Você está na página 1de 12

Engineering Structures 30 (2008) 2069–2080

www.elsevier.com/locate/engstruct

An efficient tension-stiffening model for nonlinear analysis of reinforced


concrete members
Renata S.B. Stramandinoli a,∗ , Henriette L. La Rovere b
a Civil Engineering Department, COPEL, Parana, Brazil
b Civil Engineering Department, Federal University of Santa Catarina, Brazil

Received 7 November 2006; received in revised form 16 October 2007; accepted 27 December 2007
Available online 25 February 2008

Abstract

A constitutive model for reinforced concrete elements that takes into account the tensile capacity of the intact concrete between cracks, effect
known as tension-stiffening, is proposed in this paper. In the model, the tensile stress–strain curve of concrete displays an exponential decay in the
post-cracking range, defined by a parameter that depends on the reinforcement ratio and on the steel-to-concrete modular ratio. This parameter
was derived taking as a basis the CEB tension-stiffening model. The model was implemented into a computational program that allows for
nonlinear finite element analysis of reinforced concrete beams. The numerical results obtained by the program compared extremely well with
several experimental results from simply supported beams tested under 4-point bending that displayed a dominant flexural behavior. Extension of
the model to members subjected to combined flexural and shear is also presented.
c 2008 Elsevier Ltd. All rights reserved.

Keywords: Tension-stiffening; Finite elements; Reinforced concrete beams; Nonlinear analysis

1. Introduction Several models to represent the tension-stiffening effect have


already been proposed, ranging from simple to very refined
It is well known that the intact concrete between cracks can models of great degree of complexity. One of the simplest
still carry tensile stresses after the onset of cracking in rein- models, but yet very well accepted by designers for beam
forced concrete (R/C) elements due to the bond between the re- deformation calculations, is the model of Branson [1] which
inforcing bars and the surrounding concrete. This effect, known considers an equivalent inertia for the cracked beam section.
as tension-stiffening, was neglected in the past since it does not The model developed by Cosenza [2] can also be quoted as a
significantly affect the ultimate strength of the reinforced con- simplified tension-stiffening model.
crete members. Since the 70s, however, the tensile behavior of Quite a few models that modify the constitutive equation
concrete was introduced in the analysis of load-deflection char- of steel or concrete after cracking have also been proposed
acteristics of R/C elements, and since the 80s in design code for nonlinear finite element analysis of reinforced concrete
recommendations for service load level. It is also important to structures. Among the models that modify the steel constitutive
consider tension-stiffening when evaluating the serviceability equation it can be quoted: Gilbert and Warner [3], Choi and
of existing R/C structures. The tension-stiffening effect depends Cheung [4] and the CEB manual design [5] model; and among
on several factors, such as member dimensions, reinforcement those that modify the concrete constitutive equation: Scanlon
ratio, rebars diameters, and the materials elastic modulus and and Murray [6], Lin and Scordelis [7], Collins and Vecchio [8],
strength. This effect occurs until yielding of the longitudinal Stevens et al. [9], Balakrishnan and Murray [10], Massicotte
reinforcement takes place, and it tends to increase as the rein- et al. [11]. There are more complex models based on the
forcement ratio of the member decreases. bond-slip mechanism and on localized phenomena such as
those of: Floegl and Mang [12], Gupta and Maestrini [13], Wu
∗ Corresponding address: Rua Pasteur, 413 ap.1502 – Batel, CEP: 80250-
et al. [14], Russo and Romano [15], Choi and Cheung [16] and
080, Curitiba, Paraná, Brazil. Tel.: +55 41 91742486; fax: +55 41 33523090.
E-mail addresses: renatastramandinoli@hotmail.com Kwak and Song [17]. These complex models, also known as
(R.S.B. Stramandinoli), henriette@ecv.ufsc.br (H.L. La Rovere). “microscopic models”, depend on a series of parameters that

c 2008 Elsevier Ltd. All rights reserved.


0141-0296/$ - see front matter
doi:10.1016/j.engstruct.2007.12.022
2070 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

