181

MACROSCOPIC MODELING FOR NONLINEAR ANALYSIS

OF RC STRUCTURAL WALLS

ALFONSO VULCANO
Associate Professor / Dipartimento di Strutture
Universita della Calabria
87036 Arcavacata di Rende (Cosenza), ITALY

ABSTRACT

In this paper attention is focused on RC shear-wall models which, based on a macroscopic

approach, can be efficiently incorporated in a practical nonlinear analysis of multistorey RC
buildings with structural walls. An overview of the above models is given. In order to evaluate
the effectiveness and reliability of selected wall models, analytical and experimental results with
reference to isolated walls and to a frame-wall structure are shown.

INTRODUCTION
Reinforced Concrete (RC) shear walls are very effective in providing resistance and stiffness
against lateral loads induced by earthquake and/or wind. Extensive research, both analytical and
experimental [1-15], has been carried out in order to clarify, and then to simulate, the hysteretic
behaviour of both isolated and coupled RC walls and of RC frame-wall structural systems.
Recent research has significantly improved understanding of the inelastic behaviour of such
structures, providing helpful information for the development of suitable analytical models.
Many analytical models have been proposed for predicting the nonlinear response of RC
structural walls. They can be classified into two broad groups: (a) detailed models, derived
using mechanics of solids, which are based on a detailed interpretation of local behaviour
(microscopic approach); (b) models based on a simplified idealization, which are capable of
predicting a specific overall behaviour with reasonable accuracy (macroscopic approach).
Even though the Finite Element Method offers a powerful analytical tool for simulating the
nonlinear behaviour of RC structures according to the microscopic approach [ 1 ], difficulties
arise because of the lack of completely reliable basic models and the complexities involved in
the analysis. But, very important, the computation is generally very time-consuming and
requires a large storage: thus, in practice the use of microscopic FE models is restricted to the
analysis of isolated or coupled walls.
On the other hand, macroscopic wall models, because of their relative simplicity, can be
efficiently adopted even for nonlinear analysis of multistorey RC structural systems. However,
attention must be paid to the reliability of macroscopic models, whose validity is limited to
cases such that the conditions on which the derivation of the models themselves is based would
be satisfied.
 182

In this paper the features and limitations of macroscopic wall models already proposed are
considered in the light of experimentally observed behaviour. Then, suggestions are made for
improving the effectiveness and/or reliability of selected wall models.

MACROSCOPIC WALL MODELS

Equivalent Beam Model

A current modeling for simulating the hysteretic behaviour of a RC structural wall considers the
generic wall member replaced at its centroidal axis by a line element, which, for instance in a
frame-wall structural system, is connected by rigid links to the girders. Commonly a one-
component beam model is adopted. This model consists of a flexural elastic member with a
nonlinear rotational spring at each end to account for the inelastic behaviour of critical regions;
the fixed-end rotation at any connection interface can be taken into account by a further
nonlinear rotational spring.
To simulate the propagation of the inelasticity adequately, Takayanagi and Schnobrich [2]
discretized each wall member into a suitable number of short segments. However, this requires
large computational effort, which could be a limitation in the nonlinear analysis of multistorey
structures. Moreover, to account for the inelastic shear deformation effects in a coupled wall
system, Takayanagi et al. [3] introduced additional plastic hinges at the ends of each beam
element.
The main limitation of a beam model lies in the assumption that rotations occur around
points of the centroidal axis of the wall. Thus, important features of the observed behaviour
(i.e., fluctuation of the neutral axis of the wall cross-section, rocking, etc.) are disregarded and
the consequent effects in a structural system (i.e., outriggering interaction with the frame
surrounding the wall in a frame-wall structure, etc.) are not accounted for adequately.

Equivalent Truss Model

Another modeling technique represents the wall as an equivalent truss system. On the basis of
experimental test results Hiraishi [4] introduced a non-prismatic truss member whose sectional
area was determined according to the stress along the height of the boundary column in tension.
However, the use is at the present limited to a monotonic loading, because of the difficulties in
defining the structural topology and the properties of the truss elements under a cyclic loading.
M ultiple-V ertical-Line-Elem ent Models

