Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Construction and Building Materials 25 (2011) 1963–1979

Contents lists available at ScienceDirect

Construction and Building Materials


journal homepage: www.elsevier.com/locate/conbuildmat

Evaluation of debonding failure of reinforced concrete girders strengthened


in flexure with FRP laminates using finite element modeling
Feifei Lu a, Ashraf Ayoub b,⇑
a
The Shaw group Inc., Charlotte, NC 28202, USA
b
Department of Civil and Environmental Engineering, University of Houston, Houston, TX 77204, USA

a r t i c l e i n f o a b s t r a c t

Article history: The use of Fiber-Reinforced Polymer (FRP) materials dates back to the early 1940s when they were used
Received 27 June 2008 in aerospace and naval applications. During the 1970s and early 1980s, FRP started being used in civil
Received in revised form 8 August 2010 engineering applications for new construction, but more importantly for repair and strengthening of
Accepted 13 November 2010
existing structures. However, experimental research showed that the typical failure mode of reinforced
Available online 18 December 2010
concrete (RC) structures strengthened with FRP composite materials is due to the debonding that occurs
at the interface between concrete and FRP. The bond between FRP and concrete is therefore the key factor
Keywords:
controlling the behavior of these structures since it limits the full use of the FRP strength. The paper eval-
Composites
Laminates
uates the effect of the debonding failure on the response of FRP-strengthened RC beams. A nonlinear RC
Interface beam element with bond-slip between the concrete and the FRP laminates is developed and used to ana-
Debonding lyze several test specimens and to investigate their corresponding failure mode. The model was also used
Finite element analysis to study the reduction factor of FRP tensile strength of simply supported strengthened RC girders due to
debonding failure. This reduction factor proved to be affected by several parameters: (a) the bond
strength between FRP and concrete interface; (b) the concrete strength; (c) the thickness of FRP; (d)
the modulus of FRP; (e) the width of FRP laminate; and (f) the development length of the FRP sheet. A
large number of beam specimens were analyzed in order to conduct a thorough evaluation of debonding
failure of RC beams strengthened with FRP laminates. Based on these studies, new equations that account
for the aforementioned parameters were proposed to address the reduction in FRP strength due to deb-
onding failure.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction experimental research conducted on FRP-strengthened RC girders,


the corresponding failure modes are classified in two types: (a)
Strengthening of deficient RC structures with FRP materials was flexural failure due to concrete crushing or FRP rupture; or (b) fail-
established as an efficient and economical technique for repair and ure due to debonding between the FRP laminate and concrete sur-
rehabilitation since the late 1970s. The complex behavior and fail- face, or due to end peeling, where the concrete cover in the region
ure modes of FRP composite structures were studied through near the supports peels off. The experimental work, however, indi-
extensive experimental and analytical investigations. Saadatm- cate that the most common failure mode is due to FRP debonding.
anesh and Ehsani [1,2] conducted both experimental and analytical Studies related to bond behavior between FRP and concrete were
research to address the behavior of RC beams strengthened with conducted by several researchers. De Lorenzis et al. [11] conducted
Glass Fiber-Reinforced Polymer (GFRP) plates. Chajes et al. [3], Baa- pull-out tests with different layers of CFRP sheets. Additional pull-
za et al. [4], and Nakamura et al. [5] evaluated the behavior of RC out tests were carried out by Bizindavyi and Neale [12], Sena-Cruz
girders strengthened with Carbon Fiber-Reinforced Polymer (CFRP) and Barros [13], and Yao et al. [14] under monotonic loads; and by
sheets. Arduini and Nanni [6] studied the behavior of pre-cracked Ko and Sato [15] under cyclic loads.
RC beams strengthened with CFRP beams. Shahawy and Beitelman Several analytical models to describe the bond stress-slip
[7] studied the effect of different wrapping schemes on the failure behavior of FRP–concrete interface were proposed. A bond-slip
mode. Tumialan et al. [8] and Sebastian [9] evaluated the effect of model to quantify the different bonding mechanisms and describe
cover delamination on the behavior of FRP-strengthened RC the bond characteristics of FRP sheet-concrete interfaces was pro-
girders. Thomsen et al. [10] reported that based on the observed posed by Dai [16], and Ueda and Dai [17]. A closed-form analytical
solution for the bond slip analysis of beams strengthened with full
⇑ Corresponding author. or partial FRP wraps based on beam theory with a shear deform-
E-mail address: asayoub@uh.edu (A. Ayoub). able adhesive layer was developed by Rashid and Pervaiz [18].

0950-0618/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.conbuildmat.2010.11.059
1964 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