stress–strain curve of concrete is modified to take into account


Notation the tension-stiffening effect in an average way, are more widely
used. These so-called “macroscopic” models are easier to
Ac concrete area
implement and, by being simpler than the “microscopic” ones,
Ae f effective concrete area
they can be readily applied to analyze full-scale structures.
As reinforcement area
However, most macroscopic models oversimplify the tension-
d usual effective depth
stiffening effect by considering only one equation to describe
E ccr secant elastic modulus of concrete in the post-
the post-cracking range of the tensile stress–strain curve,
cracking range
independently of the member reinforcement ratio and material
E ci elastic modulus of concrete before cracking
properties.
Es elastic modulus of the reinforcing steel
Seeking a simple model that can represent more realistically
f cc concrete compressive strength
the tension-stiffening effect and at the same time be easily
f ct concrete tensile strength
implemented into finite element programs, we introduce in this
fy reinforcing steel yield stress
work a novel tension-stiffening model.
h nominal depth of the beam
The proposed model uses an exponential decay curve to
` member length
describe the post-cracking range of the tensile stress–strain
∆` total extension of a reinforced concrete member
law of concrete. The exponential decay parameter (α) is a
of length `
function of the member reinforcement ratio (ρ) and of the
N axial force
steel-to-concrete modular ratio (n = E s /E c ), and is derived
Ncr axial force at the onset of cracking
taking as basis the CEB [5] tension-stiffening model. Gupta
Ny axial force at yielding of reinforcement
and Maestrini [13] have also utilized an exponential curve
n steel-to-concrete modular ratio
to formulate a simplified tension-stiffening model and have
sh plastic-to-elastic modular ratio
obtained good correlation with experimental results.
x neutral axis depth
A brief review of the CEB [5] model is initially presented.
α exponential decay parameter of the
Next the model is proposed for members under direct tension
tension-stiffening model
and verified by comparison with other tension-stiffening
ε strain in the R/C member
models and also with experimental results obtained from
εcr strain corresponding to concrete tensile strength
pull-out tests on R/C bars. The proposed tension-stiffening
εs1 strain in the reinforcement in State I
model is then extended to members under bending, assuming
εs2 strain in the reinforcement in State II
that all cracks are orthogonal to the reinforcement. This
εs1r strain in the reinforcement in State I correspond-
model is implemented into a finite element program called
ing to stress σsr
ANALEST that allows for nonlinear analysis of reinforced
εs2r strain in the reinforcement in State II correspond-
concrete beams and plane frames. Application of the model to
ing to stress σsr
simply supported R/C beams tested under 4-point bending in
εsm average strain in the reinforcing steel
different research laboratories are presented in the sequence. In
εy reinforcing steel yield strain
all the examples, shear deformation, stirrups contribution, and
∆εs contribution of the concrete in tension between
geometric nonlinearities are neglected. A simplified approach
cracks
to further extend the tension-stiffening model to members under
∆εsmax maximum variation between the strains εs1 and
combined bending and shear is also described. Concluding
εs2
remarks are given at the end of the work.
ρ member reinforcement ratio
ρef effective reinforcement ratio
2. The CEB manual design model [5]
ρeq equivalent reinforcement ratio
θ angle between the x-direction and the principal
The CEB model developed for reinforced concrete members
direction 1.
subjected to tension considers the tension-stiffening effect
σct concrete tensile stress
through an increase in stiffness of the reinforcement. The
σs2 stress in the reinforcement at a cracked section
cracking mechanism of a reinforced concrete member subjected
under the applied load
to uniaxial tension can be observed in Fig. 1.
σsr stress in the reinforcement calculated on the basis
An equation for determining the stress–strain curve for the
of a cracked section, where the maximum stress
reinforcement is proposed in terms of an average strain, which
in the concrete under tension is equal to f ct
lies between the strain of an uncracked section (State I) and that
of a totally cracked section (State II):
are usually difficult to obtain, requiring specific experiments for
∆`
each particular member, hence they are not usually applied to εsm = = εs2 − ∆εs (1)
full-scale problems. `
Generally, the models that modify the constitutive equation where ∆` is the total extension of a reinforced concrete member
of concrete, in which the descending branch of the tensile of length ` subjected to an axial tensile force N which is
R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080 2071

An expression for the average strain can be obtained by


substituting Eq. (2) into Eq. (1):
σsr σsr
   
εsm = εs2 − ∆εs max = εs2 − (εs2r − εs1r ) ∴
σs2 σs2
εs2r σsr σsr
    
εsm = εs2 1 − + εs1r ∴
εs2 σs2 σs2
 2 "  2 #
σsr σsr
εsm = εs1 + 1 − εs2 (3)
σs2 σs2

