Professional Documents
Culture Documents
Intermediate Crack Debonding in FRP-Strengthened RC Beams FE Analysis and Strength Model
Intermediate Crack Debonding in FRP-Strengthened RC Beams FE Analysis and Strength Model
net/publication/232809419
CITATIONS READS
114 929
4 authors, including:
Some of the authors of this publication are also working on these related projects:
Simulation Technologies of Mega-Earthquake Scenarios of Cities ( National Key Technology R&D Program) View project
Study on A Novel Earthquake and Progressive Collapse Resilient Concrete/Steel Frame View project
All content following this page was uploaded by Xinzheng Lu on 04 June 2014.
Publications
Bruce Gossett, Managing Director and SPECIAL ISSUE: Recent International Advancements in FRP Research and
Publisher Application in Construction
SPECIAL ISSUE EDITOR: Rudolf Seracino
Journals Department Editorial
Johanna M. Reinhart, Director, Journals
Jackie Perry, Managing Editor, Journals 109 Preface to IIFC Special Issue
Holly Koppel, Discussions and Closures Rudolf Seracino
Technical Papers
Production Department
Charlotte McNaughton, Director, Production 110 Large-Span Woven Web Structure Made of Fiber-Reinforced Polymer
Teresa Metcalfe, Manager, Journals Peng Feng, Lie-ping Ye, and J. G. Teng
Production 120 Mineral-Based Bonding of Carbon FRP to Strengthen Concrete Structures
Gene Sullivan, Senior Production Editor Björn Täljsten and Thomas Blanksvärd
Nancy Green, Production Editor
Rajashree Ranganathan, Production Editor 129 Bond Behavior of CFRP Bars in Simply Supported Reinforced Concrete
Xi Van Fleet, Manager, Information Beam with Hanging Region
Services Rendy Thamrin and Tetsuzo Kaku
Donna Dickert, Reprints
138 Bonding Characteristics of Fiber-Reinforced Polymer Sheet-Concrete
Interfaces under Dowel Load
Publishing Office Jianguo Dai, Tamon Ueda, and Yasuhiko Sato
Journals Department
ASCE 149 ESPI Measurement of Bond-Slip Relationships of FRP-Concrete Interface
1801 Alexander Bell Drive S. Y. Cao, J. F. Chen, J. W. Pan, and N. Sun
Reston, VA 20191-4400
161 Intermediate Crack Debonding in FRP-Strengthened RC Beams: FE Analysis
Telephone: 共703兲 295-6290
and Strength Model
E-mail: journal-services@asce.org X. Z. Lu, J. G. Teng, L. P. Ye, and J. J. Jiang
175 Study of Intermediate Crack Debonding in Adhesively Plated Beams
I. S. T. Liu, D. J. Oehlers, and R. Seracino
184 Modeling Debonding Failure in FRP Flexurally Strengthened RC Members
Using a Local Deformation Model
Scott T. Smith and Rebecca J. Gravina
192 Behavior of FRP Strengthened Reinforced Concrete Beams under Torsion
Mehran Ameli, Hamid R. Ronagh, and Peter F. Dux
201 Theoretical Model for Fiber-Reinforced Polymer-Confined Concrete
J. G. Teng, Y. L. Huang, L. Lam, and L. P. Ye
211 Performance-Based Seismic Retrofit Strategy for Existing Reinforced
Concrete Frame Systems Using Fiber-Reinforced Polymer Composites
Stefano Pampanin, Davide Bolognini, and Alberto Pavese
227 Carbon Fiber-Reinforced Polymer Strengthening and Monitoring of the
Gröndals Bridge in Sweden
Björn Täljsten, Arvid Hejll, and Gerard James
VOLUME 11 / NUMBER 2
MARCH/APRIL 2007
Abstract: Reinforced concrete 共RC兲 beams strengthened in flexure with a bonded fiber-reinforced polymer 共FRP兲 plate may fail by
intermediate crack 共IC兲 debonding, in which debonding initiates at a critical section in the high moment region and propagates to a plate
end. This paper first presents a finite-element 共FE兲 model based on the smeared crack approach for concrete for the numerical simulation
of the IC debonding process. This finite-element model includes two novel features: 共1兲 the interfacial behavior within the major flexural
crack zone is differentiated from that outside this zone and 共2兲 the effect of local slip concentrations near a flexural crack is captured using
a dual local debonding criterion. The FE model is shown to be accurate through comparisons with the results of 42 beam tests. The paper
also presents an accurate and simple strength model based on interfacial shear stress distributions from finite-element analyses. The new
strength model is shown to be accurate through comparisons with the test results of 77 beams, including the 42 beams used in verifying
the FE model, and is suitable for direct use in design.
