Compufers & Stmctures, Vol. 7, pp. 137-147. Person Press 1977.

Printed in Great Britain

ANALYSIS OF CYCLIC LOADING OF PLANE

R/C STRUCTURESt

DAVID DARWIN+
Department of CivilEngineering, University of Kansas, Lawrence, KS 66045,U.S.A.

and

DAVID A. PECKNOLDP
Department of Civil Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801,U.S.A.

(Receioed 30 November 1975)

Ahstrati-A numerical procedure for cyclic loading response of planar reinforced concrete structures is presented. A
nonlinear orthotropic stress strain law for biaxially loaded plain concrete is developed and compared with
experimental results for monotonic biaxial loading and uniaxial cyclic loading. The stress-strain law recognizes
strength and ductility changes due to biaxial stress, and strength and stiffness degradation with cycles of loading. The
stress strain law is incorporated into a finite element computer program which utilizes isoparametric quadrilaterals
with extra non-conforming deformation modes. Numerical and experimental results are presented for a
monotonically loaded shear wall-frame system and a cyclically loaded shear wall.

INTRODUCTION these models appear to be promising for the prediction of

Reliable information on strength, failure mode, ductility strength and monotonic load-deflection response. Vital
and energy absorption capacity is required for the design questions which remain unanswered relate to behavior
of important reinforced concrete structures such as under conditions of multiaxial stress reversals.
nuclear containment structures and prestressed reactor In this paper a numerical procedure is presented which
vessels, and for the development of new types of building partially fills this gap. Many structures such as slabs, thin
construction such as coupled shear wall-frame systems, shells and shear walls can be idealized as being in two
particularly for seismic loading cdnditions. In many cases dimensional states of stress. Biaxial stress is therefore a
strength under monotonic loading can be adequately most important special case. A nonlinear stress-strain law
estimated by conventional means. The major concern is for plain concrete under cyclic biaxial stresses is
prevention of strength deterioration and assurance of developed, and incorporated into a finite element numeri-
adequate ductility and energy absorption capacity under cal simulation for planar reinforced concrete structures.
repeated loadings. The stress-strain law is supported by comparisons with
The complex behavior of reinforced concrete under the available experimental data, and the structural
biaxial or triaxial cyclically varying stresses, coupled with analysis capabilities of the finite element procedure are
the lack of sufficient accumulated experience to draw demonstrated by comparisons with experimental results.
upon necessitates fundamental investigations into the
behavior of such structures. Structural testing is the BREAVIOR OF PLAIN CONCRETE
cornerstone of any investigation, but numerical simula- Some of the important and well-known characteristics
tion procedures aid in planning an experimental program of the behavior of plain concrete under general biaxial
and in interpreting the results. In addition, analytical or stress conditions are first recalled. The relevance of these
numerical models allow the extrapolation of experimental characteristics to the problem of simulation of structural
results to cover the many parameter variations which behavior is then discussed in order to provide a
cannot be experimentally investigated because of perspective from which the stress-strain law for plain
economics. In some circumstances, depending on the concrete, proposed herein, can be v&wed.
complexity of the model and the importance of the Consider first the behavior of plain concrete under
structure, a numerical model may .be used directly in the uniaxial stress conditions. Figure 1 illustrates several
design process, but models substantially simpler than important points: (1) The tensile strength ft of concrete is
those under discussion here would normally be used for an order of magnitude smaller than the compressive
this purpose. strength fe’. (2) At compressive stresses larger than about
Many recent publications deal with biaxial stress-strain 0.7 fc’, concrete begins to soften significantly. This
relations and biaxial and triaxial strength of plain behavior is apparently due to the onset of extensive
concrete [ l-91 and with numerical analysis procedures for microcracking which initiates at the aggregate mortar
reinforced concrete structures [2-4,10-171. Several of interfaces, beginning at about 0.3 f’, and eventually
extends between aggregate particles[7]. (3) At strains
beyond cc”,the strain corresponding to peak compressive
tPresented at the Second National Synposium on Computer-
ized Structural Analysis and Design at the School of Engineering stress, the uniaxial stress-strain curve exhibits a
and Applied Science, George Washington University, Washing- downward slope, or degradation of strength. (4) Behavior
ton DC., 29-31 March 1976. under cyclic compressive loading is also shown in Fig.
SAssistant Professor. 1[18]. The stress-strain curve for monotonic loading
BAssociate Professor. forms an envelope curve for cyclic loading. For small

137
138 DAvm DARWIN and DAVID A. PECKNOLD

4 Compression

Compression

Slroin

Fig. 1. Responseof concreteto uniaxialloading.