where:
εs1r is the strain in the reinforcement in State I (uncracked
section) corresponding to stress σsr ;
εs2r is the strain in the reinforcement in State II
(totally cracked section without any concrete contribution)
corresponding to stress σsr .
Fig. 1. Cracking mechanism of a reinforced concrete member subjected to Eqs. (1)–(3) were derived for pure tension, however, as
uniaxial tension: (a) reinforcement stress; (b) bond stress; (c) concrete stress indicated in the CEB Manual [5], they are also valid for flexure.
(CEB [5]).
Eq. (3) was derived assuming monotonic loading and high-bond
bars, but it was further modified to take into account cyclic
loading and the use of smooth bars.
This model presents a consistent theory to represent
the average post-cracking behavior of a reinforced concrete
member under tension. Since the proposed constitutive
equation is based on experimental results, it is also an accurate
model. However, as it can be observed from Eqs. (1)–(3), it
is difficult to be implemented into a finite element code, since
σs2 cannot be explicitly obtained from the average strain εsm .
A simplification of this model was proposed later on CEB-
FIP Model Code-90 [18] where a trilinear curve was utilized
to represent the “stress–average strain” relationship for the
reinforcement. This curve is an approximation of the original
curve shown in Fig. 2, with a bilinear branch adopted after
Fig. 2. Stress–strain curve for the reinforcement (CEB [5]).
cracking instead of a continuous curve. D’Avila and Campos
Filho [19] proposed a trilinear curve for the tensile constitutive
equation of concrete in the post-cracking range, based on this
greater than the force Ncr which produces the first crack; simplified model from CEB-FIP MC-90. A continuous curve,
∆εs represents the contribution of concrete between cracks however, is more desirable for computational implementation
which follows a hyperbolic relationship approaching the line into nonlinear finite element codes.
εs2 asymptotically for stresses in excess of σsr .
In the CEB Manual Design [5], the following expression for 3. Proposed model
∆εs , based on experimental results, is proposed:
A novel tension-stiffening model that modifies the tensile
σsr constitutive equation of concrete is proposed in the following.
∆εs = ∆εs max (2)
σs2 The model uses an explicit formulation for the concrete
stress–strain curve and thus can be easily implemented into
where:
a finite element code. Some features of the CEB manual
σsr is the stress in the reinforcement calculated on the basis design [5] model instead of the CEB-FIP MC-90 [18] model are
of a cracked section, where the maximum stress in the concrete utilized, therefore a continuous stress–strain curve is obtained
under tension is equal to f ct ; for the concrete in the post-cracking range providing numerical
σs2 is the stress in the reinforcement at a cracked section stability in nonlinear finite element analysis of R/C members.
under the applied load; and Concrete is assumed to behave like a linear-elastic material
∆εsmax is the maximum variation between the strains εs1 until its tensile strength is reached, so that a straight line
and εs2 which occurs at the beginning of the cracking process. defines initially the stress–strain curve. In the post-cracking
The definition of all these parameters can be better observed in range, an exponential decay curve is adopted until yielding
Fig. 2. of reinforcement takes place, and is defined by the following
2072 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

equation: where:

ε
 N (1 + nρ) f ct
σct = f ct e
−α εcr σs2 = and σsr =
(4) As ρ
where, in which ρ is the reinforcement ratio equal to As /(Ac ).
f ct is the concrete tensile strength and εcr is the The strain in the member after cracking is:
corresponding strain;
α is an exponential decay parameter to be determined. N
ε= (8)
In the absence of experimental results, the expression given E s As + E ccr Ac
by CEB-FIP MC-90 [18] can be used to estimate f ct : where,
 2/3 E ccr is an equivalent elastic modulus of concrete in the post-
fcc (MPa) − 8
fct (MPa) = 1.4 . (5) cracking range, defined by the secant modulus:
10
σct
An alternative way for determining an expression for α E ccr = (9)
ε
would be by adjusting the experimental results from reinforced
concrete bars subjected to direct tension, by varying the which varies according to the cracking level in the member.
specimen longitudinal reinforcement ratio (ρ) and material Substituting Eqs. (7) and (9) into Eq. (8), the stress in the
properties. Many authors have conducted parametric studies concrete can be obtained by:
to investigate the influence of ρ and of material properties N − εE s As
(fracture energy or specimen diameter-to-length ratio, tensile- σct = = σs2 ρ − εE s ρ. (10)
Ac
to-bond strength ratio, etc. . . ), on tension-stiffening. Amongst
all properties, the reinforcement ratio is the one that has shown For tracing the concrete stress–strain curve between points “a”
the greatest influence on the tension-stiffening effect (Hegemier and “b”, the values of ε are initially calculated using Eq. (6),
et al. [20]). by varying the applied force N from Ncr (the axial force at
Hence, instead of adjusting the experimental data, an the onset of cracking) up to N y (the axial force at yielding
expression for the parameter α, defined as a function of the of reinforcement). From the obtained values for ε, the stress
member reinforcement ratio (ρ) and of the steel-to-concrete in concrete is then calculated by means of Eq. (10). In the
modular ratio (n = E s /E c ), is derived in this paper, taking as descending branch where εs2 > ε y (from point “b” to point
basis the CEB [5] model. “c”), a straight line is found until ε reaches the strain at yielding
The same concept of average deformation (εsm ) and of reinforcement, where the stress in concrete drops to zero.
its definition given by Eq. (3) is adopted. The concrete Seeking an expression to determine the exponential decay
stress–strain curve is determined through the analysis of a parameter α, several concrete stress–strain curves were initially
reinforced concrete member subjected to direct tension. The traced using the procedure described above, by varying the
contribution of concrete between cracks can be observed from material properties f ct and f y , and by selecting different
the graphs displayed in Fig. 3. In this figure, point “a” values of (nρ) for fixed values of f ct and f y . An exponential
represents the onset of cracking; “b” defines the point where curve, as defined by Eq. (4), was then adjusted to each traced
the strain in the reinforcement for State II reaches the strain stress–strain curve. It was then observed that (nρ) was the most
at yield (εs2 = ε y ); and “c” the point where the strain in the important property to define the exponential decay parameter.
member reaches the strain at yield (ε = ε y ). Hence, several exponential curves were fitted to the traced
Additionally, it is assumed that the strain in the concrete stress–strain curves by varying only (nρ), and a value
reinforcement is equal to the strain in the surrounding concrete. of α was found for each corresponding value of (nρ). The
Thus, in the linear-elastic range and before the onset of curve fitting was made using the Mathcad 2001 program for
cracking, the strain in the R/C member can be determined by: normalized concrete stress (σct / f ct ) versus strain curves. Fig. 4
illustrates an example of curve fitting obtained for (nρ) = 0.2,
N yielding an exponential decay parameter α = 0.069.
ε= (6)
E s As + E ci Ac An expression for the exponential decay parameter was then
where: derived by using all the obtained values of α for different values
E s is the elastic modulus of the reinforcing steel; of (nρ), as shown in Fig. 5. The best fit curve achieved was
E ci is the elastic modulus of concrete before cracking; a third degree polynomial (COR = 0.996), described by the
As is the reinforcement area; and following equation:
Ac is the concrete area.
α = 0.017 + 0.255 (nρ) − 0.106 (nρ)2 + 0.016 (nρ)3 . (11)
After cracking, the strain in the concrete between points “a”
and “b” in the curve shown in Fig. 3(b) is calculated by Eq. (3): This tension-stiffening model was derived for R/C members
 2 "  2 # subjected to direct tension. Extension of the model to R/C
σsr σsr beams under bending, can be done by employing the effective
εc = εs = ε= εs1 + 1 − εs2 (7)
σs2 σs2 area concept that corresponds to the tensile zone in the member
R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080 2073

