Sunteți pe pagina 1din 11

Nonlinear nite element analysis of concrete structures

using new constitutive models


Taijun Wang, Thomas T.C. Hsu
*
Department of Civil and Environmental Engineering, University of Houston, Houston, TX 77204-2162, USA
Received 24 January 2001; accepted 1 August 2001
Abstract
Nonlinear nite element analysis was applied to various types of reinforced concrete structures using a new set of
constitutive models established in the xed-angle softened-truss model (FA-STM). A computer code FEAPRC was
developed specically for application to reinforced concrete structures by modifying the general-purpose program
FEAP. FEAPRC can take care of the four important characteristics of cracked reinforced concrete: (1) the softening
eect of concrete in compression, (2) the tension-stiening eect by concrete in tension, (3) the average (or smeared)
stressstrain curve of steel bars embedded in concrete, and (4) the new, rational shear modulus of concrete. The pre-
dictions made by FEAPRC are in good agreement with the experimental results of beams, panels, and framed shear
walls. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Beams; Computer code; Concrete; Constitutive laws; Finite element; Shear modulus; Shear walls; Structures; Reinforced
concrete; Constitutive model
1. Introduction
Since Ngo and Scordelis [1] applied the nonlinear
nite element method to reinforced concrete beams, this
technique has developed into an important tool for the
analysis of the complex concrete structures, such as
nuclear containment vessels, oshore platforms, long-
span bridges, shear walls in buildings, etc. This tech-
nique also played a signicant role in the research of
concrete structures by helping to understand the for-
mation and propagation of cracks and the mechanism
and process of failure. Furthermore, numerical ex-
periments can also be performed on structures that are
dicult to model in laboratory study [2]. Future deve-
lopment of the nonlinear nite element method lies
primarily in the improvements of the constitutive models
of materials. This is especially true for concrete struc-
tures where the constitutive relationships of concrete
and the embedded steel bars are still evolving [3], e.g. in
shear and torsion.
Experimental research to establish the constitutive
relationships of concrete structures has been an on-
going activity at the University of Houston over the past
13 years. Nearly a hundred full-size panels have been
tested using the universal panel tester [4]. This test fa-
cility, equipped with a servo-control system, allowed
researchers to conduct strain-controlled tests [5]. Based
on extensive panel tests, two behavioral models were
developed for the analysis of structures subjected to
shear: the rotating-angle softened truss model (RA-
STM) and the xed-angle softened truss model (FA-
STM).
The RA-STM assumes cracks will develop in a di-
rection parallel to the principal compressive stresses in
concrete elements, and the cracks will ``rotate'' to follow
the principal stresses over the entire loading history.
Because shear stresses cannot develop along the cracks,
this model requires only three constitutive laws: (1)
concrete in compression, (2) concrete in tension, and (3)
Computers and Structures 79 (2001) 27812791
www.elsevier.com/locate/compstruc
*
Corresponding author. Fax: +1-713-743-0421.
E-mail address: thsu@uh.edu (T.T.C. Hsu).
0045-7949/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S0045- 7949( 01) 00157- 2
steel bars embedded in concrete. In contrast, the FA-
STM assumes cracks will develop along the direction of
principal compressive stresses applied at initial cracking,
and the cracks will be xed at this angle thereafter. Be-
cause shear stresses can develop along such cracks, FA-
STM requires an additional constitutive law of concrete
in shear.
The advantage of FA-STM over RA-STM was that
