Professional Documents
Culture Documents
Experiments and Analytical Comparison of RC Beams Strengthened With CFRP Composites
Experiments and Analytical Comparison of RC Beams Strengthened With CFRP Composites
Experiments and Analytical Comparison of RC Beams Strengthened With CFRP Composites
www.elsevier.com/locate/compositesb
Abstract
This paper deals with strengthening, upgrading, and rehabilitation of existing reinforced concrete structures using externally bonded
composite materials. Five strengthened, retrofitted, or rehabilitated reinforced concrete beams are experimentally and analytically
investigated. Emphasis in placed on the stress concentration that arises near the edge of the fiber reinforced plastic strip, the failure modes
triggered by these edge effects, and the means for the prevention of such modes of failure. Three beams are tested with various edge
configurations that include wrapping the edge region with vertical composite straps and special forms of the adhesive layer at its edge. The
last two beams are preloaded up to failure before strengthening and the ability to rehabilitate members that endured progressive or even total
damage is examined. The results reveal a significant improvement in the serviceability and strength of the tested beams and demonstrate that
the method is suitable for the rehabilitation of severely damaged structural members. They also reveal the efficiency of the various edge
designs and their ability to control the characteristic brittle failure modes. The analytical results are obtained through the Closed-Form High-
Order model and are in good agreement with the experiment ones. The analytical and experimental results are also used for a preliminary
quantitative evaluation of a fracture mechanics based failure criterion for the strengthened beam.
q 2003 Elsevier Ltd. All rights reserved.
Keywords: A. Layered structure; B. Delamination; D. Mechanical testing; C. Analytical modeling; Reinforced concrete
involved, and it combines concepts of structural analysis CFRP strips. All the beams were loaded in a four-point
and fracture mechanics. As a result, the model is capable of bending loading scheme and were supported on a clear span
providing a quantitative description of the shear and vertical of 2.1 m. In attempt to examine the effect of the original
normal stresses that arise near the edge of the bonded strip design of the member and the influence of the diagonal
and the stress concentrations that occur near flexural and shear cracking on the behavior of the strengthened member,
shear cracks in the RC member itself. These stress two types of specimens were tested. The two type of beams
concentrations lead, in many cases, to the formation and are denoted by A and B and appear, along with their
unstable growth of an interfacial crack, to separation of the dimensions and reinforcement details, in Fig. 2. A short
bonded FRP strip, and to failure of the strengthened description of each specimen appears in Table 1. Note that
member. the type A beams were designed with a larger shear
In this paper, the behavior of five RC beams strengthened reinforcement ratio, whereas the type B beams have a larger
with externally bonded FRP strips is investigated exper- longitudinal reinforcement ratio and a reduced shear
imentally and analytically. The investigation focuses on the reinforcement ratio. Also note that beams A1 and B1 were
stress concentrations that arise near the edge of the FRP preloaded up to failure before strengthening, whereas beams
strip, the brittle and sudden failure modes triggered by these A2, A3, and B2 were strengthened and tested without
edge effects, and the means for the reduction of the edge preloading.
stresses and prevention of the delamination failure modes.
For this purpose, the various edge-anchoring designs that 2.2. Material properties
appear in Fig. 1 and include wrapping the edge of the
strengthened member with additional FRP sheet and the The concrete compressive strength tests were conducted
existence of a surplus adhesive body that forms a ‘spew- using 100 mm3 at the ages of 7 and 28 days and at the time
fillet’ at the edge of the adhesive layer are investigated. In of the experiment (87 days). These tests yield average
addition, the feasibility of rehabilitating severely damaged values (over three cubes sample in each test) of 47.8(35),
RC members and the effect of the original design (shear or 70.1(50), and 75.3(55) MPa, respectively. (The numbers in
flexural design) of the RC member on its failure mode are brackets refer to the minimum characteristic cylinder
studied. Finally, analytical results obtained through the strength based on BS EN 206-1:2001 [25]). Standard
Closed-Form High-Order (CFHO) model are compared with deformed reinforcement steel bars with a characteristic
the experimental ones, provide additional confirmation of strength of 390 MPa and an elastic modulus of 200 GPa
the analytical modeling approach, and explore further were used for the longitudinal and shear reinforcement
aspect of the structural response. The analytical and (stirrups).