Fig. 3. Tension-stiffening effect in a reinforced concrete member under tension: (a) Applied force × strain curve for the R/C member, (b) Concrete stress–strain
curve.

0.1h, the above expression for the effective area simplifies to:
b.h
Aef ∼
= . (13)
4
Therefore, for R/C beams, an effective reinforcement ratio,
expressed as
As
ρef = (14)
Ae f
must be employed in Eq. (11).

4. Comparison of the proposed model with other


Fig. 4. Curve fitting for tensile stress–strain curve of concrete obtained for tension-stiffening models
nρ = 0.2.
To verify the validity of the proposed model, correlation
studies with other tension-stiffening models are initially
conducted. The proposed model is compared to some refined
tension-stiffening models, such as those of Gupta and
Maestrini [13], Kwak and Song [17], and also the CEB [5]
model that was utilized for deriving the proposed model.
Comparison is made in terms of tensile stress–strain curves
of concrete obtained from R/C bars under direct tension by
selecting two typical values of nρ: (i) nρ = 0.2, the same
value utilized in the previous example which yielded α = 0.064
for the exponential decay parameter in the proposed model
(see Eq. (11)); and (ii) nρ = 0.4, yielding α = 0.103 from
equation. (11). The resulting normalized stress versus strain
curves obtained by the different models for the two examples
Fig. 5. Curve fitting to obtain the exponential decay parameter (α) expression. can be seen in Fig. 6.
Analyzing the graphs of Fig. 6(a) and (b) it can be observed
section. An equation for this effective area is suggested in the that, as expected, the proposed model curves approach the
CEB-FIP MC-90: ones given by the CEB [5] model. For nρ = 0.4 the two
b(h − x) curves are almost coincident, and small differences are found
Ae f = 2.5b (h − d) < (12) for nρ = 0.2, due to the fact that a continuous curve was
3
derived in the proposed model from discrete points, obtained
where, using the CEB [5] model for a given nρ. The proposed model
h is the nominal depth of the beam, d is the usual effective also displays fairly good agreement with the models proposed
depth, and x is the neutral axis depth. by Gupta and Maestrini [13] and Kwak and Song [17].
Recalling that in R/C beams under bending the relationship Next, the model is compared to two simplified models
between nominal and effective depth is usually given by h−d ∼= commonly employed in finite element analysis of R/C
2074 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

Fig. 6. Comparison of different tension-stiffening models: (a) nρ = 0.2; (b) nρ = 0.4.

Fig. 7. Comparison of tension-stiffening models: (a) Collins and Vecchio [8] and bilinear model [21]; (b) the simplified models and the proposed model for different
values of nρ.

members: Collins and Vecchio [8] and a bilinear model uses a bar of 6 m length and cross-section dimensions of
(Figueiras [21]), shown in Fig. 7(a). Three values of nρ 30 cm × 50 cm. The longitudinal steel ratio is 0.67%, and
have been selected for the proposed model application. A the material properties are as follows: f ct = 1.17 MPa; E c =
comparison of the tensile stress–strain curves given by the 10 GPa; f y = 526 MPa; E s = 197 GPa. The resulting value for
proposed and other simplified models is shown in Fig. 7(b). α from Eq. (11) is 0.049. Fig. 8 shows the stress (MPa) versus
As it can be noted from this figure, the simplified models strain (h) curves, obtained experimentally and numerically
are not able to consider the effect of different reinforcement using Eq. (4).
ratios on tension-stiffening. This fact shows the advantage of The second test was conducted by Hwang and Riskalla, apud
a model such as the present, which takes into account this Gupta and Maestrini [13], in a study of tension members. The
effect. Therefore, our model combines an accuracy comparable case chosen for comparison is example number 7, which uses a
to more refined models with the ease of implementation as other bar of 76.2 cm length and cross-section of 17.8 cm × 30.5 cm.
simplified models. The longitudinal steel ratio is 1.476%, and the materials have
the following properties: f ct = 2.62 MPa; Ec = 27.8 GPa;
5. Comparison of the proposed model with experiments —
pull-out tests f y = 469 MPa; E s = 199 GPa. The resulting value for α from
Eq. (11) is 0.043. Fig. 9 shows the stress (MPa) versus strain
To further verify the validity of the proposed model, (h) graph, obtained experimentally and numerically using Eq.
comparisons with experiments are also performed. Two pull-out (4).
tests on reinforced concrete bars have been selected. The first It can be observed from Figs. 8 and 9 that the proposed TS
one, V3, tested by Rostásy et al., apud Massicotte et al. [11], model can reproduce very well the experimental behavior. The
R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080 2075

