Engineering Computations

Tension stiffening model with increasing damage for reinforced concrete

M.H.F.M. Barros, R.A.F. Martins, C.C. Ferreira,
Article information:
To cite this document:
M.H.F.M. Barros, R.A.F. Martins, C.C. Ferreira, (2001) "Tension stiffening model with increasing
damage for reinforced concrete", Engineering Computations, Vol. 18 Issue: 5/6, pp.759-785, https://
doi.org/10.1108/02644400110393616
Permanent link to this document:
https://doi.org/10.1108/02644400110393616
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Downloaded on: 01 July 2017, At: 06:58 (PT)

References: this document contains references to 11 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 882 times since 2006*
Users who downloaded this article also downloaded:
(2003),"Flexural behaviour of concrete beams internally reinforced with GFRP rods and steel rebars",
Structural Survey, Vol. 21 Iss 4 pp. 146-157 <a href="https://doi.org/10.1108/02630800310507159">https://
doi.org/10.1108/02630800310507159</a>
(2004),"Closed form solution of optimal design of rectangular reinforced concrete sections", Engineering
Computations, Vol. 21 Iss 7 pp. 761-776 <a href="https://doi.org/10.1108/02644400410565298">https://
doi.org/10.1108/02644400410565298</a>

Access to this document was granted through an Emerald subscription provided by emerald-srm:333301 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald for
Authors service information about how to choose which publication to write for and submission guidelines
are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as well as
providing an extensive range of online products and additional customer resources and services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the Committee
on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for digital archive
preservation.

*Related content and download information correct at time of download.

 The current issue and full text archive of this journal is available at
http://www.emerald-library.com/ft

Tension stiffening model with Tension

stiffening model
increasing damage for
reinforced concrete
M.H.F.M. Barros 759
Faculty of Sciences and Technology, University of Coimbra, Coimbra, Portugal Received February 2000
R.A.F. Martins Revised February 2001
Accepted February 2001
Faculty of Engineering, University of Porto, Porto, Portugal, and
C.C. Ferreira
Faculty of Sciences and Technology, University of Coimbra, Coimbra, Portugal
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Keywords Concrete, Reinforced concrete, Localization, Numerical analysis

Abstract A model for the analysis of plain and reinforced concrete structures is developed in the
present work. In the plain concrete the localisation of the microcracking within a small band is
formulated in terms of constant fracture energy. In the reinforced concrete the model considers
the cracking localisation in the element, according to the expressions of Model Code 1990. The
objective of this model is to approximate the tension stiffening effect observed in the reinforced
concrete elements submitted to tensile stresses. The reinforcing steel is analysed with an element
overlapped to the concrete element, having equivalent nodal displacements. Numerical examples
are presented and compared with experimental and other numerical results.

1. Introduction
A reliable model for the analysis of reinforced concrete structures must take into
account the properties of the two components, namely the steel and the concrete,
and the changes in these properties due to the interaction between them. The
present model calculates the deformation of the structures by the use of the finite
element method and the non-linear behaviour due to smeared cracks appearing in
any zone of the structure. The deformation of reinforced concrete elements is also
highly influenced by the bond between the components. In case of good bond
conditions the cracks are smeared into micro-cracks. The opening of a crack
reduces the shear strength in the crack face that becomes dependent in this case
on the aggregate interlock and dowel action. This interaction effect between the
two components of the reinforced concrete is formulated by the use of two distinct
zones in the element, one termed intact and the other cracked. The dimension of
the cracked zone increases with the loading and is defined according to the
expressions of Model Code (1990). The constitutive law of this model relates to the
global stresses and strains including the contribution of the intact and the cracked
zones. Concrete is an artificial rock and similar formulations can be found in
masonry structures, such as those described in Lee et al. (1993). The reinforcing
steel is considered in the finite element formulation as an overlapped element to
Engineering Computations,
The financial support of the Program Praxis XXI by grant PRAXIS/P/ECM/12126/1998 is Vol. 18 No. 5/6, 2001, pp. 759-785.
gratefully acknowledged. # MCB University Press, 0264-4401
 EC the concrete element, with the same nodal displacements. This element has
18,5/6 rigidity only in the reinforcement direction. The model used in the concrete has
localisation of the microcracking within a small band. The width of this band may
coincide with the dimension of the element. After cracking the concrete can
release all the tensile stresses or have a softening behaviour defined by linear or
exponential laws. In the analysis of plain concrete structures, the localisation of
760 damage is formulated in terms of constant fracture energy.
The algorithms developed consider anisotropy and the continuous degradation
of the elasticity modulus of the material. In reinforced concrete structures, where
the interaction is analysed with cracked and intact zones, the cracked zone
coincides with the damaged band. As a result the softening laws used in plain
concrete can be applied in the cracked zone.
The present work is divided into two separate parts: the first one describes the
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

model for plain concrete; and the second part presents the model for reinforced
concrete with interaction. In both cases, numerical results are presented.

