Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: This paper presents a three-dimensional (3-D) meso-scale finite element (FE) model for near-surface
Received 20 March 2012 mounted (NSM) FRP strip-to-concrete bonded joints established using the general-purpose FE software
Accepted 3 December 2012 package MSC.MARC. In the FE model, elements of the order of 1 mm in size are employed. The concrete
Available online 27 December 2012
is simulated using the orthogonal fixed smeared crack model while the FRP and the adhesive are treated
as linear brittle-cracking materials. The FE model is calibrated and verified using results of well-
Keywords: documented bonded joint tests. Using the verified FE model, the failure process of NSM FRP strip-to-
Three-dimensional (3-D) analysis
concrete bonded joints is carefully studied; furthermore, the local bond stress distributions and the
Meso-scale modeling
Finite element (FE) modeling
bond-slip relationships are extracted and analyzed. This 3-D meso-scale FE model offers a powerful tool
Near surface mounted (NSM) FRP for deployment in further investigations to establish bond-slip models and bond strength models for NSM
Strip FRP strip-to-concrete bonded interfaces. While the present study is focused on NSM FRP strips, the
Bond behavior proposed modeling approach is generally applicable to NSM FRP bars of other cross-sectional shapes.
Ó 2012 Elsevier Ltd. All rights reserved.
0045-7949/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruc.2012.12.002
106 J.G. Teng et al. / Computers and Structures 117 (2013) 105–117
has shown convincingly that a reliable FE model employing ele- 2.2. Li et al.’s pull-out tests
ments of very small sizes (referred to as meso-scale elements)
can be a powerful and economical alternative to laboratory testing A total of five NSM CFRP strip-to-concrete bonded joints were
to gain a full understanding of the complex behavior of FRP-to- tested by Li et al. [6], with the bond lengths being 30 mm,
concrete bonded joints and to generate numerical results for the 100 mm, 150 mm, 200 mm and 250 mm respectively. The concrete
development of a local bond-slip relationship model. Lu et al. block had a square section of 150 mm 150 mm and a length of
[12] developed the first reliable 2-D meso-scale FE model using 350 mm, and the bonded region of the NSM strip started at
elements of the order of 0.5 mm and an orthogonal fixed smeared 50 mm away from the loaded end (Fig. 1). The groove size had a de-
crack model to simulate the local bond-slip behavior and local fail- sign width wg of 8 mm and a design height hg of 22 mm, but post-
ure process of externally bonded FRP-to-concrete bonded joints. preparation measurements indicated that the actual values were
Adopting a similar approach, this paper presents a 3-D meso-scale about 9 mm and 22 mm.
FE model for NSM FRP-to-concrete bonded joints. The present The concrete had an averaged cube compressive strength fcu of
paper is focused on the FE modeling of bonded joints between 29 MPa. The cylinder compressive strength fc can be estimated
NSM FRP strips (i.e. bars of narrow rectangular section) and con- from
crete for the simulation of failure process and the prediction of
fc ¼ 0:8f cu ð1Þ
the local bond-slip relationship, although the proposed modeling
techniques can also be employed to model the bond behavior of and the tensile strength ft can be estimated as [17]
NSM FRP bars of other cross-sectional shapes.
