Mecánica Computacional Vol XXXIV, págs.

655-679 (artículo completo)

Sebastián Giusti, Martín Pucheta y Mario Storti (Eds.)
Córdoba, 8-11 Noviembre 2016

NON-LINEAR SECTIONAL ANALYSIS OF REINFORCED CONCRETE

MEMBERS

Mauro Poliottia , Oscar Möllera,b , Juan P. Ascheria and Pablo L. Sierraa

a Instituto de Mecánica Aplicada y Estructuras (IMAE),Facultad de Cs.Ex., Ingeniería y Agrimensura,
Universidad Nacional de Rosario, Riobamba y Berutti, 2000 Rosario, Argentina,
poliotti@fceia.unr.edu.ar, http:// www.imae.fceia.unr.edu.ar
b Consejo de Investigaciones (CIUNR), Universidad Nacional de Rosario, Maipú 1065, 2000 Rosario,
Argentina, http:// www.unr.edu.ar

Keywords: Reinforced Concrete Sectional Analysis, Plastic-Damage Model, Unified Non-

linear Library.

Abstract. In the most general case 3D reinforced concrete frame are subjected to complex load states
that lead to a combination of axial and shear forces, torsion and bending moments on each frame element.
Material non-linearity produce an anisotropic behavior that leads to the coupling between all the six
efforts. In this work a sectional model and code for non-linear analysis of reinforced concrete member
developed by other authors is presented. This model is able to reproduce the complex coupled behavior
of a reinforced concrete section and can be used as a constitutive equation on each integration point of
any flexibility based frame model. In order to contribute to the improvement of the sectional model and
the implemented code, a plastic-damage model based in plasticity theory and fracture-energy damage
concept is presented and implemented to model concrete behavior. Also a unified library of non-linear
solver schemes is presented and implemented to allow the code to deal with highly non-linear problems.
Finally the code with the new constitutive law and non-linear library is tested trough a series of examples
showing the capabilities of the sectional analysis.

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

656 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

1 INTRODUCTION
Reinforced concrete frame structures can be subjected to complex dynamic or static load
states that lead to a combination of axial and shear forces, torsion and bending moments on
each frame element. Non-linear behavior of concrete and steel, specially cracking on concrete
and yielding of reinforcement produce an anisotropic response that cause the coupling between
all the six efforts on the three-dimensional case. In order to trace accurately the response of
such structures the interaction of all the efforts must be taken into account.
Traditionally to capture this phenomenon, complex 2D or 3D finite element analysis was
made, but its high computational cost is not suitable for usual engineering practice. Beam
elements are widely spread between structural engineers because of their robustness and the
time savings in modeling and post-processing steps of analysis, also because beam models
are more related to engineers reasoning and to design process prescribed by most of current
construction codes than a full 3D analysis.
Nevertheless most of beam models fail to reproduce the interaction between normal and
tangential efforts. One family of elements is worth to be mentioned from the wide range of
beam models that exists nowadays, these are fiber beam models (Taucer et al., 1991; Möller
et al., 2009; Poliotti et al., 2013) which are able to reproduce the interaction between axial
force and bending moments, in order to do this on each integration point of the element they
discretize the cross section of the beam into concrete and steel fibers each one with his uniaxial
constitutive law, and after calculating the elongation of each fiber through a proper kinematic
hypothesis (Navier-Bernoulli) they can calculate the force contribution of each fiber and the
efforts on the cross section see Fig. (1). But it is because the inter-fiber equilibrium is neglected
that they cannot reproduce the interaction between normal and tangential efforts.

Figure 1: Fiber beam model

In order to account the interaction of all possible efforts and with the section analysis in-
troduced by the fiber models as a basis, Bairán (2005); Bairán and Marí (2006a,b) introduced
a new model for sectional analysis that reproduces coupling of normal and tangential efforts.
This model was implemented on a finite element environment called TINSA (Total Interaction
Nonlinear Section Analysis). This model can be used as a constitutive equation for general
beam elements at their integration points.

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 657

First in this work the study of the sectional model TINSA is presented. Then, in order to
contribute to the improvement of the robustness of the finite element code, a coupled plastic-
damage constitutive law for concrete based in the plasticity theory with hardening variables
defined by the fracture-energy damage concept first introduced by Lubliner et al. (1989) and
then extended by Lee and Fenves (1998) is introduced and implemented in the FE code. Also a
unified library of nonlinear solvers including load, displacements, arc-length and work control
methods introduced by Leon et al. (2011) is presented and implemented in order to allow the
FE code to deal with highly nonlinear problems. Application examples and study cases are
presented in order to test the sectional model with the constitutive law and the nonlinear library.
Finally conclusion and further developments are summarized.

2 SECTIONAL MODEL - TINSA

In the following the analytic model and finite code TINSA are presented. This model con-
siders the inter-fiber equilibrium explicitly leading to the interaction between tangential and
axial efforts and constitutes a powerful tool to study the behavior of cross sections of beams
under complex load states. The model is more expensive in a computational point of view that
classical beams elements, but it can reproduce 3D complex states without the cost of a full 3D
analysis.

2.1 General assumptions

The domain that’s considered in the following formulation is an arbitrary shaped cross sec-
tion belonging to a beam-column element. The sections is defined in the y-z plane with normal
x parallel to the beam axis. The cross section is intersected by the beam axis in a point defined
by the vector r0 . The generalized displacements of the section at this point are six: three rigid
body translations (u0 , v0 , w0 ) and three rigid body rotations (θx , θy , θz ). See Fig. (2).

Figure 2: Domain, local axis and generalized displacements

The main assumption of the sectional model is that the displacement field at the cross section
can be calculated by superposition of two orthogonal fields, a "plane section" displacement field
consistent with Navier-Bernoulli’s hypothesis uP S , and a "warping-distortion" displacement
field which can reproduce out of plane warping and section’s change of shape uW . See Fig. (3).

u = uP S + uW (1)

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

658 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

As small perturbations hypothesis is assumed, the same decomposition can be straightfor-

ward made with strains.