DOI: 10.1061/共ASCE兲1090-0268共2007兲11:2共161兲
CE Database subject headings: Cracking; Bonding; Concrete beams; Concrete, reinforced; Finite element method; Fiber reinforced
polymers.
Strength Models
Teng et al. 共2003兲 developed a simple strength model for the IC
debonding failure mode of RC flexural members. This model is a
simple modification of the bond strength model of Chen and Teng
共2001兲, with the modification being based on IC debonding test
Fig. 1. Debonding failure modes in RC beams with an externally data available to them. In addition, existing design guidelines
bonded FRP plate: 共a兲 various debonding failure modes; 共b兲 共ACI 2002; fib 2001; JSCE 2001兲 also contain provisions for
FRP-stengthened RC beam after failure by IC debonding design against premature debonding failures, some of which may
be used to design against IC debonding. The ACI 440 共2002兲
approach is to directly limit the FRP strain to avoid debonding, so
which were more easily accessible and are referred to as Group A it is rather simple to apply. By contrast, JSCE 共2001兲 limits the
specimens in the paper, were used to verify the FE model. These maximum stress gradient in the FRP plate, but two key param-
42 beams were tested by Brena et al. 共2003兲; Chan and Niall eters including the fracture energy and the crack width are incom-
共2001兲; Fang 共2002兲; Garden and Hollaway 共1998兲; Grace 2002; pletely defined and the method requires an involved iterative pro-
Rahimi and Hutchinson 共2001兲; Shin and Lee 共2003兲; Teng et al. cess of analysis. fib 共2001兲 provides three alternative approaches
2003; Ye et al. 共2001兲, and Zhao et al. 共2002兲. to avoid debonding failures by limiting: 共1兲 the FRP strain, 共2兲 the
maximum stress gradient in the FRP plate, and 共3兲 the shear force
in the RC member. The first approach is similar to the strain limit
Finite-Element Analysis approach of ACI 440 共2002兲 and suggests a FRP strain limit in the
Most existing FE studies of IC debonding failures of FRP- range of 0.0065–0.0085, which is much higher than indicated by
strengthened RC beams have been conducted using a two- some of the existing tests 共Table 2兲; further, no suggestion is
dimensional model, in which the concrete is represented using given on the specific value to be used for a particular situation.
plane stress elements, whereas the FRP plate and the steel rebars The second approach is similar to the JSCE 共2001兲 approach but
are represented using bar elements. The shear bond stress-slip is significantly more complicated. The third approach is suitable
behavior of the FRP-to-concrete interface is represented using for application in design in terms of simplicity. These provisions
interface elements based on predefined bond-slip relationships. in design guidelines do not appear to either have a solid scientific
These relationships are generally derived from pull tests in which basis or have been appropriately verified by IC debonding test
the FRP-to-concrete interface is subject to shear by loading the data.
FRP plate in tension 共Chen and Teng 2001兲. Yao et al. 共2005兲 recently assessed the above-mentioned
Two approaches are available for the modeling of crack propa- strength models using their own test data and reached the follow-
gation in the FE analysis of concrete structures: The discrete ing conclusions: 共1兲 the IC debonding strength model of Teng
crack approach and the smeared crack approach. The discrete et al. 共2003兲 generally provides safe predictions of the experimen-
crack approach requires continuous remeshing in the solution pro- tal debonding strains; 共2兲 the models of ACI 440 共2002兲, fib
cess to trace the propagation of cracking, so for the realistic mod- 共2001兲, and JSCE 共2001兲, which are also models for design use,
eling of IC debonding in RC beams, the computational effort are not sufficiently safe for use in design; and 共3兲 the model of
required generally becomes intolerable as a large number of Teng et al. can become overly conservative and the scatter of its
cracks need to be traced. This approach has so far only been used predictions is large. In addition, the JSCE 共2001兲 suffers from the
to study IC debonding caused by the propagation of one or a few tediousness of an iterative process of analysis. Therefore, much
predefined cracks 共Wu and Niu 2000; Niu 2002; Monti et al. further research is needed to develop a simple and accurate IC
2003兲. Such simplified analyses however ignore the interaction debonding strength model that has a rational theoretical basis.