cycles of compressive stress (not shown) the hysteresis inhibit the formation of the microcracks in the concrete.
curves are closed loops. For large cycles of compressive Figure 3 shows the biaxial strength envelope experimen-
stress, the hysteresis loops intersect the envelope curve tally determined by Kupfer, Hilsdorf and Riisch[6].
and additional inelastic strain is accumulated during each In Fig. 4 are shown the observed strains in both
cycle, eventually leading to failure. With increasing principal stress directions plotted vs uz for the same case
numbers of cycles the slope of the hysteresis curve illustrated in Fig. 2. At fairly high stress levels, it is
decreases (stiffness degradation) and the energy dissi- evident that the effective stiffnesses are quite different in
pated, interpreted as the area within the hysteresis loop, the two directions. The difference is much greater than
decreases. can be explained by the Poisson effect. This evidence
Consider now the behavior of plain concrete under points to stress-induced orthotropic behavior under
biaxial states of stress. Three extensive experimental biaxial stress conditions.
programs have been carried out to determine the behavior
of plain concrete under monotonically increasing, propor- DISCUSSION OF MODELlNG OF BEHAVIORAL CDABACt’BBISTICS
tional biaxial loading[6,7,19]. Figure 2 shows the OF PLAINCONCRETE
observed stress-strain response[6] in the direction of c~2 The degree to which each of these characteristics
for a biaxial compression test conducted with a constant affects structural behavior depends on several factors
principal stress ratio (at/a* = 0.52). It is evident that the including structural type, nature of the loading and the
compressive stress in the orthogonal direction acts to response quantity sought. Of course, the stress-strain
increase both the strength and ductility of the concrete as response of reinforcing steel, and bond slip between
compared to the uniaxial case. This is apparently because concrete and steel play a very sign&ant role in overall
the compression in the orthogonal direction tends to structural behavior, but attention is focussed here on the
concrete. Tensile cracking of the concrete is the most
-r, significant source of nonlinearity present, and essentially
Biaxial produces a softened material with directional
u2(u, = 0,52q2)
properties[ 11,131. This effect should be included in all
cases. For monotonically loaded structures which fail by
yielding of steel reinforcing, reasonable estimates of
failure loads can be made by assuming the concrete to
behave elastically in compression. For more general
monotonic loading situations, in which the concrete may
experience high compressive stresses, a failure criterion
for the concrete in biaxial compression is needed. Several
investigators[lO, 16,17,20] have idealized concrete as
elastoplastic in compression and as an elastic brittle
material in tension. Several different yield criteria, or
biaxial strength envelopes, have been employed, usually
based on the experimental data of Kupfer, Hilsdorf and
Rtisch[6]. These procedures recognize the biaxial
strengthening effect, but not the ductility increase, or the
softening prior to reaching maximum compressive stress.
Nevertheless, models of this type are usually quite
successful in providing failure load estimates for
E
monotonic loading. More sophisticated models which
incorporate the compressive softening of concrete have
been developed recently [ 1,5,8,9], and these are capable
Fig.2. Effectof biaxialstresson concrete. of providing better estimates of deformations near
 Analvsis of cvclic loading of Diane R/C structures 139

Experimental
1(Kupfer, Hilsdorf, Riisch)

w= Te;. Strength

.)f’t

Y~+~;-~-3.65~=0

Fig. 3. Biaxial strength envelope for concrete.

I I I
-I,0 -20 -30
Strain, 0.001Win.
Fig. 4. Stress-induced orthotropic behavior of biaxiallyloaded concrete.

ultimate load, as opposed to merely predicting the stress strain model for concrete must extend past the
ultimate load itself. strain corresponding to peak stress and must represent
Despite the large expenditure of effort and the many hysteretic behavior reasonably well. None of the previ-
models which have been developed, it is not clear how ously proposed models is capable of handling this loading
sensitive the results are to some of the modeling condition. It appears to be very ditlicult in this case to
assumptions. The treatment of shear transfer in zones of extend plasticity-based models to incorporate the needed
tensile cracking is one such question. Further investiga- behavioral characteristics of plain concrete.
tion is needed on this and other points, including bond There are several serious limitations to the experimen-
slip. The authors believe that it has been demonstrated tal information available at the present time. In particular,
that many of the refinements are not necessary for the the work of Kupfer, Hilsdorf and Riisch[6], Liu, Nilson
adequate estimation of ultimate loads. However, if and Slate [7j and Nelissen [ 191involved monotonic loading
deformations near ultimate load are of concern, the only. The effects on strength and stress-strain response of
situation is less clear. Comparative studies of several of non-monotonic loading is unknown. It is well known that
the models would be valuable. effective confinement of concrete has a significant
In order to simulate cyclic loading behavior, and influence on ductility. Models of the type discussed here
particularly to assess strength deterioration, ductility and cannot analytically reproduce this effect, and must
energy absorption capacity under large load reversals, the account for it indirectly by modifying the stress-strain
140 DAVID DARWINand DAVID A. PECKNOLD