ε = Lu = LuP S + LuW = εP S + εW (2)

Where L is the linear operator:
∂
L = Lx + Lyz = Ex + Lyz
∂x
 ∂ T  ∂ ∂ T
0 0 0 0 0 0 0 0 ∂y ∂z
0 (3)
∂x
=  0 0 0 ∂
0 0  + 0 ∂
∂y
0 0 0 ∂
∂z

∂x
∂ ∂ ∂
0 0 0 0 ∂x
0 0 0 ∂z
0 0 ∂y

Figure 3: Plane Section - Warping Distortion decomposition

The same decomposition can be made with stresses if the material is linear nevertheless if
the material is nonlinear it can be proved that this decomposition is valid in an incremental way.

σ̇ = σ̇ P S + σ̇ W (4)

2.2 Navier-Bernoulli’s theory. Plane section displacement field

The basic assumption of NB beam theory is that plane sections remain plane and orthogonal
to beams axis after deformation. This kinematic hypothesis can be written as follows:

v0′ = θz ; w0′ = −θy (5)

The displacement field is defined only by the plane section field being the warping distortion
displacement field null uW = 0. The vector of plane section displacement is computed using
the generalized displacements of the section us (x).
 
u0
  v0 
1 0 0 0 −z y  


PS PS    w0 
u (x, y, z) = N (y, z)us (x) = 0 1 0 −z 0 0   (6)
θ
0 0 1 y 0 0  x
 θy = −w0′ 


θz = v0′

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 659

The resulting strains corresponding to the PS displacement field can be calculated using
Eq. (3) and taking into account Eq. (5).

εP S = B P S (y, z)es (x) (7)

Where BP S (y, z) is the strain interpolation matrix according to plain section hypothesis and
es (x) = [ε0 , γ0y , γ0z , φy , φz ]T is the generalized strain vector of the section or beam defined as
usual with ε0 being the axis elongation, γ0y and γ0z are the generalized shear deformations, φx
as the torsion curvature and finally, φy and φz being the bending curvatures. In fact only the
elongation and the three curvatures are relevant for the plane section strain calculation so the
matrix form of Eq. (7) can be rearranged in a shorter form.

εP S = B ∗ P S (y, z)e∗s (x) (8)

2.3 Warping distortion displacement field

Differently from classical approaches, this model calculates the warping distortion displace-
ment field in way to consider the complete 3D inter-fiber equilibrium equations explicitly.

div σ = LT σ = 0 (9)
The previous equation can be rewritten in its weighted residual form:
Z Z Z 
T T T T
δu L σdΩ = δu L σdS dL = 0 (10)
Ω L S

Where the term between parentheses can be interpreted as the equilibrium residual of a
differential portion of the beam.
Z Z Z
T T (i.b.p)
R(x) = δu L σdS = δu Ex σ dS − LTyz δuT σdS = G (σ ′ ) − F (σ) (11)
T T ′
S S S

The equilibrium residual, see Eq. (11), can be interpreted as the projection of equilibrium
equations on the space of displacements used that means that in order to be in equilibrium in
the displacement space, the overall residual along x has to be null.
First the displacement space is reduced to the plane section displacement space, and the
residual is computed as follows:

Z Z
PS PS PS PST T
R (x) = G ′
(σ ) − F (σ) = δuTs N ExT σ ′ dS − δuTs Lyz N P S σdS (12)
S S

Making Eq. (12) equals to zero on each cross section of the beam for any δus constitutes
the system of equilibrium equation of the beam element. Also RP S (x) = 0 represents the
equilibrium among sections of the beam. The integrity of each section meaning the equilibrium
among all the fibers in the section is not considered, that’s because we reduced the space of
possible solution only to uP S , so the rest of 3D equilibrium conditions is lost.
Tangential forces (shear forces or torsion moments), produces a stress state that tends to
dissociate or delaminate the section’s fibers, at this point its possible to differentiate the equilib-
rium at the beam level (RP S (x) = 0) the one among sections of the beam, and the equilibrium
at the section level, the one among fibers of the section. See Fig. (4).

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

660 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

Figure 4: Equilibrium at beam and section level

In order to account for equilibrium among the fibers a more general displacement field space
has to be used, but is worth to notice that projecting the equilibrium equations in the whole
admissible displacement space is equivalent to the full 3D problem and the benefits of using a
beam formulation are lost.
Now in the same way as with the plane section displacement field the residual can be com-
puted in the warping-distortion field.
Z Z
W W W WT T
′
R (x) = G (σ ) − F (σ) = δus Ex σ dS − Lyz δuW
T ′
s σdS (13)
S S
The equilibrium conditions can be resumed by the system composed by expressions (12)
and (13) where the implicit unknowns are uP S and uW . This system can be rewritten using the
stress decomposition presented in Eq. (4).



RP S (x) = GP S (σ ′ ) − F P S σ P S − F P S σ W = 0


(14)
RW (x) = GW (σ ′ ) − F W σ P S − F W σ W = 0

One way to solve this system is to impose RW (x) = 0 at each section of the beam and calculat-
ing σ W as a function of uP S .
−1 


σW = F W GW (σ ′ ) − F W σ P S = H W σ′, σP S (15)
Replacing Eq. (15) on the first equation of system (14) represents the traditional beam equi-
librium equations, but considering both components of the stress field.
Expression (15) allows to define a cross section constitutive relation. At beam level equi-
librium can be solved using frame elements formulated in the standard fashion. Its remarkable
that no fixed shaped of strains or stresses is needed, instead strain patterns are computed from
inter-fiber equilibrium in the sectional level.
The warping-distortion displacement field has to be free of body rigid movements, this dis-
placements coincide with the subset of generalized displacements contained in the PS hypothe-
sis (u0 , v0 , w0 , θx ) . A priori both fields, PS and W, can reproduced the previous displacements
modes, and uniqueness of solution is not guaranteed. This can be solved forcing uW to satisfy
following equation:
 R     
u 1 0 0  W  0
Z
 vR   0 1 0  uW  0 
 R =     
 w   0 0 1  v W dS =  0  (16)
S w
θxR 0 −z y 0