FA-STM was capable of taking into account the con-
crete contribution V
c
, induced by the shear stresses along
the cracks. The disadvantage of FA-STM, however, was
the involvement of a complex empirical constitutive law
of concrete in shear. Fortunately, this weakness was
overcome by the recent derivation of a rational, yet
simple, shear modulus of concrete [8]. As a result, the
solution algorithm of FA-STM is now as simple as that
of RA-STM. In this paper, we describe the incorpora-
tion of the set of four FA-STM constitutive laws [68]
into the computer code FEAPRC, and the comparison
of FEAPRC predictions with the experimental results of
three types of concrete structures.
2. Application of xed-angle cracks in nite element
method
2.1. Coordinate system in xed-angle model
Fig. 1 shows the coordinate system xy, 12 and
1
/
2
/
. Coordinates xy and 12 represent the directions
of the reinforcing steel bars and the principal applied
stresses at initial cracking, respectively. Angle a
1
is the
xed angle between these two coordinates. The coordi-
nate 1
/
2
/
is the post-cracking principal coordinate of
applied stresses, which has an angle a
/
1
with respect to
the xy coordinate.
The transformation of stresses from the xy coordi-
nate to the 1
/
2
/
coordinate is given by the matrix
equation as follows:
r
1
/
r
2
/
s
1
/
2
/
_
_
_
_
_
_
=
c
2
s
2
2cs
s
2
c
2
2cs
cs cs c
2
s
2
_
_
_
_
r
x
r
y
s
xy
_
_
_
_
_
_
(1)
where c = cos(a
/
) and s = sin(a
/
). The principal angle a
/
is obtained as
tan(2a
/
) =
2s
xy
r
x
r
y
(2)
The angle a
/
may be the maximum principal angle a
/
1
or
the minimum principal angle a
/
2
. From Mohr circle these
two angles can be dierentiated by the following equa-
tion:
tana
/
i
=
s
xy
r
i
r
y
in which i = 1; 2 (3)
In the xed angle theory, the 12 coordinate is the
coordinate determined by the principal applied stresses
at initial cracking. With further loading after cracking,
the principal coordinate of applied stresses will rotate
from the 12 coordinate to the 1
/
2
/
coordinate, and the
principal angle will change. Since the cracking angle is
assumed to be xed at the initial cracking angle a
1
, it is
necessary to transform the stresses in the 1
/
2
/
coordi-
nate back to the stresses in the 12 coordinate through
an angle b = a
/
1
a
1
. This stress transformation can be
accomplished by the following equation:
r
1
r
2
s
12
_
_
_
_
_
_
=
c
2
s
2
2cs
s
2
c
2
2cs
cs cs c
2
s
2
_
_
_
_
r
1
/
r
2
/
0
_
_
_
_
_
_
(4)
where c = cos(b) and s = sin(b).
2.2. Material stiness matrix
Before cracking, concrete can reasonably be assumed
as an elastic isotropic solid. After cracking, however, it
may be considered as a nonlinear, orthotropic solid. For
an orthogonal solid under plane stress or plane strain
conditions, the material stiness matrix [D] is expressed
by the 3 3 matrix in Eq. (5):
r
1
r
2
s
12
_
_
_
_
_
_
=
E
1
lE
1
0
lE
2
E
2
0
0 0 G
12
_
_
_
_
e
1
e
2
c
12
_
_
_
_
_
_
(5)
where r
1
, r
2
are stresses in the principal directions 1, 2
respectively, e
1
, e
2
are strains in the principal directions
1, 2 respectively, E
1
, E
2
are secant moduli of concrete in
the principal directions 1, 2 respectively, l is Poisson
ratio in the principal directions 1, 2 (taken as zero in FA-
STM), G
12
is secant shear modulus of concrete in the
principal plane.
The matrix in Eq. (5) contains only ve non-zero ele-
ments because the eects of shear stresses on the normal
stresses are ignored. The ve elements involve three
moduli E
1
, E
2
and G
12
, and one Poisson ratio l. The
shear modulus of cracked concrete, G
12
, has been inves-
tigated by a number of researchers over the past three
decades [913]. The two most popular shear moduli are: Fig. 1. Principal coordinate systems and the xed angle.
2782 T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791
(a) Darwin et al., [9]
G
12
=
1
4(1 l
2
)
E
1
_
E
2
2l