curve for the concrete, usually by making the downward

(1 - u2)G =; (E, + Ez-2vd(E,E& (3)
slope of the stress-strain curve (Fig. 1) less steep. In order
to accurately simulate the effects of different degrees of
confinement, more experimental information is needed. Equation (1) then takes the form

vqE,Ez 0
Ez 0 (4)
1/4(E1t Ez -

With regard to cyclic loading, the only available If incremental equivalent uniaxial strains, dcl. and de*“,
information[ll, 211,relates to uniaxial stress conditions. are defined as
Thus the development of a model for cyclically varying
biaxial stresses is subject to many uncertainties, most of
de,. =$-&l+v@)dcz]
which are shared with many monotonic stress-strain
models. Nevertheless, the authors feel that the informa-
tion which can be provided by such a model warrants the dtZu=--& [v@)d”+de2]. (5)
effort, despite the assumptions and extrapolations of
experimental data which must be made at the present
time. Schnobrich [ 141 and Scordelis [15] present com- Then eqn (4) becomes
prehensive surveys of the current techniques for model-
ing reinforced concrete structures. du, = E, de,.

STRESS-STRAINLAW FOR PLAIN CONCRETE duz = Ez dez. (6)

The general objectives in developing the constitutive dr12= G dy,2.
model for plain concrete[3] were (1) to express it in a form
which could be easily incorporated into a finite element These relations have the same form as for uniaxial stress
program, (2) to include cyclic loading, and (3) to use conditions, hence the name “equivalent uniaxial strain”
conventional parameters such as fc’, ft, lcu and Eo for for de,,, and dezu. The definition of equivalent uniaxial
which the analyst has some feel, rather than empirical strain can be restated, using eqn (6) as
constants which must be determined for each concrete
IniX.
doi
The inclusion of the cyclic loading capability and the i = 1,2
consequent need to realistically model strength and eiu = i?
stiffness degradation introduces some major difficulties. A
strain-like variable is needed to trace the deformation or its discrete equivalent,
history and determine hysteretic behavior. Rather than
attempt to extend plasticity concepts to cover this case, a
quantity called “equivalent uniaxial strain”[3] has been
introduced. This variable, which will be described
subsequently, traces the response history and is used to
control cyclic behavior. Its introduction is not essential
for the description of monotonic loading, but provides where Am = incremental change in principal stress UL
even in this case, a more compact representation of In order to further interpret these quantities, consider a
biaxial stress-strain response. linear elastic orthotropic material. The material properties
Incremental stress-strain relations for an orthotropic are then constant, and the equivalent uniaxial strains
material take the form become elU = at/E, and EZ.= ~/EL EL”therefore repres-
ents in the elastic case the strain in principal direction 1,
with the Poisson effect removed, i.e. the strain due to UI
acting alone. A similar interpretation can be made in the
inelastic case. It should be emphasized that the eiu are not
real strains; they do not transform under axis rotations as
true strains do. Further, they are accumulated in the
where VIEZ= VZEIand subscripts 1,2 denote the current principal stress directions which in general change during
principal stress axes. The material properties El, Ez, G, loading, so that ~1” for example does not provide a
vl, vz are stress dependent. This form can be easily “deformation history” in a tixed direction, but in the
incorporated into standard numerical analysis procedures, continuously changing direction corresponding to princi-
and recognizes the stress-induced orthotropic behavior pal stress ur. Nevertheless the introduction of these
mentioned previously. Two modifications to eqn (1) are quantities makes possible the use of quite realistic
now made. First, a more convenient expression results if hysteresis rules for plain concrete. The basic idea is that
an “equivalent Poisson’s Ratio”, v, is defined as once the stress-strain law has been written in a form
similar to that for the uniaxial case, stress-strain curves
v2= v,v2. (2) similar to the uniaxial stress-strain response can be used.
The determination of the stress dependent properties
Second, since no information is available on shear El, Ez, v occurring in eqn (l)-(S), which completely
modulus G, the assumption is made that it is independent characterize the proposed stress-strain law, will now be
of axis orientation. This leads to the relation]31 described. Consider first the case of monotonic loading.
 Analysisof cyclicloadingof planeR/C structures 141