2. Plain concrete
2.1 Numerical modelling
The composite material, represented in Figure 1, has a global volume, V,
divided into an intact zone Vi and a damage zone Vd. The co-ordinate axes n
and t coincide with the orthotropic directions.
The intact material with superscript (i) is represented in white and the
strains and stresses are respectively:
…"in ; "it ; nt
i
†

and
…in ; it ; nti †:

Assuming an elastic behaviour in this zone, the constitutive equation will be:
8 9 2 38 i 9
i 1=E =E 0
>
< n=" > < n >
> =
i 6 7
"t ˆ 4 =E 1=E 0 5 it : …1†
: i >
> ; : i >
> ;
nt 0 0 1=G nt

Figure 1.
Crack localisation
 The damaged material is identified by the superscript (d) and represented by Tension
dotted strips normal to the t direction. This means that cracking has happened stiffening model
in this direction. The strains and stresses in the damaged zone are respectively:
…"dn ; "dt ; nt
d
†
and
761
…dn ; dt ; ntd †:

The constitutive equation in the damaged zone, assumed to be elastic in the n

direction and defined by a function e (t) in the t direction, is therefore:
8 9 2 38 d 9 8 9
< "dn = 1=E =E 0 < n = < 0 =
ˆ 4 =E 0 5 dt
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

"d 0 ‡ e…t † : …2†

: dt ; : d; : ;
nt 0 0 1=G nt 0

In this expression,  represents the decrease in the tangential stiffness, as defined in

the smeared crack model. The function e (t) can be defined by different expressions
which aim to approximate the experimental results, as will be seen later.
This composite material is substituted by an equivalent material, which has
the following strains and stresses:
…"0n ; "0t ; nt
0
†

and
…0n ; 0t ; nt0 †:
The strains in the n direction for the two materials must be the same, that is:
"0n ˆ "in ˆ "dn : …3†
The mean stress, 0n , is computed by the following integration:
Z Z 
0 i d
n ˆ n dv ‡ n dv =V:
Vi Vd

Performing this integration by Gaussian quadrature, as the stresses are

constant within the volume corresponding to each Gauss point, the previous
equation becomes:
0n ˆ vi in ‡ vd dn …4†
with
Vi
vi ˆ
V
and
Vd
vd ˆ ;
V
 EC being
18,5/6 vi ‡ vd ˆ 1:
In a similar procedure, the mean strain and stress in the t direction are given by:
0t ˆ it ˆ dt …5†
762
"0t ˆ vi "it ‡ vd "dt : …6†
The mean tangential strain and stress, using the assumptions stated by Chen
(1982) and Yamaguchi and Chen (1990), are the following:
nt0 ˆ nti ˆ ntd …7†
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

0
nt ˆ vi nt
i
‡ vd nt
d
: …8†
After substituting Equations (1) and (2) in Expressions (3) to (8), the
constitutive relation of the equivalent material is obtained, with the following
incremental form:
8 0 9 2 38 0 9
< d"n = 1=E =E 0 < dn =
d"0t ˆ 4 =E 1=E 0
0 5 d0t : …9†
: 0 ; : ;
d nt 0 0 1=G0 dnt0

The elastic modulus of the equivalent material, E0 , and the shear modulus, G0 ,
are given by:
  1
0 1 vd d de …t †
E ˆ ‡v …10†
E dt

G
G0 ˆ : …11†
…1 vd † ‡ vd

In these expressions the value vd, ratio of the damage volume, and the function
e(t) must be defined. The damage volume, representing the volume with
microcracking and damage within the concrete, is in many cases arbitrarily
established. In the smeared crack model it is equal to the integration volume
and Bazant considered it equal to three up to eight times the maximum
dimension of the aggregate, in order to obtain a representative volume of the
heterogeneity of the concrete.
The most common expressions for e (t) are the linear, the multi-linear or the
exponential laws, as in Martins et al. (1995). The linear and exponential laws
are respectively the following:
ft E
ˆ …" "tu † …12†
ft E"tu
  ˆ ft e b…" E†
ft
…13† Tension
stiffening model
where ft and "tu are represented in Figure 2. The parameter b in Equation (13)
defines the area A limited by the " axis, the linear equation " = ft /E, and the
exponential function, such that:
A ˆ ft =b …14† 763
The area defined by the  " diagram (see Figure 2), is the energy deformation
per unit volume Gf. After the fracture mechanics concepts, the energy for crack
propagation per unit area of crack GF is a property of the material and is
defined by the following equation:
Z "tu
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

GF ˆ w Gf ˆ w  d" …15†
0

where w is the width of the damage zone, whose volume is equal to Vd, (see
Figure 1). The ratio vd, as defined by equation (4), can take the following
alternative expression (see Figure 1):
Vd w
vd ˆ ˆ …16†
V h
where h is the length of the element orthogonal to the crack.
Using the linear function in Figure 2, defined by Equation (12), the integration
(15) becomes:
 2  
ft ft ft
GF ˆ w ‡ "tu : …17†
2E 2 E

The deformation "tu can be obtained from this relation.

Using the exponential function in Figure 2, defined by Equation (13), the
integration (15) becomes:
 2 
ft ft
GF ˆ w ‡ …18†
2E b

and the parameter b can be obtained.

