ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 94-S37

A New Simplified Model for Nonlinear RC Slabs Analysis

by Michael A. Reitman and David Z.Yankelevsky

A new technique for nonlinear analysis of reinforced concrete (RC) development of material models has passed several stages
slabs under various loading programs is developed. The technique from the introduction of simplified empirical moment-curva-
is based on nonlinear finite element (FE) analysis of a grid repre- ture relationships for RC cross-section7 to the use of piece-
senting the slab. An original formulation for the representative wise linear stress-strain curves with an artificially imposed
grid allows a good quality of the representation for the most
integration rule8 to the “layered” models incorporating more
coarse mesh of the grid axes. A refined nonlinear FE model of a
RC frame under complex loading was used. The model is based on refined constitutive relations for concrete and steel.9,10
a new modification of spatial anisotropic beam finite element and These methods are very powerful and have a high level of
refined constitutive models of concrete and steel-in-concrete under generalization and capabilities for consideration of various
complex loading history. The developed simplified method is complicated factors during the loading.
capable of predicting nonlinear behavior up to the failure for Two limitations restrain further development of these
various RC slabs subjected to monotonic and complex loading approaches. The first one is the excessive computation time
(including unloading and reloading branches). Slabs with different and computer resources required even for relatively simple
boundary conditions, geometry, reinforcement, and material prop- problems and modern high-speed computers. The second
erties can be analyzed. Membrane action and influence of nonideal difficulty relates to the behavior of plain and reinforced
boundary conditions are treated. Analytical results for various RC
concrete under multiaxial stress states. Considerable efforts
slabs tested by different investigators under different programs
show a good agreement with the experimentally observed data. have been made to establish the governing relationships;
however, a number of key points still remain under discus-
sion even for slabs under monotonic loading. At the same
Keywords: beam analogy; complex loading; finite elements; nonlinear
analysis; representative grid; stress-strain diagrams.
time a series of intensive investigations has led to refined
constitutive models for RC frames, which show excellent
agreement with experiments even under complex loadings. It
INTRODUCTION would be tempting to extend and apply this accumulated,
The first method for nonlinear analysis and design of rein- developed, and repeatedly validated material to nonlinear
forced concrete (RC) slabs was the yield-line method devel- analysis of RC slabs. In fact, developers of analytical tech-
oped simultaneously by Johansen1 and Gvozdev.2 The niques for slabs should examine the analytical methods
method appeared to be a convenient and powerful tool for already established for beams and frames, taking into
limit equilibrium assessment and for proper reinforcement of consideration additional complicated effects influencing
the slab according to the possible collapse mode. However, reinforced concrete nonlinear response.
it disregards the possible development of axial forces in the This, however, requires first of all an adequate beam
slab and the resulting membrane action which considerably analogy technique which would be able to represent well
contributes to the resistance mechanism and stiffness of an (both qualitatively and quantitatively) the force and
actual slab. Another serious disadvantage is that, being based displacement parameters of a plate at all the loading stages.
on equilibrium considerations, the yield-line method does
not provide any information on the development of the struc- Beam analogies for plate analysis
ture’s displacements and internal force redistribution during Early attempts to reduce the analysis of plates to a one-
the process of loading. Many attempts have been made to dimensional problem already existed in the nineteenth
extend the basic yield-lines theory to incorporate the axial century when Grashoff and Rankine proposed their formulae
force effects.3,4,5 Several simplified techniques were based on the consideration of the plate as strips in each of
proposed to obtain a load-deflection curve, based on the two directions and requiring equal deflections in the middle
yield-line approach.4,6 However, all these extensions of the strips. The simplified method of Marcus11 also
employed ideal assumptions, which are considerably employed the concept of two orthogonal strips of unit width
different from the actual conditions, thus, failing to properly
represent the nonlinear behavior of RC slabs.
A revolution in nonlinear analysis is attributed to using ACI Structural Journal, V. 94, No. 4, July-August 1997.
Received June 19, 1995, and reviewed under Institute publication policies. Copy-
numerical techniques and, especially, to the finite element right © 1997, American Concrete Institute. All rights reserved, including the making
method. Refined techniques have been developed employing of copies unless permission is obtained from the copyright proprietors. Pertinent
discussion will be published in the May-June 1998 ACI Structural Journal if
various types of finite elements and material models. The received by January 1, 1998.