experimental results are also used for the preliminary Sika CarboDur S1212 unidirectional carbon/epoxy strips
quantitative evaluation of a fracture mechanics based failure were used for the strengthening and rehabilitation of the RC
criterion for the strengthened beam. beams. SikaWrap Hex-230c CFRP sheets were used for
wrapping the edge region of beams A3 and B2. During the
fabrication process, these sheets were saturated with
2. Experiments Sikadur-330 epoxy resin. The mechanical properties of the
CFRP composite strips, the CFRP fabrics, and the epoxy
2.1. Test specimens resin are based on the manufacturer’s datasheets and appear
in Tables 2 and 3.
Section 2 includes the examination of five 2.5-m-long The flexural elastic modulus, Ea ; and the flexural tensile
rectangular RC beams strengthened with externally bonded strength, fat ; of the epoxy adhesive were determined through
Fig. 1. Edge configurations and edge anchorage devices: (a) ‘square-edge’ configuration; (b) ‘spew-fillet’ configuration; (c) wrapping with additional
FRP sheet.
O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677 665
Fig. 2. Dimensions and reinforcement details of the tested beams: (a) type A beams; (b) type B beams.
four-point-bending tests of adhesive beams. The compres- voltage displacement transducers (LVDT) and strain
sive strength of the adhesive was evaluated through gauges, see Fig. 3. These devices are divided into five
compression tests of 1 in3 cubes compression tests. The groups as follows:
results for the adhesive beam tests are summarized in
Table 4. These results highlight the unique properties of the 1. Measurement of the vertical deflections. Three LVDTs,
adhesive that include very high tensile strength (about 10 denoted by LVDT 1, 2, and 3, were located at midspan
times higher than the tensile strength of concrete) and the and at the locations of the concentrated loads and
very low modulus of elasticity (less than 20% of the elastic monitored the absolute vertical deflections of the tested
modulus of concrete). It was also observed that the adhesive beam (Fig. 3).
specimens exhibit a linear and elastic behavior up to failure. 2. Measurement of the ‘average strain’ in the RC beam.
The results of the adhesive cube tests reveal a mean These quantities were monitored using two pairs of
compressive strength of 76.3 MPa with a coefficient of LVDTs. The first pair, denoted by 4F and 4B, was
variance of 1.22%. located at the upper face of the beam on its front and back
sides, respectively. The second pair, denoted by 5F and
2.3. Test setup 5B, was located at the level of the internal longitudinal
reinforcement on the front and back sides of the beam,
The test setup, the various monitoring devices, and their respectively. The ‘averaged strain’ is defined as the
location along the beam appear in Fig. 3. The beams were relative displacement measured by the LVDT divided by
tested using a displacement control testing machine with the distance between the measuring points.
a 500 kN hydraulic actuator. The loading rate was 3. Measurement of the strains of the internal steel rebars.
0.011 mm/s and the data was assembled using a 0.2 Hz This is achieved using two strain gauges (denoted by SG
sampling rate. The monitoring devices include linear 6 and SG 7) bonded to the internal steel rebars.
666 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677
Table 1 Table 3
Rationales and description of the various test specimens Mechanical properties of the CFRP sheets and resin used for anchoring
determine the flexural elastic modulus and the tensile wrapped with two layers of 100-mm-wide saturated CFRP
strength of the concrete. The results of beam A1 yield values sheets, see Fig. 5. The adhesive and the epoxy resin were
of 32.8 GPa and 5.6 MPa, respectively, and the those of cured for at least 7 days before loading.
beam A2 yield values of 31.4 GPa and 5.5 MPa, respect-
ively. The yielding strain of the steel reinforcement was also 2.6. Strengthened beams: results and discussion
detected and revealed values of about 2.1 £ 103 microstrain
(m1). It is important to emphasize that both beams have The results include the load versus midspan deflection
exceeded their ultimate flexural capacity, and they are curves; the measured and ‘averaged’ strains; and a descrip-
considered as totally damaged structural members. The tion of the failure mode. The beams that are not preloaded
ability to rehabilitate these beams is examined next. (A2, A3, B2) are presented first, and the rehabilitated beams,
denoted by A1p, B1p, are discussed later.