MONOTONIC LOADING corresponding to maximum compressive stress, and

If the equivalent Poisson’s ratio, v, is initially taken to & = Q/G The equivalent uniaxial strain eii. essentially
be 0.20, which agrees with the results of Kupfer, Hilsdorf removes the Poisson effect, whereas the strengthening
and Rilsch [6] for all but relatively high stress levels, initial and ductility increase effects are incorporated in uic and
estimates of equivalent uniaxial strain corresponding to lic respectively. That is, the effect of principal stress ratio
the experimental data can be made. If the observed on stress-strain response is primarily contained in (Ticand
curves, each for a different principal stress ratio, are then lti, and the effect of Poisson’s ratio is contained in the
normalized by the peak observed stress and the corres- variable liu.Thus if E, were constant, i.e. if the ductility
ponding equivalent uniaxial strain, and plotted, it is found increase and strength increase due to principal stress ratio
that they are nearly coincident and resemble uniaxial were proportional, a single curve, eqn (9), could represent
curves. This observation suggests using a single analytical the infinite variety of monotonic biaxial loading curves.
expression to describe the compressive loading portion However, in order to improve the accuracy of the
of the response (Fig. 5). The equation suggested by representation, the secant modulus Es was determined as
Saenz[22] for uniaxial compression a function of principal stress ratio from the data[3].
The values of El and E2 for a given principal stress
ratio are found as the slopes of the 01 vs llu and uz vs EZ”
curves given by eqn (9) at the current values of zlU and
L& J eic \Eir/ eZr, which are accumulated during the loading history
using eqn (8). The proposed model is compared with
is used for the biaxial case, where EO is the tangent experimental data[6] for biaxial compression, tension-
modulus of elasticity at zero stress, aie is the maximum compression and uniaxial compression in Fig. 6. One
compressive stress, eic is the equivalent uniaxial strain should expect the model to agree with the data on which it

// I / I _
lIC 4%” -‘iu
+‘iU u,+- _-__ -_----

Fig. 5. Modelfor plainconcreteunder cyclicload.

E2
= 0.52

= &52

, a = x204
-
----
Experimental
Proposed
(K, H, R)
Model 1
Strain , 0.001In./ I”.
Fig.6. Comparisonof modelwithmonotonicbiaxialloadtests.
142 DAVIDDARWINand DAVUI A. PECKNOLD

is based, however the model depends only on the load was cycled into or above the band, additional
parameters Eo, fe’, jt and EC”.Thus it should apply to permanent strain would accumulate if the peak stress was
different concrete mixes. In [3] comparisons are given maintained between cycles. For the model, the band is
with a completely independent set of experimental reduced to a single curve or “locus of common points”.
data[19], and the agreement is also very good for that As shown in Fig. 5, at low values of equivalent uniaxial
case. The model, at least the monotonic loading portion of strain unloading and reloading take place on a single line
it, appears to agree more closely with experimental results with slope Eo. At higher values of equivalent uniaxial
under a wider set of conditions than do the models strain, the reloading curve is represented by a straight line
proposed in [5] and [S]. The value of equivalent Poisson’s from the “plastic strain” point (0, ep), through the
ratio, v, is taken to be 0.2 in biaxial tension and biaxial common point. The unloading curve is approximated by
compression. A stress dependent value is used in uniaxial three straight lines: the first with slope Eo; the second
compression and tension_compression[3]. parallel to the reloading line; and the third with zero slope.
Load reversals initially follow the line with slope E.
CYCLICLOADING between the parallel unloading and reloading lines.
To realistically simulate cyclic response, which is the The location of the common points with respect to the
major emphasis of this work, a model should be capable envelope curve may be adjusted to control the number of
of accounting for (1) strength degradation (2) stiffness cycles to failure. As the locus of common points is
degradation and (3) hysteretic behavior under cycles of lowered, fewer cycles of load are required to intersect the
load. The model is based on the uniaxial data of Karsan envelope curve for a given maximum stress.
and Jima [21] since no biaxial data is available. Implicit in The energy dissipated for each cycle is controlled by
this procedure is the assumption that the same relation- the location of the turning point as shown in Fig. 5. The
ship or transformation between uniaxial and biaxial lower the turning point, the greater the energy dissipated
loading, via equivalent uniaxial strain, which was used for per cycle. The locations of common points and turning
monotonic loading, is acceptable for cyclic loading. There points, based on the experimental data, are shown in Fig.
is no experimental data to support such a contention, but 7. Simple expressions for these curves are given in [3].
it is felt to be a reasonable assumption. The model dissipates less energy than the experimental
Typical hysteresis curves from experimental specimen at cycles of low strain, which is a common
data[l8,211 and for the proposed model are shown in defect of many hysteresis models, but improves as the
Figs. 1 and 5. Straight line segments are used for the strain increases. Figures 8 shows a comparison of the
hysteresis loop. The size and shape of the loops are based model with experimental results for a load history run by
on several of the experimental findings of Karsan and Karsan and Jirsa[Zl]. The numerical model was unloaded
Jirsa [21], summarized below. An empirical relationship from the experimental envelope strains on each cycle.
was found between the strain eenon the envelope curve at
unloading, termed the “envelope strain”, and the residual
strain remaining at zero stress, epr termed the “plastic CRACKING
strain’. This relationship, The modeling of tensile cracking in concrete is
extremely important, yet will be outlined only briefly here,
$0.145 (zY+o.13 (5 since the techniques for incorporating this effect have
(10) become rather standard[24,10,11,13-171.
Cracks are assumed to form perpendicular to the
is incorporated into the model. principal tensile stress direction and are assumed to be
Karsan and Jirsa also found that there are a band of “smeared out” or evenly spaced over a region. When
points on the stress-strain plane that intluence the cracks form, the tangent modulus is reduced to zero in the
degradation of the concrete under continued cycles of principal tensile stress direction, and the tensile stress
load. If the load was cycled below the band, the previously carried in that direction is released. It is
stress-strain curve formed a closed hysteresis loop. If the important in the treatment of cyclic loading to recognize