ACI Structural Journal/July-August 1997 399

 The load distribution is calculated based on the strip stiffness
Michael Reitman specializes in development and implementation of nonlinear numer-
ical analysis and design algorithms for professional CAD software. He got his DSc properties. During nonlinear analysis these properties vary
from Reinforced Concrete Research Institute, Moscow, Russia, and did postdoctoral with the load magnitude. This would result in complications
research at Technicon, Israel Institute of Technology. Among his major areas of inter- in an iterative technique, and raise doubts about its conver-
est are nonlinear finite element analysis, analytical models of materials, design opti-
mization, and damage assessment techniques. gence. The assumption of the uniform load distribution in
each grid rectangle is a step forward comparing with the
ACI member David Z. Yankelevsky is Associate Professor of Civil Engineering at the nodal loads in most of the grid models. However, in
Technicon, Israel Institute of Technology, Haifa. He is a member of ACI committees
370, Short Duration Dynamics and Vibratory Effects, and 444, Models of Concrete nonlinear analysis of areas intersected by the yieldlines
Structures. His research interests include mechanical behavior of reinforced concrete, (zones of intensive deformations) this assumption would
earthquake engineering, and impact response of structures.
become a cause of error.
An interesting technique for development of an equivalent
representative grid was proposed by Rzhanitsyn.15 In addi-
with equal midpoint deflections. Both methods served for a
tion to orthogonal beams-strips, the grid contained diagonal
long time in the engineering practice and became a foundation
strips—two for each square cell of the axes. The stiffness
for a family of simplified techniques.
characteristics of the strips were established requiring a
Hillerborg12 proposed a lower bound design method compatibility between the in-plane deformation components
reducing the problem to a design of strips in two directions. of a grid cell of the grid and the plane stress state relations in
The method adopts plastic moment fields corresponding to the the corresponding plate. Further integration over the plate
yield-line concepts. Regions of different load distributions are depth allowed the calculation of the strip stiffnesses for the
separated by discontinuity lines (meaning the discontinuity in grid, which would represent the bending state of the plate
the assumed load distribution). Being a simple and efficient (the torsional stiffness of the strips had been neglected). The
tool, Hillerborg's strip method is very attractive for practical resulting grid, representing an isotropic plate under the
design. However, it is quite sensitive to the load distribution bending, contained the orthogonal strips, which width was
assumptions and needs additional recommendations for their about 75 percent of the distance between their axes (the cell
correct choice for different boundary conditions. size). The width of the diagonal strips was 2 times less.
Many suggestions have been made for simplified analytical Thus, only a part of the plate material contributed to the
techniques using the grid analogy approach, where grids and bending stiffness in each direction along the grid axes. The
lattices of various layouts and stiffness characteristics were rest of the stiffness was provided by the diagonal strips-
used to represent the plate for different practical purposes. beams; they also partially compensated the lack of the
However, the use of grid analogy for nonlinear simplified torsional stiffness in the grid joints. This approach, although
analysis is limited by two major factors. First, it is difficult analytically well founded, could not provide a reasonable
to represent well both the internal forces and the displace- representation for a coarse mesh. Since the stiffness distribu-
ments, or, at least, to achieve the same extent of disagree- tion in a grid cell had been derived on the basis of the rela-
ment for both. Attempts to improve the agreement between tionships of an infinitely small area, the smaller the cell was,
the predicted displacements and the displacements according the better was the quality of the plate behavior representa-
to the plate theory by increasing the number of the strips or tion. Thus, the refined mesh requirement became here even
their orientation13 increase at the same time the disagree- a more essential condition than it was in ordinary grids.
ment in terms of the internal forces, and vice versa. Second,
a relatively good quantitative agreement with the elastic
plate parameters of either the deflections or the internal SIGNIFICANCE OF THE RESEARCH
forces can be achieved only for a grid with a rather large To overcome the difficulties described above, a new
number of strips. A nonlinear technique based on such a grid formulation of the grid analogy model is presented herein
would be very laborious and inefficient even in comparison that allows, within reasonable calculations, adequate results
with a “layered” finite element technique. These two disad- concerning both the forces and the displacements for the
vantages are somewhat connected, and should be eliminated most coarse mesh. This formulation is based on intensive
in the development of a better model. analysis of various grid parameters and has been extended to
Fernando and Kemp14 proposed a technique, called the the nonlinear domain. Incorporation of a powerful finite
strip deflection method, which was intended to determine an element model for analysis of RC frames under complex
appropriate load distribution for its usage in the Hillerborg loadings has resulted in the development of a new simplified
method. A slab is subdivided into orthogonal strips, thus method for RC slabs analysis. The method was verified by
forming a set of rectangular areas. The intensity of the load comparisons with the actual behavior of various RC slabs
components in each direction is assumed to be uniform having different boundary conditions, and showed good
inside each area but can vary from area to area. An additional agreement with the experiment both for monotonic and
condition is imposed requiring compatibility of the elastic complex loading including unloading-reloading branches.
deflections at each intersection point of the strip center lines.
This condition allows the calculation of the load distribution AN OPTIMAL GRID
values in each area, which are then used in the Hillerborg FOR SIMPLIFIED PLATE ANALYSIS
strip method to design the slab reinforcement. Fernando and An introductory investigation was carried out in order to
Kemp noted the similarity between their proposed technique develop the most appropriate grid analogy as a foundation
and the analysis of a slab as a grid with zero torsional stiffness. for a nonlinear simplified analysis technique. The influence
They recommended the use of the grid analogy to calculate of various layout parameters on the quality of the represen-
the load. There was no intention to develop a technique for tation was investigated, including: the number and the layout
plate analysis, and no investigation was carried out of the orthogonal strips and their shape, the connection type
concerning an optimal layout of the strips. Two features of at the grid joint, and the appropriate load distribution. The
this technique make it inappropriate for the nonlinear analysis. intention was to find the grid which would simulate reasonably

400 ACI Structural Journal/July-August 1997

well the force-displacement state of an elastic plate with a
limited number of strips. Thus, the foundation for a
nonlinear technique may be laid for further improvements
and modifications after which analysis of actual RC slabs
would demonstrate the method’s accuracy and efficiency.