2.5. Strengthening procedure
2.6.1. Beam A2
All beams were strengthened using externally bonded This beam was tested in its basic configuration, i.e. no
CFRP strips. The strengthening procedure included surface edge devices were used, and the ‘spew-fillets’ that are
preparation, application of a priming adhesive layer, and formed spontaneously in the bonding process were
bonding of the FRP strip. The surface preparation included removed. The load – deflection curves for the strengthened
intensive cleaning, removal of unstable particles, and beam A2 and for the control beam A1 appear in Fig. 6a. The
mechanical roughening down to aggregate surface. Follow- curves reveal that the strengthening process has signifi-
ing this stage, a 1 –2 mm thick priming layer of the same cantly increased the load resistance capacity of the beam.
adhesive used for bonding the CFRP strips has been applied The failure load of the strengthened beam was 156.6 kN,
on the roughened surface and cured in the open air for 24 h. which is more than twice of the failure load of the control
Next, the FRP strips were covered with a 2– 3 mm thick beam. In addition, smaller crack widths and much smaller
adhesive layer, attached in position on the tensile face of vertical deflections were observed through the entire
each beam, and fastened with a roller. Finally, a ‘spew- loading process, see Fig. 6a. The response of the
fillet’, see Fig. 1, was formed at one edge of beam A3. The strengthened member is almost linear and elastic up to
other edge of beam A3 and both edges of beam B2 were failure and it lacks the ductile behavior that is typical for
Fig. 4. Load versus midspan deflection for the control beams: (a) beam A1; (b) beam A2.
668 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677
Fig. 6. Results of beam A2: (a) load versus midspan deflection; (b) average compressive strains in concrete; (c) tensile strains in concrete and steel rebar;
(d) strains in FRP strip at midspan; (e) strains in FRP strip at the edge; (f) debonded edge after failure.
O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677 669
toward the midspan. A photograph showing the edge A3 is similar to that of beam A2 and it lacks the inelastic
region after failure appears in Fig. 6f and reveals a local range that is typical to RC members.
damaged region within the FRP strip itself. The damage The measured and ‘averaged’ strains in the various parts
takes the form of splitting in the transverse and through the of beam A3 appear in Fig. 7b – d. These figures indicate that
thickness directions. Such damage occurs due to the edge yielding of the internal steel reinforcement occurred at a
stress concentrations that are associated with tractions strain of about 2.1 £ 103 m1 and a load level of 140 kN. It
normal to the fiber orientation of the unidirectional appears that the strain gauges SG 6 and SG 7, which were
laminate. Hence, it is recommended that additional attached to the internal steel reinforcement bars, were
reinforcement in the transverse direction and, especially, debonded or damaged in the post yielding stage due to
through the thickness (stitched laminates) would be concrete cracks intersecting them. However, the results of
considered. Application of such additional reinforcement LVDT 5F and 5B, which reflect the ‘averaged’ strains at the
can be limited to the edge regions only, thus is does not level of the internal reinforcement, provided consistent
affect the capacity of the strip in the regions subjected to readings throughout the entire loading process. Based on the
higher unidirectional tensile stresses. In addition, it can be readings of LVDT 4F and 4B and the external strain gauge
combined with other edge anchorage devices such as the SG 8, it appears that the concrete did not exceed its crushing
vertical straps used in beam B2. strain (3.5 £ 103 m1) and the FRP strip did not reach its
rupture strain (17 £ 103 m1). It means that the upgraded
2.6.2. Beam A3 beam did not exceed its potential flexural capacity due to the
In this case, different edge configurations are examined. premature failure of the edge region.