Fracture mechanics-based approaches referred to as mesoscale wy


models, utilize two-dimensional FE models with the assumption
of concrete substrate failure to simulate and characterize the local
debonding failure in the interface region. Such models were pro- V V + dV
posed by Taljsten [19], Wu et al. [20], Yuan et al. [21], Wu and Nc N c + dN c
Yin [22], Lu et al. [23], Coronado and Lopez [24,25], and Lu et al.
[26]. These models however utilize a very fine mesh with small
element sizes and are generally computationally expensive. An- M M + dM
other approach that involves simulating cracking and failure of h
the concrete adjacent to the adhesive layer following a smeared
crack model that treats cracked concrete as a continuum was pro- q
posed by Pham and Al-Mahaidi [27]. Discrete crack models to rep-
NFRP NFRP+dNFRP
resent discontinuities due to major cracks near the adhesive layer
were adopted by Yang et al. [28], Kishi et al. [29], and Niu and Wu
[30]. The finite element approach with interface elements having a dx
pre-defined bond-slip relation was used to describe the bond
Fig. 1. Reinforced concrete beam segment strengthened with FRP sheet.
behavior by Chen et al. [31], Aprile et al. [32], Wong and Vecchio
[33], Supaviriyaki et al. [34], Baky et al. [35], and Godat [36]. A crit-
ical review and assessment of different bond-slip models was pre- LT DðxÞ  LTb qqb ðxÞ ¼ w ð4Þ
sented by Lu et al. [37]. In the study, a set of three new bond-slip 2 3
models were also proposed. Empirical models to address the effect d=dx
of debonding were proposed by Chaallal et al. [38] and Khalifa et al. where DðxÞ ¼ ½N c ðxÞNFRP ðxÞMðxÞT ; L ¼ 4 5is a 3 
d=dx
[39]. d=dx
To account for debonding failure, the ACI-440 guidelines [40] 3 differential operator, q is a 1  3 differential operator, q is the
introduced a bond reduction factor km for the FRP strength. This beam width, and w ¼ ½ 0 0 wy T is a 1  3 load vector.The com-
factor equals the effective FRP stress at failure divided by the origi- patibility equations are grouped in the following matrix form:
nal FRP strength. The development of the proposed values for the
LuðxÞ  dðxÞ ¼ 0 ð5Þ
bond reduction factor km was based mainly on experimental inves-
tigations and do not account for all the parameters affecting this where uðxÞ ¼ ½ uc ðxÞ uFRP ðxÞ v ðxÞ T , and dðxÞ ¼ ½ ec ðxÞ eFRP ðxÞvðxÞT
type of failure mode. is the concrete axial displacement, uFRP is the FRP axial displace-
ment, m is the transverse displacement, ec is the concrete axial
2. Research objective and scope strain, eFRP is the FRP axial strain, and v is the curvature. L is the
differential operator defined in (4).
The objective of the research study is to determine the param- The slip between the FRP sheet and the concrete bottom fiber
eters influencing the debonding failure of FRP-strengthened RC Sb(x) is calculated as follows:
girders and to propose new guidelines for bond reduction factors
Sb ðxÞ ¼ Lb uðxÞ ð6Þ
that account for all these parameters. In order to accomplish this
task, a nonlinear finite element model capable of simulating all where Lb is as defined in (4).
possible failure modes of FRP-strengthened concrete girders is The present model uses fiber discretization to describe section
developed. The model is based on fiber discretization, and was behavior, and an interface element to model bond between FRP
added to the library of elements of the finite element program and concrete. The concrete uniaxial constitutive law used is based
FEAP described in Zienkiewicz and Taylor [41]. The model was on the model by Kent and Park [42], the steel uniaxial stress–strain
used to simulate the behavior of a large number of specimens with
different material and structural properties. A least square fit of the
results of the analyses was conducted to develop the desired bond
reduction equations. The finite element model is briefly described
in the next section.

3. Finite element formulation of FRP-strengthened RC beams

Consider first an element of length dx of an RC beam section


reinforced with one sheet as shown in Fig. 1. The equations of equi-
librium for the beam are:
Nc;x ¼ qqb ð1Þ

NFRP;x ¼ qqb ð2Þ

M xx  hqb;x ¼ wy ð3Þ

where Nc, and NFRP are the axial forces in the concrete and FRP sheet
respectively, qb is the bond stress, q is the section width, h is the
distance between the centroid of the concrete beam and the FRP
sheet, wy is the applied external load, M is the bending moment,
and a comma denotes derivation. Writing the equilibrium equations
in matrix form: Fig. 2. Bond-slip model.
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1965

law used is assumed to be elasto-plastic, the FRP is modeled as an


elasto-brittle material, while the bond between the concrete and
the FRP is represented with a three stage multi-linear model as
show in Fig. 2. This bond model was used successfully by various
researchers and proved to agree well with experimental results.
The finite element formulation is derived by considering the
weighted integral form of the equilibrium equations at a New-
ton–Raphson iteration i.
Z L
duT ðxÞ½LT Di ðxÞ  LTb qqib ðxÞ  wdx ¼ 0 ð7Þ Fig. 3b. Loading arrangement of RC beam.
0

where du(x) is a weighting function. For the sake of simplicity the


element loads w is not considered in this study. Integrating by parts Table 1
twice the first term in (7), and substituting the incremental force– CFRP properties.
deformation relation of the section and bond interface respectively Property Value Units
yields
Tensile strength 2400 MPa
Z L Z L Tensile modulus 150 GPa
i1 i i1
T
L du T
ðxÞ½ks Dd þD i1
dx þ LTb duT ðxÞ½
q kb DSib þq qbi1 dx Thickness 1.2 mm
0 0
¼ BT ð8Þ

where ks and kb are the section and bond stiffness terms respec- Table 2
tively, BT is a boundary term, D and qb are the section and bond Section and material properties.
resisting loads respectively. Property Value Units
Adopting a displacement approach by substituting assumed
As 384 mm2
displacement shape functions a(x) into (8) yields A0s 256 mm2
  fy 400 MPa
Ksi1 þ Kbi1 Dui ¼ P  Q si1  Q bi1 ð9Þ Es 210 GPa
fc0 25 MPa
RL i1
where Ki1 s ¼ 0 BT ðxÞks ðxÞBðxÞdx is the section element stiffness
matrix
RL i1
Ki1
b ¼ 0 BTb ðxÞqkb ðxÞBb ðxÞdx is the bond interface element
mm was assumed for the numerical investigation, and a bond
stiffness matrix
RL strength was assumed to equal 3.1 MPa.
Q i1
s ¼ 0 BT Di1 ðxÞdx is the section element resisting load
Fig. 4a shows the experimental load–deformation plot of the
vector
RL specimen, while Fig. 4b shows the analytical results using the pro-
Q i1
b ¼ 0 BTb ðxÞqqi1
b ðxÞdx is the bond interface element resisting
posed model. The obtained numerical results clearly match well
load vector
with the experimental ones. Fig. 5a shows the bond force distribu-
BðxÞ ¼ LaðxÞ; Bb ðxÞ ¼ Lb aðxÞ and P is the vector of applied exter-
tion along the FRP–concrete interface provided by Thomsen et al.
nal loads
[10] for different values of the CFRP plate width w. Fig. 5b shows
the corresponding analytical bond stress distribution using the
4. Validation of analytical model proposed finite element model. In both figures, all specimens reach
the maximum bond stress near the point where the RC beam was
Three test specimens were selected to confirm the validity and loaded.
accuracy of the proposed finite element model. The first is the shal- The second specimen was tested by Shahawy and Beitelman [7].
low RC beam tested by Zarnic et al. [43]. The dimension of the The dimensions and cross-sectional details of the T-shaped test
beam is 800  120 mm as shown in Fig. 3a. The CFRP plate was beam are shown in Figs. 6 and 7 respectively. Tables 3 and 4 show
100 mm wide and 1.2 mm thick. The dimensions and loading
arrangement of the beam are shown in Fig. 3b. The material prop-
erties and geometry of the specimen are given in Tables 1 and 2.
According to the mechanical properties of the epoxy resin used
in the experimental test, a bond elastic stiffness of 2384 MPa/

Fig. 3a. Section information of RC beam. Fig. 4a. Experimental response of RC beam (from [1]).
1966 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

Load-Deformation CFRP Strengthened Slab

Control Slab

Load 2F (KN)

Mid-Span Displacement (mm)

Fig. 4b. Analytical results of Zarnic beam.