Grid axes layout and load distribution

The common layout of equal strips with equidistant axes
always carries an intrinsic disadvantage, especially when
nonlinear analysis is concerned. Using such a layout, inde-
pendent of the strain distribution and of the expected failure
mode of a plate, only a considerable number of strips would
guarantee a proper representation of the nonlinear behavior.
On the other hand, an irregular layout of grid axes, which
would a priori better correspond to a general distribution of
main zones of nonlinearity in the plate, can provide the same
(or even a higher) accuracy of representation using a grid of
a smaller number of strips.
A good guide for an optimal axes layout can be obtained
by considering the yield-line scheme of the plate. Since the
yield-lines represent the location of highly strained zones, it
would be preferable that the grid axes in orthogonal direc-
tions intersect at a yield-line. Thus, the highly strained grid
joints lie along the yield-lines. Such a scheme is shown in
Fig. 1. The edges of two adjacent strips are assumed to be
located in the middle between their axes.
An appropriate load distribution is of particular impor-
tance for the correct representation of a nonlinear plate. The
authors suggest that the yield-line scheme of the plate at
failure should also govern the load distribution technique.
Indeed, since the calculated yield-line pattern represents a
collapse mode, which is close to the actual mode observed in Fig. 1—Grid layout based on the yield-line scheme and the
tests, a load distribution in the nonlinear domain should conventional cutout of the strips in two directions.
agree with the yield-line pattern. Moreover, even in the
elastic range the yield-line based load distribution appears to
be most suitable. Observation of the load distribution calcu-
lated by the strip deflection method14 shows that it practi-
cally describes the developing yield-lines. Therefore, the
load distribution according to the yield-line scheme of the
plate (Fig. 2) provides a good representation of elastic forces
distribution and may be assumed adequate for nonlinear
analysis. The assumed load distribution in the proposed
simplified model is, to a certain extent, the return to the Hill-
erborg strip method approach with the yield-lines serving as
the load discontinuity lines. However, here the analysis of
the whole representative grid is the objective, not the consid-
eration of a few separate characteristic bands. The uniformly
distributed load is concentrated along a strip axis (not neces-
sarily a strip center line) exactly according to the yield-line
scheme. Fig. 2 shows how the load, gathered from a rectan-
gular subarea, results in uniformly distributed load on an axis
in the corresponding direction. The load, gathered from a
triangular subarea (near a yield-line intersection) becomes a Fig. 2—Load distribution on the grid axes according to the
linearly distributed load on an axis. yield-line scheme of the plate.

Connection at the grid joint without any connection at the joints thus forming two separate
The proper choice of the connection type at a grid joint is systems of the orthogonal strips, and a grid with deflection
the pivot question which determines, to a large extent, the compatibility only at the joint (called hereafter “grid with
quality of the analytical model. A special study of this factor nails”). Parameters such as strip layout and number and the
influence has been carried out by means of comparative load distribution were also varied, and their effect on the
analyses of various simply supported and clamped square quality of the plate representation was studied. The study
and rectangular elastic plates under a uniformly distributed compared the deflection and internal bending moments at the
load. Three alternatives were considered for every grid: a central point of the plate and, in plates with clamped edges,
grid with a fully clamped connection at the joint, a grid it also compared bending moments at the edge centers.

ACI Structural Journal/July-August 1997 401

Table 1—Divergence of analysis results, provided by various representative grids, from the plate theory
solution for elastic square plate with all edges clamped, percent
Full connection No connection “Nail”-type
Load Parameters 3x3 5x5 7x7 3x3 5x5 7x7 3x3 5x5 7x7
Center deflection –54.0 –46.0 –33.0 +2.52 +37.2 +36.9 +34.5
Uniform Moment in
distribution midpoint –53.9 –42.9 –32.2 +18.4 +48.4 +47.9 +46.3
Moment at edge –29.6 –22.5 –17.4 –19.1 +11.2 +16.9 +15.8
Center deflection –59.3 –48.1 –34.2 +87.0 +99.0 +102.3 +18.1 +30.5 +32.0
According to Moment in –56.6 –43.0 –31.7 +90.9 +114.8 +123.8 +32.9 +45.2 +46.3
Y-lines midpoint
Moment at edge –35.4 –27.3 –20.8 +56.9 +60.3 +61.2 –4.65 +8.80 +11.0

