Structural Journal I May-June 1989
 Moment, M

--

~ .....,.-
p(EI).

¢r
Curvature, ¢>

Fig. 15-Domainfor damaging load of positive sense

D, = L: ( a/ ;;+
n+ n.-)
+ a,-~- (16)
-
k/ =
}
-+
N;
L: k,j
is the average stiffness during N,+
1 I I N, j=l
cycles up to load level i
where M 1j = M1~ - U - 1)~/ is the moment reached after
j cycles up to load level i (Fig. 16)
= indicator of displacement or curvature level The definition of Eq. (17) requires some explana-
N, M- Mfi = number of cycles to cause f a1·1 ure at
I I tion. As Fig. 16 illustrates, the energy that is dissipated
~M, during a single cycle up to a given level i decreases for
curvature level i successive cycles. That means the damage increments
n, number of cycles actually applied at curvature also decrease. In a constant-amplitude loading se-
level i quence, the first load cycle will cause more damage
a1 = damage modifier than the last one. Therefore, the a,-factor decreases as
load cycling proceeds, being a function of the stiffness
and + and - are indicators of loading sense. ratio.
When counting load cycles n1, only those cycles that As an example, consider the two load histories of
enter the shaded area shown in Fig. 15 are considered, Fig. 6(a) and 6(b). The expected moment-curvature re-
i.e., load cycles No. 1 and 2 are not counted when sponses to these different load histories are shown
computing the damage index D,. This can be justified qualitatively in Fig. 17(a) and (b), respectively. Even
on the grounds that load cycle No. 1 can hardly incur though the second cycle in both cases involves the same
damage during the closing of existing cracks, whereas load level, the energy dissipated in the second cycle
load cycle No. 2 can cause only a negligible amount of (and the incremental damage incurred) is different, be-
damage. cause the damages due to the preceding load cycles are
The loading-history effect is captured by including different. As Fig. 17 illustrates, the factor ¢/ +
the damage modifier a 1, which, for positive moment ¢/_ /2¢,+ is necessary to normalize the damage incre-
loading, is defined as ments in the case of changing load amplitudes. Thus, if
the same moment level m = m were reached in the
1

two cases depicted, the factor for case (b) appears to be

L: k;j
j=l ¢,+ + ¢;+_! greater than that for case (a), which must be closely re-
nt x k/ 2¢/ (17) lated with the amount of the dissipated energy.
For negative loading, the damage modifier is defined
¢/ + ¢,+_1 similarly
2¢,+ "'
i1 k;j r/J;- + ¢;-_I
n,- x k,- 2¢,- (18)
where
c/>;- + ¢;-_1
kt = M;j is the stiffness during the jth cycle up to load
¢/ 2¢,-
n1- [ M,j
level i
ACI Structural Journal I May-June 1989 267
 Moment, M

MG_

..i.+
-'I'J

Curvature, ¢;
I

Fig. 16-Strength drop due to cyclic loading

=s
Q) Load Cycle No. 1
0

- - - - - - - -~

m m'

Curvature, ¢ Curvature, ¢

b)

.6-A'B'C' _ ~: +~i-1
AABC - 2~: if m = m'
Fig. 17-Energy dissipation for different load histories

where results 2•8•23 •24 have been reproduced numerically. Agree-

ment between numerical and experimental results was
in general excellent. Fig. 18 and 19 represent examples
that are typical for the kind of agreement achieved.
1 n' (More comparisons are given in Reference 1.) There-
lC,~ ~ k-ij
= N-
i
~
J= I
fore, the new model can serve as a reliable tool for pre-
dicting earthquake-induced damage in reinforced con-
M;j = M,-; - (j - 1)D.M,-
crete members. However, because of the shortage of
To illustrate the accuracy with which the proposed pertinent low-cycle fatigue experimental data, the cali-
mathematical model can simulate hysteretic response of bration of some of the model parameters cannot be
reinforced concrete members, many experimental considered final.
268 ACI Structural Journal I May-June 1989
 ...-
~ a) Experiment
:;;1
..........

-1

-Z

-z~---L--~-o~.~~~-L----0~...
25,_-L---fo----L-~~~2~5---L--,A~~L-__J

Load-Point Deflection (in.)

35.

28.

21.
b) Analysis
14.

7.

o.
-7.

-14.

-21.

-28.