4. In the beam designation of Table 5, the first letter indicates that


the beam was a control specimen (C), fully wrapped (W) or par-
tially wrapped (P); while the second part of the beam nomencla-
ture indicates the number of layers of the fabric. Grade 60
reinforcing steel was used as the internal reinforcement with yield
stress and elastic modulus of 441 MPa (64 ksi) and 203 MPa
(29,500 ksi) respectively. The system is composed of unidirectional
dry carbon material formed by weaving individual yarns into a fab-
ric that has a manufacturer reported ultimate tensile strength of
3.65 GPa (530 ksi) and an elastic modulus of 231 GPa (33,500
ksi). The sheets created during the weaving process result in a
2 mm thick material with individual yarn density of 0.43/mm
(11/in.). Each individual yarn has a cross-sectional area of pure car-
bon equal to 0.45 mm2 (700  106 in.2).
Figs. 8a and 8b show the experimental results of the test spec-
imens, while Fig. 9 shows the corresponding analytical results. In
Fig. 5a. Bond force distribution by Thomsen et al (2004) (from [1]).
general, the analytical results match well with the experimental
ones, and in both cases, it was shown that the partially wrapped
specimens exhibited less ductility and strength than full wrapped
the details of the material properties of the specimens and FRP specimens. For the fully wrapped specimens, however, as the lay-
respectively. Table 5 lists the static test programs. As shown in ers of the FRP increase, the ultimate moments of the strengthened
Table 5, the number of biaxial fabric layers varies between 0 and beam increase.

Ultimate Bond stress 3.1 MPa

W=50 mm
Bond Stress (MPa)

W=75 mm

W=100 mm

W=150 mm

W=300 mm

Distance along the beam (mm)

Fig. 5b. Bond force distribution using proposed model.


F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1967

Table 3
Section and material properties.

Property Value Units


A 95,380 mm2
As 1135 mm2
A0s 284 mm2
Sb 4.9  106 mm3
St 10.8  106 mm3
I 1.5  109 mm4
fy 441 MPa
Es 203 GPa

Table 4
CFRP properties.

Property Value Units


Tensile strength 3.65 GPa
Tensile modulus 231 GPa
Fig. 6. Section information of Shahawy specimen. Filament diameter 7 lm
Filament/yarn 12,000
Ultimate elongation 1.4 %

Fig. 10 shows the bond stress distribution along the beam for
the partially wrapped specimen P-2L5-A. From the figure, the peak
Table 5
value of bond stress equals to 2.3 MPa and appears near the point
Details of specimens and test setup.
where the RC beam was loaded.
The third specimen was tested by Arduini et al. [44]. The dimen- Specimen Mn (KN m) Reinforcement description fc0
sions and loading arrangement of the test beams are shown in Specimen Mn (KN m) Reinforcement description fc0
Fig. 11. Beams B1–B3 were selected for this study. Beam B1 is C-OL5 189.8 Control specimen, no carbon 35.9
the control specimen, Beam B2 is strengthened with one-ply of P-2L5 190.6 Two layers on bottom of stem 37.2
unidirectional carbon FRP sheets with thickness of 0.17 mm cover- 1 Layer 211.4 One layers on full stem 35.9
ing the soffit of the beam, and Beam B3 is strengthened with three 2 Layer 259.9 Two layers on full stem 35.1
3 Layer 282.5 Three layers on full stem 35.1
plies of the same thickness. Another specimen B4 was tested by the
4 Layer 305.1 Four layers on full stem 35.1
researchers and is similar to specimen B3, except that it had a
fourth ply with fibers in the transverse direction wrapped around
the lateral faces. This specimen was not included in this study
due the additional shear strengthening provided by the transverse 5. Evaluation of parameters affecting bond reduction factor
sheets that was not considered in this work. The concrete compres-
sive strength fc0 used is 30 MPa, the steel yield stress is 340 MPa, the In order to assess the effect of the different parameters influenc-
modulus of the CFRP sheets is 400 GPa and its rupture stress is ing the behavior and failure mode of FRP-strengthened girders, a
3000 MPa, and the bond strength is 4.5 MPa. series of numerical studies on two representative specimens is
The experimental load–deformation plot is shown in Fig. 12. undertaken using the proposed finite element model. The selected
The failure mode of Beam B1 was due to concrete crushing after specimens are shown in Figs. 14a and 14b. The specimens are ana-
steel yielding. Beam B2 failed due to FRP rupture in the middle lyzed using different material and geometric properties. For each
span region after steel yielding, while Beam B3 failed due to FRP case, the failure mode is identified as to whether it is due to con-
debonding. The analytical load–deformation plot for the specimens crete crushing, FRP rupture or debonding. If failure is due to deb-
is shown in Fig. 13. Good correlation was observed between the onding, the value of the bond reduction factor km, defined as the
experimental and analytical plots although the ultimate strength ratio of the FRP stress at debonding ffe to the FRP strength ffu, is
was slightly overestimated since the numerical model did not ac- evaluated. The parameters evaluated in this study are: (a) the span
count for the loading–reloading cycles performed in the tests. More length-to-depth ratio, (b) the bond strength, (c) the concrete com-
importantly, the analytical model was able to accurately predict pressive strength, (d) the FRP modulus, and (e) the FRP sheet
the proper failure mode. Concrete crushing was observed for Beam width, thickness and development length. To evaluate the effect
B1, FRP rupture was detected for Beam B2 as shown in Fig. 13, of each parameter, the values of this parameter were varied be-
while the bond strength was reached for Beam B3 at a displace- tween its potential upper and lower limits, while all other param-
ment of around 8.2 mm. eters were kept constant and equal to a typical case. The typical

Fig. 7. Loading arrangement of Shahawy specimen.