Figure 2.
Linear and exponential
laws
 EC The linear Equation (12) with the parameter "tu obtained previously becomes:
18,5/6 ft E
ˆ 2 …wft " 2GF †: …19†
wft 2EGF

The exponential Equation (13) with parameter b obtained previously gives the
764 following result:
h i1
GF f t
wft 2E …" fEt †:
 ˆ ft e …20†

The last expression is applied in the descending branch of the curve in Figure 2,
i.e. satisfying the following inequality:
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

ft 4GF
" : …21†
E wft
If the strains " do not fall within this domain, that is, for:
4GF
">
wfct
a cut-off in the exponential is considered and no stress retention is possible. With
this simplification the error is very small, since the area neglected represents only
2 percent of the exact area defined by the exponential.
The differentiation d/d" of the linear law (19) and the exponential law (20)
gives, respectively, the following results:
 
d 1 2GF 1
ˆ …22†
d" E wft2

and
  h i 1

d GF ft 1 wfGFt ft
2E …" fEt †
ˆ ft e : …23†
d" wft 2E

The elastic modulus E0 of the equivalent material is obtained substituting in

Equation (10) the value of de/d by the inverse of the previous expressions.
Considering the linear softening law, the equivalent material modulus, E0 , is as
follows:
 
0 1 2GF 1
E ˆ : …24†
E hft2

This expression was obtained by Yamaguchi and Chen (1990). According to

this expression, the objectivity in the finite element discretization is attained
because the equivalent material, with modulus E0 , depends only on the
dimension of the element in the direction normal to the crack. The fracture
 energy per unity of area can be kept as a material property, satisfying the Tension
fracture mechanic concept. After Yamaguchi, the dimension of the crack band stiffening model
width, w, is not important, because it does not affect the value of E0 . The
transverse modulus G0 depends on the band width that can be considered
equal to the element dimension (w = h) in Expression (11), like a particular
case of the crack band formulation. In fact, the independence between the
modulus E0 and the band width w is verified only for the linear softening law. 765
In the general case, and specifically in the exponential law, this independence
does not exist.
The equivalent material modulus, E0 , for the exponential law is obtained by
substituting into Equation (10) the value de/d given by the inverse of
Equation (23), that is:
    
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

0 1 w GF EGF ft
1 1
E ˆ exp ft E" : …25†
E 2Eh hft2 wft 2

The parameter GF has been calculated by experimental methods. Ozbalt

and Bazant (1996) suggest that the band width w must be a function of the
maximum aggregate size and also a function of the state of stress and
strain.

2.2. Numerical applications

2.2.1 Uniaxial traction. A square plate with dimensions 1.0  1.0m2 and 0.008m
in thickness, under uniaxial traction, is divided into one finite element. The
material properties are summarised in Table I.
The constitutive law of the damage material was defined by the linear or by
the exponential relations, as represented in Figure 2. The analysis with the
finite element method gives the stresses and strains in the cracked zone
represented in Figure 3, that are coincident with the constitutive laws.
In the model the consideration of different widths w of damage zone in the
finite element is available. In Figure 4 are represented the load-displacements in
the softening branch, corresponding to different damage localisations equal to
w/h = 1/4, w/h = 1/2 and w/h = 1/1.
These results show the advantage of this model that permits the localisation
of the damage without a change of the finite element division.
2.2.2. Notched beam. A notched beam with dimensions 2,000  200 
50mm3, tested by Petersson in 1981, is simulated. The two finite elements
topology, represented in Figure 5, are employed: a coarse mesh with 18
elements and a fine mesh with 34 elements. The material properties are

E (MPa) 24,500
 0
Table I.
ft (MPa) 1.5 Material properties of
GF (Nm/m2) 200 the panel
 EC
18,5/6

766
Figure 3.
Stress-strain curve
corresponding to an
energy density GF =
200Nm/m3
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 4.
Load-displacements
curves corresponding to
different localisation

presented in Table II. The shear retention factor, , varies linearly from 0.01,
when the crack starts, to zero, when the crack is completely opened (no tensile
stress in the crack).
The two meshes are used in the computation of load-displacement curves
employing the linear softening law. With the exponential law only the finer
mesh is used. In the numerical computation with the finer mesh it was admitted
that the crack is smeared across the element width, w/h = 1, and with the
coarse mesh the crack band is developed in a half element, w/h = 0.5. The
results are plotted in Figure 6 and they can be compared with the experimental
ones that fall within the dashed zone of the same Figure.
The load-displacement curve obtained by Rots et al. (1984), using a linear
softening, is also plotted. The curves corresponding to the linear softening with
the fine and coarse meshes are almost coincident and fit the upper limit of the
experimental zone. The curve obtained with the exponential law corresponds to a
lower limit.
 Tension
stiffening model

767
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 5.
Finite element meshes of
Petersson's beam

E (MPa) 30,000
 0.15
Table II.
ft (MPa) 3.33 Finite element meshes
GF (Nm/m2) 124 of Petersson's beam

Figure 6.
Load-displacement
curves of Petersson's
beam
 EC 2.2.3. Double cantilever beam. Figure 7 represents the structure tested by Sock
18,5/6 et al. (1979) and Figure 8 the finite element division used in this numerical
simulation.
The material properties are presented in Table III. In this example the crack
is smeared in the element, permitting the comparison with Rots et al.'s (1984)
numerical results.
768 The load-displacement diagrams at the point A are plotted in Figures 9 and
10, for two different values of the fracture energy, respectively, GF = 350N/m
and GF = 250N/m.
In Figure 9, for GF = 350N/m, the maximum load was 55.45KN with the
exponential law and 57.18KN with the linear law. These curves approximate
the experimental results, especially the linear softening, up to the displacement
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 7.
Structure tested by Sock
et al. (1979) (a) geometry
of the specimen with
opposite forces at A and
scheme of the loading
apparatus; (b) detail of
the notch; (c) transversal
section

Figure 8.
Finite element division
 near 0.7mm. The analysis stopped there due to tangential stresses in the Tension
elements near the load P. If a pre-stress PS was applied, as in the experimental stiffening model
apparatus, the analysis could be continued.
In Figure 10, for GF = 250N/m, the maximum load is smaller than 50KN. The
linear softening does not give good results.
The exponential law seems to fit the experimental curve, although with a 769
smaller maximum load. This example shows the important role of the energy
GF in the model.