2
fc 8 3
ft ¼ 1:4 ð2Þ
10
2. Bond tests and FE modeling
The groove-filling material was a two-component epoxy adhesive
2.1. Bond test methods with a mixing ratio of 2 (resin):1 (hardener) by weight. The tensile
strength and elastic modulus of the adhesive averaged from five
Two popular bond test methods have been adopted by tensile coupon tests were 42.6 MPa and 2.62 GPa respectively. The
researchers to investigate the bond behavior of FRP bars near-sur- Poisson’s ratio of the epoxy is assumed to be a typical value of
face mounted to concrete: (1) the beam pull-out test, and (2) the 0.35. The CFRP strips had a thickness of 2 mm and a width of
direct pull-out test. The beam pull-out test for NSM FRP bars was 16 mm. The ultimate tensile strength and elastic modulus of the
derived from the pull-out bending test for assessing the bond char- CFRP strips were 2068 MPa (according to the manufacturer) and
acteristics of conventional steel bars and was used to study the 151 GPa (deduced from readings of strain gauges installed on the
bond behavior of NSM FRP bars by Nanni et al. [15]. The direct exposed part of the NSM CFRP strip in the bonded joint tests),
pull-out test is conducted on a concrete block embedded with an respectively. Two CFRP strips were bonded together using the
NSM FRP bar and is usually conducted under displacement control. groove filling adhesive, forming a compound strip whose total
The direct pull-out test setup has three main variations: traditional thickness was approximately 5 mm (i.e., 4 mm of CFRP plus about
one-side direct pull-out test e.g. [4,6], two-side direct pull-out test 1 mm of adhesive), whereas the width was still 16 mm. The use
e.g. [16], and C-shaped block direct pull-out test e.g. [3]. In the of a compound strip instead of a normal strip is to allow the instal-
present study, the NSM CFRP strip-to-concrete bonded joint spec- lation of strain gauges between the two strips of a compound strip
imens analyzed are from Li et al. [6], in which the one-side direct so that the strain gauges would not interfere with the interfacial
pull-out test approach (Fig. 1) was adopted. These specimens were behavior. It should be noted that the elastic modulus of the com-
selected for analysis because all needed details of the tests includ- pound strip for FE modeling in this study was modified according
ing the measured strain distributions of the FRP strip are readily to its real thickness as follows: Ef = 151 4/5 = 120.8 MPa. The Pois-
available to the authors. son’s ratio of the CFRP compound strip was assumed to be 0.2. The
effect of using a nearly zero Poisson’s ratio (0.002) was also ex-
plored before choosing this Poisson’s ratio for use in the parametric
study; it was found that such a small Poisson’s ratio led to load–
displacement curves which are almost identical to those obtained
with a Poisson’s ratio of 0.2 with the ultimate loads differing by less
than 1%. Details of the specimen are listed in Table 1.
Table 1
Bond test specimen details.
20mm Bonded region 50mm that failure of the bonded joint is determined by the material
strengths of the three constituent materials. In theory, cohesion
Displacement failure can occur in any of the three constituent materials, but
50mm
existing tests on NSM FRP-to-concrete bonded joints have indi-
cated that in most cases, cohesion failure occurs in the concrete
50mm as the adhesive normally has a much high tensile strength than
the concrete. In practical applications, it should also be ensured
that failure is governed by cohesion failure in concrete through
appropriate surface preparation and through the use of an appro-
Fig. 2. Boundary conditions of the FE model.
priate adhesive as otherwise the strength of concrete cannot be
fully utilized. Cohesion failure in FRP is the least likely and is not
considered in the present study, but the possibility of cohesion fail-
end of the concrete block (Fig. 1) were simulated by setting the ure in the adhesive is considered in the present study.
corresponding displacements to zero in the FE model. The numer- Consequently, the modeling of concrete is of particular impor-
ical specimen was loaded by imposing a displacement at the end of tance. As mentioned earlier, the so-called meso-scale FE approach
the CFRP strip. was first proposed for externally bonded FRP-to-concrete bonded
The 3-D, eight-node, first-order, isoparametric solid element joints to investigate the interfacial debonding failure process and
with full Gaussian integration available in MSC.MARC was used the local bond-slip relationship by Lu et al. [12]. It should be noted
to model the concrete block, the CFRP strip and the adhesive. The that the meso-scale FE analysis here is different from the meso-
FE mesh in and near the bonded region was finer than elsewhere. mechanical model in which the concrete is treated as a three-phase
After a convergence study to be presented later, the final mesh of composite including the cement matrix, aggregate and matrix-
the FE model was determined as follows: (1) the element size for aggregate interface e.g. [19,20]. Instead, a meso-scale FE model still
the FRP strip was 0.83 mm in its thickness direction (3 elements), treats the concrete as a homogenous material and thus reflects the
1.23 mm in its height direction (13 elements), and 1.25 mm in the average response of concrete, but uses very small elements to cap-
longitudinal direction; (2) the element size for the adhesive was ture the complex process of cracking of the concrete.