1968 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

Fig. 8a. Experimental results for Shahawy specimen.

Fig. 8b. Experimental results for Shahawy specimen.

case for Section I consists of: concrete strength of 25 MPa, FRP


350 modulus of 150 GPa, bond strength of 3 MPa, FRP thickness of
1.2 mm, ratio of FRP width to section width of 0.25, and ratio of
300 length of FRP sheet to span length of 0.284. The typical case for Sec-
tion II consists of: concrete strength of 25 MPa, FRP modulus of
Applied Moment (kN-m)

250
150 GPa, bond strength of 3 MPa, FRP thickness of 1.5 mm, ratio
of FRP width to section width of 0.25, and ratio of length of FRP
sheet to span length of 0.208. In addition, for each case evaluated,
200
two bond stress-slip models were used. The first model, referred to
as bond model I, is an elasto-brittle model; and the second, re-
150
ferred to as bond model II, is a piece-wise trilinear model similar
Control
P-2L5-A to the one shown in Fig. 2. In all analyses performed, only failure
100 due to debonding is considered, while failure due to end peeling
1 Layer
2 Layer is not considered since the finite element formulation does not ac-
50 3 Layer count for shear deformations.
4 Layer
0
0 20 40 60 80 100 120 5.1. Span length to depth ratio
Midspan Deflection (mm)
Typically, for FRP-strengthened beams with very large spans,
Fig. 9. Analytical results of Shahawy specimen. the failure is not due to debonding, but rather to either FRP rupture
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1969

Fig. 10. Bond force distribution of Shahawy specimen.

280
P/2 P/2 Beam B1
240 Beam B2
Beam B3
200
400 Load (kN)
160

1100 300 1100 120


150 150
80

A’s=2φ13 40

0
0 2 4 6 8 10 12
400 350
Mid-span Deflection (mm)

Fig. 13. Analytical load–deformation of Arduini et al (1997) specimens.


As=3φ13
300
or concrete crushing. This is due to the fact that for beams with
Dimensions in mm
large spans, the flexural behavior dominates the response. In this
Fig. 11. Arduini et al (1997) test beams. case, the bond strength has little effect on the behavior. To identify
the critical span after which bond strength does not affect the

Fig. 12. Experimental load–deformation of Arduini et al (1997) specimens (from [44]).


1970 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

Fig. 14a. Section I and geometry information of RC beam.

Fig. 14b. Section II and geometry information of RC beam.

response, numerical analyses of beams with different span-to- approximately 20, the failure of the beams was due to debonding.
depth values were carried out. As shown in Fig. 15a, two different Therefore, debonding failure occurs when L/d < 20. For the case of
curves illustrate the effect of the span length-to-depth ratio (L/d) bond model II, as shown in Fig. 15b, debonding failure occurs when
on the km factor for two different concrete strengths, 25 MPa and L/d is less than 12; otherwise the failure is due to concrete crush-
35 MPa, and using bond model I. For both curves, except for the last ing. Therefore, the critical span length-to-depth ratio L/d < 20 for
point on the curves which corresponds to an L/d value of bond model I, while L/d < 12 for bond model II.
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1971

Ratio of ffe/ffu

f'c=25MPa debonding failure


f'c=35MPa debonding failure
f'c=25MPa concrete failure
f'c=35MPa concrete failure

L/d

Fig. 15a. Relationship of km factor and L/d for bond model I.

f'c=25MPa debonding failure

f'c=35MPa debonding failure

f'c=25MPa concrete failure

f'c=35MPa concrete failure

L/d

Fig. 15b. Relationship of km factor and L/d for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
2.0 2.5 3.0 3.5 4.0 4.5 5.0
Bond Stress between FRP and Concrete (MPa)

Fig. 16a. Relationship of bond stress and value of ffe/ffu for bond model I.
1972 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

1.00

0.90

0.80

0.70
Ratio of ffe/ffu
0.60

0.50

0.40

0.30

0.20

0.10

0.00
2.0 2.2 2.4 2.6 2.8 3.0
Bond Stress between FRP and Concrete (MPa)

Fig. 16b. Relationship of bond stress and value of ffe/ffu for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
25.0 30.0 35.0 40.0 45.0
Concrete Compressive Strength (MPa)

Fig. 17a. Relationship of concrete strength and value of ffe/ffu for bond model I.

5.2. Bond strength The previous analysis is repeated taking into account the bond
ductility, by adopting the trilinear bond stress-slip model II. The
There are several factors that can affect the bond stress between hardening slope was assumed to equal 1%, and the slip at the start
the FRP and concrete surface, such as the adhesive properties and and end of the softening branch were assumed to equal 0.12 mm
the surface preparation. The ultimate bond stress is the key param- and 0.3 mm respectively. Fig. 16b shows the average value of ffe/
eter affecting the potential debonding failure, and typically varies ffu for different bond strength values ranging from 2 MPa to
between 2 MPa and 6 MPa as evidenced from experimental pull- 3 MPa. As in the previous case, the value of ffe/ffu is linearly increas-
out tests. In order to assess the effect of the ultimate bond stress ing as the bond strength increases. When the bond strength is lar-
on the km factor, numerical studies covering this range were under- ger than 3 MPa, concrete failure occurs before debonding failure.
taken on the representative specimens. Fig. 16a shows the average
value of ffe/ffu as a function of the bond strength using bond model 5.3. Concrete compressive strength
I. As the bond strength increases, the value of ffe/ffu also increases
up to a corresponding bond strength value of 5 MPa. At that bond The effect of the concrete compression strength on the bond
strength value and beyond, failure was due to concrete crushing reduction factor was also examined. Different values of concrete
rather than debonding. strengths were used, ranging between 20 MPa and 45 MPa.
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1973

1.00

0.90

0.80

0.70
Ratio of ffe/ffu
0.60

0.50

0.40

0.30

0.20

0.10

0.00
25.0 30.0 35.0 40.0 45.0
Concrete Compressive Strength (MPa)

Fig. 17b. Relationship of concrete strength and value of ffe/ffu for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
80 120 160 200 240 280 320
Modulus of FRP (GPa)

Fig. 18a. Relationship of FRP modulus and value of ffe/ffu for bond model I.