E (MPa) 40,000
 0
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

ft (MPa) 4 Table III.

GF (Nm/m2) 250; 350 Material properties of
 0    0.0001 the cantilever

Figure 9.
Load-displacement
curves for GF = 350N/m

Figure 10.
Load-displacement
curves for GF = 250N/m
 EC 3. Reinforced concrete
18,5/6 3.1. Numerical modelling
Reinforced concrete structures have a complex behaviour due, in part, to the
interaction of the two materials. The bond between the concrete and the
reinforcement allows stresses in the steel to be transmitted to the surrounding
concrete, as shown by the stress distribution in a concrete tie (see Figure 11).
770 This Figure suggests the existence of two zones in the tie: an intact zone,
denoted by i, with volume Vi, and a damage zone denoted by d, with volume Vd.
The damage zone has a variable dimension increasing with the load.
The mean strain in the concrete tie, obtained in experimental results, is
represented in the structural codes by different laws. According to CEB (1990)
the mean strain in the tie, "sm, is given by the following expression:
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

"sm ˆ …1 †"is ‡ "ds …26†

where "is and "ds are the strains in the steel, respectively in the intact and
damage zones. The parameter  defines the dimension of the two zones (see
Figure 12), and is given by:
sr
 ˆ 1 1 2 … d †2 …27†
s
where 1 and 2 are parameters that depend on the bond properties of the
reinforcement as well as on the existence of repeated loading. The steel
stress ds corresponds with the strain "ds , and sr is the stress in the steel
when the first crack appears; both of these stresses are calculated in a
cracked section.

Figure 11.
Concrete tie with tensile
stress distributions in
concrete, in steel and
bond stresses
 In the model developed in this work, the two zones in the finite element have Tension
different constitutive laws and the dimension of each zone is calculated stiffening model
according to Equations (26) and (27). In a finite element approach this
behaviour can be estimated by an element of steel and concrete divided into
two different zones, as represented in Figure 12.
The intact zone is composed of a volume of steel, Vis , and a volume of
concrete, Vic ; similarly, the damage zone is composed of a volume of steel Vds 771
and a volume of concrete Vdc . We have therefore:
Vi ˆ Vis ‡ Vic

Vd ˆ Vds ‡ Vdc :
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

In the intact zone, the incremental constitutive laws in the steel and in the
concrete are respectively given by:
fgis ˆ ‰DŠ is f"gi …28†

fgic ˆ ‰DŠic f"gi : …29†

The [D]is and [D]ic matrices contain the elastic properties of steel and concrete,
that, in the bidimensional case and with the steel in one direction coincident
with the reference axis, can be written as follows:
2 3
1 0 0
‰DŠis ˆ Es 4 0 0 0 5 …30†
0 0 0

2 3
1 v 0
Ec 4
i
‰DŠc ˆ v 1 0 5 …31†
1 v2 0 0 1 v
2

where Es and Ec represent the Young's modulus of steel and concrete respectively; 
represents Poisson's ratio. The vector f
"gi contains the incremental strains in the

Figure 12.
Finite element with
intact and damage zones
 EC intact zone, and it is the same for the steel and the concrete due to the compatibility
18,5/6 of the deformation in both materials. The f
gis and f
gic vectors contain the
incremental stresses obtained, respectively, in the steel and in the concrete.
In the intact zone, i, the incremental mean stress f
gi can be written as
follows:
Z Z 
772 i 1 i i
fg ˆ fgs dv ‡ fgc dv : …32†
V Vs Vc

The stresses, f
gis and f
gic , given by Expressions (28) and (29), are
constant at each Gauss point, allowing an easy calculation of the mean stress
given by Expression (32).
Denoting by  s and  c the proportional volumes (s ˆ Vs =V; c ˆ Vc =V), the
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

above expression, after integration, is transformed into the constitutive law of

zone i, given by:
h i
fgi ˆ s ‰DŠis ‡ c ‰DŠic f"gi : …33†

The mean strain is obtained by the inversion of this equation, that is:
 i
f"gi ˆ D 1 fgi …34†

where ‰D 1 Ši is the inverted matrix given by:

‰DŠi ˆ s ‰DŠis ‡c ‰DŠic : …35†

The damage zone constitutive law can be derived by a similar procedure,

giving the following result for each material:
fgds ˆ ‰DŠ ds f"gd …36†

fgdc ˆ ‰DŠdc f"gd …37†

and the final damage zone constitutive law, zone d, is:

h i
fgd ˆ ‰DŠd f"gd ˆ s ‰DŠds ‡c ‰DŠdc f"gd : …38†

The mean strain in the damage zone, equal in both materials as mentioned
before, becomes:
 d
f"gd ˆ D 1 fgd : …39†
In Expression (38), the matrix [D]ds has the form of Equation (30), when the steel
remains elastic or, after the yield stress is attained, the incremental stress
ds
in steel direction is calculated through the effective plastic strain,
" p,s in the
same direction:
ds = Hs
" p,s. The matrix [D]dc has the form:
 2 3
1=E =E 0
1
Tension
‰DŠdc ˆ 4 =E 1=E0 0 5 …40† stiffening model
0 0 1=G0
where the equivalent modulus E0 is a function of the concrete damage, given by
Expressions (24) or (25). We consider that the crack localisation is already
restricted to the damage zone, where the cracks are supposed to be uniformly 773
distributed. In this way, the E0 and G0 computations, in part I, are simplified
due to: vd=1 (respectively Expressions (10) and (11) ). The band width w, is
equal to the element dimension, giving also a simplification in Expressions (22)
and (23).
In Equation (38) it is supposed that cracks are normal to the steel. If the cracks
direction is oblique to the direction of the reinforcement, the corresponding
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

rotation of axis is necessary.

In a bidimensional state of stress and incremental form Equation (26)
becomes:
f"g ˆ ‰T1 Š f"gi ‡ ‰TŠ f"gd …41†

with [T1] and [T] matrices defined by:

2 3 2 3
1  0 0  0 0
‰T1 Š ˆ 4 0 0 05 ; ‰TŠ ˆ 4 0 1 0 5: …42†
0 0 0 0 0 1

Using the definition of mean strains of the intact and damage zones,
Expressions (34) and (39) respectively, the mean strain of the element, Equation
(41), becomes a function of mean stress in each material, f
gi and f
gd,
that is:
 i  d
f"g ˆ ‰T1 Š D 1 fgi ‡ ‰TŠ D 1 fgd : …43†

Since the equilibrium of stress must be insured, the stresses f
gi and f
gd
are equal to the mean stress f
g in the element, or:

fg ˆ fgi ˆ fgd : …44†

The constitutive law of the composed element can be written in the following
short form:
 
f"g ˆ D 1 fg …45†

and matrix [D±1] obtained by equation (43), giving:

 1  i  d
D ˆ ‰T1 Š D 1 ‡ ‰TŠ D 1 : …46†
 EC The constitutive relation defined by Equation (45) is non-linear due to the [D]
18,5/6 matrix non-linearity, Equation (46), that results from [T] matrix through
parameter d and [D±1]d matrix. This [D±1]d matrix contains the non-linear
behaviour of the steel and softening of concrete.
In the finite element method usual procedure this matrix is used in the
stiffness calculation by the expression:
774 Z
‰KŠ ˆ ‰BŠT ‰DŠ‰BŠdv …47†
v

where [B] contains the derivatives of the shape functions. The resolution of the
finite element equation gives the incremental nodal displacements {
d}, that
are the addition of {
d}i and {
d}d, respectively in the intact and damage
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

zones:

fdg ˆ fdgi ‡ fdgd : …48†

The incremental displacements in each zone, {
d}i and {
d}d, are given by:

fdgi ˆ ‰T1 Š fdgi1 and fdgd ˆ ‰TŠ fdgd1 …49†

where f
dgi1 and f
dgd1 are the element displacements assumed to be

whole intact or whole damaged. These displacements can be calculated
considering the stiffness matrix of the intact zone [K]i and the damage zone
[K]d given by:
Z Z
i d
‰KŠi ˆ ‰BŠT ‰DŠ ‰BŠ dv and ‰KŠd ˆ ‰BŠT ‰DŠ ‰BŠ dv: …50†
v v

Using the incremental load vector {
f} , the displacements f
dgi1 and
f
dgd1 are obtained:

fdgi1 ˆ ‰K 1 Ši ffg and fdgd1 ˆ ‰K 1 Šd ffg: …51†

These expressions are used in the computation of the stresses and strains in
each zone.

3.2. Computational algorithms

In the incremental iterative process of non-linear problems, the following
values are known at the end of each converged load increment:
. the stresses in both zones and materials, fgic ; fgis ; fgdc ; fgds , or only
the mean stress, {};
. the vector {}, that contains the dimension of the damage zone in the
volume of integration, corresponding to each Gauss point;
. the nodal displacements, {d} .
 The computation of vector {} can be made through the mean stress {} , if the Tension
whole load is supported by the steel and no tensile stress exists in the concrete. stiffening model
In that case the result of the stresses, in the volume of integration of the Gauss
point, supported by the steel, is used directly in the CEB Expression (27). If the
tension stiffening in concrete is considered, the {} computation is made with
the steel stress fgds .
The principal steps of the algorithm, after applying the new increment load 775
vector {
f}, are the following:
(1) Calculation of the incremental load vector {
f}, global load vector {f},
and beginning of the iterative process, ``j'', with {
fj} = {
f} in the first
iteration;
(2) Computation of global stiffness [Kj], Equation (47), and incremental
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

nodal displacements {
dj} by:

f
dj g ˆ ‰Kj 1 Šf
fj g …52†

(3) In the first iteration, the stiffness matrix of the intact zone [K]i, the
incremental nodal displacements, f
dj gi1 , and the stresses normal to the
cracks, fj gic ; fj gis , corresponding to the elastic behaviour, can be
calculated by the equations:
Z
i i
‰KŠ ˆ ‰BŠT ‰DŠ ‰BŠ dv …53a†
v