E
1
E
2
_ _
(6)
(b) Vecchio et al., [13]
G
12
=
E
1
E
2
E
1
E
2
(7)
Recently, Zhu et al. [8] derived a rational, yet very
simple shear modulus as follows:
G
12
=
r
1
r
2
2 e
1
e
2
( )
(8)
The shear modulus G
12
in Eq. (8) is rational because it
was derived from the equilibrium and compatibility
conditions. Eq. (8) was adopted in FA-STM to signi-
cantly simplify the solution algorithm and to improve its
accuracy.
The three expressions of shear moduli (Eqs. (6)(8))
were installed in FEAPRC to be used for comparison
later.
2.3. Constitutive relationships in FA-STM
2.3.1. Concrete in compression
The average (or smeared) stressstrain curve of
concrete in compression [8,1416] is shown in Fig. 2 and
is expressed as
r
2
= ff
/
c
2
e
2
fe
0
_ _
_

e
2
fe
0
_ _
2
_
e
2
=fe
0
61 (9)
r
2
= ff
/
c
1
_

e
2
=fe
0
1
4=f 1
_ _
2
_
e
2
=fe
0
> 1 (10)
where f
/
c
is the cylinder compressive strength of concrete;
e
0
, concrete strain at maximum compressive stress, taken
as 0.002; and f, softened coecient. The constant 4 in
Eq. (10) replaces the old constant 2 in the original FA-
STM [1416]. This revision was made in 1999 [8,17] to
take care of the long plateau after peak point observed
in the strain-control tests of panels using the servo-
control system.
The softened coecient f in Eqs. (9) and (10) is
f =
5:8

f
/
c
_
1

1
400e
1
g
/
_ 60:9 (f
/
c
in MPa) (11)
g =
q
y
f
yy
r
y
q
x
f
xy
r
x
(12)
where q
x
, q
y
are reinforcement ratios in the x and y di-
rections, respectively; f
xy
, f
yy
, yield stress of steel in the x
and y directions, respectively; and r
x
, r
y
, applied stresses
in the x and y directions, respectively. The symbol g
/
in
Eq. (11) is expressed by g in Eq. (12) or its reciprocal,
whichever is less than unity. The g
/
values are limited to
a range of 0:2 < g
/
< 1. In the descending portion of the
concrete stressstrain curve (Eq. (10)) the lowest stress
value was taken as 0:2ff
/
c
to avoid the potential nu-
merical problem in calculation.
2.3.2. Concrete in tension
The average (or smeared) stressstrain curve of
concrete in tension [14,18,19] is shown in Fig. 3 where
the ascending and descending branches are given as:
r
1
= E
c
e
1
e
1
6e
cr
(13)
r
1
= f
cr
e
cr
e
1
_ _
0:4
e
1
> e
cr
(14)
where E
c
is modulus of elasticity of concrete; f
cr
is
cracking stress of concrete; and e
cr
is cracking strain of
concrete. If the sharp change at the cracking strain e
cr
Fig. 2. Softened compressive stressstrain curve of concrete. Fig. 3. Average tensile stressstrain curve of concrete.
T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791 2783
cause any numerical diculty, a short plateau at the
peak point can be installed.
2.3.3. Concrete in shear
The average (or smeared) stressstrain relationship of
concrete in shear was originally derived empirically by
Pang and Hsu [6] and Hsu and Zhang [7]. Now that a
rational and simple shear modulus has been derived by
Zhu et al. [8], Eq. (8) is being used in the FA-STM.
2.3.4. Mild steel
The average (or smeared) stressstrain curve of mild
steel bars embedded in concrete, shown in Fig. 4, is
expressed by two straight lines [8,14,18]:
f
s
= E
s
e
s
e
s
6e
n
(15)
f
s
= f
y
0:91 (
_
2B) 0:02
_
0:25B
e
s
e
y
__
e
s
> e
n
(16)
where f
s
and e
s
are the average (or smeared) stress and
strain of mild steel bars, respectively; f
y
and e
y
are the
yield stress and strain of bare mild steel bars, respec-
tively; E
s
is the modulus of elasticity of steel bars; and
e
n
= e
y
(0:93 2B). The parameter B is given by B =
(f
cr
=f
y
)
1:5
=q, where q is the reinforcement steel ratio and
P0.5%. f
cr
is the cracking strength of concrete given by
f
cr
= 0:31