that display dominant bending behavior, with orthogonal cracks


to the longitudinal reinforcement. Either the Hognestad [24]
parabola or the constitutive equations given by the CEB-FIP
MC-90 [18] can be employed for the compressive stress–strain
law of concrete. Confinement provided by stirrups can also
be considered, by using a modified compressive law (Mander
et al. [25]). The proposed tension-stiffening model is utilized
to represent the tensile behavior of concrete, with the slight
modification in the stress–strain curve shown in Fig. 3(b): the
concrete stress drops to zero directly when the strain in the
member reaches the strain at yield of reinforcement, ε = ε y ,
hence a straight line orthogonal to the horizontal axis, between
points b and c, defines the cut-off. This modification simplifies
the numerical implementation and does not significantly alter
Fig. 8. Comparison between numerical and experimental results for V3
member tested by Rostásy et al. the overall behavior of the R/C member. It is assumed that
the reinforcing steel is an elasto-plastic material, modeled by a
bilinear stress–strain curve. To avoid convergence problems and
oscillations in the numerical iterative process, a parabolic curve
is fitted between the elastic and plastic branches of the bilinear
stress–strain curve, between 0.8 and 1.2ε y . Strain hardening of
the reinforcing steel may or may not be considered, through the
use of a coefficient sh , which is the plastic-to-elastic modular
ratio (sh = 0 for perfectly plastic steel). Either the classic
Newton–Raphson iterative Methods (load control) or the Arc-
length Method of Riks in conjunction with Newton iterative
procedures can be employed to solve the nonlinear equilibrium
equations. The convergence criterion is based on the ratio of
the norm of residual forces to the norm of applied forces and
reactions. Several parametric studies have been conducted by
La Rovere et al. [26] to investigate the effect of mesh size,
Fig. 9. Comparison between numerical and experimental results for # 7 number of layers for element section discretization, variation
member tested by Hwang and Riskalla.
in material properties, etc., in the flexural behavior of R/C
model shows only a slightly stiffer behavior at the end of the members.
tests, near the onset of yielding in the reinforcing steel. Some conclusions from these studies are reproduced here:
(1) no dependence on mesh size was observed when tension-
6. Finite element model for R/C beams stiffening is included, showing objectivity of results; (2) fewer
elements are needed to model the members when the internal
The proposed tension-stiffening model was introduced into node with axial degree of freedom is included; (3) 12 layers
a nonlinear finite element model, which was then implemented are usually sufficient to discretize the member section; (4)
into a computational program called ANALEST. The program the tension-stiffening parameter α usually plays a significant
allows for nonlinear analysis of reinforced concrete beams, and role in the post-cracking response of the member; and (5) no
both geometric and material nonlinearity can be considered. significant difference is observed when different equations (the
A 2D finite element bar with two external nodes and three Hognestad parabola or the equation from CEB-FIP MC-90) are
degrees of freedom each, and one internal node with only employed to describe the compressive constitutive behavior of
one axial degree of freedom is employed (Mari, [22]). As concrete.
shown by Chan [23], the internal node needs to be included
to properly capture the variation of the neutral axis position 7. Comparison of FE analyses with experiments — beam
due to material nonlinearities. Three Gauss points have been tests
used for the numerical integration of the stiffness matrix and
the restoring force vector along the length of the element. To investigate the reliability of the finite element model and
The fiber model is employed, discretizing the element section ANALEST program, comparisons with experimental tests on
into layers, with each layer assumed to be under a uniaxial reinforced concrete beams have been performed. Since the FE
stress-state. The finite element model allows for the complete model is limited to members that display dominant bending
response of a reinforced concrete member under bending, behavior, beams subjected to 4-point bending that displayed
from its elastic phase, passing by cracking of the concrete a typical flexural failure, without significant development of
and yielding of reinforcement, until collapse is reached. The shear cracks, were selected for the comparison. Four beams
model neglects shear deformation and is limited to members tested by different researchers from different countries were
2076 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

Fig. 10. Tested beam (VRE) geometry, load application and support positions (Ferrari [27]).