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 661

This condition is in fact the orthogonality between uP S and uW . And also it can be inter-
preted as imposing the projection of uR on any virtual variation of the rigid displacement to be
null.
The warping distortion field is assumed to have the following form:

uW = U (y, z)â (17)

Where U is a matrix of functions in the section space weighted by a coefficient vector â,
defining the distortion of the section. This vector is independent of the x coordinate.
Calculating the residual Eq. (13) on this subspace and using the definition of the plane section
strains in expression (8) one obtains:
Z  Z 
W T T ∗P S ∗′ T ∗P S
R = U Ex DB dS es − Lyz (U ) DB dS e∗s
S Z S
(18)
− Lyz (U )T DLyz (U ) dS â = 0
S
= A1 e∗s ′ − B1 e∗s − C1 â = 0
Where D is the material constitutive matrix. By the previous expression the coefficient
vector â can be expressed as a function of the generalized strain vector of the section relevant
to the plane section hypothesis and its derivatives ξ ∗ .
The distortion displacement field is an implicit function of the generalized strains relevant
to the PS hypothesis of the beam. Also uW is determined from information on the same x
coordinate and non-local information is obtained through the derivatives of the generalized
strains.

uW = U (y, z)â = U (y, z) C1−1 A1 e∗s ′ − B1 e∗s = U (y, z)H1 ξ ∗ (19)

2.4 Local solution of the warping distortion field

The warping distortion field presented in the previous section is defined through the deriva-
tives of the generalized strains of the section which is a non-local information. In order to
make uW only a function of the complete vector of generalized strains es , in other words to
express the displacement field only with local information, a relationship between e∗s and e∗s ′ is
needed. This can be done thanks to an additional equilibrium condition and to the fact that the
components of e∗s ′ are not independent.
First of all ε0 ′ and φx ′ can be calculated from the conditions that Nx ′ = 0 and Mx ′ = 0,
because they can be condensed in the cross section domain. Also both derivatives of curvature
are related to the corresponding generalized shear deformations γ0y and γ0z . Taking this into
account one may write the following relationships.

uW = A (es ) (20)

εW = LA (es ) = B W es (21)
Z Z
PS PS W
σ = Ddε ; σ = DdεW (22)

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

662 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

2.5 Generalized stresses. Section’s stiffness matrix

The calculation of generalized stresses is made through the Virtual Work Principle as follows:

Z Z Z 
δeTs ss dL = δεTs σdS dL
L L S
Z Z Z  (23)
PST WT
⇒ δeTs ss = δεTs σdS = δeTs B σdS + B σdS
S S S
Z Z
PST T
ss = B σdS + B W σdS = sPs S + sW
s (24)
S S
To calculate the sections stiffness matrix of the cross section, linearization of expression (24)
has to be made: Z Z
PST T
ṡs = B D ε̇dS + B W D ε̇dS (25)
S S
After decomposition of strains and using expression (21) and (7) the section’s stiffness
matrix results:
Z h i
PST PS PST W WT PS WT W
Ks = B DB + B DB + B DB + B DB dS (26)
s

2.6 Numerical implementation

The model presented in the previous sections is implemented in a finite element environment
for sectional analysis where the cross section is discretized into bidimensional elements (quad-
rangular or triangular elements can be used) with first order Lagrange polynomials as interpola-
tion function and Gauss-Legendre quadrature as integration rule that represent the solid matrix
of the section. Also 2 node linear Lagrange elements representing transverse reinforcements
is used. Finally a point element representing longitudinal reinforcement intersecting the cross
section is used. This three elements are the minimum required to model a typical reinforced
concrete section. See Fig. (5).

Figure 5: Elements for sectional discretization

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 663

3 PLASTIC-DAMAGE MODEL FOR CONCRETE

Four material constitutive models where originally implemented in the FE code. Three of
them are 3D constitutive models to represent material behavior of the solid matrix of the cross
section, an additional uniaxial constitutive law is implemented to represent material behavior of
either transverse or longitudinal reinforcements. In order to contribute to the improvement of
the finite element code and to extend his capabilities, a continuous concrete constitutive model
is presented and implemented.
The constitutive model implemented in the finite element code is the one presented by Lee
and Fenves (1998) that is in fact an extension to the cyclic loading case of the Barcelona Model
presented by Lubliner et al. (1989). This is a plastic-damage model based in the plasticity theory
that introduces two damage internal variables using the concept of fracture-energy damage, also
it introduces two stiffness degradation scalar variables based in isotropic continuum damage
mechanics and in the effective stress concept. The model uses a new yield surface as a criterion
and a different plastic potential surface in order to match experimental data, this leads to a
non-associative flow rule that causes a non-symmetric tangent stiffness matrix.

3.1 Main features

The main aspects of the model are summarized in the following. First of all, as usual in
plasticity theories, the strain is decomposed into the elastic and plastic parts.

ε = εe + εp
εe = E −1 σ (27)
σ = E : (ε − εp )
Where E is the elastic stiffness four rank tensor. Also, introducing the concept of effective
stresses from continuum damage mechanics, it can be seen that stresses can be mapped into
effective stress through a four rank tensor D, but the hypothesis of isotropic degradation is
made so this can be simplified as follows:
1 1
σ̄ = D : σ = I:σ= σ = E0 : (ε − εp )
(1 − D) (1 − D) (28)
E = (1 − D)E0
In (28) the scalar degradation of the elastic stiffness variable D has a value between zero,
where no degradation had place, and one where full degradation is assumed (0 ≤ D < 1). Also
in the previous expression E0 is the initial elastic stiffness tensor.
Another important aspect of any plasticity model is the evolution of the plastic strains. This
is made as usual by means of the normality rule, so a plastic potential function is needed. Is
worth to mentioned that to improve the numerical implementation, the elastoplastic response is
separated, by means of the effective stress concept, of the degradation process so this plastic
potential function has to be defined in the effective stress space.
∂φ
(σ̄) ε˙p = λ̇ (29)
∂ σ̄
Where λ is the plastic consistency parameter, and φ is the plastic potential function.
In order to take into account the different damage states another internal variable is intro-
duced, κ. The components of this 2D vector variable are the damage in tension and in compres-

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

664 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

sion parameters κ (κt , κc ), the evolution of this variable is given by:

κ̇ = λ̇H (σ̄, κ) (30)

A yield criterion is needed to complete the elastoplastic model. This yield surface is able to
evolve during the damage process so it will be seen that the damage variable plays the role of
the hardening variable in classic plasticity models. Here hardening has to be understand in a
broad sense including softening as well as hardening processes. Eq. (31).