f
/
c
(MPa)
_
.
When a steel bar is embedded in concrete and starts
to yield at the cracks, the stresses in the steel bars be-
tween the cracks will be less than the yield stress at the
cracks, because part of the tensile force is resisted by the
concrete. Using the smeared crack concept, the steel
stresses are then averaged along the steel bar traversing
several cracks. The resulting average (or smeared) steel
stress at rst yield will obviously be less than the local
yield stress of a bare bar at the cracks. As shown in Fig.
4, the average (or smeared) stress of the steel bar at
initial yielding, f
n
, is lower than the yield stress of the
bare steel bars, f
y
. The dierence between f
n
and f
y
de-
pends on the parameter B which is a function of the steel
ratio, q, and the ratio f
cr
=f
y
.
The use of the average (or smeared) steel stresses
(Eqs. (15) and (16)), in combination with the average (or
smeared) concrete stresses (Eqs. (13) and (14)), allows
the tension stiening of steel bars by concrete to be
considered and the deformations of the steelconcrete
composites to be correctly evaluated. At the same time,
the use of average (or smeared) stresses in steel and
concrete avoids the unwarranted increase of the rst
yield strength of steelconcrete composites associated
with the use of stresses of the bare steel bars [19].
3. Program ``FEAPRC''
The nite element analysis program for reinforced
concrete (FEAPRC) used in this paper to predict the
behavior of reinforced concrete structures is a modied
version of nite element analysis program (FEAP).
FEAP is described in the book ``The Finite Element
Method'' by Zienkiewicz and Taylor [20], and is a gen-
eral-purpose program containing about 1000 subrou-
tines and more than 100,000 lines. As it stands, however,
FEAP is not applicable to reinforced concrete structures
because no constitutive relationships of concrete and
steel bars were included. The computer code FEAPRC
was developed by installing the relevant material models
of reinforced concrete into FEAP. This extension of
FEAP for application to reinforced concrete structures
involves the modications of 26 subroutines and the
addition of 41 new subroutines.
The subroutines in FEAP that are modied are listed
as follows: easble.f, emat2d.f, elas1d.f, elmlib.f, estrsd.f,
lnam.f, inpt2d.f, modl1d.f, modlsd.f, pform.f, pgauss.f,
pmacr.f, pmacr1.f, pmacr2.f, pmacr3.f, prtdis.f, prtrea.f,
pstr2d.f, resid2d.f, sld2d1.f, sld2d3.f, solid2d.f, sparse.f,
ster2d.f, strn2d.f and trussnd.f.
The 41 new subroutines in FEAPRC are summarized
into 12 groups ((a) to (c)) according to their functions.
The number in brackets is the number of subroutines in
the group: (a) Categorizing the seven cases of stresses as
shown in Tables 1(3), (b) Calculating stresses in the
seven cases(7), (c) Reviewing stresses in the seven
cases(7), (d) Storage and retrieval of material mo-
duli(5), (e) Storage and retrieval of strains(5), (f)
Storage and retrieval of stresses(5), (g) Transforma-
tion of matrices(3), (h) Nonlinear stinesses of con-
crete and steel(2), (i) Choice of shear moduli(1), (j)
Constitutive relationship of steel bars embedded in
concrete(1), (k) Calculation of principal strains(1),
(l) Judging the limit states(1).
Fig. 4. Average stressstrain curve of steel bars embedded in
concrete.
2784 T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791
Important modications to FEAP regarding material
properties and nonlinear solution techniques are given
below:
3.1. Material properties
3.1.1. Failure surface
The classication of constitutive relationships de-
scribed in Eqs. (9)(14) must be guided by an interactive
failure criterion for concrete. Various proposals have
been made to describe the failure strength characteristics
of concrete [2127]. Among these proposals, the most
popular was Kupfer's failure surface which provides
the biaxial strength of concrete under tensiontension,
compressioncompression and tensioncompression
(Fig. 5). Kupfer's failure surface was installed and im-
plemented in FEAPRC.
3.1.2. Various stress states in nonlinear procedure
As mentioned previously, concrete before cracking is
considered as an elastic isotropic solid. After cracking, it
is treated as an elastico-plastic solid having orthotropic
characteristic. Since cracking of concrete is the primary
source of the nonlinear behavior of reinforced concrete
structures, it is necessary to check the stress states of
every element and to judge the state of cracking at each
loading increment and at every iterative step. This in-
formation is needed to calculate the changes of concrete
moduli in order to update the material stiness matrix in
the nonlinear procedure.
The biaxial stress states of every element can be
divided into seven cases as shown in Table 1: (1)
uncracked tensionuncracked tension, (2) uncracked
tensionuncrushed compression, (3) uncrushed com-
pressionuncrushed compression, (4) cracked tension
cracked tension, (5) cracked tensionuncracked tension,
(6) cracked tensionuncrushed compression and (7)
crushed compressioncrushed compression. With these
seven cases of stress condition, one can judge the de-
velopment of cracks, the redistribution of stresses in
cracked elements, the changes of elastic moduli, and the
unbalanced nodal forces.
3.1.3. Limit state criteria
The limit state criteria depend on many factors:
characteristics of structures, properties of composite
materials, boundary conditions, loading path, etc. Al-
though much research has been done in this eld, it is
still not possible to propose a general criterion that is
suitable for all cases. For reinforced concrete beams,
panels and structural systems, the limit state criteria
may involve ultimate tensile strain, ultimate compressive
strain, extent of cracked or yielded areas, severity of
deformation, magnitude of deection, etc. According to
our own experimental results and others tests, the ulti-
mate tensile strain e
ut
and the ultimate compressive
strain e
uc
of concrete will be used as the limit state cri-
teria as follows:
e
ut
= 10e
cr
(17a)
e
uc
= 4e
0
(17b)
where e
cr
is the tensile cracking strain of concrete and e
0
is the strain at maximum compressive stress of concrete.
Table 1
Seven stresses states in nonlinear procedure
Case Stress states Elastic
moduli
Uncracked tensionun-
cracked tension
r
1
< r