Fig. 11. Tested beams (VT1 and VT2) geometry and reinforcement; load application and support positions (Beber [28]).

selected, taking also as basis a certain variation on the α Two analyses were performed with ANALEST program for
parameter (ranging from 0.037 to 0.094), in order to validate each example, one considering the proposed tension-stiffening
the proposed tension-stiffening model. Larger values of α model and the other without tension-stiffening consideration.
parameter need not to be considered, since when α > 0.1 the In all analyses, geometric nonlinearities are neglected and the
tension-stiffening effect becomes very small. Newton–Raphson Method (tangent stiffness) was employed.
The beams analyzed are: Confinement provided by stirrups is disregarded, with the
– VRE tested by Ferrari [27], at the Federal University of Hognestad parabola being adopted for the compressive
Santa Catarina, Brazil. constitutive law of concrete.
– VT1 and VT2 tested by Beber [28], at the Federal The material properties used in the finite element models are
University of Rio Grande do Sul, Brazil; condensed in Table 1. Comparison between FE analyses and
– VB6 and VC3 tested by Juvandes [29], at the University of experiments are shown in Figs. 14–17, for the beams VRE, VT1
Porto, Portugal; and VT2, VB6 and VC3, respectively, in terms of total applied
Figs. 10–13 illustrate the beam dimensions, reinforcing vertical load versus mid-span vertical displacement graphs.
detailing, the tests set-up showing the load application and A close agreement between numerical and experimental
support positions, and the finite element meshes adopted for results is observed for the beam VRE, in both the elastic
the beam models. and the post-cracking range of the beam. The yield load
R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080 2077

Fig. 12. Tested beam (VB6) geometry and reinforcement; load application and support positions (Juvandes [29]).

Fig. 13. Tested beam (VC3) geometry and reinforcement; load application and support positions (Juvandes [29]).

Table 1
Material properties utilized in the finite element analyses of beam examples

Concrete Tension-stiffening Steel


Beam f cm (MPa) f tm (MPa) εo n ρeff (%) α φ (mm) f y (MPa) E s (GPa) s.h.
VRE 30.7 2.95 0.0020 6.19 1.33 0.037 6 767.5 210 0.016
8 545.8 210 0.01
VT1/VT2 33.5 2.62 0.0020 6.39 1.50 0.040 6 738 214.8 0.016
10 565 214.8 0.000
VB6 37.9 2.90 0.0023 5.37 4.47 0.072 3 192 174 0.001
8 497 195 0.0042
VC3 20.7 1.60 0.0020 9.00 3.80 0.093 12.5 507 184.6 0.0014

predicted analytically was 34 kN, while in the experiment reinforcement and large deflections of the beam near to failure,
the measured value was between 33 and 35 kN. When no as it can be observed from Fig. 14; however, the instruments
strain hardening is considered, no convergence of the iterative have been removed from the specimen prior to failure, thus
procedure could be achieved in the analysis after yielding of hindering the measurement of ultimate displacement.
the steel reinforcement, either using the Newton–Raphson or For the beams VT1 and VT2, it can be observed from
the Arc-length Method. By adopting a small strain hardening Fig. 15 that the FE model can capture very well the ascending
coefficient for the longitudinal reinforcement, sh = 0.01, branch of the curve and approaches well the experimental
three more load increments of 0.5 kN could be applied after curves after the onset of cracking when tension-stiffening of
yielding, reaching an ultimate displacement of 30 mm. For concrete is considered. The numerical model shows a slightly
the analysis without considering the tension-stiffening effect stiffer response until a total applied load of 30 kN is reached,
(NO T.S.), a much more flexible response is observed, however but beyond that load a close agreement to the experimental
yielding of reinforcement could be captured and an horizontal curves is observed. For the analysis without tension-stiffening
load-displacement threshold is displayed, as it can be seen in (NO T.S.) consideration, the finite element model shows a
Fig. 14. In the experimental test, Ferrari reported yielding of much more flexible response as compared to the experimental
2078 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

Fig. 14. Comparison between numerical (ANALEST) and experimental results Fig. 17. Comparison between numerical (ANALEST) and experimental results
for the beam VRE. for the beam VC3.

numerical value obtained at failure by the program ANALEST


was 46 kN.
An excellent agreement can be observed from Fig. 16
by comparison between the numerical analysis and the
experimental test on the beam VB6. The numerical model
is only a little stiffer at the beam elastic range, before the
onset of cracking, region more susceptible to instrumentation
imprecision. For the analysis without considering tension-
stiffening (NO T.S.), a more flexible response is again observed,
showing the importance of considering the tension-stiffening
effect in the beam post-cracking behavior. The finite element
model predicted very well the ultimate load, but a smaller
value for the corresponding displacement was obtained by the
ANALEST program.
Fig. 15. Comparison between numerical (ANALEST) and experimental results For the beam VC3, the comparison between analysis and
for the beams VT1/VT2. experiment shows an excellent agreement in the elastic range
(see Fig. 17), but beyond that, for loads higher than 25
kN, the numerical model becomes slightly stiffer than the
experimental model. In this example, the analytical curves
obtained considering and not considering tension-stiffening
(NO T.S.) are very close, since for this beam the elastic modulus
of concrete is low, yielding a high value for the modular ratio n,
which results in a high value for the α parameter, and therefore
the tension-stiffening effect becomes very small.