Table 1 presents the results of this comparative study for a line scheme at the failure, and the “nail”-type connections at
square slab with all the edges clamped. This problem is the joints were chosen as the most favorable solution having
symmetrical with three main parameters characterizing the a potential to serve as a foundation for a simplified nonlinear
stressed-deformed state. The axes distribution based on the technique. Certain modifications still should be introduced
symmetric yield-line scheme coincides in this particular case in order to improve the quantitative agreement of the
with the evenly distributed mesh. Two variants of load distri- provided results with the elastic plate solution. The objective
bution on the strips were under investigation: the uniform of these modifications is to minimize the error caused by
distribution on all the strips (half the load in each direction) choice of the representative grid.
and the proposed yield-line based distribution (Fig. 2).
Table 1 shows that the grids with fully clamped joints An improved cutout of strips
provide a stiffer behavior for a smaller number of strips The authors’ attention was drawn to a consistent contra-
because of an inadequately high torsional stiffness. The diction in the proposed grid (Fig. 1 and 2). This is the contra-
results approach gradually to the theoretical plate state with diction between the strip cutout and the load distribution on
increasing of the number of strips, though the improvement them at the vicinity of the yield-line intersections. There are
of the representation quality is quite slow. The agreement zones where the width of the acting load, being gathered to
with the plate state parameters remains unacceptable even act on a strip, is greater than the width of the strip itself.
for the 7x7 mesh of strips for both types of load distribution. Compare, for example, strip 2 in Figures 1 and 2. While its
The grid without connections actually represents a system edge is located in the middle between the axes 1 and 2, the
of strips that are cut out of the plate and analyzed separately, width of the load, being gathered from the triangular loading
neglecting any interaction between them. The results in this area near the yield-line intersection, reaches the whole
case strongly depend on the assumed load distribution. The distance between these axes. Such a formulation, therefore,
fair agreement obtained under the uniform distribution (half leads to overloading of these parts of the structure.
the load in each direction) proves again the validity of the To eliminate this contradiction, it was proposed to cut out
characteristic strips approach the foundation for many the strips in a slightly different way to satisfy a condition that
simplified engineering techniques. Unfortunately, this the strip would always be at least as wide as the width of the
model is of no use for purposes of nonlinear analysis since it load area acting on it. The strip axis scheme remains
does not allow any redistribution process to occur between unchanged. If the overloading occurs, the width of a strip is
different zones in the plate. increased in its overloaded central part so that the strip edge
The “grid with nails” solution appears to be more flexible coincides with the adjacent axis. The new cutout of the strips
than the plate state in all the analyzed cases (because of the is shown in Fig. 3. When axis 2 in X-direction (the longer
lack of constraints at the joints). Different tendencies can be strips) is intersected by Y-axis 1, the width of the X-strip 2 is
observed, however, for the two load distributions. When the increased so that its edges coincide with adjacent X-axes 1 and
load is uniformly distributed over all the grid strips, the 3. This leads to the corresponding decrease in the adjacent
results approach slowly but gradually to the plate solution strips width (X-strips 1 and 3). After the intersection of X-axis
with refinement of the axes mesh. The apparent inconsis- 2 by Y-axis 5 the widths of all the X-strips return to their initial
tency concerning the moment at the edge center is attributed values. The same technique is applied to Y-strips 2 and 4.
to its averaging over the strip width (the distribution over the This change in the strip width distribution has surprisingly
strip width is assumed to be uniform). The total moment caused a considerable improvement in the analytical results,
reaction at the clamped edge gradually decreases with the which yielded a much better agreement with the plate
increase of the number of strips. The load distribution behavior (Table 2). In all the examined cases it has led to a
according to the yield-line scheme (Fig. 2) yields the opposite substantially improved prediction of the plate state values
tendency. The more refined the mesh is, the more flexible is simultaneously for all the comparison parameters. In many
the grid solution, and the greater is the disagreement with the studies performed by the authors the coarse grid “with nails”
theoretical plate solution. This qualitative tendency (which consisting of the strips of variable width always appeared to be
has been consistently obtained in all the analyzed cases, for the best plate representation, providing a quantitative agree-
both square and rectangular plates with various boundary ment with the plate parameters. It therefore has been chosen as
conditions) represents a peculiar case: the most coarse mesh the representative grid for further nonlinear analysis.
provides the best representation of the plate state. Extensive validation of the proposed model on various
Due to this peculiarity, the most coarse mesh of the axes, actual RC slabs has shown its ability for a proper prediction of
subjected to the load distribution corresponding to the yield- the nonlinear behavior of slabs with various boundary condi-

402 ACI Structural Journal/July-August 1997

Table 2—Divergence of analysis results, provided by
various representative grids with different cutout
strips, from the plate theory solution for an elastic
square plate with all edges clamped, percent
3 x 3 grid with “nail”-type
connections in nodes
Strips of constant Strips of variable
Load Parameters width (Fig. 1) width (Fig. 3)
Center deflection 18.05 0.04
According Moment in midpoint 32.9 –4.60
to Y-lines
Moment at edge –4.65 0.08

Table 3—Comparison of the total moment volumes

for a square simply supported slab carrying
uniformly distributed load*
Method V/Vmin
Minimum weight solution 1.00
Strip deflection method:
3 strips in each direction 1.45
4 strips in each direction 1.39
5 strips in each direction 1.37
Yield line theory,
isotropic reinforcement 1.74
no top steel
Hillerborg’s strip method,
5 bands in each direction 1.60
Proposed representative grid
3 strips of variable width in each direction 1.15
*Fernando and Kemp (1978)