about 0.5 mm in its thickness direction (4 elements), 1.23 mm in
the height direction (13 elements in the strip height region), and 3.2. Modeling of concrete
1.25 mm in the longitudinal direction; and (3) the element size
for the concrete was 1.25 mm within the central zone of 30 mm 3.2.1. Modeling approaches for cracked concrete
from the plane of symmetry, 60 mm from the top surface of the In the present study, the cracking of concrete was simulated
concrete block, and the bonded region plus 10 mm in each direc- using the orthogonal fixed smeared crack model available in the
tion in the longitudinal direction. The maximum side length of con- general-purpose software package MSC.MARC [18], and a very fine
crete elements in the remaining parts was 2.5 mm. The mesh FE mesh, which simulates cracking by changing the constitutive
adopted for specimen CS-150 is shown in Fig. 3, which includes relationship, in contrast to the discrete crack concrete model which
an overall view of the mesh and a detailed view near the loaded treats a crack as a geometry entity e.g. [21]. It allows one crack at
end. one integration point for one-dimensional elements, two orthogo-
nal cracks for two-dimensional elements and three orthogonal
3. Constitutive laws and material properties cracks for three-dimensional elements.
Besides the orthogonal fixed smeared crack model, two other
3.1. General popular approaches for modeling cracked concrete are the multi-
directional smeared crack model and the rotating smeared crack
Failure of a NSM FRP-to-concrete bonded joint may occur in one model [22]. In the multi-directional smeared crack model, when
of the three constituent materials (FRP, adhesive and concrete) or the angle between the maximum principal tensile stress and the
at one of the two bi-material physical interfaces (FRP/adhesive existing crack direction exceeds a certain value (referred to as
and concrete/adhesive interfaces). The former is commonly re- the threshold angle hereafter), new cracking occurs perpendicular
ferred to as cohesion failure while the latter is commonly referred to the direction of the current maximum principal stress. The rotat-
to adhesion failure. Adhesion failure can be and should be avoided ing smeared crack model can be divided into two main types [22].
through the proper surface preparation of the two adherends so The first type is called the non-coaxial rotating smeared crack
108 J.G. Teng et al. / Computers and Structures 117 (2013) 105–117
Concrete
Adhesive
FRP strip
model, in which the directions of the principal stresses can be dif- erations, the multi-directional model and the rotating smeared
ferent from those of the principal stains; the second type is called crack model have seldom been incorporated into general-purpose
the coaxial rotating smeared crack model, in which the directional software packages which are capable of handling complicated
alignment between the principal stresses and the principal strains 3-D nonlinear analysis as well as heavy computation. On the con-
is forced. The concept of the rotating smeared crack model has trary, these software packages usually include the orthogonal fixed
been criticized in that the rotation of the crack lacks physical smeared crack model. If a good shear retention factor model is em-
meaning and some parameters need to be determined empirically ployed and the element size is made small enough, accurate FE
[12]. The multi-directional smeared crack model is thought to be modeling of crack appearance and propagation is possible using
the most advanced in theory, but the threshold angle for the the orthogonal fixed smeared crack model. Indeed, this meso-scale
appearance of new cracking is difficult to define, and numerical FE method based on the orthogonal fixed smeared crack model has
convergence is usually a great challenge. Due to the above consid- been successfully used by the authors’ group in the FE modeling of
J.G. Teng et al. / Computers and Structures 117 (2013) 105–117 109
externally bonded FRP-to-concrete joints [12], and was thus also where r0 and e0 are the maximum stress and the corresponding
adopted in the present study. strain respectively, and are assumed to be the compressive cylinder
The smeared crack model for concrete was first introduced by concrete strength (fc) and 0.002 respectively following Chen [26]; eu
Rashid [23] and since then has been adopted by numerous and ru are the ultimate strain and the corresponding stress respec-
researchers. One of the intrinsic properties of concrete is its tively, and are assumed to be 0.0038 and 0.85fc following Hognestad
tension-softening behavior (i.e. the stress–strain curve exhibits a [29] respectively; If the compressive strain in the concrete reaches
descending branch after cracking), which leads to the well-known 0.0038, the concrete fails by crushing and the stress drops to zero;
mesh sensitivity issue. To mitigate the mesh sensitivity problem, E0 is the initial elastic
pffiffiffimodulus
ffi and is calculated according to ACI-
Bazant and Oh [24] proposed the crack band model, which relates 318 [30]: E0 ¼ 4730 fc ; and Es is the secant modulus at the maxi-
the element size to the tension-softening behavior of concrete mum stress: Es = r0/e0.