between crack propagation and structural response and thus can-
not provide realistic predictions of IC debonding failures. Interfacial Stresses and IC Debonding Process
The smeared crack approach has been more widely used in the
FE modeling of IC debonding failures 共e.g., Wong and Vecchio Existing research has shown that the interfacial shear stresses in a
2003; Wu and Yin 2003兲. In this approach, a crack is evenly FRP-strengthened RC beam can be divided into two types: Those
distributed as tensile straining over a representative zone of con- due to the shear force in the beam 关Fig. 2共a兲兴 and those due to the
crete, and the cracking behavior is simulated using an appropriate opening-up of flexural cracks 关Fig. 2共b兲兴, respectively. Fig. 2共a兲
constitutive law for concrete. This approach does not require re- shows a small segment of the beam in the flexural-shear zone,
meshing, but is unable to predict strain/slip concentrations near a where the interfacial shear stresses between the FRP plate and the
S8-50-0 400 800 50 330 100 50 1.2 0 0 15a 4⌽8 653 15a 53.36 165 2940 NA 35.82 15.22 NA Yes
S8-50-1 400 800 50 330 100 50 1.2 0 0 15a 4⌽8 653 15a 53.36 165 2940 NA 41.13 17.48 NA Yes
Gao et al. 共2001兲 BB1 250 350 150 150 200 75 0.22 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 15.2 80.70 20.18 NA No
BB2 250 350 150 150 200 75 0.22 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 13.1 78.70 19.68 NA No
BB3 250 350 150 150 200 75 0.22 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 15.9 87.90 21.98 NA No
BB4 250 350 150 150 200 75 0.44 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 11.51 86.40 21.60 NA No
BB5 250 350 150 150 200 75 0.44 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 12.13 93.20 23.30 NA No
BB6 250 350 150 150 200 75 0.44 2⌽8 531 27 2⌽10 531 38 35.70 235 4200 13 96.80 24.20 NA No
Xiong et al. 共2001兲 CF1 300 600 0 120 150 80 0.11 2⌽6 346 25a 2⌽12 569 25a 15.68 252 3652 14.93 51.39 15.42 NA No
GF1 300 600 0 120 150 80 0.08 2⌽6 346 25a 2⌽12 569 25a 15.68 17 350 17.6 47.58 14.27 NA No
a
CF3 200 400 0 80 120 50 0.11 2⌽6 346 25 2⌽12 346.7 25a 15.68 252 3652 11.93 30.45 6.09 NA No
Ye et al. 共2001兲 BM0 500 1380 0 400 200 300 0.111 6⌽6 372.5 25 6⌽14 372.5 25 40.24 235 3550 103.46 117.04 80.76 15000 Yes
Yi and Huang B 06 300 650 50 100 180 100 0.22 2⌽6 324 25a 3⌽8 324 25a 32.64 235 3550 NA 43.30 15.16 NA No
共2001兲 B 08 300 650 50 100 180 100 0.33 2⌽6 324 25a 3⌽8 324 25a 34.00 235 3550 NA 37.70 13.20 NA No
Fang 共2002兲 B1 200 470 80 150 200 150 0.11 2⌽6 301 15 3⌽8 288 15 24.96 235 3500 10.87 65.00 17.88 7834 Yes
B2 200 470 80 150 200 150 0.11 2⌽6 301 15 3⌽8 288 15 24.96 235 3500 11.20 67.07 18.44 7100 Yes
B3 200 470 80 150 200 150 0.11 2⌽6 301 15 3⌽8 288 15 24.96 235 3500 10.70 66.55 18.30 5868 Yes
Zhao et al. 共2002兲 LL-3 200 550 50 150 250 100 0.166 2⌽12 350.7 15 2⌽12 350.7 15 20.16 240 3550 8.56 96.88 29.06 NA Yes
LL-4 200 550 50 150 250 100 0.166 2⌽12 350.7 15 2⌽12 350.7 15 33.12 240 3550 7.5 91.75 27.53 NA Yes
LL-5 200 550 50 150 250 100 0.083 2⌽12 350.7 15 2⌽14 381.2 15 20.16 240 3550 8.9 117.35 35.21 NA Yes
a
Not provided in the original source; assumed to be 25 mm for beams and 15 mm for slabs; L1= distance from midspan to loading point; L2= distance from loading point to FRP plate end;
L3= distance from FRP Plate end to support; bc= beam width; f y= yield strength of tension rebars; Es= Young’s modulus of tension rebars; f c⬘= cylinder compressive strength of concrete; E0= initial Young’s
modulus of concrete; h= beam height; b f =FRP plate width; t f = FRP plate thickness; f ⬘y = yield strength of compression rebars; Es⬘= Young’s modulus of compression rebars; a⬘=cover to centers of
compression rebars; a=cover to centers of tension rebars; E f = Young’s modulus of FRP plate; f f = tensile strength of FRP plate; du= deflection at ultimate load; Pu= ultimate load; M u= ultimate bending
moment; and IC f = FRP strain at IC debonding.