Common Points

C
+% Eic 46” -Qiu

+y

Fig. 7. Control points for hysteresis loops.

 Analysisof cyclicloading of plane R/C structures 143

-1.2 I I I
Specimen Strength
= 0.85 f ‘c Specimen AC3-IO
-l.O- f’c = 5010 psi

- Test (Karsan, Jirsa)

---- Proposed Model

Fig. 8. Comparisonof modelwithuniaxialcyclicloadtests.

crack closing and crack formation in more than one had occurred. The structure tangent stiffness is updated at
direction[3]. every iteration, which may appear to be an unnecessarily
Several procedures for modifying the shear modulus expensive procedure. Many variations of the basic
in regions of tensile cracking have been numerical procedures are possible, differing, for example,
proposed[l2,17,20]. It appears to be unacceptable in in the frequency with which the structural stiffness is
some cases to simply reduce the shear modulus to zero updated or the number of iterative cycles carried out
when cracking occurs[ll, 131.Hand et al. [12]found that a within a load increment. The choice is largely a matter of
“shear retention factor” was necessary to preserve efficiency and economics, and it appears that for
numerical stability in some cases. In the proposed model, monotonic loading, a combined incremental-iterative
when the tangent modulus in one direction is set to zero a procedure, with stiffness update every few load incre-
reduced but non-zero value for the shear modulus ments, is a good compromise. For cyclic loading the
automatically results. possibility of divergence of the procedure due to loading
Reinforcing steel is modeled as an elastoplastic with a low value of stiffness, followed by unloading with a
strain-hardening material. Again, the technique for much larger stiffness led the authors to elect the safe, but
incorporating the steel is more or less standard by now, perhaps too expensive, method described previously in
and wilI not be discussed. Bond slip is not included in the which the stiffness is updated not only at every load
proposed model, although it could be if deemed neces- increment but for every cycle of iteration within the
sary. increment. Details of the numerical procedure are
available in [3].
NUMERICAL PROCEDURE AND COMPUTER PROGRAM The computer program was written in double-precision
The proposed constitutive model has been incorporated in FORTRAN IV for the IBM 360-75 system at the
into a finite element computer program[3] suitable for the University of Illinois. The program uses 240 kilobytes of
cyclic loading analysis of planar reinforced concrete core storage, and is capable of handling problems with
structures. Some brief comments will be made regarding about 25 or 30 elements, each with 3 x 3 or 3 x 4 meshes of
the numerical procedures employed and the experience of Gaussian integration points. About 20 variables must be
the authors in using this program. stored for each Gaussian integration point which accounts
The finite element employed in the program is a for the relatively large core storage required. Relatively
four-noded isoparametric quadrilateral, developed by coarse meshes are satisfactory because of the improved
Wilson et d.[23]. The element optionally includes four efficiency of the element due to the additional non-
extra non-conforming deformation modes to soften the conforming modes. The deformation history is tracked at
element and improve its behavior in flexure. The 9 (3 x 3) or 12 (3 x 4) locations within each element, so the
non-conforming modes were used in all the numerical extent of zones of cracking and yielding can be quite
examples reported herein. Element tangent stiffnesses are accurately determined with rather large elements. Typical
formed numerically, using Gaussian integration. execution times for monotonic loading cases with roughly
A combined incremental-iterative numerical 25 elements, 100 degrees of freedom, and the loading
procedure [3] is used. Increments of load or displacement divided into 10 increments, are on the order of 5-6min.
are applied and residual nodal loads representing the One cyclic loading case reported in [3] and [7j, involved 24
effect of inelastic behavior are redistributed to the elements, 133 degrees of freedom, 3 x3 Gaussian
structure until a specified convergence tolerance is integration meshes, and 2.5 cycles of load, each divided
reached. It was found necessary to use both a residual into 10 increments. CPU time for this problem was 20 min
force criterion and an incremental displacement criterion on the IBM 360-75. A great deal of this time can be
to achieve satisfactory convergence within a load or attributed to the updating of the structural stiffness at
displacement increment. Typically, the residual force every iteration, so that improvements in economy are
criterion governed, when substantial inelastic behavior certainly possible. However, a limiting factor in this type
144 DAVID DARWINand DAVID A. PECKNOLD