Three-Vertical-Line-Element Model (TVLEM): This model was originally proposed by

Kabeyasawa et al. [7] on the basis of the experimentally observed behaviour of a seven-storey
RC frame-wall structural system. The model, shown in Fig. la, idealized a wall member as
three vertical line elements with infinitely rigid beams at the top and bottom floor levels: two
outside truss elements represented the axial stiffnesses Ki and K 2 of the boundary columns,
while the central element was a one-component model consisting of vertical, horizontal and
rotational springs concentrated at the base with stiffnesses Kv, KHand K^, respectively. The
model was intended to simulate the deformation of the wall member under a uniform
distribution of curvature. The hysteretic behaviour of the elements constituting the wall model
was simulated by adopting empirical assumptions on the basis of the experience coming from
experimental tests.
The axial-stiffness hysteresis model (ASHM) in Fig. lb was proposed by the above
authors in order to simulate the axial force-deformation relationship of the three vertical line
elements. Many empirical assumptions were made. For instance, as regards the skeleton curve,
when the axial force changed direction from compression to tension, the stiffness was reduced
to 90% of its initial value under compression (Kt=0.90Kc); once tensile yielding occurred at
point Y, the stiffness was reduced to 0.1% of the above initial value (Kh=0.001Kc). Two
parameters, a=0.9 and (3=0.2, were introduced to describe the unloading stiffness degradation
 183

Kr and the stiffness hardening point P, respectively. However, as already shown in Ref. [11], a
value a<0.687 should be assumed in order to obtain in any case a realistic unloading path.
The origin-oriented hysteresis model (OOHM) in Fig. lc was used for both the rotational
and horizontal springs at the base of the central vertical element. The stiffness properties of the
rotational spring were defined by referring to the cross-section of the central panel only,
disregarding in this manner the displacement compatibility with the boundary columns. With
regard to the stiffness properties of the horizontal spring, the trilinear skeleton curve was
defined on the basis of empirical formulae.
Although the original TVLEM contains many empirical assumptions, its merit is that it can
be considered as the early proposed macroscopic model accounting for the fluctuation of the
neutral axis of the wall cross-section: this is particularly important for an adequate simulation of
phenomena experimentally observed for RC frame-wall structures (e.g., outriggering effect due
to the interaction of the walls with the surrounding frames).

<J>m
FO RCE. F Y (D y t, F y ) M (D m’ Fm )
(t e n s io n ) v / _ _ __
/ic ------------- K h

tf/] / y _ ( D x ,F ,„ - F y) _

KCVh fU f f y/ / j/ ' D E F(eOx tRe nMs ioA nT )I O N . D

il
it
Y ' (D y e , -F y ) £
f ' P (D p - F p)

Y " ( 2 D y c , -2 F y ) / K r = K c (D y, / D m) «

D p = D y C + p ( D x - D yc)

(a) Idealization of a Wall Member

(Kabeyasawa et al. [7]) (b) Axial-Stiffness-Hysteresis Model (Ref.[7])

(c) Hysteresis Model for Horizontal and (d) Modification of the Model in Fig. lb
Rotational Springs in Fig. la (Ref. [7]) (Vulcano and Bertero [10-11])

Fig. 1 - Three-Vertical-Line-Element Model and Its Modification

 184

Modified Three-Vertical-Line-Element Model (MTVLEM); In order to limit as much as

possible the empirical assumptions, in Refs. [10-11] the TVLEM was modified by replacing the
ASHM with the two-axial-element-in-series model (AESM) shown in Fig. Id: the element 1
was a one-component model to represent as a whole the axial stiffness of the column segments
in which the bond was still active, while the element 2 was a two-component model to represent
the axial stiffness of the remaining segments of steel (S) and cracked concrete (C) for which the
bond was almost completely deteriorated. The AESM was intended to idealize the main features
of the actual hysteretic behaviour of the materials and their interaction (yielding and hardening
of the steel, concrete cracking, contact stresses, bond degradation, etc.). Even though refined
constitutive laws could be assumed for describing the hysteretic behaviour of the materials and
their interaction, in the mentioned references very simple assumptions (i.e., linearly elastic
curve for the element 1 , bilinear curve and linearly elastic curve in compression neglecting
tensile strength, respectively for S and C components of the element 2) were made in order to
carry out a first check of the effectiveness and reliability of the proposed model.
Moreover, it was observed that, in order to account for the diplacement compatibility
between the central panel and the boundary columns of the wall, a softening skeleton curve
should be assumed for describing the response of the rotational spring of the TVLEM.