0
E
1
= E
0
r
2
< r

0
E
2
= E
0
Uncracked tensionun-
crushed compression
r
1
< r

0
E
1
= E
0
r
2
< r

E
2
= r
2
=e
2
Uncrushed compres-
sionuncrushed com-
pression
r
1
< r

E
1
= r
1
=e
1
r
2
< r

E
2
= r
2
=e
2
Cracked tension
cracked tension
r
1
Pr

0
E
1
= r
1
=e
1
r
2
Pr

0
E
2
= r
2
=e
2
Cracked tensionun-
cracked tension
r
1
Pr

0
E
1
= r
1
=e
1
r
2
< r

0
E
2
= E
0
Cracked tensionuncru-
shed compression
r
1
Pr

0
E
1
= r
1
=e
1
r
2
< r

E
2
= r
2
=e
2
Crushed compression
crushed compression
r
2
> r

E
1
= 0
E
2
= 0
r

0
: cracking strength of concrete; r

0
: ultimate compressive
strength of concrete.
Fig. 5. Failure surface by Kupfer and Gerstle [24].
T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791 2785
When the ultimate strain values are reached, the
calculation procedure will be terminated and all the in-
formation at the limit state will be printed out. Detailed
discussion of the limit state criteria for framed shear
walls is given in a separate paper [28].
3.2. Nonlinear solution technique
3.2.1. Concrete and steel elements
The concrete body in program FEAPRC is modeled
by 4- or 8-node quadrilateral isoparametric 2D elements.
The reinforcing bars are represented by discrete pin-
connected truss elements. A truss element is subjected
only to axial deformations and forces, and has two de-
grees of freedomat the nodal points of concrete elements.
In the thickness direction, the steel bars at the same layer
are idealized as a plate with the same width as the
thickness, and with the same cross section as these bars.
In view of the discrete representation of the steel
bars, it would be more convenient to move the steel bars
to the edge of the concrete elements where the node
points are located. The dierence caused by this relo-
cation of steel bars can be compensated for by adjusting
the cross sectional area so that the moments of inertia
are the same. The equivalent cross section, F
eq
, is shown
in Fig. 6 and is calculated as follows:
F
eq
=
a
2
0
a
2
F
0
(18)
where F
0
is the original cross section of steel bar; while a
0
and a are the original distance and the relocated dis-
tance, respectively, of steel bars measured from the
center line of steel cross section to the neutral axis of
concrete element.
In a shear panel or in the panel of a framed shear
wall, the steel mesh is idealized as being smeared over
the whole panel and modeled by 2D solid elements. For
a mesh with equal percentages of steel in the two di-
rections, the mesh is treated as a thin layer of homoge-
neous steel plate pinned to the concrete elements at the
nodes. For a mesh with unequal percentages of steel in
the two directions, it is treated as a combination of two
thin layers of orthogonal steel plates, each is imagined to
smear only in one direction. These two thin steel layers
of steel occupy the same space and are pinned at the
same nodes of concrete elements.
3.2.2. Unbalanced nodal forces
The load was applied in increments to the structure.
For each load increment, we used the modied Newton
Raphson method with iteration procedure to achieve
convergence. In every iterative step the calculated ele-
ment stresses were converted to a set of equivalent nodal
forces R
eq
and then compare to the set of applied
nodal forces R
ap
. The resulting unbalanced nodal
forces R
un
was then calculated as:
R
un
= R
ap
R
eq
(19)
The equivalent nodal forces R
eq
in Eq. (19) was
obtained from the well known formula, R
eq
=