Fe = (σ, ft , fc ) ≤ 0 (31)
Where (ft , fc ), the uniaxial strength functions, are explicit functions of (κt , κc ) and they
control the evolution of the yield surface. Introducing the stiffness degradation variable D, see
Eq. (28) which is also an explicit function of κ, the yield surface Eq. (31) can now be rewritten
in the effective stress space and also as function of the damage variable, which plays the role of
hardening variable.

F (σ̄, κ) ≤ 0 (32)
The elastoplastic response of the plastic-damage model is described in terms of the effective
stresses, the plastic strains and the damage variable by equations (28), (29), (30) and (32).
Additionally loading/unloading conditions are needed to represent the cyclic behavior. These
are the Kuhn-Tucker optimization conditions and the consistency condition.

λ̇ ≥ 0
λ̇F = 0 (33)
F ≤0

Finally the total stresses are computed after evaluating the degradation variable D,

σ = [1 − D (κ)] σ̄ (34)

3.2 Damage variable - Evolution of internal variables

The damage variable in cohesive-frictional materials can be defined as a relative measure of
the energy dissipated during an inelastic process.
For simplicity the damage variable is presented in the for a uniaxial process ℵ, either in
compression or tension ℵ ∈ (t, c). Assuming that from the stress-strains diagrams obtained
through uniaxial tests the stress-plastic strains diagrams can be constructed, see Fig.6, and that
the area under these curves is assumed to be finite, the damage variable can be defined as
follows:
Z p
1 ε
κℵ = σℵ (εp ) dεp (35)
gℵ 0
Z ∞
gℵ = σℵ (εp ) dεp (36)
0

The quantities (gt , gc ), in expression (36), cannot be only material properties. It is well
known that frictional solids tend to concentrate the dissipation of energy in an infinitesimal

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 665

Figure 6: Uniaxial curves (σ, εp ) in tension and compression

volume. In order to treat this phenomena by means of a continuous solid mechanic theory, it
is necessary to introduce a characteristic length that will depend on the FE mesh size. The
definition of this length was explained in detail by Oliver (1989) and by Bazant (1976, 1983).
So both are now defined as:
Gℵ
gℵ = (37)
lch
Where lch is the characteristic length of the mesh and Gℵ is assumed to be a material property.
In the tension case can be interpreted as the specific fracture energy defined as the energy
required to form a unit area of crack. And in the compression case, only the post peak part of
the curve is mesh-sensitive so it can be decomposed in two parts, one mesh-independent and
other mesh-dependent.
Taking into account (35) the damage variable in a uniaxial process can be written:
1
fℵ (κℵ ) ε˙p
κ˙ℵ = (38)
gℵ
To extend the uniaxial version of the evolution of damage to the multiaxial case, the scalar
plastic strain rate in the general case can be calculated as follows:

p 
 p
ˆ ε̇ˆmax
ε˙p = δtℵ r σ̄ ˆ ε̇ˆmin
+ δcℵ 1 − r σ̄ (39)
Where δ is the Kronecker’s delta, ε̇ˆmax
p
and ε̇ˆmax
p

are the minimum and maximum eigenvalues
ˆ
of the plastic strain tensor. The weight factor r σ̄ depends on the eigenvalues of the effective
stress tensor and its equal to zero for triaxial compression and equals to one for triaxial tension.
Evolution laws in expressions (29) and (30) can be modified into relations in the principal
space of effective stresses and strains.


ˆ p = λ̇ ∂φ σ̄
ε̇ ˆ
ˆ
∂ σ̄ (40)


κ̇ = λ̇Ĥ σ̄ ˆ, κ

3.3 Yield surface and plastic potential function

The yield surface used in this model was first introduced by Lubliner et al. (1989) as an im-
provement of classical yield functions for frictional materials as the Mohr-Coulomb or Drucker-

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

666 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

Prager criteria. Fig.7

1 h p i
F (σ, κ) = αI1 + 3J2 + β (κ) hσ̂max i − cc (κ) ≤ 0 (41)
1−α
Where I1 = tr (σ) , J2 = 12 s : s, α and β are dimensionless parameters and σ̂max is the largest
eigenvalue of the stress tensor.
The α parameter in (41) is computed in order to match experimental data specially to con-
trol the enhancement in the yield stress that concrete exhibits in biaxial compression, so α is
determined by comparing the uniaxial with the biaxial yield stress.
The parameter was firstly introduced as a constant by Lubliner et al. (1989) for monotonic
loading cases, and then it was defined as function of the damage variable by Lee and Fenves
(1998) in order to extend the model to the cyclic loading case. This parameter allows the model
to adjust the relationship between the uniaxial strength in compression and in tension.