tions up to the failure—both under monotonic and complex

loading programs. The scheme with the strips of constant
width (Fig. 1) was also checked in the nonlinear analyses of Fig. 3—Modified cutout of the strips in two directions
RC slabs. It has shown a much worse agreement with the corresponding to the assumed load distribution
experimental data for slabs in bending, underestimating their
ultimate load and predicting a much greater flexibility, and has A refined nonlinear finite element had been developed16,17
failed to predict the nonlinear behavior when axial forces for analysis of RC planar and spatial frames under simple and
develop within the slab (see Example 1). The scheme with the complex loading programs (including combinations and super-
strips of variable width (Fig. 3) has described well various positions of cyclic, reversed, nonproportional loadings, etc.).
cases with controlled in-plane load and horizontal restraints. A new constitutive model of the concrete behavior under
Another assessment of the quality of the elastic plate repre- a general complex loading program had been suggested by
sentation by the proposed grid may be obtained on the basis of an extensive analysis of test data. It is based on a refined
the moment volume. The total moment volume is calculated analytical description of complete stress-strain curve of
as the integral of the principal moments' absolute values over concrete which main parameters at any loading stage depend
the entire area of the plate. It can be considered as an integral on the corresponding parameters at the previous stage and
characteristic of the internal force distribution and a relative the change in the loading program. Thus, the information
measure of the total weight of the required reinforcement. about the initial properties of concrete and their accumulated
Fernando and Kemp14 presented the moment volume values change during complex loading can be represented at any
for a square slab on simple supports carrying uniformly stage. Refined dependencies for the concrete diagram
distributed load, which had been calculated according to parameters under various loading programs had been estab-
different methods (Table 3). The analysis of the proposed grid lished by extensive processing and numerical simulation of
with 3 strips of variable width in each direction yields a experimental data. These dependencies are capable of repre-
moment volume which is 15 percent above the minimum senting many complicated effects that occur in concrete,
weight solution. The improvement concerning this parameter hysteretic behavior, strength deterioration, etc. Being intro-
is encouraging regarding the proposed method since no duced into the constitutive material model they have allowed
special attention was paid to its optimization. Obviously, it is good prediction of the concrete behavior under different
a result of an overall improvement in plate representation. loading programs.
A number of additional factors influencing concrete
NONLINEAR FINITE ELEMENT MODEL behavior in a reinforced concrete structure are taken into
Once the appropriate representative grid is chosen, it may account by means of a proposed “diagram transformation”
be incorporated in a nonlinear model for its analysis as a RC technique. The idea is that a certain factor influence may be
frame. The more refined and adequate the model is represented in an integrated manner in the form of the stress-
(concerning proper prediction of the reinforced concrete strain curve transformation. A separate study based on
response), the more accurately the proposed simplified tech- extensive numerical simulation of observed structural
nique would follow the RC slab nonlinear behavior. behavior in a wide range of cases had allowed formulation of

ACI Structural Journal/July-August 1997 403

particular transformation techniques for the concrete stress- general complex loading program. The analytical descrip-
strain diagram. The use of those transformed diagrams in the tion of concrete stress-strain diagram consists of parabolic
global nonlinear analysis of RC structures allows consider- parts with parameters continuously dependent on previous
ation of the given factor influence during a loading. Efficient load history. The unloading and reloading branches are also
diagram transformation technique had been developed to parabolic, exhibiting hysteresis effects and strength
consider such effects as strain gradient, confinement, local decrease. Particular effects, which influence concrete
change of concrete strength near supports, strain rate, and behavior in the structure, are considered by means of a
time-dependent effects. proposed diagram transformation technique. Usage of the
Influence of cracking on the structural behavior is consid- stress-strain diagram after special transformations in
ered using a special formulation of stress-strain diagrams nonlinear analysis allows the consideration of effects such as
both for concrete and reinforcing steel after the crack appear- strain gradient, confinement of concrete, and partly time
ance. Those diagrams relate local stresses in a cracked dependent effects. Cracking, bond deterioration, and tension
section with strains averaged over the element length. They stiffening are taken into account by a special “steel-
are obtained by the transformation of the corresponding embedded-in-concrete” diagram. It is obtained for any
original diagrams of the unreinforced concrete and the bare complex loading by transformations of the bare steel stress-
steel. Introducing those transformed diagrams of the strain diagram.18
“cracked concrete” and of the “steel-embedded-in-concrete” The analytical model was validated repeatedly on various
into a global nonlinear analysis provides a good simulation reinforced concrete structures from numerous experimental
of the structural behavior. investigations under different loading programs. Even for
The use of such diagrams also allows the elimination of quite complex loading it showed fine agreement of the results
the known conflict. Now we can carry out the analysis in with experimentally observed behavior at all the loading
terms of averaged strains, which directly define the displace- stages. A computer program based on the developed analytical
ments, but the equilibrium equations are formulated in the model has been successfully exploited for over four years
most dangerous cracked sections. providing proper predictions of nonlinear behavior of RC bar
A finite element implementation of the analytical model is structures from columns and continuous beams to multistory
based on a developed 3D anisotropic beam finite element spatial frameworks under simple and complex loadings.
with an additional displacement field “due to shear.” The In the present paper this nonlinear FE model of RC frame
finite element has nine degrees of freedom in each node. Six has been used to analyze the grid representation for predic-
are of the traditional spatial beam element, and three addi- tion of PC slab nonlinear behavior. Peculiarity of the
tional degrees, one along each axis, represent the integrated proposed formulation is the “nail”-type connections in the
“due to shear” additions to the linear nodal displacements. joints of the grid requiring common deflections only. The
The introduced additional “shear” field relates to the tradi- connections of this type are introduced into the structure’s
tional displacement field by the extended material matrix global stiffness matrix in the form of constraint equations
and also by consideration of overall boundary conditions. relating the displacements degrees of freedom for corre-
Such a separation of the “bending” and the “shear” degrees sponding nodes of the orthogonal beams (strips).
of freedom eliminates a need to impose initially relations
between the classic and the additional displacements. Those VALIDATION OF THE PRESENT MODEL WITH
relations differ from those described in the elasticity theory EXPERIMENTAL DATA
and are difficult to predict in the case of complex loading Validation of the proposed simplified nonlinear model
history of a RC element. The proposed approach suggests has been accomplished through analysis of various PC
recalculation of these relations for different stages of a slabs with different boundary conditions, which had been
loading depending on its history. This allows a much better tested by different authors. Specimens with controlled in-
consideration of the shear influence on the displacements plane load or horizontal restraints were also investigated to
and the internal force distribution under various loading check if the “membrane effect” would be predicted properly.
programs. The most complex verification of the model’s abilities has
The material matrix of a generally anisotropic material is been carried out for a PC rectangular slab with nontrivial
calculated by numerical integration over the cross-section horizontal restraints tested under a complex loading
area. The element stiffness matrix was derived assuming the program including unloading and reloading. Three demonstra-
material matrix coefficients’ variation along the element tive examples follow.
length. The off-diagonal coefficients of the material matrix
contain information about the actual neutral axis location at Example 1
any stage of loading and, thus, allow the consideration of the Twin RC square slabs on simple supports were tested by
additional axial forces in statically indeterminate RC struc- Aghayere and MacGregor.19 The slabs were 1.83 m (6 ft) in
tures. The distorted analytical scheme influence is treated by span, 65 mm (2.55 inch) thick, and had two layers of rein-
the element stiffness matrix transformation taking into forcement in each direction. The layer areas were 225 mm2/m
account the consistent geometric stiffness matrix in special (0.106 in.2/ft) and 257 mm2/m (0.121 in.2/ft).
form (due to the existence of the additional displacement The material properties were the following: fc = 32.3 MPa
field). The authors supposed that the repeatedly validated (4.68 ksi), Eb = 23,100 MPa (3350 ksi), ft = 2.8 MPa
ability of the model to treat properly the axial force-related (0.406 ksi), fsy = 504 MPa (73 ksi), fsu = 670 MPa (97.2 ksi),
effects in statically indeterminate frames would serve a foun- Es = 197,300 MPa (28,615 ksi), and εsu = 0.0455.
dation for the “membrane effect” consideration in RC slabs The first slab was subjected to transverse load by nine
with horizontal restraints. concentrated forces distributed over its area (an equivalent
Extensive analysis of numerous test data had suggested a transverse pressure was calculated at each step). The
new constitutive model for concrete behavior under a observed ultimate bearing capacity was 58.6 kPa (8.5 psi),