Fig. 17a shows the effect of the concrete strength on the value of ffe/ case of bond model I. The value of ffe/ffu tends to increase as the
ffu for the case of bond model I. From the figure, the ratio of ffe/ffu FRP modulus increases from 80 GPa to 200 GPa. Beyond this value,
increases as fc0 slightly increases up to a value of 35 MPa, after it was observed that bond failure typically occurs near the support,
which the value of ffe/ffu remains constant. For values of fc0 less than and the corresponding value of ffe/ffu tends to remain almost con-
25 MPa, failure was found to be typically due to concrete crushing. stant. Fig. 18b shows the same plot but using bond model II. For
Fig. 17b shows the same plot for the case of bond model II. The plot cases where the FRP modulus ranges between 80 GPa and
reveals a similar trend, where the ratio of ffe/ffu increases as fc0 in- 150 GPa, failure was typically due to concrete crushing. For values
crease up to a value of 40 MPa, then remains constant. of the FRP modulus ranging between 150 GPa and 300 GPa, value
of ffe/ffu tends to linearly increase with the increase of the FRP
modulus.
5.4. FRP modulus

The modulus of FRP is directly related to the type of FRP. Three 5.5. Thickness of FRP sheet
types of FRP are widely used, namely, aramid, glass, and carbon.
For these three types of FRP the difference in their respective mod- The thickness of FRP plays an important role regarding the
uli is quite large ranging between 80 GPa and 300 GPa. Fig. 18a behavior and failure mode of strengthened RC beams. The thick-
shows the effect of the FRP modulus on the value of ffe/ffu for the ness of FRP also affects the bond strength. As the thickness of
1974 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

0.9

0.8

Ratio of ffe/ffu 0.7

0.6

0.5

0.4

0.3

0.2

0.1

0
80 120 160 200 240 280 320
Modulus of FRP (GPa)

Fig. 18b. Relationship of FRP modulus and value of ffe/ffu for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
0.5 1.0 1.5 2.0 2.5 3.0
Thickness of FRP (mm)

Fig. 19a. Relationship of thickness of FRP and value of ffe/ffu for bond model I.

FRP increases, the bond strength also increases. In this study, the by the contact area between the FRP sheet and concrete surface.
thickness of FRP was assumed to vary between 0.5 mm to The simplest way to increase the contact area between the FRP
2.8 mm. As shown in Fig. 19a for the case of bond model I, the va- sheet and concrete is to increase the width of the FRP sheet. In this
lue of ffe/ffu decreases sharply at first with the increase in the thick- study, the width of FRP was assumed to vary from 10% to 100% of
ness of FRP up to a value of about 1 mm. Beyond this point, the the section width.
value of ffe/ffu tends to decrease slowly. For the case of bond model Fig. 20a shows the relationship between the average value of ffe/
II, failure was due to concrete crushing for thicknesses less than ffu and the ratio of the width of the FRP sheet to the section width
1.5 mm. For thickness values greater than 1.5 mm, the value of ffe using bond model I. When the width of FRP sheet exceeds 42% of
/ffu decreases almost linearly as the thickness of FRP increases as the section width, concrete crushing occurs before debonding fail-
shown in Fig. 19b. ure. For cases where the width ranges between 10% and 42% of the
section width, the value of ffe/ffu increases linearly with the in-
5.6. Width of FRP sheet crease in the width of FRP sheet. The same trend is also observed
for the case of bond model II as shown in Fig. 20b, except that con-
The width of the FRP sheet is a key factor affecting the failure of crete crushing failure is observed when the width of FRP sheet ex-
FRP-strengthened beams. The bond stresses are directly influenced ceeds 28% of section width.
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1975

0.9

0.8

0.7
Ratio of ffe/ffu
0.6

0.5

0.4

0.3

0.2

0.1

0
0.5 1.0 1.5 2.0 2.5 3.0
Thickness of FRP (mm)

Fig. 19b. Relationship of thickness of FRP and value of ffe/ffu for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
0.10 0.20 0.30 0.40 0.50
Ratio of FRP width to Section width

Fig. 20a. Relationship of ratio of FRP width to section width and value of ffe/ffu for bond model I.

5.7. Development length of FRP sheet support for cases where L’ is less than 0.65. The value of ffe/ffu also
increases as L’ increases up to 0.65, as shown in Fig. 21b. Beyond
To distinguish between different FRP sheet lengths, an FRP this point, the value of ffe/ffu appears to remain constant.
development length L0 is introduced, and defined as the ratio of
the distance from the end of FRP sheet to the point where the load 6. Proposed new equations for bond reduction factor
is located to the shear span. To ensure adequate bond between the
FRP and concrete surface, sufficient development length of FRP is The numerical studies performed in the previous section using
needed. The value of L0 used in this study ranges between 0.44 the proposed finite element model were used to propose a new
and 0.95. expression for the bond reduction factor that accounts for the ef-
Fig. 21a shows the effect of L’ on the average ratio of ffe/ffu using fect of the different parameters discussed earlier. A least square
bond model I. For values of L’ up to 0.79, debonding failure typi- fit approach was used to develop the following equation, assuming
cally occurs near the end support, while for L0 values between the bond to be elasto-brittle up to the bond strength value, which
0.79 and 0.95, the debonding failure occurs where the beam was is referred to as bond model I:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
loaded. In general, the value of ffe/ffu increases as L’ also increases.
km ¼ a  Ef  t 0:8  fc0  s  w  L02:4 L=d < 20 ð10Þ
For bond model II, debonding failure was observed near the
1976 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

0.9

0.8

0.7
Ratio of ffe/ffu
0.6

0.5

0.4

0.3

0.2

0.1

0
0.10 0.15 0.20 0.25 0.30
Ratio of FRP width to Section width

Fig. 20b. Relationship of ratio of FRP width to section width and value of ffe/ffu for bond model II.