-35.
-0.75 -o.5o -0.25 0.00 0.25 0.50 0.75

Fig. 18-Experimental and analytic load-deformation curves for Beam S2-3 tested
by Hwang 8 (1 in. = 2.54 em; 1 kip = 4.448 kN)

CONCLUSIONS An accurate determination of damage is essential for

A new damage model has been proposed that is be- meaningful nonlinear dynamic analysis of concrete
lieved to be more rational and which takes into account structures, because the damage index is closely tied to
factors that are ignored in earlier models (such as the the residual strength reserve of a member after it has
loading sequence). It is based on a thorough study of undergone large inelastic load cycles. The principal
the many factors that can contribute to damage of shortcoming of the model is the small number of test
reinforced concrete members. Some model parameters data on which it is based. Thus, it is important that ex-
were calibrated against the few available experimental perimental investigations be undertaken, especially to
results. For example, the rate of strength deterioration determine the relationship between load (deformation)
and damage modifiers were thus determined. level and energy-dissipation capacity (as shown sche-
ACI Structural Journal I May-June 1989 269
ASCE, V. 108, STI, Jan.l982, pp. 1-88. RC Beams under Cyclic Load," Journal of Structural Engineering,
8. Hwang, T. H., " Effects of Variation in Load History on Cyclic ASCE, V. 112, No. 8, Aug. 1986, pp. 1829-1846.
Response of Concrete Flexural Members," PhD Thesis, Department 17. Gosain, Narendra K.; Brown, Russell H.; and Jirsa, James 0.,
of Civil Engineering, University of Illinois, Urbana, 1982. "Shear Requirements for Load Reversals on RC Members," Pro-
9. Krawinkler, H., and Zohrei, M., "Cumulative Damage in Steel ceedings, ASCE, V. 103, No. ST7, July 1977, pp. 1461-1476.
Structures Subjected to Earthquake Ground Motions," Computers 18. Lybas, J., and Sozen, M. A., "Effect of Beam Strength and
and Structures, V. 16, No. 1-4, 1983, pp. 531-541. Stiffness on Dynamic Behavior of Reinforced Concrete Coupled
10. Oliveira, C. S., "Seismic Risk Analysis for a Site and a Met- Walls," Civil Engineering Studies, Structural Research Series No.
ropolitan Area," Report No. EERC-75-3, University of California, 444, University of Illinois, Urbana, July 1977.
Berkeley, Aug. 1975, 45 pp. 19. Banon, H., and Veneziano, D., "Seismic Safety of Reinforced
II. Kasiraj, Iyadurai, and Yao, James T.P., "Fatigue Damage in Concrete Members and Structures," Earthquake Engineering and
Seismic Structures," Proceedings, ASCE, V. 95, ST8, Aug. 1969, pp. Structural Dynamics (Chichester), V. 10, 1982, pp. 179-193.
1673-1692. 20. Park, Young-Ji, and Ang, Alfredo H-S., "Mechanistic Seismic
12. Reitherman, R., "Review of Earthquake Damage Estimation Damage Model for Reinforced Concrete," Journal of Structural En-
Methods," EERI Earthquake Spectra, V. I, No. 4, Aug. 1985, pp. gineering, ASCE, V. lll, No.4, Apr. 1985, pp. 722-739.
805-847. 21. Nishigaki, T, and Mizuhata, K., "Experimental Study on Low-
13. Seed, H. B.; Idriss, I. M.; and Dezfulian H., "Relationships Cycle Fatigue of Reinforced Concrete Columns," Report No. 328,
Between Soil Conditions and Building Damage in the Caracas Earth- Japanese Architectural Society, Tokyo, June 1983, pp. 112-122.
quake of July 29, 1967," Report No. EERC-70-2, University of Cal- 22. Roufaiel, Magdy S. L., and Meyer, Christian, "Analytical
ifornia, Berkeley, Feb. 1970, 102 pp. Modeling of Hysteretic Behavior of R/C Frames," Journal of Struc-
14. "Effects Prediction Guidelines for Structures Subjected to tural Engineering, ASCE, V. I 13, No. 3, Mar. 1987, pp. 429-444.
Ground Motion," URS/Blume, J. A. and Associates, San Francisco, 23. Atalay, M. B., and Penzien, J., "The Seismic Behavior of
1975. Critical Regions of Reinforced Concrete Components as Influenced
15. Whitman, R. V.; Cornell, C. A.; Vanmarcke, E. H.; and Reed, by Moment, Shear and Axial Forces," Report No. EERC-75-19,
J. W., "Methodology and Initial Damage Statistics," Report R73-57, University of California, Berkeley, Dec. 1975, 76 pp.
No. ST380, Department of Civil Engineering, Massachusetts Insti- 24. Popov, E. P.; Bertero, V. V.; and Krawinkler, H., "Cyclic Be-
tute of Technology, Cambridge, 1972. 00 pp. havior of Three RC Flexural Members with High Shear," Report No.
16. Darwin, David, and Nmai, Charles K., "Energy Dissipation in EERC-72-5, University of California, Berkeley, Oct. 1972, 78 pp.

ACI Structural Journal I May-June 1989 271