These configurations include the formation of a ‘spew-fillet’ The failure of beam A3 was also triggered by the shear
at one end of the adhesive layer and wrapping the other end and vertical normal stress concentration near the edge of the
with additional CFRP sheets, see Figs. 1 and 5. The load bonded strip, yet, a different mode of failure was observed
versus midspan deflection curve of beam A3 appears in here. In this case, the sudden failure occurred due to
Fig. 7a and it is compared with the curve of the control beam detachment of the FRP strip along with the concrete layer
A1. These curves show that the ultimate load here is beneath the internal reinforcement level. In other words, the
178 kN, which is 2.36 times larger than the ultimate load of vertical strap located at one edge and the ‘spew-fillet’ at the
the control beam. It is also seen that the behavior of beam other edge prevented the interfacial delamination failure
Fig. 7. Results of beam A3: (a) load versus midspan deflection; (b) average compressive strains in concrete; (c) tensile strains in concrete and steel rebar; (d)
strains in FRP strip at midspan.
670 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677
2.6.3. Beam B2
In this beam, both edges were wrapped with a 100-mm-
wide carbon sheet. In addition, the beam was cast with a
reduced shear reinforcement ratio and a larger longitudinal
reinforcement ratio. The load versus midspan deflection
curve for this beam is presented and compared with the
response of the control beam in Fig. 9a. The ultimate load
capacity of the strengthened beam was 187.4 kN, which is
about 72% higher than the ultimate load of the control beam.
The cracking load was 35 kN and the yielding load was
Fig. 8. Beam A3 after failure.
about 155 kN. These values are 59 and 55% higher than the
cracking and yielding loads of the control beam. This effect
observed in beam A2. However, while the vertical strap is attributed to the compressive force induced in the RC
strengthened the entire section in the vertical direction and member as a result of the composite action with the FRP
prevented failure due to the peeling stresses, the spew-fillet strip. The compressive stresses neutralize the tensile stresses
could only contribute to the stress transfer mechanism at the in the concrete, reduce the tensile force in the steel rebars,
adhesive-concrete interface. As a result, the vertical normal and postpone their yielding.
peeling stresses are transferred through the adhesive- The developments of the measured and ‘averaged’
concrete interface into the concrete member itself and the strains appear in Fig. 9b –d. These curves indicate that
failure is shifted to the critical section at the level of the the potential flexural capacity of the strengthened beam
internal reinforcement. In other words, the ‘spew-fillet’ has not been utilized, as both the concrete and the FRP
provides a solution of the stress transfer mechanism in the did not reach their ultimate strains. Also, Fig. 9c and d
localized level, while the use of the vertical strap provides point out that the strains in the FRP strip are somewhat
Fig. 9. Results of beam B2: (a) load versus midspan deflection; (b) average compressive strains in concrete; (c) tensile strains in concrete and steel rebar; (d)
strains in FRP strip at midspan.
O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677 671
Fig. 11. Results of beam A1p: (a) load versus midspan deflection (loading and reloading); (b) average compressive strains in concrete; (c) tensile strains in
concrete and steel rebar; (d) strains in FRP strip at midspan.
672 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677
the process has rehabilitated the damaged beam and even cases, the delamination failure of the retrofitted beam is
improved its original load carrying capacity. The ultimate sudden, brittle, and without any warning. Thus, it empha-
load of the rehabilitated beam was 138 kN, which is 83% sizes the necessity of efficient edge anchorage devices to
higher than its failure load in the initial loading stage. It is avoid the edge delamination failure. Along with the
also seen that the process has increased the stiffness of the delamination failure, a local crushing of the concrete
damaged beam. On the other hand, the load versus occurred within the constant moment region. This obser-
deflection curve clearly shows that the retrofitted beam vation is also reflected by the average compressive strains at
lacks the ductile behavior that characterizes the control the upper face of the beam that exceeded the ultimate value
beam. Hence, in spite of the significantly improved of 3.5 £ 103 m1 (Fig. 11b). It means that even if edge
capacity, the lack of ductility and the significantly reduced anchorage devices had been used and the edge delamination
deflections prior to failure requires that such retrofitted failure had been prevented, a classical ductile failure
structural members would be designed and used with extra mechanism could not develop. However, such failure
caution and appropriate factors of safety. Additional results mechanism due to compressive crushing of the concrete
in terms of the tensile and compressive strains within the should be preferred over the brittle and hazardous
rehabilitated beam appear in Fig. 11b –d. These curves delamination failure modes.