 i  1 i
dj 1 ˆ K …fj † …53b†

 i  i
j c ˆ ‰DŠic ‰BŠ dj 1 …53c†

 i  i
j s ˆ ‰DŠis ‰BŠ dj 1 …54†

where [D]ic , [D]is and [D]i are given by Equations (30), (31) and (35),
respectively;
(4) Calculation of the stresses in the damage zone with a subiterative
looping ``k''. The load vector in the first subiteration is f
fjk g ˆ f
fj g:
. The incremental and global displacements corresponding to the
damage behaviour are given by:
n od h id n o
dkj ˆ Kj 1 fjk …55†
1

n od n od n od
dkj ˆ dkj 1 ‡ dkj …56†
1 1 1
 EC . The incremental and total strains and stresses in the damage zone
18,5/6 are obtained with the following computations:
n od n od n od n od
"kj ˆ ‰BŠ dkj ; "kj ˆ ‰BŠ dkj …57†
1 1

776 n od   n od n od   n od
d d
kj ˆ Dj c "kj ; kj ˆ Dj s "kj …58†
c s

n od n od n od n od n od n od
kj ˆ kj 1 ‡ kj ; kj ˆ kj 1 ‡ kj …59†
c c c s s s
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

When the first crack appears in a Gauss point, the corresponding

steel stress fgds is kept in memory, value sr in the CEB Expression
(27), and the computation of {jk } is made:
. the equivalent load vector is computed by:
Z n od Z n od
frg ˆ ‰BŠT kj dvs ‡ ‰BŠT kj dvc …60†
vs s vc c

If {r} 6ˆ {f} with a prescribed error, the subiterative process has to

continue. A new load vector is computed by:
n o
fjk ˆ frg ffg …61†

and the process returns to point (i). If the subiterative process

converged, the final fjk g is calculated;
(5) In the iterative process the load vector {r*} is calculated, with a new
stiffness matrix [Kj] by:
fr g ˆ ‰KjŠf
djg …62†

If {r*} 6ˆ {f} with a prescribed error, the iterative process has not
converged and the procedure goes to step (b). When the iterative process
converges, the global displacements are computed by:
fdjg ˆ fdgj 1 ‡f
ji g …63†

In the numerical examples presented, two algorithms are used. In

algorithm SR there is no tensile stress in the concrete after cracking. In
CR algorithm there is a tension stiffening diagram in the concrete
damage zone, defined either by an exponential law, denoted by CRE, or
by a linear law, denoted by CRL.
 3.3 Numerical application Tension
3.3.1. Reinforced concrete panel. The model is tested in a square panel stiffening model
represented in Figure 13. The panel, with 1.00  1.00  0.08m3, is under
uniform tensile stress. The thickness of the finite element is divided into four
layers of concrete and two steel layers. The properties of the steel and the
concrete are summarised in Table IV.
The value of GF was chosen for comparison of the model with the results 777
obtained by the smeared crack approach, where the tension stiffening effects
were simulated by a tensile retention in the concrete with a great maximum
ultimate strain value.
In Figure 14 are plotted the panel load-displacement curves, obtained with
the model and two different values of parameter  ˆ 1 2 ( = 1.0 and  = 0.5),
defined by Expression (27). These results, obtained with the algorithm SR
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

(where there is no retention of tensile stress in the concrete damage zone), are a
good approximation to Leonhardt's (1979) experimental results with  = 0.37
percent. The linear elastic response of steel (without concrete) is also plotted in
this figure.
Figure 15 represents the evolution of the damage zone dimension with
increasing load,  parameter, for two  values ( = 1.0 and  = 0.5).
In Figure 16 are plotted the load-displacement curves, considering tensile
stress retention in the concrete after cracking, with a liner softening ±
algorithm CRL.
For comparison, the curves without tensile stress retention are also plotted.
The differences appear in the propagation zone, where CRL algorithm

Figure 13.
Reinforced
concrete panel

Ec (MPa) 24,500
Es (MPa) 206,000 Table IV.
Material properties
ft (MPa) 1.5 of the reinforced
GF (Nm/m2) 1,875 concrete panel
 EC
18,5/6

778

Figure 14.
Load-displacement of
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

the panel without

softening, algorithm SR

Figure 15.
Evolution of the damage
zone, parameter , with
the load

Figure 16.
Comparison of the
algorithms SR and CRL,
for two values of 
 corresponds to a greater hardening. In the stabilised cracking zone there are no Tension
differences between SR and CR algorithms. stiffening model
The differences between concrete linear and exponential softening laws can
be observed in Figure 17, for  = 0.5 in Equation (27).
The response in the propagation zone is harder in the linear algorithm CRL than
in exponential CRE, and the opposite behaviour is observed in the stabilised
zone. 779
In Figure 18, the algorithms SR and CRL are compared with the smeared
crack model represented by the continuous line. In the smeared crack model the
softening law admitted to the concrete corresponds to a value of GF = 1,875N/
m. The smeared crack model can be considered as a damage model, where the
damage zone is coincident with the dimension of the element. This Figure
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

shows that, after the opening of the crack (corresponding to zero stress in the
concrete), the smeared crack model curve coincides with the deformation of the
steel. In the present model, there is always some concrete contribution to the
mean deformation of the panel, usually termed tension stiffening. It is
interesting to observe that the smeared crack model approaches the CRL
algorithm in the cracking propagation zone.
3.3.2. Beam. In Figure 19 there is represented a beam experimentally tested
by Duddeck, whose results are given in Barros et al. (1990) and Figueiras
(1983). Two finite element meshes are employed in the simulation: the mesh
represented in Figure 19(b) with the layers of Figure 19(f) in A analysis; the
mesh represented in Figure 19(c) with the layers of Figure 19(e) in B analysis.
The meshes, used in the analysis of the beam, are represented in Figures 19(a)
and (d).