1.00

0.90

0.80

0.70
Ratio of ffe/ffu

0.60

0.50

0.40

0.30

0.20

0.10

0.00
0.40 0.50 0.60 0.70 0.80 0.90 1.00
L'

Fig. 21a. Relationship of ratio of FRP length to shear span and value of ffe/ffu for bond model I.

with s and fc0 represent the bond strength, and concrete compressive shows the parameters that were used in this numerical study.
strength respectively in MPa, Ef is the FRP modulus in GPa, t is the The results revealed that when using bond model II, failure was
FRP thickness in mm, L’ is the ratio of the length of the FRP sheet typically governed by concrete crushing, except for L/d values less
within the shear span to the length of the shear span, and w is than 8. In this case, shear governs the behavior rather than flexure,
the ratio of width of FRP plate to width of concrete section. The fac- an issue that is beyond the scope of this work.
tor of a is a constant that equals 0.00066. The analysis using bond model I resulted in the following
Similar to the analysis using an elasto-brittle bond model, a expression for the bond reduction factor:
least square fit approach was also used to develop an expression
for the case of a ductile bond behavior, referred to as bond model km ¼ a  E0:35
f  t0:7  ð0:13s þ 0:08Þ  ½ðw  0:5Þ5 þ 0:44 L=d < 12
II, which resulted in the following equation:
ð12Þ
km ¼ a  E0:3
f  fc00:4  ðs=tÞ0:6  L00:05  w0:15 L=d < 12 ð11Þ
where a equals to 0.422.
In this case, the factor a equals to 0.0276. In order to assess the practicality of the new expressions, they
The analysis conducted in section 5 for rectangular sections was were used in a comparative study with the current ACI-440 speci-
repeated for the T-section of Shahawy and Beitelman [7]. Table 6 fications. The study is discussed in the next section.
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1977

0.9

0.8

0.7
Ratio of ffe/ffu
0.6

0.5

0.4

0.3

0.2

0.1

0
0.40 0.50 0.60 0.70 0.80 0.90 1.00
L'

Fig. 21b. Relationship of ratio of FRP length to shear span and value of ffe/ffu for bond model II.

7. Evaluation of new equations for bond reduction factor Table 6


Numerical parameters for T-section.
According to the guidelines of ACI-440, the expression for the
Case fc0 t (mm) Ef (MPa) B (MPa) W L (m)
km factor is given by:
8 Case fc0 t (mm) Ef (MPa) B (MPa) W L (m)
nEf t f
< 601e ð1  360;000 Þ 6 0:90 for nEf t f 6 180; 000
fu Concrete strength 20 1.20 150 3.00 1.00 1.40
km ¼ SI ð13Þ 25 1.20 150 3.00 1.00 1.40
: 1
ð90;000Þ 6 0:90 for nEf tf > 180; 000
60efu nEf t f 30 1.20 150 3.00 1.00 1.40
35 1.20 150 3.00 1.00 1.40
The term km, expressed in (13) is a factor no greater than 0.90 that 40 1.20 150 3.00 1.00 1.40
may be multiplied by the rupture strain of the FRP laminate to reach Thickness of CFRP 25 0.50 150 3.00 1.00 1.40
a strain limitation to prevent debonding. The number of plies n used 25 1.00 150 3.00 1.00 1.40
in this equation is the number of plies of FRP flexural reinforcement 25 1.20 150 3.00 1.00 1.40
at the location along the length of the member where the moment 25 2.00 150 3.00 1.00 1.40
25 2.50 150 3.00 1.00 1.40
strength is being computed. This term recognizes that laminates
with greater stiffnesses are more prone to delamination. Thus, as Modulus of FRP 25 1.20 80 3.00 1.00 1.40
25 1.20 100 3.00 1.00 1.40
the stiffness of the laminate increases, the strain limitation becomes
25 1.20 120 3.00 1.00 1.40
more severe. For laminates with a unit stiffness n  Ef  t f greater 25 1.20 150 3.00 1.00 1.40
than 180,000 N/mm, km limits the force in the laminate as opposed 25 1.20 200 3.00 1.00 1.40
to the strain level. This effectively places an upper bound on the to- 25 1.20 300 3.00 1.00 1.40
tal force that can be developed in an FRP laminate, regardless of the Bond stress 25 1.20 150 2.00 1.00 1.40
number of plies. The width of the FRP laminate is not included in 25 1.20 150 2.50 1.00 1.40
25 1.20 150 3.00 1.00 1.40
the calculation of the unit stiffness, n  Ef  t f , because an increase
25 1.20 150 3.50 1.00 1.40
in the width of the FRP results in a proportional increase in the bond 25 1.20 150 4.00 1.00 1.40
area. The km term is only based on a generally recognized trend and
Width of FRP 25 1.20 150 3.00 0.55 1.40
on the experience of engineers practicing the design of bonded FRP 25 1.20 150 3.00 0.66 1.40
systems. The ACI-440 committee is currently developing new spec- 25 1.20 150 3.00 0.77 1.40
ifications for bond reduction factor (ACI-440.2R-08). Accordingly, 25 1.20 150 3.00 0.88 1.40
the FRP strain level at which debonding may occur, efd, is evaluated 25 1.20 150 3.00 1.00 1.40

as follows: Length of FRP bond 25 1.20 150 3.00 1.00 1.40


sffiffiffiffiffiffiffiffiffiffiffi 25 1.20 150 3.00 1.00 1.30
fc0 25 1.20 150 3.00 1.00 1.20
efd ¼ 0:41 6 0:9efu SI ð14Þ
nEf t f