confirm that the reloading process is not associated with
renewed cracking and yielding points. Hence, the initial 2.6.4. Beam B1p (reloading of beam B1 after rehabilitation)
tensile strength of the concrete and the pre-yielding The results of the initial loading and the reloading stages
contribution of the steel reinforcement should be disre- are presented in terms of the load versus midspan deflection
garded in the case of rehabilitating members that endured curves in Fig. 12a. The ultimate load resistance of the
such progressive damage. retrofitted beam was 151.8 kN, which is about 40% higher
The failure mode observed in this case is similar to that than its originating failure load. Hence, the rehabilitation
of beam A2, which has also been tested without edge process regained and even increased the original load
anchorage devices. This mode consists of the formation of carrying capacity of the damaged beam. Note that the
an interfacial crack at the upper face of the adhesive layer percentage increase in the load is significantly smaller
and its unstable growth toward midspan. As in the previous than that for beam A1p, whereas the absolute values of
Fig. 12. Results of beam B1p: (a) load versus midspan deflection (loading and reloading); (b) average compressive strains in concrete; (c) tensile strains in
concrete and steel rebar; (d) strains in FRP strip at midspan.
O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677 673
the ultimate loads of the two strengthened beams are of the fracture energy of the strengthened member. The exper-
same order (138 kN for A1p and 151.8 kN for B1p). This imental data and the analytical predictions are compared
effect is explained by the larger longitudinal reinforcement using the results of beams A2, A3, and B2 and the reloaded
ratio and the increased ultimate load of beam B1 in the beams A1p and B1p. Beam A2, A3, and B2 are used for the
preloading stage (109 kN for beam B1 and 75.5 kN for comparison of the overall load – deflection behavior and the
beam A1). The increased reinforcement ratio has only minor strains in the FRP strip. Beams A2, A1p and A2p, which
influence on the response of the strengthened beams and on have been tested with a ‘square-edge’ configuration at the
its modes of failure, hence the ultimate load of the retrofitted edge of the adhesive layer, are used for studying
beams are similar. The tensile and compressive strains in the the development of the edge stresses and the criteria for
rehabilitated beam appear in Fig. 12b –d and exhibit an the edge failure.
almost linear behavior up to the failure of the beam. Again, The analytical results are calculated using the nonlinear
it is attributed to the absence of renewed cracking and CFHO model [22]. This model is somewhat simplified here,
yielding points. and the predicted behavior is limited to three load levels,
The failure of beam B2p was associated with the unstable namely: the cracking load; the yielding load of the steel
growth of a horizontal crack at the interface between the reinforcement; and to the ultimate load, that are observed
priming and the adhesive layers. The crack initiated at one experimentally. In the case of the reloaded beams, the
end of the bonded FRP strip and propagated horizontally in tensile strength of the concrete and the pre-yielding stage of
an unstable manner, leading eventually to separation of the the internal reinforcement are disregarded. The analytical
strip and failure of the beam. The complete separation of and experimental load versus midspan deflection curves and
the FRP strip allows the examination and assessment of the load versus strain in the FRP strip at midspan curves appear
quality of the bonding in terms of existence of disbonded in Fig. 13. Fig. 13a, c and e demonstrate that the differences
regions. The examination of the inner side of the between the measured and the calculated load versus
delaminated strip reveals that the bonding and fastening deflection curves are small and are within the range of the
methods used here could not prevent the formation of air accepted engineering accuracy. It is also seen that the
bubbles and disbonded regions. Hence, the development of polygonal approximation using four significant points
alternative techniques that would ensure proper application provides a satisfactory description of the behavior of the
of the bonded strip, as well as the elaboration of reliable and structure. The comparison of the load versus the strain in the
yet practical non-destructive tests for the assessment of the FRP strip curves reveals that the analytical and experimen-
quality of the bonding, are called for. tal curves of beam A3 are in good agreement (Fig. 13d).