Table 2. Comparison of Results from Tests, FEA, and Strength Model
Test FEA Design model
Specimen du Mu IC
f du Mu IC
f Mu IC
f
Data source name 共mm兲 共kN·m兲 共mm兲 共kN·m兲 共kN·m兲
Triantafillou and Plevris 共1992兲 B4 — 6.78 — 6.28 5,393
B5 — 7.01 — 6.28 5,393
B6 — 6.41 — 6.89 4,428
B7 — 5.86 — 6.89 4,425
B8 — 8.56 — 8.41 2,713
Garden and Hollaway 共1998兲 Can-05 — 9.06 — 10.50 7.61 7,796 6.74 7,957
Can-06 — 8.75 — 19.50 8.42 8,494 6.86 8,177
Can-07 — 8.95 — 19.50 8.42 8,494 6.86 8,177
Tumialan et al. 共1999兲 A1 — 77.53 — 92.54 10,871
A2 — 90.42 — 98.54 7,581
A7 — 91.70 — 91.55 10,171
C1 — 82.22 — 87.47 14,585
Beber et al. 共2001兲 VR5 27 40.01 — 33.23 5,208
VR6 33 39.15 — 33.23 5,208
VR7 — 48.62 — 37.27 3,839
VR8 — 48.55 — 37.27 3,839
VR9 19 50.74 — 40.29 3,146
VR10 18.9 53.64 — 40.29 3,146
Ceroni and Prota 共2001兲 A2 51.7 6.94 — 6.62 9,148
A3 39.3 7.20 — 6.55 8,963
Chan and Niall 共2001兲 S6-50-0 — 12.66 — 45.00 14.20 6,743 12.53 7,446
S6-50-1 — 13.11 — 45.00 14.20 6,743 12.53 7,446
S8-50-0 — 15.22 — 50.00 18.28 6,945 16.13 7,446
S8-50-1 — 17.48 — 50.00 18.28 6,945 16.13 7,446
Gao et al. 共2001兲 BB1 15.2 20.18 — 19.12 9,640
BB2 13.1 19.68 — 19.12 9,640
BB3 15.9 21.98 — 19.12 9,640
BB4 11.51 21.60 — 21.48 6,451
BB5 12.13 23.30 — 21.48 6,451
BB6 13 24.20 — 21.48 6,451
Rahimi and Hutchinson 共2001兲 A4 32 23.21 — 25.89 21.29 5,164 21.32 5,629
A5 31.3 23.70 7,200 25.89 21.29 5,164 21.32 5,629
A6 23.1 22.28 — 22.64 25.17 4,759 23.69 4,501
A7 27.9 26.48 — 22.64 25.17 4,759 23.69 4,501
A8 31.9 24.45 — 25.89 21.29 5,164 21.32 5,629
A9 33 23.96 — 25.89 21.29 5,164 21.32 5,629
B3 38.7 20.70 9,700 34.62 17.68 6,544 18.03 8,176
B4 38.1 19.69 — 34.62 17.68 6,544 18.03 8,176
B5 30.3 26.14 — 22.65 25.18 4,761 23.69 4,501
B6 28.3 26.10 5,500 22.65 25.18 4,761 23.69 4,501
B7 33.3 22.16 — 28.46 19.51 6,514 19.09 7,179
B8 34.2 23.10 — 28.46 19.51 6,514 19.09 7,179
Xiong et al. 共2001兲 CF1 14.93 15.42 — 13.86 7,948
GF1 17.6 14.27 — 12.55 37,318
CF3 11.93 6.09 — 6.29 7,931
Ye et al. 共2001兲 BM0 103.46 80.76 15,000 108.00 78.60 14,043 76.31 14,063
Yi and Huang 共2001兲 B 06 — 15.16 — 12.97 7,047
B 08 — 13.20 — 14.35 5,797
RC beam are caused by the shear force Q directly. These interfa- outside the pure bending zone. Laboratory tests on RC beams in
cial stresses are referred as s hereafter: they define the overall four-point bending showed that IC debonding was commonly ini-
distribution of interfacial shear stresses and should be zero in a tiated under one of the two concentrated loads 共e.g., Fang 2002;
pure bending zone of a beam. Fig. 2共b兲 shows the opening-up of Rahimi and Hutchinson 2001兲. This is because at this location,
a flexural crack in a RC beam and the constraint provided by a both the shear force and the bending moment are at their maxi-
FRP plate bridging the crack. In this case, the interfacial shear mum values, so both the shear force-induced interfacial shear
stresses developed between the FRP plate and the RC beam near stress s and the crack-induced interfacial shear stress c are
the flexural crack, which are hereafter referred to as c, define the maximized.