based on the fracture energy concept. The crack band model was
adopted in the present study. 3.2.4. Tension-softening curve
Three tension-softening curves were considered in this study:
3.2.2. Yield and failure surfaces (a) the linear softening curve [31], (b) the bilinear softening curve
The yield surface proposed by Buyukozturk [25] (Eq. (3)) with [32] and (c) the exponential softening curve [33]. The linear soften-
the associated flow rule was adopted in this study. The failure sur- ing curve is based on the assumption that the tensile stress reduces
face adopted is a combination of Buyukozturk’s [25] model (Eq. (3)) linearly to zero at a crack opening displacement of w0 (Fig. 5a); the
and the maximum tensile stress criterion (Eq. (4)). The yield sur- bilinear softening curve is based on the assumption that the tensile
face and the failure surface employed in this study are the criteria stress reduces linearly to 1/3 of the ultimate tensile stress at a
recommended by MSC.MARC [18] for modeling concrete. crack opening displacement of 2/9w0 firstly and then reduces line-
pffiffiffi arly to zero at w0 (Fig. 5b); and the exponential softening curve is
3J 2 þ 3vr I1 þ nI21 ¼ ðr Þ2 ð3Þ based on the assumption that the softening branch follows the fol-
pffiffiffipffiffiffiffi
f ðI1 ; J 2 ; hÞ ¼ 2 3 J 2 cos h þ I1 3f t ¼ 0 ð4Þ lowing exponential function [33] (Fig. 5c):
" 3 #
where I1 and J2 are the first invariant of stresses and the second rt w 6:93ww w ð6:93Þ
⁄ ¼ 1 þ 3:0 e 0
10 e ð7Þ
pffiffiffirespectively; n, v and r are material
invariant of deviatoric stresses ft w0 w0
constants and equal to 0.2, 3 and r ^ =3 respectively [25]; r
^ is the
yield stress of concrete, taken to be f3c in the present study according where w is the crack width (i.e. crack opening displacement),
to Buyukozturk [25] when Eq. (3) is used to define the yield surface and w0 is the crack width at the complete release of stress or
and the compressive strength fc when Eq. (3) is used to define fail- fracture energy. The w0 value for these three tension-softening
ure surface; ft is the tensile strength of concrete; and h is the angle models is different. It can be related to the tensile fracture en-
of similarity. ergy Gft as w0 = 2Gft/ft for the linear softening curve, w0 = 3.6Gft/
This approach is similar to that used by Chen [26], in which the ft for the bilinear softening curve and w0 = 5.4Gft/ft for the expo-
Mohr–Coulomb criterion and the maximum tensile stress or strain nential softening curve respectively, based on the following
criterion were combined to simulate the behavior of concrete. relationship:
Z
3.2.3. Uniaxial compressive stress–strain curve rt dw ¼ Gft ð8Þ
The following uniaxial compressive stress–strain curve which
was modified by Elwi and Murray [27] from Saenz’s curve [28] The tensile strength of concrete ft (Eq. (2)) and the tensile fracture
was adopted in this study (Fig. 4): energy Gft (Eq. (9)) proposed by CEB-FIP [17] were used in this study
E0 e 0:7
rc ¼ 2 3 ð5Þ fc
E0
1 þ R þ Es 2 e0 ð2R 1Þ ee0 þ R ee0
e Gft ¼ ð0:0469D2a 0:5Da þ 26Þ ð9Þ
10
in which where Da is the maximum aggregate size.
E0 =Es ðr0 =ru 1Þ 1 The unloading and reloading curves of the tensile stress-crack
R¼ ð6Þ opening displacement relationship follow the linear path that
ðeu =e0 1Þ2 ðeu =e0 Þ
passes through the origin of the coordinate system as shown in
Fig. 5.
bcr ecr
2 4.1. Mesh convergence
Gcr ¼ nt
2
3:8fc1=3 ð13Þ
w2 þD Chen and Pan [42] and Lundqvist et al. [10] conducted 3-D lin-
(5) Kang’s shear stress–slip model [38] ear elastic FE analyses to study stress concentration in respectively
pffiffiffiffi external FRP-to-concrete and NSM FRP-to-concrete bonded joints.
Gcr ¼ ð0:543w0:585 þ 0:199Þ fc ðbcr Þ0:72 ðecr
nt Þ
0:28
ð14Þ Teng et al. [43] conducted 2-D linear elastic FE analyses to study
stress concentration near plate ends in RC beams strengthened
(6) Vecchio and Lai’s shear stress–slip model [39] with an externally bonded plate. In their FE models, very small ele-
pffiffiffiffi 2
ments were used (e.g. the smallest elements used near the plate
bcr ecr
fc
nt ends were 0.1 mm in size). Although stress singularity points also
Gcr ¼ 2 2 24w
ð15Þ
d þD 0:31 þ Da þ16 exist in the present problem in the elastic range, it is not necessary
pffiffiffi to employ very small elements as in these two previous studies for
0:5 fc =ð0:31þD24w Þþfcu =30
where d ¼ 1:8w0:8 þð0:234w 0:707 0:2Þf .
a þ16
cu
two reasons: first, this study is more concerned with the failure
J.G. Teng et al. / Computers and Structures 117 (2013) 105–117 111
σ σ
σ max σ max
ε ε
− σ max − σ max
60 Table 2
Predicted ultimate loads and corresponding displacements for specimen CS-150.