Figure 7: Yield surface in the Westergaard space

Is worth to notice that the yield surface presented is open in the direction of hydrostatic com-
pression. This is equivalent to assume that no yielding is reached for hydrostatic compression.
It can be seen that the yield surface has a conic shape in the effective stress space, in the π plane
also it describes a three vertex curve, and also has a vertex in the uniaxial compression point and
in the biaxial tension point. All this singularities produce a multiple definition of the normal
vector of the yield surface which is needed for the definition of the tangent stiffness tensor.
The plastic flow rule in expression (40) states that the plastic flow direction is defined by
the normal vector of the plastic potential surface. Lee and Fenves (1998) proposed the linear
version of the Drucker-Prager criteria, but is well known that this surface has a singularity at the
triaxial tension point, so Omidi and Lofti (2010) used a hyperbolic version of Drucker-Prager
surface in order to avoid this singularity.
q
φ (σ̄) = (ε1 αp ft0 )2 + 2J2 + αp I1 (42)
Where ε1 is a constant that adjusts the eccentricity of the hyperbolic function from the apex
in the linear function. The αp parameter controls the dilatancy of the model and is held constant
during the whole analysis. The choice of a non-associative plasticity is made in order to fit
properly experimental results, using two different functions one for the yield surface and one
for the plastic potential function will cause the non-symmetry of the stiffness tensor. Finally the
plastic potential function presented in (42) is a continuous and smooth function where the flow
direction is always uniquely defined.

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 667

3.4 Stiffness degradation

The stiffness degradation caused by the microcracking either in tension or compression is
treated separately from the elastoplastic response. In the cyclic loading case the phenomenon
of crack opening and closing during unloading or reloading process has to be represented by
the model. See Fig.(8).

D = D (σ̄, κ) = 1 − [1 − st (σ̄) Dt (κt )] [1 − sc (σ̄) Dc (κc )]

0 ≤ (st , sc ) ≤ 1

(43)
st (σ̄) = 1 − wt r σ̄ ˆ



sc (σ̄) = 1 − wc 1 − r σ̄ ˆ
Where wt and wc are the stiffness recovery factors that will be chosen in order to match the
experimental data. In the case of concrete this factors are wt = 0 and wc = 1.

Figure 8: Stiffness recovery factors

3.5 Integration and numerical implementation - Stiffness matrix

The numerical integration of the plastic-damage model is made through time discretization of
the evolution equations and the backward-Euler method is used, this method is unconditionally
stable and has first order precision. It can be seen also has the generalized trapeze method with
the α parameter equal to one.
The numerical algorithm in Fig.(9) is a strain-driven discrete problem. Each new step will
have has an input the last converged step values of the problem variables {εpn , κn , Dn }. The
last input of the algorithm is the strain tensor of the new step εn+1 .
The stress computation can be arranged in a three step calculation method. First an elastic
predictor is made by means of the trial effective tensor, the second step is a plastic corrector
step and finally in an uncoupled way a damage corrector is introduced. In order to solve the
stress computation in a more efficient way a spectral returning algorithm based on the spectral
decomposition of the effective stresses is used. This is done because of the form of this yield
function, it can be seen in (41) that the function includes principal stress terms in addition of
the tensor invariants so in this cases is more efficient to work with the eigenvalues matrix.

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

668 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

Figure 9: Constitutive model flow chart

Both evolution equations of the internal variables (40) are solved by means of the backward-
Euler method, and the nonlinear problem in the evolution of the damage variable is solved by a
Newton-Raphson solver.
Finally, in nonlinear problems the rate of convergence of the nonlinear solver depends strongly
on the tangent stiffness matrix that’s being used. Lee and Fenves (1998) presented the contin-
uum tangent stiffness of the model, and later Lee and Fenves (2001) presented an algorithmic
tangent stiffness consistent with the stress updating algorithm described in the previous section.
Both matrices were analyzed but in this work a numerical stiffness matrix was implemented
because of the simplicity in the implementation.
The numerical stiffness matrix is calculated once the updating stress algorithm reached the
new converged state. Each column of the stiffness matrix is calculated introducing a perturba-
tion on the corresponding term of the strain tensor.
q
εen+1 = {ε1n+1 , · · · , εqn+1 + h, · · · , ε6n+1 }T (44)

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 669

Where h is a numerical perturbation that is determined as multiple of the machine precision.

With this new strain tensor, and using the updating stress algorithm a new perturbed stress is
calculated
 q
 q
εen+1 ; εn+1 ; κn+1 ; Dn+1 → U pdating stress algorithm → σ
e n+1 (45)
Finally the q-column of the tangent matrix is calculated as:
1 q


Eq =
e n+1
σ − σn+1 (46)
h
This procedure is repeated for each component of the strain tensor and the stiffness matrix is
finally the concatenation of the six columns calculated with expression (46).

4 UNIFIED LIBRARY OF NONLINEAR SOLVERS

The TINSA code for sectional analysis has two different ways to solve the nonlinear problem.
A Newton-Raphson method as load control method and a displacement control method are
implemented as well as a substepping strategy in order to improve the convergence of the solver.
In order to improve the ability to treat highly nonlinear problems a new library of nonlinear
solvers presented by Leon et al. (2011) is implemented in the TINSA code. This library consists
in a unified framework that allows to introduce different nonlinear solution schemes with the
same algorithm.
The unified framework is presented in the following. Four methods are implemented using
this framework the load control method (Newton-Raphson), the displacement control method,
the arc-length control method and the work control method. Is worth to mention that even only
this four methods were implemented the unified scheme allows the implementation of other
control methods.

4.1 Unified nonlinear framework

The unified nonlinear formulation is a usual incremental-iterative procedure where the load
is applied in incremental steps and iterations are performed at each incremental step until the
convergence criterion is satisfied. See Fig.(10).

Figure 10: Incremental-iterative procedure

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

670 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

In structural nonlinear analysis the unbalanced force vector to be minimized in i-incremental

step and in the j-iteration can be written as follows:


rji = pij − q uij (47)
Where pij , is the vector of applied external forces and qji is the vector of internal forces
which is a function of the displacements uij . In the incremental-iterative procedure the loads and
displacements are computed by adding to the previous converged step the incremental updates
at the current j–iteration.

uij = ui−1 + ∆uij

(48)
pij = pi−1 + ∆pij
On each iteration the incremental updates presented in (48) are calculated in a similar way by
adding the contribution of the previous iteration and the iterative updates at the current iteration.