e
_
v
B [ [
T
r dv, where B [ [ is the geometric matrix of
elements. After each iteration step, the unbalanced no-
dal forces should become small enough to meet the
convergence condition. If not, the load increment was
reduced and the iteration process resumed. This process
continued until the failure of the structure.
3.2.3. Dynamical storage/retrieval of historical data
In the FEAP program the current functions for
storing and retrieving historical data are limited. Since
the nonlinear analysis of concrete structures depends
strongly on the historical path, a special dynamical
storage/retrieval system for historical data was estab-
lished in the FEAPRC program. This system was rela-
tively independent from the original FEAP program in
order to avoid the interference between the FEAP pro-
gram and the special storage/retrieval process. Using this
system, any component of stresses and strains, as well as
any material nonlinear parameter at any Gauss point of
any element, could be stored and retrieved at any
loading step.
3.2.4. Additional input and output
A new NAMELIST function was installed to le the
additional inputs for FEAPRC. This additional input
le for nonlinear analysis of concrete structures includes
the basic material parameters of concrete; special cal-
culation parameters of nonlinear analysis procedure; the
selected calculation plan and the additional output
demands. The additional outputs include the displace-
ments and loads at indicated nodal points; the stresses
and strains at assigned Gauss points of elements; the
information of steel yielding and concrete cracking; the
various material moduli (E
1
; E
2
; G; l); the softened co-
ecients at given Gauss points; and the limit state cri-
teria. These additional output functions would facilitate Fig. 6. Equivalent cross-section of steel bars.
2786 T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791
the use of a post-processor so that the whole picture of
nonlinear analysis results can be presented expedi-
tiously.
4. Comparison with experimental results
The applicability and accuracy of the program
FEAPRC are illustrated by comparing the predicted
results to the test results of exural beams [29], shear
panels of normal and high strength concrete [15,16], and
framed shear walls [30].
4.1. Bresler/Scordelis beam test
The simply supported reinforced concrete beam with
mid-point load tested by Bresler and Scordelis [29] is
shown in Fig. 7a. The longitudinal reinforcement con-
sists of four steel bars with a total area of 2580 mm
2
(4
in.
2
), an ultimate strength of 917 MPa and an elastic
modulus of 191,400 MPa. The concrete has a compres-
sive strength of 24.5 MPa, and an elastic modulus of
21,300 MPa. The failure load was measured to be 258.1
kN.
Four-node bilinear isoparametric elements are se-
lected to model the concrete, while 2-node truss elements
are used for modeling the reinforcement. Perfect com-
patibility of deformation between concrete and rein-
forcement is assumed. The meshes of concrete and steel
bars are shown in Fig. 7b and c, respectively. The total
mesh consists of 70 quadratic elements for concrete, 20
linear truss elements for steel bars and a total of 88
nodal points.
The loaddisplacement curves at the mid-point of the
beam span in Fig. 8 show excellent agreement between
predictions and test results. The calculated ultimate load
is 253.2 kN, about 1.9% lower than the test value of
258.1 kN.
4.2. Shear panels tested at University of Houston
Nearly 100 reinforced concrete panels have been
tested at the University of Houston since 1988. These
panels were subjected to in-plane shear and normal
stresses as shown in Fig. 9a. The testing panels had a size
of 1397 1397 mm in plane, and a thickness of 178 mm.
They were made of normal strength concrete (42 MPa),
mediumhigh strength concrete (65 MPa) and high
strength concrete (100 MPa).
Two panels are chosen to illustrate the validity of the
program FEAPRC. These were normal-strength con-
crete panel A3 tested by Pang and Hsu [15] and high-
strength concrete Panel VA1 tested by Zhang and Hsu
[16]. The material properties of these two panels are
given as follows:
Fig. 7. Flexural beam by Bresler and Scordelis [29].
Fig. 8. Loaddeection curves at mid-span of beam.
Specimen f
/
c
(MPa) e
0
E
c
(MPa) q
x
; q