local fluctuations in the interfacial shear stresses in an FRP- Unless strong additional anchorage measures are provided in
strengthened RC beam and exist near each crack both within and the flexural-shear zone, IC debonding, once initiated, will propa-
gate quickly to a FRP plate end. During this process of debonding
propagation, there may be a small load increase, followed by an
eventual sudden drop. The initiation of debonding propagation
and the attainment of the peak load 共i.e., ultimate load兲 are two
important instances of an IC debonding process, and the loads
corresponding to these two instances are usually very close to
each other. In many laboratory tests, the debonding propagation
process is so quick that a distinction between the two instances
can hardly be made, but the ultimate load and the corresponding
deflection can generally be captured. In developing the simple
strength model presented later in this paper, the interfacial stress
Fig. 2. Two types of interfacial shear stresses: 共a兲 shear distribution at the initiation of debonding propagation was taken
force-induced; 共b兲 crack-induced as the basis. This has two important advantages: 共1兲 the interfacial
was employed to obtain the bond-slip relationship of the FRP-to- one order smaller than the thickness of the concrete layer that
concrete interface in a pull test by Lu et al. 共2005a兲, which led to debonded. The elements representing the FRP plate were directly
excellent results. Further details of the mesoscale FE approach connected to those representing the concrete, without an adhesive
can be found in Lu et al. 共2005b兲. layer in between. This treatment is believed to be appropriate as
for commonly used adhesives, the amount of interfacial slip con-
tributed by the adhesive layer is a small portion of the total slip
Mesoscale FE Analysis of Plain Concrete Beams
and debonding generally occurs in the concrete rather than in the
In the present study, the mesoscale FE method was used to find adhesive layer 共Lu et al. 2005b兲. The general purpose FE package
out whether the bond-slip behavior of the interface in a FRP- MSC.MARC 共2003兲 was used to establish the mesoscale model.
strengthened RC beam, with cracking in the tensile concrete, can The FE load–deflection curve is compared with those from the
be predicted by a bond-slip model derived from pull tests. Two two tests in Fig. 4. The mesoscale FE results are seen to be in
identical FRP-strengthened plain concrete beams under three- close agreement with the test results in general, given that the two
point bending tested by Wu and Yin 共2003兲 were chosen for nominally identical tests delivered significantly different results
analysis by the mesoscale approach. The beams had a cross sec- 共Fig. 4兲. The crack pattern, as indicated by the contours of crack
tion of 100 mm⫻ 150 mm and a clear span of 900 mm. It was strain 共Lu et al. 2005b兲, which is defined as the total tensile strain
symmetrically bonded with a 750 mm long carbon FRP 共CFRP兲 minus the elastic strain at each Gauss point, is shown in Fig. 5共a兲.
sheet having a nominal thickness of 0.11 mm and a Young’s This crack pattern is also in close agreement with that given in
modulus of 230 MPa based on this nominal thickness. The con- Wu and Yin 共2003兲. The accuracy of the mesoscale model is thus
crete of the beams had a cylinder compressive strength of verified.
24.6 MPa, an elastic modulus of 25 MPa and a Poisson’s ratio of
0.15 共Wu and Yin 2003兲. The tensile strength of concrete was
Bond-Slip Models
found to be 2.6 MPa using a relationship in GB-50010 共2002兲.