∆uij = ∆uij−1 + δuij

(49)
∆pij = ∆pij−1 + δpij
Taking (48) into account int the residual vector (47) can be rewritten as:


rji = pi−1 + ∆pij − q ui−1 + ∆uij (50)
Finally the equation to be solved at the j–iteration of i–increment that governs the response
of the structure is:

i
Kj−1 δuij = pij − qj−1
i
(51)
i
Where Kj−1 , is the tangent matrix of the structure.
Introducing into equations (48) and (49) the load factor parameter λ and the reference load
vector p̄, the iterative form of the external applied load vector and the governing equation of the
system can be rewritten as follows:

pij = pi−1 + ∆pij−1 + δλij p̄

i
(52)
Kj−1 δuij = rj−1
i
+ δλij p̄
Equation (52) is a set of N equations, being N the degrees of freedom of the system, with
N + 1 unknowns: the N displacements δuij and the load parameter δλij . In order to solve the
system an additional constrain equation must be added.

aij · δuij + bij δλij = cij (53)

The system of N + 1 unknowns and N + 1 equations formed by (52) and (53) can be sum-
marized in a matrix form as follows:
 i    i 
Kj−1 −p̄ δuij rj−1


i T i = (54)
aj i
bj δλj cij
The new system’s matrix presented in (54) is no longer symmetric and also the bandwidth has
been modified from the original tangent stiffness matrix. This may cause a low computational

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 671

efficiency when solving the system by means of traditional linear solvers. To overcome this
problem the decomposition of the displacement iteration update is introduced.

δuij = δλij δup ij + δur ij

i
Kj−1 δup ij = p̄ (55)
i
Kj−1 δur ij = rj−1
i

Finally the load parameter which is needed to compute the total displacement for the j–iteration
at i–increment, can be calculated using the constraint equation (53) and the displacement de-
composition (55) as follows:

cij − aij · δur ij

δλij = i (56)
aj · δup ij + bij
The benefit of using this unified framework is the fact that each nonlinear solution scheme
differs only in the constraint equation and they can be solved by the same algorithm where only
the calculation of the parameters of equation (53) and consequently the calculation of the load
factor are different for each scheme.
The algorithm is implemented in the TINSA code following the flow chart in Fig.(11), addi-
tionally for each nonlinear solution scheme a function to calculate the load factor parameter is
implemented. Is remarkable that the code allows the simple implementation of other nonlinear
solutions by means of this “function” structure.

Figure 11: Unified scheme flow chart

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

672 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

Four different control methods are implemented under the form of the unified framework
previously presented. The load control method or Newton-Raphson methods where the external
loads are computed at the first iteration of each increment and held constant in the remaining
iterations. The displacement control method which uses a fixed displacement component as the
control parameter to trace the structural response. The arc-length control method that considers
simultaneous iterations on the load and displacements variables, the method is based in the
constraint of the solution to an arc-length. And finally the work control method which uses a
constant work increment thought the iterations of an increment, this method was introduced to
overcome to the unit problem in the arc-length method.
Each nonlinear scheme has its own advantages and weakness so the user may choose be-
tween the different schemes depending on the nonlinear problem to solve. For example the load
control method fails to reproduce load limits points and softening parts, while the displacement
can capture load limit points and softening responses but it fails to near displacements limits
points. The arc-length method is able to reproduce complex nonlinear problems with load and
displacement limit points or snap-back phenomena but it has units problems because it uses the
displacement vector where the displacement and rotation components has different orders of
magnitude, which may cause incorrect changes in the sense of the response.

5 APPLICATION EXAMPLES
In order to show the capabilities of the TINSA code with the new plastic-damage model and
the new library of nonlinear solvers three analysis of reinforced concrete sections are presented
in this chapter. The first case shows the influence of the concomitant shear force in the moment-
curvature curve of a rectangular reinforced concrete section. Finally the experimental results of
a prestressed concrete beam are compared with the numerical prediction of TINSA.

5.1 Moment-curvature of a reinforced concrete section with shear forces

Traditionally in the nonlinear analysis of reinforced concrete frames the nonlinear response
of the frame elements is characterized by the moment-curvature (M − φ) curve, so several
beam models uses this curve as an input. Also this curve allows the simply identification of
characteristic points and properties of the section as the cracking and yielding points and the
ductility of the section.
Is well known that the concomitance of axial forces has an important influence on the
(M − φ) curve, a moderate axial compression reduces the sections ductility and increases the
resisting moment. This phenomenon can be captured by traditional fiber beams element based
in the NB kinematic hypothesis. In the other hand the influence of a concomitant shear force is
not clear and models that neglect the inter-fiber equilibrium are not capable of reproducing this
phenomenon.
The (M − φ) curves of a rectangular reinforced concrete section are obtained with TINSA
using the plastic-damage model and the displacement control method for different levels of
shear force which is quantified as the shear span to effective depth ratio (M/V d).
The geometry, reinforcement and discretization of the concrete section can be seen in Fig.(12).
Also material properties of concrete, the longitudinal and transverse steel reinforcement are
summarized in Table (1) and Table (2).
The obtained curves for shear span ratios of 2, 3, 4 and 5 are presented in Fig.(13). Also the
deformed shape of the section showing the warping of the cross-section and the distribution of
the tension damage variable for the shear span ratio of 5 are presented in Fig.(14).

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 673

Figure 12: Cross-section - Geometry discretization

E0 = 25600N/mm2 Gt = 0.133N/mm
ν= 0.2 Gc = 13.3N/mm
fc′ = 28.2N/mm2 Det = 0.51
f0c = 16.9N/mm2 ec =
D 0.38
ft = 1.75N/mm2 αp = 0.25
f0b
f0c
= 1.16 ε1 =0.1 lch = 35mm
Table 1: Concrete properties

Es = 200000M P a
Longitudinal
Ep = 200M P a
Reinforcement
fy = 442M P a
Es = 200000M P a
Transverse
Ep = 200M P a
Reinforcement
fy = 400M P a
Table 2: Reinforcement properties

Figure 13: Moment-curvature for different shear levels

It can be seen that the shear has an influence in several nonlinear characteristics of the section,
the yielding moment is clearly affected by the concomitant shear force, also both post-cracking
and post-yielding stiffness are reduced when shear forces increases. Fig.(15) shows the distri-
bution of normal stresses at the same curvature level for the shear-span ratio of 5 and 2, it can be
seen that the axial stress is almost null in the cracked zone for the shear-span ratio of 5 and in the
case of the shear-span ratio of 2 where the shear influence is higher the cracked zone presents

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

674 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

Figure 14: Warping - Distribution of κt for M/V d = 5

Figure 15: σxx distribution at φ = 6x10−3 for M/V d = 5 and M/V d = 2

compression in the cracked zone this is because the concomitant shear effort that produces this
stresses. Finally it can be seen in the shear-span ration of 2 a reduction of ductility.