However, discrepancies between analytical prediction and
the strains experimentally measured by the strain gauges are
3. Analytical comparison observed in beams A2 and B2 (Fig. 13b and f). Furthermore,
in beam A2 the model predicts smaller strains than these
In this section, analytical results obtained using the measure by the strain gauge while in beam B2 the predicted
CFHO approach are compared with the experimental ones. strains are higher than the ones experimentally measured. In
The analytical models appear in Refs. [21 – 24] and have order to clarify this phenomenon, the analytical curves are
been verified using both FE analysis [21], and experimental also compared to the averaged strains measured by LVDT
results that are available in the literature [22]. The 5F and 5B, which are expected to be of similar magnitude to
comparison of the experimental and analytical results has the averaged strains in the FRP strip. Such comparison
two primary objectives. The first one is to provide further reveals fair agreement between the analytical and the
confirmation of the analytical model. The second objective experimental quantities in all examined beams. This effect is
is to reveal and to examine additional data such as the shear explained by the sensitivity of the strain gauges to localized
and vertical normal stress concentrations near the edge of effects such as localized cracking of the RC member,
the bended FRP strip. These localized stress distributions roughness of the bonded surface, flows in the bonding
are essential for understanding the behavior of the tested process, formation of delaminated areas, or localized
beams, yet they cannot be detected using conventional disturbance caused by the device itself. While the strain
monitoring techniques. In addition, the experimental results, gauges are strongly influenced by such effects, the averaged
along with the analytical CFHO and fracture mechanics strains determined by normalizing the relative displace-
approach of Rabinovitch and Frostig [23], are used for the ments and the analytical model do not monitor the response
preliminary quantitative evaluation of a failure criterion for of the structure in such resolution (unless intended to).
the strengthened system. This approach is based on the Hence, the combination of the data detected by the strain
fracture mechanics concept of the Energy Release Rate gauges and evaluated using the readings of the LVDTs
(ERR) compared with the specific fracture energy of the provides a reliable picture of the strains in the examined
system and replaces the traditional stress based failure structure and establishes a reliable basis for the conclusions
criterion. Hence, the results of the experimental study are drawn upon the data. This observation also reflects on the
used here for a preliminary assessment of the specific types of non-destructive techniques that should be adopted
674 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677
Fig. 13. Analytical and experimental results for beam A3: (a) load versus midspan deflection; (b) load versus strain at the FRP strip at midspan.
for health monitoring of the strengthened structure and concentrations near the edge of the bonded strip, and their
suggests that combining techniques may be advantageous. development with the imposed load [21,22]. The results for
The results presented in Fig. 13 reveal that the analytical the three beams that were tested without edge anchorage
model can describe the overall behavior of the strengthened devices (beams A2, A1p, and A2p) are summarized in
beam throughout the entire loading process. Furthermore, it Table 5. The development of the edge stresses in beam A2
can also provide a description of the shear and peeling stress appear in Fig. 14 and the distributions of the shear and
Table 5
Experimental and analytical results for beams failed in edge delamination
Beam Pult (kN) wmax (mm) szz at the interfaces (MPa) t (MPa) Jult ¼ Gc (N/m)
Fig. 15. Edge effects near the cut-off section of beam A2 under its failure load: (a) shear stresses in the adhesive layer (b) vertical normal stresses in the
adhesive-concrete and adhesive-FRP interfaces (c) longitudinal strains in the FRP strip.
676 O. Rabinovitch, Y. Frostig / Composites: Part B 34 (2003) 663–677