These beams were relatively small and had no steel reinforce- Using the results from the mesoscale FE model described previ-
ment, so their mesoscale model is computationally less demand- ously, the interfacial bond-slip curves in this FRP-strengthened
ing than one for a more realistic RC beam. plain concrete beam were examined at different locations along
In the mesoscale model of these two beams, the smallest the interface. These results showed that for the parts of the inter-
elements for the concrete had a side length of 0.5 mm, which is face outside the major flexural crack zone, the crack pattern of the
冑
冦
s
max if s ⱕ s0 共1a兲
= s0
maxe−␣共s/s0−1兲 if s ⬎ s0 共1b兲
where
s0 = 0.0195w f t 共1d兲
w = 冑 2.25 − b f /bc
1.25 + b f /bc
共1e兲
1
␣= 共1f兲
Gf 2
−
maxs0 3
G f = 0.3082w冑 f t 共1g兲
where b f ⫽width of the FRP plate 共mm兲; bc⫽width of the concrete
member 共mm兲; f t⫽tensile strength of concrete 共MPa兲;
G f ⫽interfacial fracture energy 共MPa mm兲; and and s⫽shear
bond stress 共MPa兲 and the interfacial slip 共mm兲, respectively. It
should be noted that three bond-slip models of different levels of Fig. 6. Bond-slip curves for the interface from mesoscale FE model
sophistication were proposed by Lu et al. 共2005a兲 and the one of the beams of Wu and Yin 共2003兲
given above is their “simplified model,” which is nevertheless as
accurate as their “precise model” for FRP plates bonded to RC
beams with commonly used adhesives. This bond-slip relation- FE Modeling of IC Debonding
ship is hereafter referred to as “Bond-Slip Model I.”
For the part of the interface inside the major flexural crack
zone, the crack pattern near the interface 关Fig. 5共c兲兴 is different General
from that observed in a pull test. Around a major flexural crack, Although the IC debonding failure of the plain concrete beam
some root-shaped cracks exist. This difference in crack pattern discussed in the preceding section was successfully simulated
implies different failure mechanisms. As it was difficult to obtain using the mesoscale FE approach, this approach is computation-
a stable bond-slip curve from the numerical results for a single ally too demanding for the simulation of a large number of large
point along the interface in the major flexural crack zone, the RC beams with a length of several meters. Using the bond-slip
average bond-slip response over a certain length obtained was models defined earlier, a conventional FE approach can be used
examined. Two averaging lengths, 15 and 20 mm, respectively, for large RC beams, in which elements of normal sizes are used.
were used, as they correspond to interface element widths of 30 Normal element sizes refer to those commonly used in conven-
and 40 mm that are most frequently used in the conventional FE tional FE analyses, which are several times the maximum aggre-
simulations 共i.e., FE analyses using elements of normal sizes兲 gate size.
presented later in the paper. The resulting average bond-slip re- The general purpose FE package MSC.MARC 共2003兲 was also
sponses for these two averaging lengths are shown in Fig. 6共b兲. It used for the conventional FE simulations. The concrete, the rebars
can be seen that both curves are composed of a curved ascending and the FRP plate were modeled with the plane stress quadrilat-
branch and a highly brittle descending branch. These bond-slip eral element 共Element 3 in MSC.MARC兲, the four-node plane
responses can be closely described by the bond-slip model of Lu rebar element 共Element 143 in MSC.MARC兲 and the truss ele-
et al. 共2005a兲 with a brittle postpeak branch 关Fig. 6共b兲兴. That is, ment 共Element 9 in MSC.MARC兲, respectively.
冑
冦
s
max if s ⱕ s0 共2a兲 Modeling of Concrete
= s0
Concrete in compression was treated as an elastic–plastic material
0 if s ⬎ s0 共2b兲
using the yield function proposed by Buyukozturk 共1977兲 with an
Here, max and s0 are still given by Eqs. 共1c兲 and 共1d兲. Eq. 共2兲 is associated flow rule, which was already included in MSC.MARC
hereafter referred to as “Bond-Slip Model II.” 共2003兲. It was also assumed that concrete is a linearly elastic
冋 冉 冊 冉 冊册
冦 冣
2
0 2 − if ⱕ 0 共3a兲
0 0
=
冋 冉 冊册
0 1 − 0.15
− 0
u − 0
if 0 ⬍ ⱕ u 共3b兲
冦
The FRP was modeled as a linearly elastic material with brittle
E 0 if ⱕ cr = f t/E0 共4a兲
failure when its tensile strength is reached. As the FRP plate was
− cr very thin with a very small bending stiffness, it was modeled
= ft u if cr ⬍ ⱕ cr u
共4b兲
cr − cr using truss elements.