5.2 Sectional analysis of a prestressed concrete beam

One section belonging to the first of eight beams of an experimental campaign of prestressed
reinforced concrete beams carried out at the Laboratori de Tecnologia d’Estructures - Univer-
sitat Politècnica de Catalunya, is analyzed using TINSA with the plastic-damage model, and
numerical prediction is compared with experimental results. In the following the test character-
istics and numerical model are presented.
The test consist on a simply supported prestessed reinforced concrete beam of double T
cross-section with a point load applied by a hydraulic jack. See Fig.(16).

Figure 16: General geometry of the test (dimensions in cm)

The analyzed section corresponds to the one with highest shear effort. Fig.(17) shows the
detailing of reinforcement which is composed of active and passive steel for the longitudinal

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 675

reinforcements and passive steel is used for the transversal reinforcement.

Figure 17: Reinforcement detail of the analyzed cross section - FE mesh

The numerical model uses the finite element mesh presented in Fig.(17) where the concrete
solid matrix is meshed using quadrangular elements, the transverse reinforcement is modeled
with linear elements and longitudinal reinforcement both passive and active are represented by
point elements. In order to introduce the prestress force, in the point element that corresponds
to the prestress wire an initial pre-strain is introduce and by means of material properties the
corresponding force is applied.
The material properties were obtained by compression and Brazilian-tension tests of cylin-
drical specimens casted at the same time as the beam. Also tension tests of the reinforcement
bars were carried out in order to obtain the steel properties. Table(3) and table (4) summarizes
the material properties used in the numerical model.

E0 = 28000N/mm2 Gt = 0.142N/mm
ν= 0.2 Gc = 14.2N/mm
fc′ = 40N/mm2 Det = 0.51
f0c = 24N/mm2 ec =
D 0.38
ft = 3.18N/mm2 αp = 0.25
f0b
f0c
= 1.16 ε1 =0.1 lch = 11.3mm
Table 3: Concrete properties

Es = 200000M P a
Passive
Ep = 20M P a
Reinforcement
fy = 550M P a
Es = 190000M P a
Active Ep = 0M P a
Reinforcement fy = 1690M P a
fpu = 1860M P a
Table 4: Reinforcement properties

The load applied by the hydraulic jack was hold at the different load levels and loading-
unloading cycles were made before the load was increased monotonically until failure that was
registered at 412 KN in the hydraulic jack that corresponds to a shear effort of 313 KN in the

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

676 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

analyzed section. The cracking started with vertical cracks due to bending near the point of
application of the load with the increment of the load cracks turned and inclined propagation
took place. The tested beam presented a shear failure mode with the corresponding yielding
process of the stirrups and diagonal cracking.
The ultimate numerical load is obtained as the predicted maximum shear force obtained by
TINSA plus the vertical component of the prestress force.

VU = VT IN SA + PW IRE sin α = 377KN (57)

This value is a 20% higher than the experimental one. A better prediction can be made ad-
justing the dilatancy parameter, but is an improvement compared with other constitutive models
that predicted ultimate forces 38% higher than the experimental value.
Figure (18) presents the numerical moment-curvature curve and the shear-generalized shear
deformation curve obtained with TINSA. It can be seen that the response presents a softening
in shear and a snap back in the moment curvature-curve.

Figure 18: Numerical M − φ - Numerical V − γ

Figure (19) presents the evolution of the tension damage variable for different load levels.
The cracking appears at the web of the cross-section and as load increases the damage is prop-
agated in the web. At the ultimate load level it can be seen that the damage is concentrated
in the web of the section which is completely damaged and the head remains with no damage.
This numerical prediction corresponds with the experimental behavior because cracking in the
analyzed section is a diagonal cracking that affects the web of the cross section and at the end
of the tests no crack was identified at the head of the beam.

Figure 19: Evolution of the tension damage variable κt

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 677

Figs. (20) and (21) presents the distribution of the stresses at different load levels. At early
stages the distribution of the axial stress is linear and the tangential stress xz is parabolic in the
web of the beam as can be predicted by classical theories. Once cracking begins a redistribution
takes place and at the end of the test it can be seen that only the head has axial stresses and is
the only part of the concrete section that still resists.

Figure 20: Evolution of σx distribution

Figure 21: Evolution of τxz distribution

Figs. (22) and (23) shows the evolution of the stresses in the stirrups and the plastic de-
formation respectively. For low load levels the transverse reinforcement steel behaves linearly
and after cracking and redistribution the portion of the shear forces resisted by the stirrups in-
creases. Finally it can be seen that plastic deformation of the stirrups begins in the vertical bars
and is extended to almost the total length of the stirrup at the end of the test. This behavior is
coincident with the one observed in the experiment.

Figure 22: Evolution of stresses in the stirrups

Figure 23: Evolution of plastic deformation in the stirrups

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

678 M. POLIOTTI, O. MOLLER, J.P. ASCHERI, P.L. SIERRA

The test presented shows the usefulness of TINSA in the analysis of reinforced concrete
sections under complex load states, it can reproduce the shear failure mode that most of classical
beam models are not capable to do. Also the plastic-damage model used as constitutive law has
a good agreement with the experimental result in spite that it can be improved by a better
calibration of the material parameters.