404 ACI Structural Journal/July-August 1997

while the maximum center displacement of 95 mm (3.74 in.)
was reported (the complete load-deflection curve was not
presented). Analysis of the proposed 3x3 grid has resulted in
the ultimate load of 53.5 kPa (7.76 psi) and the corre-
sponding center deflection of 102 mm (4.02 in.). The agree-
ment between the experimental values and the analysis
results may be considered reasonable.
Another, identical, slab was subjected to a combined
action of in-plane and transverse loads. The in-plane
compressive load acted in one direction and was uniformly
distributed along the slab edges. It was applied incrementally
up to the level of 765 kN/m (52.42 klb/ft) and maintained
constant as the transverse load increased from zero to the
maximum value. One essential factor related to the test setup
should be noted. The in-plane load acted along the mid-plane
of the slab prior to applying the transverse load. As the edge
of the slab rotated, the line of action moved toward the
bottom of the plate due to a geometry of a half-cylinder used
to apply the load to the specimen. This effect has been
analytically considered by means of an additional condition,
namely, the change of the location of the in-plane action line Fig. 4—RC slab behavior under combined action of in-plane
is assumed to be equal at any loading stage to the upward and transverse loads (Aghayere and MacGregor [1990]).
displacement of the slab edge.
Nonlinear analysis of the representative 3x3 grid has led to
the slab behavior prediction which compares well to its
actual behavior (Curve 1 in Fig. 4). The analysis has discovered
that the effect of downward displacement of the in-plane
load action line is most important and the analysis is quite
sensitive to this factor. An attempt to analyze the slab,
assuming the in-plane load always acting along the mid-
plane, has resulted in the predicted ultimate load of 22.5 kPa
(3.26 psi), which considerably underestimates the actual
bearing capacity.
For comparison, the two test slabs have been analyzed
using the same grid with strips of constant width (Fig. 1).
The predicted ultimate load, for the first slab subjected to
transverse load only, was 40.5 kPa (5.87 psi), and the corre-
sponding center point deflection 128 mm (5.04 in.). The
predicted nonlinear behavior of the second slab (Fig. 4,
curve 2) does not agree with the experimental results.
The test performed by Aghayere and MacGregor,19 with
constant in-plane load, relates to wall-beams subjected to
transverse load rather than to slabs. An in-plane load field
appears in reinforced concrete slab because of horizontal
restraints at its edges. Both the values of the in-plane forces Fig. 5—Load-deflection behavior of simply supported RC
and their trajectories vary during loading, thus making the slabs tested by Muspratt (1970).
behavior more complicated and difficult to predict. The
influence of this additional “membrane effect” factor Load-deflections prediction of the slab up to failure, using
determines, to a large extent, the bearing capacity and a 3x3 grid and assuming ideal simple supports with no hori-
ultimate deformability of the slab. zontal restraints, is shown in Fig. 5 (curve 0). The agreement
at the initial stage of the loading is reasonable but then the
Example 2 observed behavior becomes stiffer than the predicted. The
A square reinforced concrete slab was tested by Muspratt.20 experimental ultimate load is considerably underestimated in
The slab span was 3.05 m (10 ft) and the effective depth of the the analysis. It is noteworthy that the analytical results are in
cross-section was 57 mm (2.25 in.). The slab reinforcement good agreement with a theoretical failure load calculated
area was 259 mm2/m (0.122 in.2/ft) in each direction. The using the yield-line technique (the dashed line in Fig. 5).
properties of the concrete and the reinforcing steel: fc = 28 MPa Muspratt could not explain the increased bearing capacity of
(4.06 ksi), fsy = 310 MPa (45 ksi), fsu = 450 MPa (65.3 ksi), the slab and made an attempt to attribute it to a “tensile
and Es = 185,000 MPa (26,831 ksi). membrane action” in simply supported slabs.
Pneumatic loading to simulate a uniformly distributed The authors suppose, however, that the actual cause is the
load condition was achieved by inflating an enclosed test setup, which failed to provide the ideal representation of
membrane acting on the underneath side of the slab. A steel a “rolling simple support.” Instead, a partial horizontal
support frame along the slab perimeter had supposed to restraint was developed along the slab perimeter (especially
represent simple supporting conditions. at the higher load levels), which affected its behavior. An