Multi-Component-in-Parallel Model (MCPM) : In Ref. [11] it was emphasized the

opportunity of obtaining a more refined description of the flexural behaviour of the wall from
one or both the following approaches: (a) modification of the geometry of the wall model to
gradually account for the progressive yielding of the steel; (b) use of more refined laws, based
on the actual behaviour of the materials and their interaction, to describe the response of the two
elements in series constituting the AESM.
In Ref. [12] a new wall model was proposed by following both the above approaches
(Fig. 2). The flexural response of a wall member was simulated by a multi-uniaxial-element-in-
parallel model with infinitely rigid beams at the top and bottom floor levels: the two external
elements represented the axial stiffnesses Ki and K2 of the boundary columns, while two or
more interior elements, with axial stiffnesses K3,...., Kn , represented on the whole the axial
and flexural stiffnesses of the central panel. A horizontal spring, with stiffness KH and
hysteretic behaviour described by the OOHM already mentioned, simulated the shear response
of the wall member. The relative rotation A(j>m was intended around the point placed on the
central axis of the wall member at height ch. A suitable value of the parameter c could be
selected on the basis of the expected curvature distribution along the inter-storey height h: for
instance, 0<c<l, if the curvature presented tha same sign along h.
A modified version of the AESM above shown was proposed to describe the response of a
uniaxial element (Fig. 2b). Analogously to the original AESM, the two elements in series were
still representative of the axial stiffness of the column segments in which the bond remained
active (element 1 ) and those segments for which the bond stresses were negligible (element 2).
Unlike the AESM, also the element 1 consisted of two parallel components to account for the
mechanical behaviour of the uncracked concrete (C) and the steel (S); a suitable law for the
dimensionless parameter X defining the length of the two elements provided with an accurate
description of the measured tension-stiffening effect. Refined constitutive laws were adopted to
idealize the hysteretic behaviour of the materials and the tension-stiffening effect.
Exactly, the stress-strain relationship proposed by Bolong et al. [16] was assumed for
cracked concrete, because accounting for the contact stresses due to the progressive closure of
cracks (Fig. 3a). The stress-strain relationship in Fig. 3b and a set of rules under a generalized
load history, both proposed in Ref. [12], were adopted for uncracked concrete. The stress-
strain relationship originally proposed by Giuffre and Pinto [17] and later implemented by
Menegotto and Pinto [18] was adopted to describe the hysteretic response of the reinforcing
steel (Fig. 3c); in order to avoid the storage of all parameters required for a generalized load
history to retrace all previous reloading curves which were left incomplete, a set of simple rules
was used.
Under monotonic tensile loading the tension-stiffening effect was taken into account by
calculating the value of the dimensionless parameter X such that the tensile stiffness of the
 185

uniaxial model in Fig. 2b would be equal to the actual tensile stiffness of the uniaxial RC
member which was intended to be idealized:
/ (1-X)h ‘S ~e.s

| e c,A c+ EsA! + E A / ~ h em

where EctAc and ESASare the axial stiffnesses in tension of the concrete and of the steel,
respectively, while the ratio zJZm of the steel strain in a cracked section to the current average
strain for the overall member was evaluated by the empirical law proposed by Rizkalla and
Hwang [19]. Under cyclic loading it was assumed that, along an unloading path from a tensile
stress state, the value of X was kept constantly equal to the value corresponding to the
maximum tensile strain previously attained; if this maximum strain was exceeded during a
tensile reloading, the value of X was updated as for the case of monotonic tensile loading.

A w m = w m . w m .j

A<f>m- *^m-l A v m = v m . v m -i

element 1

—element 2

(a) Idealization of a Wall Member (b) Uniaxial Element Model

Fig. 2 - Multi-Component-in-Parallel Model (Vulcano et al. [12])

Fig. 3 - Material Constitutive Laws Adopted for Model Components in Fig. 2b

 186

Modified Multi-Component-in-Parallel Model (MMCPML As it will be shown in the

next section, the accuracy in predicting the flexural response of the wall by the MCPM is
excellent when the constitutive laws in Fig. 3 are adopted for the modified-AESM components
in Fig. 2b, even by assuming few uniaxial elements. However, the above constitutive laws are
very sophisticated; to improve the effectiveness of MCPM without renouncing a reasonable
accuracy, the use of simplified (more schematic) constitutive laws could be suitable as well.
This is confirmed in recent studies conducted by Fajfar and Fischinger [14], who introduced
simplified hysteretic rules to describe the response of both the vertical and horizontal springs
(Fig. 4). A further variant of the MCPM was applied in a study of Fischinger at al. [15], who,
in order to reduce the uncertainty in the assumption of a suitable value for the parameter c, used
a stack of many elements which were placed one upon the other.