of analysis is the number of variables which must be mesh consisted of 20 elements with 12 (3 x 4) Gaussian
stored, effectively limiting the size of structure which can
integration points each. The load was applied in 9
be solved without using auxiliary storage and drastically increments and load control was used. Material properties
increasing cost. are given in [17]. Figure 10 shows the computed
load-deflection response of the wall, together with
EXAMPLES experimental results and the results obtained by
Two examples are presented here to indicate the type of Yuzugullu and Schnobrich[l7]. The results in [17] were
results which can be obtained with the numerical obtained with a finite element grid of 64 quadrilateral
procedure. The first is a shear wall-frame system[l’l, 241 elements, using an elastoplastic material idealization for
loaded monotonically, and the second is a cyclically the concrete, and special link elements to connect the
loaded shear wall [25]. shear wall to the frame. Yuzugullu and Schnobrich
The shear wall-frame system tested by Umemura, obtained a good match for the yield strength but their
Aoyama and Liao[24], designated Al, is shown in Fig. 9. model proved to be much too stiff thereafter. The
The test specimen consisted of two structural systems proposed model gives a low value for “yield” as
cast together and tested like a simply supported beam. It compared to the experimental but provides a reasonable
was loaded to approximately 100 metric tons, unloaded, match with the overall force-deflection response. Deflec-
and then reloaded to failure. The unloading cycle was not tions at other locations and comparisons of computed and
considered in the numerical solution. The finite element experimental crack patterns are given in [3].

COLUMN ;

I
CoLUMN i

A -A SECTION

B-8 SECTION

Fig. 9. Shearwall-framesystem A-l.

- - - - Proposed Model

Vertical Deflection, 8L,R, mm

Fig. 10. Computed and experimental load deflection curves for shear wall-frame system A-l.
 Analysisof cyclicloadingof planeR/C structures 145

A number of shear walls have recently been tested by results of as good quality. Second, the failure is flexural
the Portland Cement Association [25] as part of a program and so does not present a very demanding problem for
aimed at developing design criteria for structural walls simulation, at least for monotonic loading. However,
used in earthquake resistant designs. One such wall, some indication of the validity of the model for predicting
designed Bl, is shown in Fig. 11. cyclic behavior is provided by considering this example.
This particular wall has several shortcomings as an Experimentally, the load was applied in increments,
illustrative example. First, the wall is fairly slender and with three complete cycles at each increment, until
thus behaves essentially as a vertical cantilever. A much yielding occurred. Loading was then controlled by
simpler model, utilizing a discretization into vertical deflections in 1 in. increments, again with three complete
strips, together with beam theory, would probably yield cycles at each increment. Flexural cracking was observed
at a load of 26 kits. Yielding occurred at a load of
45.1 kips and a top deflection-of 0.56ins. Buckling of
vertical column reinforcement occurred by the end of the
3 in. cycles. The maximum load carried by the wall was
61 kips. The initial portion of the envelope of the cyclic
load-deflection curves is shown in Fig. 12. The envelope
did show strength degradation beyond about 4 in. deflec-
tion although this is not shown in Fig. 12. It is noted here
, that displacement control is a more reasonable loading

1 scheme to use since this more closely approximates the

experimental situation.
A mesh of 20 finite elements (4 horizontally by 5
vertically) was used with a 3 x 3 Gaussian integration. The
properties of the reinforcing bars (Fig. 11) were:
6 mm bars, fy = 75.5 x lo’ psi, E = 32.5 X 10”psi; No. 4
bars, fy = 65.2 x ld psi, E = 28.3 x O6psi; D3 ties, fy =
68.7 x lo’ psi, E = 33.0 x 106psi. The wall was cast in
{ ;;;;;;;+:;“2?;:;,‘;,, several lifts with slightly varying concrete properties in
each. For the purposes of the numerical solution, the
/ properties used in the lower half of the wall were
fc’ = 7685 psi, ft = 730 psi, EO= 4.08 x lo6 psi. The proper-
ties used in the top half of the wall were jc’ = 7030psi,
fr = 750 psi, EO= 4.04 x lo6 psi. Several runs were made to
illustrate the effects of load control, displacement control
and boundary conditions on the numerical solution.
Figure 12 shows the envelope of the cyclic loading curves