8. Extension of the TS model to 2D constitutive models

In order to extend the TS model to allow for the use in 2D


constitutive models, by taking into account the angle between
the cracks and the reinforcing bars, a simplified approach is
introduced. The tensile constitutive equation for concrete is
Fig. 16. Comparison between numerical (ANALEST) and experimental results defined in the principal direction 1, hence in Eq. (4) σc1 should
for the beam VB6. be used instead of σct , and the α parameter is calculated from
Eq. (11) by using an equivalent reinforcement ratio, ρeq , defined
curves. The onset of yielding of reinforcement was accurately
captured by the finite element model, corresponding to a total by:
applied load of 44 kN. However, the post-yielding response ρeq = ρx cos2 θ + ρ y sin2 θ (15)
of the beams and the ultimate displacement could not be
measured experimentally, since the instruments have been where: ρx is the reinforcement ratio in the x-direction;
removed from the specimens to avoid damage. The ultimate ρ y is the reinforcement ratio in the y-direction and
total load measured experimentally was 47 kN, while the θ is the angle between x-direction and principal direction 1.
R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080 2079

already been extended for planar members under combined


flexural and shear (Stramandinoli [33]), by using a Timoshenko
beam element in conjunction with the 2D smeared and rotating-
cracked model, proposed by Collins and Vecchio [8], and the
modified TS model proposed in this work. Very good results
have then been obtained by comparison with experimental
testing on beams and planar frames that showed significant
shear distress. For those examples of beams under combined
bending and shear where flexural cracks are dominant, the
Bernoulli beam element still gives very good results in the post-
Fig. 18. 2D cracking parameters. cracking range, at service loads. The ANALEST program is
currently been extended to also allow for the analysis of 3D
These parameters are illustrated in Fig. 18. reinforced concrete frames.
With this modification the TS model can also be applied
to beams (or planar elements) subjected to combined bending Acknowledgments
and shear. This modified TS model was implemented more
recently into the ANALEST program, by considering the The authors gratefully acknowledge the scholarship granted
Timoshenko beam theory in the FE formulation, which takes by Conselho Nacional de Desenvolvimento Cientı́fico e
shear deformation into account. Each section layer is then Tecnológico (CNPq) to the first author of this paper. Thanks
considered under a biaxial stress state, and the 2D constitutive are also due to researchers Juvandes, L.F.P.; Beber A.J. and
model developed by Collins and Vecchio [8] is adopted. This Ferrari, V.J. for making available the experimental data used
Timoshenko beam element is more suitable for members where in this work.
significant shear cracks develop, and it has shown excellent
correlation with experimental results from 3-point bending tests References
on beams with low transverse reinforcement ratio (Bresler and
[1] Branson DE. Design procedures for computing deflection. ACI Journal
Scordelis [30]) and also from plane frame tests (Vecchio and 1968;65(8):730–42.
Balopolou [31]; Ernst et al. [32]). [2] Cosenza E. Finite element analysis of reinforced concrete elements in a
cracked state. Computers and Structures 1990;36(1):71–9.
9. Conclusions [3] Gilbert RI, Warner RF. Tension stiffening in reinforced concrete slabs.
Journal of the Structural Division ASCE 1978;104(2):1885–900.
[4] Choi CK, Cheung SH. A simplified model for predicting the shear
In this work a novel tension-stiffening model for reinforced response of reinforced concrete membranes. Thin-Walled Structures
concrete members is developed. An exponential decay curve is 1994;19:37–60.
utilized for the concrete tensile stress–strain curve in the post- [5] CEB. Cracking and deformation. Bulletin d’information N ◦ 158. Paris,
cracking range. The exponential decay parameter (α) is defined France; 1985.
[6] Scanlon A, Murray DW. Time-dependent reinforced concrete slab
as a function of the reinforcement ratio (ρ) and of the steel- deflections. Journal of the Structural Division 1974;100(8):1911–24.
to-concrete modular ratio (n), and is derived taking as a basis [7] Lin CS, Scordelis AC. Nonlinear analysis of RC shells of general form.
the CEB [5] tension-stiffening model. The model is initially Journal of the Structural Division ASCE 1975;101(3):523–38.
validated by comparison with other tension-stiffening models [8] Collins MP, Vecchio FJ. The modified compression-field theory for
and also with experimental pull-out tests on R/C bars. The reinforced concrete elements subjected to shear. ACI Journal 1986;83(2):
219–31.
proposed tension-stiffening model is implemented into a finite [9] Stevens NJ, Uzumeri SM, Collins MP. Analytical modeling of reinforced
element program, named ANALEST, that allows for nonlinear concrete subjected to monotonic and reversed loadings. Report. Canada:
analysis of reinforced concrete beams that display a dominant University of Toronto; 1987.
flexural behavior. The program is then applied to analyze [10] Balakrishnan S, Murray DW. Concrete constitutive model for NLFE
several beams tested under 4-point bending in different research analysis of structures. ASCE Journal of Structural Engineering 1988;
114(7):1449–66.
laboratories. Comparison between numerical and experimental [11] Massicote B, Elwi AE, MacGregor JG. Tension-stiffening model for
results, in terms of load-displacement curves, showed very good planar reinforced concrete members. ASCE Journal of Structural
agreement. The analyses showed that the tension-stiffening Engineering 1990;106(11):3039–58.
effect plays an important role in the post-cracking behavior, [12] Floegl H, Mang HA. Tension stiffening concept based on bond slip.
especially for the beams with α ≤ 0.072. The proposed tension- Journal of the Structural Division ASCE 1982;108(12):2681–701.
[13] Gupta A, Maestrini SR. Tension-stiffness model for reinforced concrete
stiffening model proved to be very efficient allowing both, bars. ASCE Journal of Structural Engineering 1990;116(3):769–91.
easy implementation and numerical stability and, additionally, [14] Wu Z, Yoshikawa H, Tanabe T. Tension stiffness model for cracked
representing the concrete tension-stiffening effect in reinforced reinforced concrete. ASCE Journal of Structural Engineering 1991;
concrete members with different reinforcing and modular 117(3):715–32.
ratios. The proposed tension-stiffening model can also be [15] Russo G, Romano F. Cracking response of RC members subjected to
uniaxial tension. ASCE Journal of Structural Engineering 1992;118(5):
extended to planar members when the crack angle with respect 1172–90.
to the longitudinal rebars is different from 90◦ by means of [16] Choi CK, Cheung SH. Tension stiffening model for planar reinforced
an equivalent reinforcing ratio. The ANALEST program has concrete members. Computers and Structures 1996;59(1):179–90.
2080 R.S.B. Stramandinoli, H.L. La Rovere / Engineering Structures 30 (2008) 2069–2080