6 CONCLUSIONS
The sectional model presented is able to reproduce the total interaction between the different
efforts that takes place in beams under complex load states and with complex material behavior
as reinforced concrete frame structures. This is done by considering the full set of equilibrium
equations differently from classical beam formulations which neglect the inter-fiber equilib-
rium. The decomposition of the problem into the structural level and sectional level allows the
sectional model to be independent of the frame formulation and can be used as a constitutive
equation on each integration point of the beam.
In order to improve the capabilities of the sectional model a thermodynamically consis-
tent constitutive law for concrete was implemented and presented. This law, based in classical
plasticity theory, introduces a new yield surface which has as hardening variables two scalar
variables defined by means of the fracture-energy damage concept. Stiffness degradation is
considered uncoupled of the elastoplastic response which allows a simple numerical implemen-
tation. Also a non-associative flow rule is used to match experimental data.
A unified library of nonlinear solvers was introduced and implemented in the FE code to
allow the analysis of highly nonlinear problems. This unified library enables the implementation
of several control methods by means of a general constraint equation. Four control methods
were implemented. This library allows the user to choose the control method that best suits the
problem to be analyzed, also more than one method can be used during a single analysis which
improves the convergence and the speed of the process.
The FE code with the new plastic-damage constitutive model and the unified library of non-
linear solvers was tested with two application examples. First a simple analysis of the moment-
curvature response of a rectangular reinforced concrete section with different concomitant shear
force was made. This example shows the influence of the concomitant shear force, which pro-
duces a reduction in the yielding moment, it also modifies the ductility of the section and the
stiffness of the section. This phenomenon cannot be reproduced with most of classical beam
elements. Finally a complex double-T prestressed reinforced concrete section was analyzed and
numerical response was compared with experimental results, good agreement was achieved by
means of the presented model, and complex behavior was predicted, shear failure was captured
and the damage field corresponded to the cracking path observed in the experiment, also the
contribution and behavior of the compression head was captured which is very important when
the web is fully damaged but the section still resist thanks to compression head.
Further work can be done to improve the presented model . In regards to the TINSA model
an improvement of the plane-section hypothesis can be made allowing the definition of subsec-
tions that can act independently of the rest of the section, this will allow the model to concentrate
curvatures in just one part of the section, this improvement comes from experimental observa-
tions of T and double-T sections where the head acts as an independent beam after the web
reached a full damage state. When it comes to the constitutive law and its implementation a
better tangent stiffness matrix can be implemented to improve the convergence of the plastic-
damage constitutive law, also the dilatancy control can be modified to better suit experimental
data. Finally, in regards to the nonlinear library, new control methods can be implemented to

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar

 Mecánica Computacional Vol XXXIV, págs. 655-679 (2016) 679

offer other possibilities to the user when it comes to control of the nonlinear solution process.

7 ACKNOWLEDGMENTS
This work has been carried out in the context of the master thesis presented by the first author
as partial fulfillment of the requirements to obtain the degree of Master in Engineering Science
of the Université Pierre et Marie Curie and École Nationale des Ponts et Chaussées, France.
This thesis was carried at Departament D’Enginyeria de la Construcció, Universitat Politèc-
nica de Catalunya under the supervision of professor Jesús Bairán. The firs author express his
gratitude to the scholarship program BECAR of the Argentinian Government for the financial
support and to professors Bairán,J. and Oller, S. for their valuable suggestions and comments.

REFERENCES
Bairán J. A non-linear coupled model for the analysis of reinforced concrete sections under
bending, shear, torsion and axial forces. Ph.D. thesis, Departament D’Enginyeria de la Con-
strucció, Universitat Politècnica de Catalunya, Barcelona, 2005.
Bairán J. and Marí A. Coupled model for the non-linear analysis of anisotropic sections sub-
jected to general 3d loading. part 1: Theoretical formulation. Computers & Structures,
84:2254–2263, 2006a.
Bairán J. and Marí A. Coupled model for the non-linear analysis of anisotropic sections sub-
jected to general 3d loading. part 2: Implementation and validation. Computers & Structures,
84:2264–2276, 2006b.
Bazant Z. Instability, ductility and size effect in strain-softening concrete. J. Eng. Mech. ASCE,
102(2):331–344, 1976.
Bazant Z. Coment on orthotropic models for concrete and geomaterials. J. Eng. Mech. ASCE,
109:849–865, 1983.
Lee J. and Fenves G. Plastic-damage model for cyclic loading of concrete structures. Journal
of Engineering Mechanics, 124(8):892–900, 1998.
Lee J. and Fenves G. A return-mapping algorithm for plastic-damage models: 3d plane stress
formulation. International Journal for Numerical Methods in Engineering, 50:487–505,
2001.
Leon S., Paulino G., Pereira A., Menezes I., and Lages E. A unified library of nonlinear solution
schemes. Applied Mechanics Review, 64, 2011.
Lubliner J., Oliver J., Oller S., and Oñate E. A plastic-damage model for concrete. Int. J. Solids
Structures, 25(3):299–326, 1989.
Möller O., Quiroz L., and Rubinstein M. Desempeño numérico de elementos de barra en análisis
no lineal estático y dinámico. Mecánica Computacional, AMCA, XXVIII:587–611, 2009.
Oliver J. A consistent characteristic lenght for smeared cracking models. International Journal
for Numerical Methods in Engineering, 28:461–474, 1989.
Omidi O. and Lofti V. Finite element analysis of concrete structures using plastic-damage
model in 3d implementation. Ing. J. Civil Engineering, 8(3):187–203, 2010.
Poliotti M., Möller O., and Ascheri J. Modelo numérico de elemento de barras discretizado en
fibras para columnas sismorresistentes. Mecánica Computacional, AMCA, XXXII:807–827,
2013.
Taucer F., Spacone E., and Filippou F. A fiber beam-column element for seismic response
analysis of reinforced concrete structures. Earthquake Engineering Research Center, Report
EERC 91-17, 1991.

Copyright © 2016 Asociación Argentina de Mecánica Computacional http://www.amcaonline.org.ar