-r
obtained experimentally, together with four monotonic
Base Block loading curves obtained using the numerical solution
procedure. These correspond to both load and displace-
ment control solutions for two different boundary
IO’ conditions at the lower boundary, simulating the effect of
the base block. The “roller” boundary condition referred
Fig. 11. ShearwallB-l. to in Figs. 12 and 13 signifies that all of the boundary

__-D____-

1 Displacemeni
Control

P, 8 -

n
--- Fixed
----- Rollers

I
“0 I 2

Deflection 8, inches

Fig. 12. Computedandexperimentalmonotonicloaddeflectioncurvesfor shear wall B-l.

CAS Vol. 7, No. 1-J

146 DAVID DARWIN and DAVID A. PECKNOLD

Fig. 13. Computed and experimental cyclic load deflection curves for shear wall B-l.

nodes along the base, except one, are not restrained in the those presented. It is felt, however, that the proposed
horizontal direction. This boundary condition was investi- model does represent a worthwhile first step in treating
gated since there was as much as 0.2 in slip measured at cyclic loading. Before the procedure can be fully
the base construction joint during the experiment. evaluated, a much greater variety of situations must be
The closest match with the experimental envelope is studied including shear failures and cases of significant
obtained for the displacement control with relative motion strength degradation. Comparative studies of the various
allowed at the base (Fig. 12). The displacement control models proposed in the literature on a series of carefully
solutions fall below the corresponding load control selected problems and additional experimental informa-
solutions as is typically the case[4]. Two cyclic loading tion are needed before further advances in modelling
histories were run. These do not exactly correspond to the capability can be made.
experimental conditions since the many small cycles run
experimentally prior to the large excursions were too Acknowledgements-A portion of this paper is based on the first
expensive to simulate numerically. Instead, two and author’s doctoral dissertation[3], which was carried out in the
one-half cycles to displacements of +l in. were run under Department of Civil Engineering of the University of Illinois at
displacement control for the two boundary conditions Urbana-Champaignwith the aid of a National Science Foundation
Graduate Fellowship and a Scholarship from the Exxon Corpora-
discussed previously. These are shown in Fig. 13. The
tion. Partial financial support was also provided under NSF Grant
numerically obtained hysteresis loops appear larger than GI-29934. Portions of the numerical calculations given herein
the observed. The numerical model with the horizontal were performed on the Honeywell 635 System of the University
roller boundary condition reaches an unexpectedly low of Kanas Computation Center.
maximum load on the second positive cycle. This is
probably due to the fact that as more material resistance is
REFFXENCFS
lost the lack of restraint at the boundary is felt even more.
The true effect of the slip at the construction joint lies 1. A. C. T. Chen and W. F. Chen, Constitutive relations for
concrete. J. Engr. Mech. Dia., ASCE lOl(EM4), 465-481
somewhere between the two extremes shown in Fig. 13. (Aug. 1975).
These results emphasize once again that careful attention 2. J. J. Connor and Y. Same, Lecture notes on tinite element
must be paid to boundary and loading conditions in analysis of physically nonlinear systems. Inter. Symp.
attempting to simulate experimental results[3,4]. Crack Discrete Methods in Engng, Milan (Sept. 1974).
patterns determined numerically generally show quite 3. D. Darwin and D. A. Pecknold, Inelastic model for cyclic
good agreement with those observed experimentally biaxial loading of reinforced concrete. Cioil Engineering
although they sometimes indicate more extensive crack- Studies, SRS No. 409, University of Illinois at Urbana-
ing than observed[3]. Additionally, it is noted that in the Champaign, Urbana, Illinois (July 1974).
numerous examples studied to date, the omission of bond 4. D. Darwin and D. A. Pecknold, Analysis of RC shear panels
under cyclic loading. J. Struct. Div., AXE 102(ST2)(Feb.
slip has not appeared to be a major problem. 1976).
5. H. B. Kupfer and K. H. Gerstle, Behavior of concrete under
SUhiMARY AND CONCLUSIONS
biaxial stresses, J. Engr. Mech. Div., ASCE 99(EM4),852866
A finite element procedure for simulating response of (Aug. 1973).
cyclically loaded planar reinforced concrete structures 6. H. Kupfer, H. K. Hilsdorf and H. Riisch, Behavior of
has been presented. The procedure is intended as tool for concrete under biaxial stresses, J. ACI, Pm. M(8), 656-666
improving understanding of structural behavior, particu- (Aug. 1%9).
larly those properties relevant to earthquake-resistant T. C. Y. Liu, A. H. Nilson and F. 0. Slate, Stress-strain
response and fracture of concrete in uniaxial and biaxial
design.
compression. J. AU, Pm. 69(5), 291-295 (May 1972).
The two examples presented as well as other cases T. C. Y. Liu, A. H. Nilson and F. 0. Slate, Biaxial
presented previously [3] show reasonable agreement with stress-strain relations for concrete. J. Strct. Div., ASCE
experimental results and appear to indicate that the model 98(ST5), 1025-1034(May. 1972).
is capable of providing useful information. K. M. Romstad, M. A. Taylor and L. R. Herrman, Numerical
For monotonic loading many of the available numerical biaxial characterization for concrete. J. Engr. Mech. Div.,
models are capable of providing results comparable to ASCE lOO(EMS),935-948 (Oct. 1974).
 Analysis of cyclic loading of plane R /C structures 147