Figure 17.
Comparison of the
algorithms SR, CRL and
CRE for  = 0.5
 EC
18,5/6

780
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 18.
Comparison of
the smeared crack
model with the
algorithms SR, CRL

In A analysis the plane stress finite elements simulate the longitudinal and
transverse reinforcement by an algorithm developed in Barros et al. (1990): an
elasto-plastic behaviour for concrete and steel, and a smeared crack model for
cracked concrete. The material properties used are given in Table V.
Figure 20 shows the displacement in the vertical direction under increasing
applied load. In this Figure the curves correspond to experimental results, an
associated and a non-associated plastic behaviour with the Drucker-Pager yield
surface and Von Mises criterion for plastic potential. The crack pattern is
represented in Figure 21 with the points where steel undergoes plastic
deformations, signalised by circles, for a load near the collapse.
In B analysis the structure is divided into thick Ahmad plate elements and
the smeared crack model is used in the analysis of concrete under tensile
stresses. Concrete in compression is assumed to have a linear elastic
behaviour. Figure 22 represents the load-displacement diagram of a section at
middle span for B analysis with seven Ahmad thick plate elements and
variable number of layers. The objective of this analysis is to find the
minimum number of layers that gives a good approximation to the
experimental results and then use it in the proposed model. The cross-section
with eight layers seems to represent sufficiently well the experimental results
and approximates the load-displacement numerical results obtained by
Figueiras (1983), where the same element is combined with the smeared crack
and the elasto-plastic models.
From the comparison between the experimental results and the models
employed in the analysis it can be concluded that concrete cracking governs the
beam behaviour up to loads near the collapse. Up to P load equal to 12KN the
analysis and the experimental results are very similar. Near 12KN the plastic
 Tension
stiffening model

781
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 19.
(a) Geometry of the
beam tested by
Duddeck; (b) Finite
element mesh with plane
stress elements;
(c) Finite element mesh
with seven Ahmad thick
plate elements;
(d) Cross-section of the
real beam; (e) Layers
used in mesh (c);
(f) Layers used in mesh (b)
 EC deformation of steel begins producing the opening of cracks and being
18,5/6 responsible for the increasing of deformation. The plastic deformation of
concrete under compression influences the behaviour of the beam for loads
greater than 15KN, that is close to the collapse. As a result non-linearity in

Concrete (KN, cm) Steel

782
Ec 1,666 Es 19,600
 0 Est1 2,800
Uniaxial tensile strength: Est2 500
ft 0.167 Yield stress:
Table V. Uniaxial comp. strength: fy 49
Material properties of fc 3.2 Uniaxial stress:
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

the reinforced concrete  0.5 ft 57.4

beam "tu 0.002

Figure 20.
Load-displacement
diagram of the section at
middle span with
smeared crack and
elasto-plastic models,
using plane stress
elements

Figure 21.
Crack pattern with
smeared crack and
elasto-plastic models,
using plane stress
elements
 Tension
stiffening model

783
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Figure 22.
Load-displacement
diagram of the section at
middle span with
smeared crack using
Ahmad thick plate
elements

concrete cracking and linear behaviour in compression are considered to test

the reinforced concrete model presented in this paper.
One of the objectives of reinforced concrete models is to simulate the
stiffening induced in the cracked concrete by the steel, due to the bonding
between both materials. It is convenient to remember that, in the smeared crack
model, this stress stiffening is taken into account by stress retention after the
beginning of the crack. The stress-strain diagram of the concrete is composed
of two branches: the elastic one, before cracking; and a linear descending
branch, after cracking. This smeared crack model corresponds to the model
proposed in Part I if no localisation is considered and if the fracture energy is
defined per unit volume of material. In contrast, the proposed model does not
consider the hardening of cracked reinforced concrete by the modification of
one of the materials' constitutive law. The behaviour of concrete and steel is
simulated through the constitutive law of each material. The interaction
between both materials is given by the CEB expression for the mean strain in
the tie. The load-displacement diagrams corresponding to the proposed
interaction model obtained by the SR algorithm and the experimental results
are represented in Figure 23. The interaction model proposed in this work is a
good approximation to the experimental results for P load between 12KN and
15KN where, as has already been said, is cracking and plastic deformation of
the steel that has the main influence in the non-linear behaviour of the beam.
For loads greater than 15KN and near collapse the approximation could be
improved, if the non-linear behaviour of concrete under compression was
considered.
 EC
18,5/6