ACI Structural Journal/July-August 1997 405

 restraints has provided results which compare well to the
observed data (see Example 1). Obviously, due to the better
representation of the assumed boundary conditions by the
test setup, the influence of possible partial horizontal
restraints was practically eliminated and the observed data
was “free” from this effect.
Two main conclusions may be drawn from the present
example. First, even a slight horizontal restraint is of particular
importance, and it may considerably affect the behavior of a
reinforced concrete slab both quantitatively and qualita-
tively. Second, the proposed simplified technique of
nonlinear analysis is an analytical tool capable of considering
this factor properly in cases where boundary conditions
differ from the ideal ones.

Example 3
There exist many experiments on RC slabs, with various
schemes of horizontal restraints, that provide a lot of data for
analytical models validation. However, in most of the
reported tests the boundary conditions were not represented
in a sufficiently accurate manner. Analyses of many tested
slabs have shown that the load-deflection curve of the actual
behavior often lies somewhere between analytical curves
corresponding to full horizontal restraint and to no-restraint
conditions. This represents the differences between ideal and
actual horizontal restraint conditions. An extensive analysis
of test data from various investigations dealing with hori-
zontal restraints of RC slabs was carried out by Christiansen
Fig. 6—Rectangular RC slab with horizontal restraints under and Frederiksen.21 Beyond a state-of-the-art review, they
complex loading program (Christiansen and Frederiksen also processed and analyzed main results of these tests. In
[1983]). their own experiment they proposed one of the most optimal
types of restraint, which reduced considerably an error intro-
indirect indication to this suggestion emerges from the fact duced by test setup.
that the predicted in-plane outward displacements in the A slab tested by Christiansen and Frederiksen21 was rect-
centers of the edges were from the very beginning consistently angular in plan with dimensions 2.2x2.5 m (7.22x8.20 ft)
greater than those observed in the test. and 40 mm (1.57 in.) thick, with vertical supports along its
The partial horizontal restraint assumption has been edges. A test zone—an interior slab of 1x2 m (3.28x6.56 ft)
checked by a numerical experiment. First, ideal horizontal in plan—also had vertical supports along its perimeter. The
supports were imposed along the slab perimeter, that, loading on the test interior slab consisted of 16 equal point
evidently, provided much stiffer results than the observed loads uniformly distributed along its area.
behavior (Fig. 5, curve 1). Then these ideal in-plane supports The horizontal restraints were provided by the external parts
were replaced by distributed elastic springs. The springs’ of the slab surrounding the test area. This zone (of the same
stiffness, which provided a solution close to the solution thickness) had additional reinforcing bars, which bounded the
obtained for ideal supports (difference less than 5 percent test area and were also located along the outer perimeter of the
concerning any parameter), was referred to as the “full hori- slab (Fig. 6). Both the test area and the edge zone were rein-
zontal restraint” stiffness. forced by two layers of square net corresponding to 102 mm2/
A number of analyses were then carried out considering m (0.048 in.2/ft), the material properties were: fc = 13.5 MPa
different values of the stiffness of the spring supports, (1.96 ksi), Ec = 20,300 MPa (2944 ksi), fsy = 330 MPa (47.9
modeling different levels of partial horizontal restraint. It has ksi), fsu = 362 MPa (52.5 ksi), and εsu = 0.099.
been found that the experimentally observed behavior of the The test program included initial loading deep into the
slab can be well predicted taking into account the existence nonlinear domain, and then almost complete unloading
of a slight horizontal restraint which is quantitatively only followed by reloading.
0.25 percent to 0.5 percent of the full restraint (curves 2 and The nonlinear analysis of the slab has been carried out
3 in Fig. 5, respectively). using a 3x5 grid representing the test zone. The first attempt
Observation of the test setup data, reported in reference to assume ideal horizontal restraints along the perimeter of
20, shows that there may exist a partial restraint of this the internal slab led to a much stiffer behavior than that
nature. It can be attributed to friction and other resistance observed in the experiment (Fig. 6, curve 1). The cause was
effects in the contact between the slab surface and the that in-plane displacements along the test zone perimeter had
support frame. These effects were not constant during the not been fully restrained. Actually, the partial horizontal
loading (the experimental curve compares to the predictions restraints at the perimeter had been determined by stiffness
with different restraint stiffness values at various stages). It properties of the edge zone.
is noteworthy that in Reference 1, where rolling supports Four additional strips, representing the edge zone, were
were used to represent simple supporting, the nonlinear anal- added to the grid located around the test zone grid. A special
ysis assuming the complete absence of the horizontal condition was imposed which required compatibility of in-