J
(a) Vertical Spring Rules (b) Horizontal Spring Rules
Fig. 4 - Modified Hysteretic Behaviour Rules for Springs of the Wall Model in Fig. 2a
(Fajfar and Fischinger [14], Fischinger et al.[15])

RESPONSE OF MACROSCOPIC MODELS

In order to check the effectiveness and reliability of macroscopic wall models, in the course of
previous studies [10-13] an extensive numerical investigation was carried out with reference to
isolated walls tested under monotonic and cyclic loadings by Vallenas et al. [5] at the University
of California at Berkeley (UCB) (Fig. 5) and to a 7-storey RC frame wall structure, which was
tested, within the framework of a joint U.S.-Japan research project [6], at Tsukuba (full-scale
pseudo-dinamic testing) by Kabeyasawa et al. [7] and at the UCB (l/5th-scale dynamic testing,
on earthquake simulator) by Bertero et al. [8] (Fig. 6). Herein some meaningful results are
shown and critically discussed.
In Figs. 7a and 7b analytical results obtained by adopting, respectively, the TVLEM and
the MTVLEM, are compared with the experimental results. Exactly, the analytical curves in
Fig. 7a have been obtained by assuming for the degradation parameter a the value 0.9
suggested in Ref. [7] and the limit value 0.687. Both the analytical curves, but particularly that
corresponding to a=0.9, exhibit mechanical degradation. On the contrary, the improvement
obtained by the MTVLEM, even by assuming very simplified laws for the AESM components
in Fig. Id, is evident, also in comparison with the response obtained by a F.E. analysis.
A considerable improvement in the prediction of the wall flexural behaviour can be attained
by the MCPM. This is clearly shown in Figs. 8a and 8b, in which analytical and experimental
curves are compared for specimens 5 and 6 in Fig. 5b, subjected to monotonic and cyclic
 187

loadings, respectively. All the analytical curves represented by a full line in Fig. 8a have been
obtained by assuming 4 vertical uniaxial elements (n=4) and different values of the parameter c.
Apart from a slight discrepancy, the correlation of the experimental curve and the analytical
curve corresponding to c=0.4 can be considered good; it becomes excellent when adopting the
values c=0.4 and n=8 (see dashed line). The correlation is very good also with reference to the
curves in Fig. 8b, where the analytical curve represents the displacement-controlled flexural
response. Analogous results, which are omitted for sake of brevity, have been obtained for
specimens 3 and 4 in Fig. 5a. However, it should be noted that under high shear stresses,
likewise TVLEM, difficulties are met for accurately describing by MCPM experimentally
observed flexural and shear displacement components. This is clearly shown in Fig. 8c, with
reference to specimen 5, by assuming different values of the parameter c and a relatively
detailed discretization (n=8) .
Finally, in order to show the suitability of the MCPM for incorporation in the nonlinear
analysis of multistorey structures, the test structure in Fig. 6 has been modeled as shown in
Fig. 9, that is by idealizing all the structural members as line beam elements (LGM model) or,
alternatively, the wall only by the MCPM (WGM model); a suitably reduced value of the initial
flexural stiffness of the beam elements, idealized by a two-component model, has been
assumed. In Fig. 10 experimental and analytical curves are compared with reference to the
above structure. Exactly, in Figs. 10a and 10b shears (Vy, Vw) and overturning moments
(My, Mw), respectively, evaluated at the base of the overall structure or of the wall are shown
against the roof horizontal d i s p l a c e m e n t T h e experimental curves were obtained by Bertero
et al. [8] as envelope of the results of dynamic testings, whereas the analytical curves have been
obtained by a static nonlinear analysis for a uniform loading distribution.
As it can be observed, both the analytical models, but particularly LGM, underestimated
the ultimate strength of the structure, even though a uniform loading distribution has been
assumed for both of them: this, as noted in Ref. [13], should imply some overestimation of the
ultimate strength, if the actual loading distribution in the dynamic testing and some mechanical
degradation due to cyclic loading are taken into account. On the other hand, the underestimation
of the ultimate strength by both the analytical models can be mostly ascribed to the fact that the
strain rate to obtain the constitutive curves of the materials, on which the mechanical
characteristics of the analytical models have been settled, was considerably lower than that
recorded during the dynamic testing. However, it should be noted that the LGM model, in
which the rotation of the central wall is described as occurring around the axis of the wall itself,
gives rise to a wrong description of the rocking and consequent spatial-interaction effects.
Thus, an appreciable increase of the axial compressive force in the central wall has been
observed when using WGM model, whereas the axial force for the wall remained practically
equal to the gravity axial force when using the LGM model. A further consequence has been
that very different values of the ductility demand have been obtained for both longitudinal and
transverse girders by adopting the two analytical models.