784

Figure 23.
Load-displacement
diagrams of the section
at middle span with SR
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

algorithm and other

solutions

4. Conclusions
A model for the analysis of reinforced concrete structures is developed. The
model considers two zones in the element, one termed the intact zone and the
other the cracked zone. The dimension of the cracked zone increases with the
loading and is defined according to the expressions of Model Code (1990). The
constitutive laws relate the global stresses and strains including the
contribution of the intact and the cracked zones. The model used in the concrete
has localisation of the microcracking within a band. In the analysis of plain
concrete structures, the localisation of damage is formulated in terms of
constant fracture energy.
The reinforced steel is considered in the finite element formulation as an
element overlapped to the concrete element, with the same nodal
displacements. This element has stiffness only in the direction of the
reinforcement. After cracking the concrete can release all the tensile stresses or
have a softening behaviour defined by linear or exponential laws.
The algorithms developed consider anisotropy and the continuous
degradation of the elasticity modulus of the material. In reinforced concrete
structures, where the interaction is analysed with cracked and intact zones,
the cracked zone coincides with the damage band. As a result the softening
laws used in plain concrete can be applied in the cracked zone. Several
numerical examples are presented and compared with experimental and other
numerical results. In general good agreement is obtained. The examples
solved show that the crack propagation energy definition is very important in
the results.
References
Barros, M.H.F., Ahmad, S.H. and Marques, C.S. (1990), ``A non-associated elasto-plastic model for
concrete'', Proceedings of the Int. Conf. on Computer Aided Analysis and Design of Concrete
Structures, Ed. Bicanic, Mang.
 Ceb-Fip, Model Code (1990), Design Code, Bulletin d'Information, No. 203-5. Tension
Chen, W.F. (1982), Plasticity in Reinforced Concrete, McGraw-Hill, New York, NY. stiffening model
Figueiras, J.A. (1983), Ultimate Load Analysis of Anisotropic and Reinforced Concrete Plates and
Shells, PhD thesis, University College of Swansea, Swansea.
Lee, J.S., Pande, G.N. and Middleton, J. (1993), ``Numerical modelling of cracking in masonry
structures'', in Pande, G.N. and Middleton, J. (Eds), Proceedings of the 2nd Int. Symp. on
Comp. Meth. Struc. Masonry 2, Swansea. 785
Leonhardt, F. (1979), ConstrucËoÄes de Concreto, Interciencia, Rio de Janeiro.
Martins, R.A.F., Barros, M.H.F.M. and Ferreira, C.C. (1995), Analysis of Concrete Structures Using
Damage Models, EPPCMESC, Macau, pp. 1141-7.
Ozbolt, J. and Bazant, Z.P. (1996), ``Numerical smeared fracture analysis: non-local
microcrack interaction approach'', Int. J. Numerical Methods in Engineering, Vol. 39
No. 4, pp. 635-61.
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

Rots, J.G., Kuster, G.M.A. and Blaauwendraad, J. (1984), ``The need for fracture mechanics
options in finite element models for concrete structures'', Proceedings of the Int. Conf. on
Computer Aided Analysis and Design of Concrete Structures, September, Split, pp. 19-32.
Sock, C., Baron, J. and FrancËois, D. (1979), ``MeÂcanique de la rupture appliqueÂe au beÂton
hydraulique'', Cement and Concrete Research, Vol. 9, pp. 641-8.
Yamaguchi, E. and Chen, W.F. (1990), ``Cracking model for finite element analysis of
concrete materials'', Journal of Engineering Mechanics, June, Vol. 116 No. 6, ASCE,
pp. 1242-60.
 This article has been cited by:

1. Viktor Gribniak, Gintaris Kaklauskas, Lluis Torres, Alfonsas Daniunas, Edgaras Timinskas, Eugenijus
Gudonis. 2013. Comparative analysis of deformations and tension-stiffening in concrete beams reinforced
with GFRP or steel bars and fibers. Composites Part B: Engineering 50, 158-170. [CrossRef]
2. Ardalan Sherafati, Reza Farimani, Atorod Azizinamini. 2013. Effect of Concrete Slab on Shear Capacity
of Composite Plate Girders under Positive Moment. Journal of Bridge Engineering 18:2, 89-98. [CrossRef]
3. V. Gribniak, G. Kaklauskas, S. Idnurm, D. Bačinskas. 2010. Finite Element Mesh Size Effect on
Deformation Predictions of Reinforced Concrete Bridge Girder. The Baltic Journal of Road and Bridge
Engineering 5:1, 19-27. [CrossRef]
4. H.Q. Wu, R.I. Gilbert. 2009. Modeling short-term tension stiffening in reinforced concrete prisms using
a continuum-based finite element model. Engineering Structures 31:10, 2380-2391. [CrossRef]
5. S. Khalfallah. 2006. Cracking analysis of reinforced concrete tensioned members. Structural Concrete 7:3,
111-116. [CrossRef]
Downloaded by Cornell University Library At 06:58 01 July 2017 (PT)

6. Hyo-Gyoung Kwak, Do-Yeon Kim. 2006. Cracking behavior of RC panels subject to biaxial tensile stresses.
Computers & Structures 84:5-6, 305-317. [CrossRef]
7. Hyo-Gyoung Kwak, Do-Yeon Kim. 2004. Tension stiffening effect of RC panels subject to biaxial stresses.
Computers and Concrete 1:4, 417-432. [CrossRef]