50 I
H CS-150
G
40 CS-100
FE prediction (kN)
Load (kN)
30 F
E Test CS-30 CS-250
20 Bazant and Gambarova’s model [34]
D CS-200
Walraven and Reinhardt’s model [35]
C Yamada and Aoyagi’s model [36]
Okamura and Maekawa’s model [37]
10
B Kang’s model [38]
Vecchio and Lai’s model [39]
A
0
0 0.2 0.4 0.6 0.8
Displacement at the loaded end (mm)
Test (kN)
Fig. 10. Load–displacement curves obtained with six different shear retention
models. Fig. 11. Predicted ultimate load versus test ultimate load.
J.G. Teng et al. / Computers and Structures 117 (2013) 105–117 113
on the top surface in a segment of the bonded joint from FE Fig. 13 shows the distributions of cracks (represented by values
analysis. of the maximum principal cracking strain) in the concrete at
J.G. Teng et al. / Computers and Structures 117 (2013) 105–117 115
0.20 Pu (Point B)
0.06 Pu (Point A) 0.85 Pu (Point G)
3500 FE result 0.72 Pu (Point F)
3000 Solid lines are FE 0.59 Pu (Point E)
predictions 0.46 Pu (Point D)
2500 corresponding to adjacent 0.33 Pu (Point C)
test curves 0.20 Pu (Point B)
2000 5 0.06 Pu (Point A)
1500
1000
500
0
0 50 100 150
Distance from the free end (mm)
0
0 50 100 150
Fig. 14. Comparison of strain distributions in the FRP between test and FE analysis.
(Note: The strain distributions from FE analysis were adjusted for clearer compar- Distance from the free end (mm)
isons in accordance with the ultimate load ratio between test and FE analysis,
which is 0.96). Fig. 15. Bond shear stress distributions from FE analysis.
116 J.G. Teng et al. / Computers and Structures 117 (2013) 105–117
0.59 Pu (Point E) 0.46 Pu (Point D) Okamura and Maekawa’s model [37] was found to lead to
0.33 Pu (Point C) 0.20 Pu (Point B) accurate results and was adopted in the proposed FE model.
0.06 Pu (Point A)
(3) In the present study, the FE model predicted no cracks in the
adhesive; this is because the adhesive had a much higher
5 tensile strength and a much lower elastic modulus than
the concrete. This result is in agreement with the experi-
mental observation that in most tests, failure occurred in
the concrete provided the adherend surfaces were appropri-
ately prepared and the adhesive was sufficiently strong.
(4) The load–displacement curves and the strain distributions in
the NSM FRP strip from FE analysis are in close agreement
0
0 50 100 150 with the test results. The FE local bond stress distribution
is significantly different from the test distribution and this
Distance from the free end (mm)
discrepancy is due to the limited number of strain gauges
Fig. 16. Bond shear stress distributions deduced from strain measurements. installed along the bond length of the FRP strip in the test.
(5) The present study has been focused on NSM FRP strips. The
proposed modeling approach has the potential for applica-
10 tion to NSM FRP bars of other cross-sectional shapes.
(6) The proposed FE model provides a powerful and economical
alternative to laboratory testing to gain a full understanding
Bond shear stress (MPa)
FRP Composites in Civil Engineering (CICE 2010), 27–29 Sept., Beijing, 2010, p. [29] Hognestad E. A Study of Combined Bending and Axial Load in Reinforced
45–50. Concrete Members, University of Illinois Engineering Experiment Station,
[14] Li XQ, Chen JF, Lu Y. Meso-scale modelling of FRP-to-concrete bond behaviour Bulletin Series No. 399, 1951.
using LSDYNA. In: Proceedings of fifth International Conference on FRP [30] ACI-318 2008. Building Code Requirements for Structural Concrete and
Composites in Civil Engineering (CICE 2010), Beijing, China, 27–29, September, Commentary, ACI 318, Farmington Hills, MI.