10. V. Cervenka and K. H. Gerstle, Inelastic analysis of for the determination of the behavior of a shear wall frame
reinforced concrete panels-I. Theory, IABSEPubL, 31, (ll), system. J. ACI, Proc. 76(7), 474-479 (July 1973).
31-45 (1971);Part II: Experimental verification and applica- 18. B. P. Sinha, K. H. Gerstle and L. G. Tulin, Stress-strain
tion, 32, (ll), 25-39 (1972). relations for concrete under cyclic loading. J. ACI, Proc.
11. A. Echiverria-Gomez and W. C. Schnobrich, Lumped 61(2), 195-211(Feb. 1964).
parameter analysis of cylindrical prestressed concrete reactor 19. L. J. M. Nelissen, Biaxial test& of normal concrete. Heron
vessels. Civil Engng Studies, SRS No. 340, University of Netherlands 18(l), (1972). -
Illinois at Urbana-Champaign, Urbana, Illinois (Dec. 1%8). 20. C. S. Lin and A. C. Scordelis, Nonlinear analysis of RC shells
12. F. R. Hand, D. A. Pecknold and W. C. Schnobrich, Nonlinear of general form. J. Struct. Dia., ASCE lOl(ST3), 523-538
layered analysis of RC plates and shells. J. Struct. Div., AXE (Mar. 1975).
99(Sl7), 1491-1505(July 1973). 21. I. D. Karsan and J. 0. Jima, Behavior of concrete under
13. Y. R. Rashid, Analysis of prestressed concrete pressure comuressive loadinns. J. Struct. Div.. ASCE 95QTl2). . II
vessels. Nuclear Engngand Design 7(4),334-344(Apr. 1968). 2543-2563(Dec. 1969).
14. W. C. Schnobrich, Finite element determination of nonlinear 22. L. P. Saenz, Disc. of equation for the stress-strain curve of
behavior of reinforced concrete plates and shells. Symp. concrete, by Desayi and Krishnan, .J. AU, Proc. 61(9),
Nonlinear Techniques and Behavior in Structural Analysis, 1229-1235(Sept. 1964).
Transport and Road Research Laboratory, Department of the 23. E. L. Wilson et al., Incompatible displacement models. In
Environment, U.K. (1974). Numerical and Computer Methods in-structural Mechanics
15. A. C. Scordelis, Finite element analysis of reinforced concrete (Edited bv S. J. Fenves. A. R. Robinson. N. Perrone and W. C.
structures, presented at Speciaky Conf Finite Element S&nob&h). Academic Press, New York (1973).
Method in Civil Engineering, l-2 June 1972.Sponsored by the 24. H. Umemura, H. Aoyama and M. Liao, Studies of reinforced
Bridge and Structural Division of the Engineering Institute of concrete shear wall and framed masonry shear walls. Res.
Canada and McGill University, Held at Montreal, Canada. Rep., University of Tokyo, Tokyo, Japan (June 1964).
16. M. Suidan and W. C. Schnobrich, Finite element analysis of 25. J. E. Carpenter et al., Structural walls in earthquake resistant
reinforced concrete. J. Stnccr. Diu., ASCE 99(STlO),2109- structures. Experimental Program Progress Rep., Construc-
2122 (Oct. 1973). tion Technology Laboratories, Portland Cement Association
17. 0. Yuzugullu and W. C. Schnobrich, A numerical procedure (Aug. 1975).