406 ACI Structural Journal/July-August 1997

plane displacements of their nodes with those of the corre- 5. Braestrup, M. W. “Dome Effects in RC Slabs: Rigid-Plastic Analysis,”
sponding nodes of the grid. The nonlinear analysis of the Journal of the Structural Division, ASCE, Vol. 186, 1980, pp. 1237-1253.
modified scheme has allowed the calculation of the results 6. Technical Manual: Structures to Resist the Effects of Accidental
Explosions, Dept. of the Army, the Navy and the Air Force, No. 5-1300,
that compare excellently to the actual behavior (Fig. 6,
Washington, D.C., 1969.
curve 2). After the proper analytical scheme was developed, 7. Jofriet, J. C., and McNeice, G. M. “Finite Element Analysis of Rein-
the nonlinear model for IQC frames, in combination with the forced Concrete Slabs,” Journal of the Structural Division, ASCE, Vol. 97,
proposed grid analogy, has led to a fine prediction of the No. ST3, 1971, pp.785-806.
slab’s nonlinear behavior, not only under monotonic loading 8. May, I. M., and Ganaba, T. H. “Full Range Analysis of Reinforced
but also under unloading and reloading. Concrete Slabs Using Finite Elements,” International Journal Num. Meth.
in Engrg., V. 26, No. 4, 1988, pp. 973-985.
CONCLUSIONS 9. Lin, C. S., and Scordelis, A. C. “Nonlinear Analysis of RC Shells of
General Form, Journal of the Structural Division, ASCE, V. 101, No. ST3,
A new simplified nonlinear model for the analysis of
1975, pp. 525-538.
reinforced concrete slabs with different boundary conditions
10. Hinton, E.; Rahman, H. H. A.; and Zienkiewicz, O. C., “Computa-
under monotonic and complex loading has been developed. tional Models for Reinforced Concrete Slab Systems,” IABSE Colloquium
The model is based on the proposed grid analogy and the non- on Advanced Mechanics of Reinforced Concrete, Final Report, Delft, 34,
linear finite element model of RC frames under complex 1981, pp. 303-313.
loading. The model has been validated on experiments with 11. Marcus, H., Die vereinfachte berechnung biegsamer platten, second
various slabs subjected to different conditions, and has shown edition, Springer-Verlag, Berlin, 1929.
good agreement with the test data at all loading stages. 12. Hillerborg, A. “A Theory of Equilibrium for Reinforced Concrete
The proposed grid with the strips of variable width and the Slabs,” Betong, V. 41, 1956, pp. 171-182.
load distribution based on the yield-line scheme, and the 13. Cope, R. J., and Clark, L. A. Concrete Slabs, Elsevier Applied
Science Publishers Ltd., Essex, England, 1984.
“nail”-type connections in the joints allows calculation of
14. Fernando, J. S., and Kemp. K. O. “A Generalized Strip Deflection
results, which agree well with the plate theory (in elastic Method of Reinforced Concrete Slab Design,” Proceedings of the Institute
formulation), using the most coarse meshes (3 to 5 strips in of Civil Engineers, V. 65, No. 2, 1978, pp. 163-174.
each direction). 15. Rzhanitsyn, A. R., Structural Mechanics, Vyshaya Shkola, Moscow,
The combination of this modified grid analogy with the 1982.
highly developed nonlinear finite element model of RC frame 16. Karpenko, N. I.; Mukhamediev, T. A.; and Reitman-Sapozhnikov,
has resulted in the simplified, rather efficient technique, capable M. A. “Finite Element Analysis of Reinforced Concrete Bar Struc-
of predicting the RC slab behavior under various loading tures,” Structural Mechanics and Analysis of Structures, Gosstroy,
Moscow, No. 2, 1990, pp.55-61.
programs. The membrane action has been taken into account.
17. Reitman, M. A., “Analysis of Statically Indeterminate Reinforced
The proposed technique has allowed not only the consideration
Concrete Bar Structures, Considering Nonlinear Properties of the Materials
of effects related with nonideal boundary conditions and partial and Histories of Short-Term Loadings,” DSc Thesis, Reinforced Concrete
horizontal restraint, but also the investigation of their influence Research Institute, Moscow, 1990.
on the actual RC slabs behavior. 18. Reitman, M. A., “Predicting Behavior of Cyclically Loaded RC
Structures,” Discussion, Journal of Structural Engineering, ASCE, V. 118,
REFERENCES No. 2, 1993.
1. Johansen, K. W., Brudlinieteorier. Jul. Gjellerups Forlag, Copenhagen, 19. Aghayere, A. O., and MacGregor, J. G. “Tests of Concrete Plates
1943 (Translated by Cement and Concrete Association, London, 1962). under Combined In-plane and Transverse Loads,” ACI Structural Journal,
2. Gvozdev, A. A., Analysis of Bearing Capacity Using the Limit Equi- V. 87 No. 6, Nov.-Dec. 1990, pp. 615-622.
librium Method, Gosstroyizdat, Moscow, 1949. 20. Muspratt, M. A., “Destructive Tests on Rationally Designed Slabs,”
3. Wood, R. D., Plastic and Elastic Design of Slabs and Plates, Thames Magazine of Concrete Research, V. 22, No. 70, 1970, pp. 25-36.
& Hudson, London, 1961. 21. Christiansen, K. P., and Frederiksen, V. T. “Experimental Investiga-
4. Krylov, S. M., Internal Forces Redistribution in Statically Indetermi- tion of Rectangular Concrete Slabs with Horizontal Restraints,” Materials
nate Structures, Stroyizdat, Moscow, 1964. and Structures, V. 16, No. 93, 1983, pp. 179-192.

ACI Structural Journal/July-August 1997 407