The effective FRP stress, ffe, in case failure is governed by debonding is equal to 2400 MPa. While the original ACI-440 equation accounts
is: for Ef  t f , the proposed equations account for additional parame-
ffe ¼ Ef efd ð15Þ ters, such as width of FRP, bond stress between FRP and concrete,
concrete strength and length of FRP sheet. The comparison with
Fig. 22 presents a comparison between the proposed expressions for the original ACI-440 specifications reveals that using an elasto-brit-
rectangular sections and the expressions given by ACI-440. In order tle bond model produces bond reduction factors lower than those of
to conduct the comparison, a typical condition is assumed where ACI-440, and using a ductile bond model produces factors higher
the FRP modulus is equal to 150 GPa, and the FRP tensile strength than those proposed by ACI-440. While the new ACI specifications
1978 F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979

km Factor
ACI-440.2R-02
ACI-440.2R-08
Bond Model II
Bond Model I

km

n*Ef*t

Fig. 22. Comparison of proposed expressions for rectangular sections with ACI-440 guidelines.

km Factor

ACI-440-02
ACI-440-08
Bond Model I
km

n*Ef*t
Fig. 23. Comparison of proposed expressions for T-sections with ACI-440-02 guidelines.

account for the concrete compressive strength, they still do not ac- strengthened with FRP laminates. Based on these studies, new
count for the other parameters. The new guidelines appear to pro- equations for the bond reduction factors have been proposed. In
vide very conservative km values that are much lower than both, addition, the following conclusions can be drawn from the study:
the ones provided by the earlier specifications, and the proposed
ones using either bond model I or bond model II. From the preced-  The original ACI-440 expression for bond reduction factors
ing discussion, accounting for bond ductility following the use of seems to be on the conservative end if compared to a proposed
bond model II appears to provide in general a more efficient use expression for rectangular sections derived using a trilinear
of FRP than what the ACI-440 guidelines propose. bond stress-slip model, while it appears to be on the un-conser-
Fig. 23 shows the same plot for T-sections using the proposed vative end if compared to an expression derived using an elasto-
expression for bond model I. As can be seen, the proposed expres- brittle bond model. If bond ductility is considered, it appears
sion yields values very close the one proposed by the original ACI- that a more efficient use of FRP could be allowed than what is
440-02, and larger than that proposed by ACI-440-08. Again, the currently being proposed. The ACI-440 expression however
new guidelines seem to provide conservative km values for girders yields values very close to a proposed expression for T-sections
with T-sections. derived using an elasto-brittle bond model.
 The new proposed ACI-440 guidelines provide very conserva-
tive km values for rectangular sections that are smaller than
8. Summary and conclusions both, the one derived using an elasto-brittle bond model, and
that derived using a ductile bond model.
The paper analyzes the effect of the material and geometric  The new proposed ACI-440 guidelines provide conservative km
properties on the response and failure modes of FRP-strengthened values for T-sections that are smaller than the one derived using
RC beams. A nonlinear RC beam element model with bond-slip be- an elasto-brittle bond model.
tween the concrete and the FRP laminates was developed and used  The proposed ACI-440 expressions for bond reduction factors
to analyze a large number of beam specimens in order to conduct a does not take into account several factors that can directly
thorough evaluation of bond design specifications for RC beams affect the bond behavior and possible debonding failure, such
F. Lu, A. Ayoub / Construction and Building Materials 25 (2011) 1963–1979 1979