2134 i

0.64 4V
1 t0.644V 0 . 522v| jo .5 2 2 V

434 kN I |4 3 4 kN 299 kN | |2 9 9 kN
+ ~ \ Vn

91 4 run
\ 0.104V
914 mm
1 V .0 9 7 V
1181 mm
I r §1
1000 . 1200 u n it in mm

(a) Specimens 3,4 (b) Specimens 5,6

Fig. 5 - Test Walls (Vallenas et al. [5]) Fig. 6 - Test Structure (Bertero et al. [8])
 188

Fig. 7 - Analytical and Experimental Curves for Specimen 4

if / / ' /
//>
li
w

A. (mm)
lex.

(a) Specimen 5 (b) Specimen 6 (n=4; c=0.4) (c) Top Displacement Comps.

Fig. 8 - Analytical and Experimental Curves for Test Walls in Fig. 5b

Fig. 9 - Pseudo-Three-Dimensional Model of the Test Structure in Fig. 6

 189

(a) Base Shears for Overall Structure and (b) Base Overturning Moments for Overall
Wall Vs. Roof Displacement Structure and Wall Vs. Roof Displacement
Fig. 10 - Analytical and Experimental Curves for Test Structure in Fig. 6

CONCLUSIONS

Features and limitations of models idealizing a RC structural wall have been discussed in the
previous sections. As mentioned, wall models based on a macroscopic approach are more
effective than microscopic FE models for the purpose of incorporation in the nonlinear analysis
of multistorey structures. Equivalent beam and truss models present many limitations. Indeed, a
beam model, although suitable for its simplicity, is not capable of simulating the fluctuation of
the neutral axis of the wall cross-section, giving rise to a wrong description of the interaction
with the other structural members. On the other hand, the implementation of a truss model
meets with difficulties in defining the properties of the truss elements, particularly under cyclic
loading.
Multiple-vertical-line-element models prove to be the most suitable for the purpose
mentioned above, particularly because they are capable of accounting for fluctuation of the
cross-section neutral axis. As shown, a multi-component-in-parallel model accurately predicts
the flexural response, even assuming the minimum of uniaxial elements (n=4) with the
advantage of a limited computational effort. The accuracy in predicting the flexural response
depends, rather than assuming a greater number of uniaxial elements, on the choice of a suitable
value of the parameter c according to the expected distribution of curvature along the wall
member. By assuming c=0.4 for the test walls examined in this paper and refined constitutive
laws for the uniaxial elements, the correlation of analytical and experimental results has been
excellent. In order to improve the effectiveness of the model, the use of simplified, yet
reasonably accurate, constitutive laws can be adequate as well; moreover, the uncertainty in
defining the parameter c can be overcome by placing more elements one upon the others.
However, under high shear stresses, the OOHM gives only approximate description of the
shear hysteretic response; thus, the prediction of the shear and flexural displacement
components becomes difficult. Therefore, improvement of the wall model is needed by revising
the OOHM and/or by introducing some relation between flexural and shear responses, which
have been independently described. Further improvements can be pursued by making the wall
model capable of simulating other observed phenomena at the present ignored (e.g., fixed-end-
rotation at the base of the wall, etc.) and by a better calibration of the parameters affecting the
response of the models (e.g., c) on the basis of integrated analytical and experimental research.
 190