0 if ⬎ cr
u
共4c兲
where f t⫽tensile strength of concrete. If the tensile strength of User-Defined Interface Element
concrete f t was not reported for a test beam, it was obtained from Through the user subroutine facility of MSC.MARC 共2003兲, a
cu 共MPa兲 according to the Chinese code 共GB-50010
f t = 0.395f 0.55 user-defined interface element was defined for the FE model to
2002兲. The ultimate tensile strain when the tensile stress reduces represent the bond-slip behavior of the FRP-to-concrete interface
to zero is denoted by cr u
and was calculated from the fracture in an FRP-strengthened RC beam. The details of this interface
I
energy of concrete G f . According to the model code of CEB-FIP element are shown in Fig. 7. This interface element has four
共1993兲, for normal concrete, the fracture energy GIf of concrete is nodes but no thickness, with the two upper nodes connected to the
冉 冊 0.7
concrete and the two lower nodes connected to the FRP plate. At
fc each end of the interface element, the interaction between the two
GIf = ␣ 共4d兲
10 nodes is represented by two springs: the shear spring with a se-
cant stiffness Ku and the normal 共tension-compression兲 spring
where ␣ = 0.03 for concrete with normal aggregate. The assump-
with a secant stiffness Kv.
tion of a uniform distribution of the crack strain over the crack
The secant stiffness of the normal spring Kv is defined as
bandwidth bcr leads to
follows:
2GIf 1. If a compressive force develops in the normal spring, Kv is
cr
u
= 共4e兲 infinite and
f tbcr
2. If a tensile force develops in the normal spring, Kv is infinite
The shear behavior of cracked concrete was treated in two differ- if no debonding has occurred but is equal to zero if debond-
ent ways. For a RC beam with steel stirrups, a constant shear ing has occurred.
retention factor  = 0.4 was used. In a RC beam without steel For a given interfacial shear slip, the force induced in the shear
stirrups, the shear cracks may be relatively wide. As the interlock- spring is
ing effect of cracked concrete decreases with the crack width, a
constant shear retention factor seems no longer suitable. There- Lb f
F= 共6兲
fore, the variable shear retention model of Al-Mahaidi 共Walraven 2
and Reinhardt 1981兲 was chosen to represent the shear behavior
of cracked concrete, in which where L⫽length of the interface element; b f ⫽width of the FRP
plate; is found from Eqs. 共1a兲–共1g兲; and s⫽difference in hori-
0.4 zontal displacement between the upper and lower nodes. The se-
= 共5兲 cant stiffness Ku can be derived easily if the bond-slip curve is
cr
nn/共200 ⫻ 10−6兲
known.
where crnn is the direct crack strain normal to the crack, defined as Based on the dual local debonding criterion and the bond-slip
the total direct strain normal to the crack minus the elastic strain. models described previously, two situations arise in a FE analysis.
w = cr
nnL 共7兲
If w is larger than 2s0, then debonding due to crack opening-up is
deemed to occur; the shear stress in the concerned interface ele-
ment drops to zero following Eq. 共2b兲. If w is smaller than 2s0,
the crack-induced debonding does not occur and the interface
element is treated in the same way as an interface element con-
nected to a concrete element that does not contain a flexural
crack.
FE Results
General
Altogether, the 42 Group A specimens were analyzed using the
FE model described in the preceding section and the results are
compared with the test results in this section. In particular, de-
tailed comparisons are given for Beam B6 tested by Rahimi and
Fig. 14. IC debonding strains: FE analysis versus strength model Fig. 16. Normalized debonding strain versus bond length
冑
model predicts test results closely. The predictions of the debond-
= 0.228冑E f t f
4E f t f ing moment are slightly more conservative than those for the FRP
Lee = 共8兲
max/s0 debonding strain. This phenomenon is believed to be due to the
use of the plane section assumption in evaluating the moment
Based on these idealized stress blocks, the total axial force in the
from the plate strain, which ignores the slip between the FRP
FRP plate at the loaded section is given by
plate and the concrete when IC debonding occurs. As a result,
T= 冉 c,maxLee s,maxLd
2
+
2
bf 冊 共9兲
the curvature and thus the moment at the critical section are
underestimated.