y
f
xy
; f
yy
E
s
(MPa)
A3 41.62 0.0019 24,632 1.789% 446 199,810
VA1 95.10 0.00245 31,309 1.200% 445 181,000
T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791 2787
To model the panels, we used 49 4-node bilinear
isoparametric elements for concrete mesh as shown in
Fig. 9b. The meshes for steel bars in x and y directions
are modeled as in concrete. Therefore, the total mesh
consists of 147 quadratic elements and 192 nodal points.
The predicted values for Panels A3 and VA1 are
compared to the test values in Figs. 10 and 11, respec-
tively. It can be seen that the agreements are excellent.
The predicted shear strength of 7.60 MPa for Panel A3 is
only 0.8% less than the test value of 7.66 MPa. The
predicted shear strength of 6.03 MPa for Panel VA1 is
only 4.0% less than the test value of 6.29 MPa.
Fig. 11 also includes the predicted curves using the
two shear moduli given by Eqs. (6) and (7). It can be
seen that the predicted curve using the rational shear
modulus G1 (Eq. (8)) agrees best with the test curve. The
predicted curve using G2 modulus, (Eq. (6)) is higher
than that using G1 modulus, (Eq. (8)) while the pre-
dicted curve using G3 (Eq. (7)) is much lower than that
using G1.
4.3. Framed shear walls tested at University of Houston
The predictions of FEAPRC was also compared to
the observed behavior of 14 framed shear walls tested
recently by Gao [30] at the University of Houston. A
typical specimen FSW4 is chosen for illustration. The
dimensions of this specimen and the arrangements of
steel bars are shown in Fig. 12a. The mesh for the
concrete part of framed shear wall is given in Fig. 12b.
The basic material behavior of specimen FSW4 is: q
p
=
0:55%, q
f
= 1:11%, e
0
= 0:002, f
/
c
= 49:51 MPa, f
cr
=
2:73 MPa, f
y;p
= 419 MPa, f
y;f
= 425 MPa, E
c
= 34; 145
MPa, E
s;p
= 187; 544 MPa, E
s;f
= 216; 082 MPa. The
subscript p in the symbols means panel, while the sub-
script f represents frame.
Specimen FSW4 was subjected to reversed cyclic
loading, which resulted in a set of hysteretic loops. The
envelopes of the hysteretic loops in the positive and
negative directions were used to compare with the pre-
dicted curves. The predicted (solid) curve of normalized
shear stresses vs. total drift angle is compared to the two
Fig. 9. Shear Panels A3 by Pang and Hsu [15] and VA1 by
Zhang and Hsu [16].
Fig. 10. Shear stress versus shear strain curves of Panel A3.
Fig. 11. Shear stress versus shear strain curves of Panel VA1
using three shear moduli.
2788 T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791
experimental curves (dotted) in Fig. 13. It is obvious that
the predicted curve agrees very well with the two tested
curves. Detailed analysis of 9 framed shear walls is given
in a separated paper [28].
5. Conclusions
(1) By modifying the computer code FEAP, a new
nite element program FEAPRC was developed which
is useful for predicting the behavior of reinforced con-
crete structures. The modications include the addition
of new constitutive models of materials established for
the FA-STM, as well as the installation of a special
storage/retrieval system for historical data, and an input
and output system.
(2) The FEAPRC program took into account the
four characteristics of cracked reinforced concrete,
namely, the softened stressstrain relationship of con-
crete in compression, the stiening eect of concrete
in tension, the average (smeared) stressstrain curve of
steel bars embedded in concrete, and the rational shear
modulus.
(3) The behavior of beams, shear panels and framed
shear walls predicted by FEAPRC was found to agree
very well with the observed behavior. The shear panels
were made of normalstrength concrete as well as high-
strength concrete.
(4) The rational and simple shear modulus, as dened
in Eq. (8), provided reliable and accurate predictions.
Acknowledgements
This research was supported by two grants CMS-
9711084 and CMS-9713707 from the National Science
Fig. 12. Framed shear wall FSW4 by Gao [30].
T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791 2789
Foundation. These generous supports are gratefully
acknowledged.
References
[1] Ngo D, Scordelis AC. Finite element analysis of rein-
forced concrete beams. J Am Concr Inst 1967;64(3):152
63.
[2] Schnobrich WC. Behavior of reinforced concrete members