REFERENCES
1. American Society of Civil Engineering, "Finite Element Analysis of Reinforced Concrete",
State-of-the-Art Report, ASCE, New York, 1982.
2. Takayanagi T. and Schnobrich W.C., "Nonlinear analysis of coupled wall systems",
Earthquake Engineering and Structural Dynamics". Vol. 7, 1979.
3. Takayanagi T., Derecho A.T. and Corley W.G., "Analysis of inelastic shear deformation
effects in reinforced concrete structural wall systems", Nonlinear Design of Concrete
Structures. CSCE-ASCE-ACI-CEB Int. Svmp.. University of Waterloo, Canada, 1979.
4. Hiraishi H., "Evaluation of shear and flexural deformations of flexural type shear walls",
Procs. 4th Joint Tech. Coord. Committee. U.S.-Japan Coop. Earth. Research Program.
Building Research Institute, Tsukuba, Japan, 1983.
5. Vallenas J.M., Bertero V.V. and Popov E.P., "Hysteretic behaviour of reinforced concrete
structural walls", Report No. UCB/EERC-79/20. Earthquake Engineering Research
Center, University of California, Berkeley, 1979.
6. U.S. Members of JTCC Group on RC Building Structures.,"U.S.-Japan research: seismic
design implications", J. of Struct. Eng.. ASCE, Vol. 114, No.9, Sept. 1988.
7. Kabeyasawa T., Shioara H. and Otani S., "U.S.-Japan cooperative research on R/C full-
scale building test - Part 5: discussion on dynamic response system", Procs. 8th
W.C.E.E.. Vol. 6, S. Francisco, 1984.
8. Bertero V.V., Aktan A.E., Charney F.A. and Sause R., "U.S.-Japan cooperative
research program: earthquake simulation tests and associated studies of a l/5th-scale
model of a 7-story reinforced concrete test structure", Report No. UCB/EERC-84/05.
University of California, Berkeley, 1984.
9. Fajfar P. and Fischinger M., "Nonlinear seismic analysis of RC buildings: implications of
a case study", European Earthquake Engineering. No. 1, 1987.
10. Vulcano A. and Bertero V.V., "Nonlinear analysis of R/C structural walls", Procs. 8th
E.C.E.E.. Vol. 3, Lisbon, 1986.
11. Vulcano A., and Bertero V.V., "Analytical models for predicting the lateral response of
RC shear walls: evaluation of their reliability", Report No. UCB/EERC-87/19,
University of California, Berkeley, 1987.
12. Vulcano A., Bertero V.V. and Colotti V., "Analytical modeling of R/C structural
walls", Procs. 9th W.C.E.E..Vol. VI, Tokyo-Kyoto, Japan, 1988. See also: Colotti V.
and Vulcano A., "Behaviour of RC structural walls subjected to large cyclic loads" (in
Italian), Procs. Giornate A.I.C.A.P. (Italian Association of Reinforced and Prestressed
Concrete Structures), Vol. 1, Stresa, Italy, 1987.
13. Vulcano A. and Colotti V., "Analytical modeling of RC frame-wall structural systems",
Procs. 9th E.C.E.E.. Vol. 10-B, Moscow, 1990.
14. Fajfar P. and Fischinger M., "Mathematical modeling of reinforced concrete structural
walls for nonlinear seismic analysis", EURQDYN'90. Euro. Conf. on Struct. Dyn.,
Bochum, Germany, 1990.
15. Fischinger M., Vidic T., Selih J., Fajfar P., Zhang H.Y. and Damianic', "Validation of a
macroscopic model for cyclic response prediction of R.C. walls", Procs. 2nd Int. Conf.
on Computer Aided Analysis and Design of Concrete Structures. Zell am See, 1990.
16. Bolong Z., Mingshun W. and Kunlian Z., "A study of hysteretic curve of reinforced
concrete members under cyclic loading", Procs. 7th W.C.E.E.. Vol. 6, Instanbul, Turkey,
1980.
17. Giuffre A. and Pinto P.E., "The behaviour of reinforced concrete under strong-intensity
cyclic loading" (in Italian), Giomale del Genio Civile. N.5, 1970.
18. Menegotto M. and Pinto P.E., "Method of analysis for ciclycally loaded reinforced concrete
plane frames including changes in geometry and nonelastic behavior of elements under
combined normal force and bending", Procs. IABSE Svmp. on Resistance and Ultimate
Deformabilitv of Structures Acted on bv Well-Defined Repeated Loads. Lisbon, 1973.
19. Rizkalla S.H. and Hwang L.S., "Crack prediction for members in uniaxial tension", ACI
Journal. Nov.-Dec., 1984.