A comparison of FRP strains from 21 tests and the predictions
where Ld should strictly be the distance 共mm兲 from the loaded of the FE model and Eq. 共14兲 is shown in Fig. 16. For a more
section to the end of the cracked region or the plate end if the insightful comparison, the FRP strain is normalized with respect
plate is terminated within the cracked region, because in the un- to max / 冑E f t f based on Eq. 共14兲, whereas the bond length Ld is
cracked region of the beam, the interfacial shear stress is very normalized with respect to 冑E f t f . It can be seen that the predic-
small. However, to avoid the extra work involved in determining tions of Eq. 共14兲 match the overall trend of the test data and the
the cracked region of the beam, Ld is simply taken to be the FE results well and pass through the middle of the scatter of the
distance 共mm兲 from the loaded section to the end of the FRP plate test data and the smaller scatter of the FE results.
as this simplification leads to little error in the IC debonding load
predicted using Eq. 共9兲.
As the maximum value of the crack-induced interfacial shear Conclusions
stress c,max and the maximum value of the shear force-induced
interfacial shear stress s,max are both located at the loaded sec- This paper has presented a new smeared crack approach for the
tion, IC debonding is deemed to occur when finite-element prediction of IC debonding failures in FRP-
c,max + s,max = max 共10兲 strengthened RC beams. This new approach overcomes the major
deficiency of all existing finite-element models in which the effect
Hence, if a relationship between the two maximum interfacial of slip concentration near a flexural crack in the concrete on the
shear stresses can be found, the total axial force in the FRP plate behavior of the FRP-to-concrete interface is not properly cap-
can be easily calculated at the critical section for IC debonding. tured. In this proposed method, a dual local debonding criterion is
A factor ␣ is introduced here to represent the ratio between the employed to monitor local debonding along the interface, so that
maximum shear force-induced interfacial stress s,max and the local debonding in the major flexural crack region can be accu-
total interfacial stress at the critical section max. That is rately described. In addition, two different bond-slip models are
s,max = ␣max 共11a兲 used for the major flexural crack zone and the rest of the beam
respectively. Comparisons with test results showed that the pro-
posed finite-element model can accurately predict IC debonding
c,max = 共1 − ␣兲max 共11b兲
failures. Based on interfacial stress distributions predicted by the
The strain in the FRP plate at the critical section IC
f when IC finite-element model, the paper has also presented a simple
debonding occurs can then be obtained from strength model for IC debonding failures. The new strength model
is shown to be accurate through comparisons with test results.
T 共1 − ␣兲maxLee + ␣maxLd
IC
f = = 共12兲
Eftfbf 2E f t f
In Eq. 共12兲, the only unknown is ␣. By comparing the prediction Acknowledgments
of Eq. 共12兲 with the corresponding FE result, the value of ␣ was
obtained for each of the 42 beams in Specimen Group A. The The writers are grateful for the financial support received from
values of ␣ deduced from FE results vary mainly with the Lee / Ld The Hong Kong Polytechnic University through its Area of
ratio, as shown in Fig. 13. A regression of these ␣ values led to Strategic Development 共ASD兲 Scheme for the ASD in Urban
the following equation: Hazard Mitigation and the Natural Science Foundation of China
共National Key Project No. 50238030兲.
␣ = 3.41Lee/Ld 共13兲
Substitution of Eqs. 共8兲 and 共13兲 into Eq. 共12兲 yields the following
equation for predicting the IC debonding strain of the FRP plate References
at the critical section:
American Concrete Institute 共ACI兲. 共2002兲. “Guide for the design and
IC 冑
f = 0.114共4.41 − ␣兲max/ E f t f 共14兲 construction of externally bonded FRP systems for strengthening con-
crete structures.” ACI 440.2R-02, ACI Committee 440, Detroit.
The predictions of Eq. 共14兲 are compared with the FE debonding Aprile, A., Spacone, E., and Limkatanyu, S. 共2001兲. “Role of bond in RC
strains for Specimen Group A in Fig. 14 and a close agreement beams strengthened with steel and FRP plates.” J. Struct. Eng.,
can been seen. The FRP strain from Eq. 共14兲 can be used in a 127共12兲, 1445–1452.