as width of FRP, bond stress between FRP and concrete, and [19] Taljsten B. Strengthening of concrete prisms using the plate-bonding
technique. Int J Fract 1996;82:253–66.
length of FRP sheets. The original expression did not also
[20] Wu Z, Yuan H, Yoshizawa H, Kanakubo T. Experimental/analytical study on
account for concrete compressive strength. Furthermore, the interfacial fracture energy and fracture propagation along FRP–concrete
expressions do not identify the critical span-to-depth ratio. interface. Fracture mechanics for concrete materials, ACI International SP-
 New expressions for the bond reduction factor that account for 201; September 2001. p. 133–52.
[21] Yuan H, Wu ZS, Yoshizawa H. Theoretical solutions on interfacial stress
the parameters identified in the previous item have been transfer of externally bonded steel/composite laminates. J Struct Mech Earthq
proposed for both rectangular and T-sections. These expres- Eng JSCE 2001;18(1):27–39.
sions, however, were based on analysis of two specimens with [22] Wu Z, Yin J. Fracturing behaviors of FRP strengthened concrete structure. Eng
Fract Mech 2003;70(10):1339–55.
specific geometry for rectangular sections and one specimen [23] Lu XZ, Teng JG, Ye LP, Jiang JJ. Meso-scale finite element model for FRP sheets/
for T-sections. Further numerical studies on specimens with dif- plates bonded to concrete. Eng Struct 2005;27(4):564–75.
ferent geometries and cross sections are needed in order to con- [24] Coronado CA, Lopez MM. Damage approach for the prediction of debonding
failure on concrete elements strengthened with FRP. J Compos Constr
firm the accuracy of these expressions. 2007;11(4):391–400.
[25] Coronado CA, Lopez MM. Experimental characterization of concrete epoxy
interfaces. J Mater Civil Eng 2008;20(4):303–12.
[26] Lu XZ, Teng JG, Ye LP, Jiang JJ. Intermediate crack debonding in FRP
strengthened RC beams: FE analysis and strength model. J Compos Constr
References 2007;11(2):161–74.
[27] Pham HB, Al-Mahaidi R. Finite element modelling of RC beams retrofitted with
[1] Saadatmanesh H, Ehsani M. RC beams strengthening with GFRP plates: CFRP fabrics. In: Proceedings 7th international RILEM symposium on non-
experimental study. J Struct Eng ASCE 1991;117(11):3417–31. metallic (FRP) reinforcement for concrete structures (FRPRCS-7) Kansas City,
[2] Saadatmanesh H, Ehsani M. RC beams strengthening with GFRP plates: Missouri, USA; 2005. p. 499–514.
analytical and parametric studies. J Struct Eng ASCE 1991;117(11):3432–55. [28] Yang ZJ, Chen JF, Proverbs D. Finite element modelling of concrete cover
[3] Chajes MJ, Thomson TA, Januszka TF, Fin W. Flexural strengthening of concrete separation failure in FRP plated RC beams. Constr Build Mater
beams using externally bonded composite materials. Constr Build Mater 2003;17(1):3–13.
1994;8(3):191–201. [29] Kishi N, Zhang G, Mikami H. Numerical cracking and debonding analysis of RC
[4] Baaza IM, Missihoun M, Labossiere P. Strengthening of reinforced concrete beams reinforced with FRP sheet. J Compos Constr 2005;9(6):507–14.
beams with CFRP sheets. In: Proceedings 1st international conference on [30] Niu H, Wu Z. Numerical analysis of debonding mechanisms in FRP
composites in infrastructure ICCI, Tucson, AZ; 1996. p. 746–59. strengthened RC beams. Comput-Aided Civil Infrastruct Eng
[5] Nakamura M, Sakai H, Yagi K, Tanaka T. Experimental studies on the flexural 2005;20(5):354–68.
reinforcing effect of carbon fiber sheet bonded to reinforced concrete beam. In: [31] Chen JF, Yang ZJ, Holt GD. FRP or steel plate-to-concrete bonded joints: effect
Proceedings 1st international conference on composites in infrastructure. ICCI, of test methods on experimental bond strength. Steel Compos Struct
Tucson, AZ; 1996. 760–73. 2001;1(2):231–44.
[6] Arduini M, Nanni A. Behavior of pre-cracked RC beams strengthened with [32] Aprile A, Spacone E, Limkatanyu S. Role of bond in RC beams strengthened
carbon FRP sheets. J Compos Constr ASCE 1997;1(2):63–70. with steel and FRP plates. J Struct Eng ASCE 2001;127:1445–52.
[7] Shahawy M, Beitelman T. Static and fatigue performance of RC beams [33] Wong RSY, Vecchio FJ. Towards modeling of reinforced concrete members
strengthened with CFRP laminates. J Struct Eng ASCE 1999;125(6):613–21. with externally bonded fiber reinforced polymer composites. ACI Struct J
[8] Tumialan G, Serra P, Nanni A, Belarbi A. Concrete cover delamination in RC 2003;100(1):47–55.
beams strengthened with FRP sheets. In: Proceedings 4th international [34] Supaviriyakit T, Pornpongsaroj P, Pimanmas A. Finite element analysis of FRP-
symposium on FRP for reinforcement of concrete structures. American strengthened RC beams. Songklanakarin J Sci Technol 2004;26:497–507.
Concrete Institute, Detroit; 1999. p. 725–35. [35] Baky HA, Ebead UA, Neale KW. Flexural and interfacial behavior of FRP
[9] Sebastian W. Significance of midspan debonding failure in FRP-plated concrete strengthened reinforced concrete beams. J Compos Constr 2007;11(6):629–39.
beams. J Struct Eng ASCE 2001;127(7):792–8. [36] Godat A, Neale KW, Labossiere P. Numerical modeling of FRP shear
[10] Thomsen H, Spacone E, Limkatanyu S, Camata G. Failure mode analyses of strengthened reinforced concrete beams. J Compos Constr 2007;11(6):640–9.
reinforced concrete beams strengthened in flexure with externally bonded [37] Lu XZ, Teng JG, Ye LP, Jiang JJ. Bond-slip models for FRP sheet/plate-to-concrete
fiber-reinforced polymers. J Compos Constr ASCE 2004;8:123–31. interfaces. Eng Struct 2005;27:938–50.
[11] De Lorenzis L, Miller B, Nanni A. Bond of FRP to concrete. ACI Mater J [38] Chaallal O, Nollet MJ, Perraton D. Strengthening of reinforced concrete beams
2001;98:256–64. with externally bonded fibre-reinforced-plastic plates: design guidelines for
[12] Bizindavyi L, Neale KW. Transfer lengths and bond strengths for composites shear and flexure. Can J Civil Eng 1998;25(4):692–704.
bonded to concrete. J Compos Constr ASCE 1999;3(4):153–60. [39] Khalifa A, Gold WJ, Nanni A, Aziz A. Contribution of externally bonded FRP to
[13] Sena-Cruz, JM, Barros JAO. Bond behavior of carbon laminate strips into shear capacity of RC flexural members. J Compos Constr ASCE
concrete by pullout-bending tests. In: Proceedings of the international 1998;2(4):195–203.
symposium bond in concrete – from the research to standards; 2002. p. [40] ACI Committee 440. Guide for the design and construction of externally
614–21. bonded FRP systems for strengthening concrete structures. (440.2R-02 2002)
[14] Yao J, Teng JG, Chen JF. Experimental study on FRP-to-concrete bonded joints. (440.2R-08 under review 2008). American Concrete Institute, Farmington
Composites: Part B 2005;36:99–113. Hills, Michigan.
[15] Ko H, Sato Y. Bond stress-slip relationship between FRP sheet and concrete [41] Zienkiewicz OC, Taylor RL. The finite element method volume 1. Basic
under cyclic load. J Compos Constr ASCE 2007;11(4):419–26. formulation and linear problems. 4th ed. London: McGraw Hill; 1989.
[16] Dai J. Interfacial models for fiber reinforced polymer (FRP) sheets externally [42] Kent DC, Park R. Flexural members with confined concrete. J Struct Eng ASCE
bonded to concrete. Ph.D. Dissertation, Hokkaido University, Japan; 2003. 1971;97(ST7):1969–90.
[17] Ueda T, Dai J. Interface bond between FRP sheets and concrete substrates: [43] Zarnic R, Gostic S, Bosiljkov V, Bokan V. Improvement of bending load-bearing
properties, numerical modeling and roles in member behaviour. Prog Struct capacity by externally bonded plates. In: Dhir RK, Henderson NA. editors.
Eng Mater 2005;7:27–43. Proceedings creating with concrete, Thomas Telford, London; 1999. p. 433–42.
[18] Rasheed HA, Pervaiz S. Bond slip analysis of fiber-reinforced polymer- [44] Arduini M, Di Tommaso A, Nanni A. Brittle failure in FRP plate and sheet
strengthened beams. J Eng Mech ASCE 2002;128:78–86. bonded beams. ACI Struct J 1997;94(4):363–70.

You might also like