by nite element. Third Asian-Pacic Conference on
Computational Mechanics. 1618 September, Seoul,
Korea, 1996. p. 313.
[3] Chen WF, Ting EC. Constitutive models for concrete
structures. J Engng Mech (ASCE) 1980;106(EM1):119.
[4] Hsu TTC, Belarbi A, Pang XB. A universal panel tester.
J Test Eval (ASTM) 1995;23(1):419.
[5] Hsu TTC, Zhang LX, Gomez T. A servo control system
for universal panel tester. J Test Eval (ASTM) 1995;
3(6):42430.
[6] Pang XB, Hsu TTC. Fixed-angle softened-truss model for
reinforced concrete. Struct J Am Concr Inst 1996;
93(2):197207.
[7] Hsu TTC, Zhang LX. Nonlinear analysis of membrane
elements by xed-angle softened-truss model. Struct J Am
Concr Inst 1997;94(5):48392.
[8] Zhu RH, Hsu TTC, Lee JY. Rational shear modulus for
smeared-crack analysis of reinforced concrete. Struct J Am
Concr Inst 2001;98(4):44350.
[9] Darwin D, Pecknold DA. Inelastic model for cycle biaxial
loading of reinforced concrete. In: Civ Engng Stud
(Structural Research Series No 409). Urbana, ILL: Uni-
versity of Illinois; 1974.
[10] Cervenka V. Constitutive model for cracked reinforced
concrete. J Am Concr Inst 1985;82(6):87782.
[11] Criseld MA, Wills J. Analysis of R/C panels using
dierent concrete models. J Engng Mech Div (ASCE)
1989;115(3):57897.
[12] Hu H, Schnobrich WC. Nonlinear analysis of plane stress
state reinforced concrete under short term monotonic
loading. In: Civil Engng Stud (Structural Research Series
No 539). Urbana, ILL: University of Illinois; 1988.
[13] Vecchio FJ, Collins MP. Nonlinear nite element analysis
of reinforced concrete membranes. Struct J Am Concr Inst
1989;86(1):2635.
[14] Hsu TTC. Unied theory of reinforced concrete. Boca
Raton, FL: CRC Press; 1993.
[15] Pang XB, Hsu TTC. Behavior of reinforced concrete
membrane elements in shear. Struct J Am Concr Inst
1995;92(6):66579.
[16] Zhang LX, Hsu TTC. Behavior and analysis of 100 MPa
concrete membrane elements. J Struct Engng (ASCE)
1998;124(1):2434.
[17] Hsu TTC, Zhu RRH. Post-yield behavior of reinforced
concrete membrane elementsthe HSU/ZHU ratios. Pro-
ceedings Volume, USJapan Joint Seminar on Post-Peak
Behavior of Reinforced Concrete Structures Subjected to
Seismic Loads, vol. 1. Tokyo/Lake Yamanaka, Japan,
1999. pp. 4360; also, modeling of inelastic behavior of RC
structures under seismic loads. ASCE Publication; 2001.
pp. 139157.
[18] Belarbi A, Hsu TTC. Constitutive laws of concrete in
tension and reinforcing bars stiened by concrete. Struct J
Am Concr Inst 1994;91(4):46574.
[19] Hsu TTC, Zhang LX. Tension stiening in reinforced
concrete membrane elements. Struct J Am Concr Inst
1996;93(1):10815.
[20] Zienkiewicz OC, Taylor RL. The nite element method.
4th ed. London: McGraw-Hill; 1991.
[21] Kupfer HB, Hilsdorf HK, Rusch H. Behavior of concrete
under biaxial stresses. J Am Concr Inst 1969;66(8):65666.
Fig. 13. Normalized shear stress versus total drift angle of FSW4.
2790 T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791
[22] Liu TCY, Nilson AH, Slate FO. Biaxial stress strain
relations for concrete. J Struct Div (ASCE) 1972;98(5):
102534.
[23] Nelissen LJM. Biaxial testing of normal concrete, vol
18(1). Delft, Netherlands: Heron; 1972.
[24] Kupfer HB, Gerstle KH. Behavior of concrete under
biaxial stresses. J Engng Mech Div (ASCE) 1973;99(4):
85366.
[25] Cedolin L, Crutzen YRJ, Dei Poli S. Triaxial stress strain
relationship for concrete. J Engng Mech Div (ASCE)
1977;103(3):42939.
[26] Murry DW, Chitnuyanondh L, Wong C. Implementation
of an elastic plastic concrete relationship. Comput Meth
Appl Mech Engng 1980;23:3557.
[27] Chen WF. Plasticity in reinforced concrete. 1st ed. New
York: McGraw-Hill; 1982.
[28] Wang TJ, Gao XD, Hsu TTC. Nonlinear nite element
analysis of framed shear walls. Transaction, 16th Interna-
tional Conference on Structural Mechanics in Reactor
Technology (SM:RT 16), Washington DC, August 2001,
Paper H1233, 18.
[29] Bresler B, Scordelis AC. Shear strength of reinforced
concrete beamsseries II. SESM Report No. 64-2, Uni-
versity of California at Berkeley, 1964.
[30] Gao XD. Framed shear walls under cyclic loading. Ph.D.
Dissertation. Department of Civil and Environmental
Engineering, University of Houston, Houston, TX, De-
cember 1999.
T. Wang, T.T.C. Hsu / Computers and Structures 79 (2001) 27812791 2791

S-ar putea să vă placă și