2010, p. 494–498. [31] Hillerborg A, Modeer M, Petersson PE. Analysis of crack formation and crack
[15] Nanni A, Bakis CE, Boothby TE. Test methods for FRP-concrete systems growth in concrete by means of fracture mechanics and finite elements. Cem
subjected to mechanical loads: state of the art review. J Reinf Plast Compos Concr Res 1976;6(6):773–82.
1995;14(6):524–58. [32] Petersson PE. Crack Growth and Development of Fracture Zone in Plain
[16] Yan X, Miller B, Nanni A, Bakis CE. Characterization of CFRP bars used as Concrete and Similar Materials, Report TVBM-1006, Division of Building
near-surface mounted reinforcement. In: Proceedings of Eighth Inter- Materials, Lund Institute of Technology, Lund, Sweden, 1981.
national Structural Faults and Repair Conference, Edinburgh, UK (CD-ROM) [33] Hordijk DA. Local Approach to Fatigue of Concrete, Ph.D thesis, Delft University
1999. of Technology, 1991.
[17] CEB-FIP. Model Code 90, Lausanne, Switzerland, 1993, 1990. [34] Bazant ZP, Gambarova P. Rough cracks in reinforced concrete. J Struct Div ASCE
[18] MSC.MARC. User’s Manual. MSC. Software Corporation, 2 MacArthur Place, 1980;106(4):819–42.
Santa Ana, California 92707, USA, 2005. [35] Walraven JC, Reinhardt HW. Theory and experiments on the mechanical
[19] Teng JG, Zhu WC, Tang CA. Mesomechanical model for concrete – Part II: behavior of cracks in plain and reinforced concrete subjected to shear loading.
Applications. Mag Concr Res 2004;56(6):331–45. Heron 1981;26(1A):1–68.
[20] Zhu WC, Teng JG, Tang CA. Mesomechanical model for concrete – Part I: Model [36] Yamada K, Aoyagi Y. Shear transfer across crack. In: Proceedings, JCI Second
development. Mag Concr Res 2004;56(6):313–30. Colloquium on Shear Analysis of RC Structures, October, 19–28, 1983.
[21] Yang ZJ, Chen JF. Finite element modelling of multiple cohesive discrete crack [37] Okamura H, Maekawa K. Nonlinear analysis and constitutive models of
propagation in reinforced concrete beams. Eng Fract Mech reinforced concrete. Tokyo: Gihodo-Shuppan; 1991.
2005;72(14):2280–97. [38] Kang QL. Finite element analysis for reinforced concrete. Beijing: China Water
[22] Lu XZ, Jiang JJ, Teng JG, Ye LP. Finite element simulation of debonding in FRP- Power Press; 1996 (in Chinese).
to-concrete bonded joints. Constr Build Mater 2006;20(6):412–24. [39] Vecchio FJ, Lai D. Crack shear-slip in reinforced concrete elements. J Adv Concr
[23] Rashid YR. Ultimate strength analysis of prestressed concrete pressure vessels. Technol 2004;2(3):289–300.
Nucl Eng Des 1968;7:334–44. [40] Rots JG. Computational Modelling of Concrete Fracture, Ph.D thesis, Delft
[24] Bazant ZP, Oh BH. Crack band theory for fracture of concrete. Mater Struct University of Technology, The Netherlands, 1988.
1983;16(93):155–77. [41] Yu T, Teng JG, Chen JF. Chapter 54: Mechanical properties of FRP composites.
[25] Buyukozturk O. Nonlinear analysis of reinforced concrete structures. Comput In: ICE Manual of Construction Materials. London: Thomas Telford Ltd.; 2009.
Struct 1977;7:149–56. p. 41–647.
[26] Chen WF. Plasticity in reinforced concrete. New York: McGraw-Hill; 1982. [42] Chen JF, Pan WK. Three dimensional stress distribution in FRP-to-concrete
[27] Elwi AA, Murray DW. A 3D hypoelastic concrete constitutive relationship. J Eng bond test specimens. Constr Build Mater 2006;20(1–2):46–58.
Mech Div ASCE 1979;105(4):623–41. [43] Teng JG, Zhang JW, Smith ST. Interfacial stresses in reinforced concrete beams
[28] Saenz LP. Discussion of equation for the stress–strain curve of concrete, by bonded with a soffit plate: a finite element study. Constr Build Mater
Desayi and Krishnan. J Am Concr Inst 1964;61:1229–35. 2002;16(1):1–14.