[17] Kwak HG, Song JY. Cracking analysis of RC members using polynomial [26] La Rovere HL, Chimello AA, Stramandinoli RSB. Modelo de Elementos
strain distribution function. Engineering Structures 2002;24:455–68. Finitos de Barra (2D) para Análise Não Linear de Vigas de Concreto
[18] CEB-FIP Model Code 90. Design code. London (England); 1993. Armado. Paper presented at XXIV CILAMCE (Iberian Latin American
[19] D’Avila VMR, Campos Filho A. Estudo Comparativo entre as Formas congress on computational methods in engineering). Ouro Preto, Minas
Distribuı́da e Incorporada de Representação da Fissuração em Peças de Gerais, Brazil; 2003.
Concreto Armado via Método dos Elementos Finitos. Paper presented [27] Ferrari VJ. Reforço à Flexão em Vigas de Concreto Armado com
at V Simpósio EPUSP sobre Estruturas de Concreto. São Paulo (Brazil); Manta de Fibras de Carbono: Mecanismos de Incremento de Ancoragem.
2003. Disseration. Florianópolis (Brazil): Federal University of Santa Catarina;
[20] Hegemier GA, Murakami H, Hageman LJ. On tension stiffening in 2002.
reinforced concrete. Mechanics of Materials 1985;4(2):161–79. [28] Beber AJ. Avaliação do Desempenho de Vigas de Concreto Armado
[21] Figueiras JA. Practical approach for modeling the nonlinear response of Reforçadas com Lâminas de Fibra de Carbono. Dissertation. Porto Alegre
RC shells. Computational Modeling of Reinforced Concrete Structures (Brazil): Federal University of Rio Grande do Sul; 1999.
1986;217–53. [29] Juvandes LFP. Reforço e Reabilitação de Estruturas de Betão Usando
[22] Mari AR. Nonlinear geometric, material and time dependent analysis of Materiais Compósitos de “CFRP”. Ph.D. thesis. Portugal: University of
three dimensional reinforced and prestressed concrete frames. Report. Porto; 1999.
Berkeley (California): University of California; 1984. [30] Bresler B, Scordelis AC. Shear strength of reinforced concrete beams.
[23] Chan EC. Nonlinear geometric, material and time dependent analysis ACI Journal 1963;60(1):51–74.
of reinforced concrete shells with edge beams. Ph.D thesis. Berkeley, [31] Vecchio FJ, Balopolou S. On the nonlinear behavior of reinforced concrete
(California): University of California; 1982. frames. Canadian Journal of Civil Engineering 1990;17:698–704.
[24] Hognestad E. A study of combined bending and axial load in reinforced [32] Ernst GC, et al. Basic reinforced concrete frame performance under
concrete members. Bulletin Series 1951, 399:128. Urbana (Illinois): vertical and lateral loads. ACI Journal 1973;70(4):261–9.
University of Illinois. [33] Stramandinoli RSB. Modelos de elementos finitos para análise não linear
[25] Mander JB, Priestley JN, Park R. Theoretical stress-strain model for fı́sica e geométrica de vigas e pórticos planos de concreto armado.
confined concrete. ASCE Journal of Structural Engineering 1988;14(8): Ph.D. thesis. Florianópolis (Brasil): Federal University of Santa Catarina;
1804–26. 2007.

Você também pode gostar