Bond of Frp Laminates to Concrete

Fiber reinforced polymer (FRP) laminates are being successfully used for strengthening of existing reinforced concrete structures. Bond of the FRP reinforcement to the concrete substrate is of critical importance for the effectiveness of the technique. In this project, flexural test specimens were prepared to address some of the factors expected to affect bond, namely, bonded length, concrete strength, number of plies (stiffness), sheet width and, to a limited extent, surface preparation. Results are presented and discussed in this paper. A shear lag approach, along with a simple shear model for the evaluation of the slip modulus, is used to model the strain distribution at moderate load levels. Finally, expressions of the peeling load and the effective bond length are presented. A design equation is proposed to calculate the effective FRP ultimate strain to be used in design to account for bond-controlled failure. INTRODUCTION Fiber reinforced polymer (FRP) laminates are being successfully used for strengthening of existing reinforced concrete (RC) and prestressed concrete (PC) structures. Bond of the external FRP reinforcement to the concrete substrate is of critical importance for the effectiveness of the technique, since it is the means for the transfer of stresses between concrete and FRP in order to develop composite action. Some research efforts carried out so far on this topic are outlined in the following. Chajes et al. (1996) studied the bond and force transfer mechanism in composite material De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 2 plates bonded to concrete, using a single -lap shear test specimen. A first set of tests was performed to investigate the effect of surface preparation, type of adhesive and concrete strength on the average bond strength. A second set of tests was used to study the force transfer from the composite plates to the concrete. Two failure mechanisms were observed: direct concrete shearing beneath the concrete surface and cohesive-type failure, depending on the type of adhesive. Test results showed that (1) surface preparation of the concrete can influence the bond strength, (2) if the failure mode of the joint is governed by shearing of the concrete, the ultimate bond strength will be proportional to the square root of f'c, (3) there is an effective bond length for a joint beyond which no further increase in failure load can be achieved. A group of researchers conducted a study on the effect of the type of concrete surface preparation on the bond of carbon FRP (CFRP) sheets (Yoshizawa et al, 1996). The specimen used in these tests was a concrete prism with CFRP sheets applied to two opposite sides. The specimen was tested in tension, causing direct shear to be placed on the sheets. The concrete surface of the specimens was prepared by either water jet or sandblasting. It was found that the water jet doubled the capacity of the specimen as compared to sandblasting. The bonded length of the CFRP sheet was determined to have little effect on the ultimate load of the specimen. Another group of researchers studied the effect of test method and quality of concrete on the bond of CFRP sheets (Horiguchi and Saeki, 1997). Three different test methods, namely, shear test, flexural test, and direct tensile test were investigated. The tensile test produced the largest average bond strength, followed by the bending test. The lowest average bond strengths were found in the shear test. Three failure modes were observed: shearing of the concrete, delamination and FRP rupture. When the compressive strength of the concrete was low, less than 3600 psi (25.3 MPa), then failure occurred in the concrete. Delamination De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 3 occurred when the compressive strength was high or when the shear-type test was conducted. FRP fracture was observed in the bending tests with higher strength of concrete. Bond strength increased as the concrete compressive strength increased. The bonded length of the CFRP sheet had minimal effect on the ultimate load. Brosens and Van Gemert (1997) performed direct shear tests on two concrete prisms connected with 3 layers of CFRP on two opposite sides. Their findings showed that an increase in bonded length increases the failure load. This is contrary to findings of other researchers. However, they did find that the influence of bonded length decreases at longer lengths. They concluded that for computational purposes a linear bond stress distribution in the FRP sheet may be assumed. Another study on the bond mechanism of CFRP sheets was conducted by Maeda et al. (1997). The variables in this testing were bonded length, number of layers of FRP, and type of FRP sheet. Results of the tests showed that, as the stiffness of the fiber sheet increases, the ultimate load increases. For bonded lengths above approximately 4 inches (100 mm), the ultimate load did not change, implying the existence of an effective bond length. Xie and Kharbari (1997) used a specially designed peel test to characterize the bond strength between a carbon fiber/epoxy composite and a concrete substrate. The interfacial fracture energy was measured for various peel angles and peel rates. When the peel angle increased, or the peel rate decreased, the fracture energy at the interface was found to increase. Most failures happened as cohesive failures in the adhesive. A finite-element analysis was performed to evaluate the strain energy release rate in the peel test specimen for various crack propagation paths: (1) cohesive in concrete (2) adhesive-concrete interface (3) cohesive in adhesive (4) composite-adhesive interface. It was found that path 4 yielded the lowest strain energy release rate, therefore, it is the most possible failure path. However, the difference between paths 2, 3 and 4 is not significant. Cohesive debonding in the concrete substrate is De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 4 the least possible failure mode. However, for a small peel angle the difference is small and cohesive failure in the concrete is more possible than for a larger peel angle. Täljsten (1994, 1997) carried out shear tests on concrete prisms with steel and CFRP bonded plates. Results were compared with Volkersen's theory for lap joints and the comparison showed that the theory can predict the shear stress in the joint fairly well for moderate load levels. Failure in the joint always occurred by shearing of the concrete and test results seemed to indicate that there exists a critical strain level at which the concrete starts to fracture. It was also found that a critical anchor length exists above which longer lengths do not contribute to the ultimate load. Bizindavyi and Neale (1999) performed an experimental and analytical investigation on transfer lengths and bond strengths of composite laminates bonded to concrete. The observed modes of failure were shearing of the concrete beneath the glue line and rupture of the composite coupon. An analytical model based on a shear lag approach was developed. A comparison of analytical to experimental results was reasonably good for loads less than the initial cracking load, especially for specimens with 1-ply bonded joints. Volnyy et al. (1999) investigated bond between concrete and CFRP plates to be used as connections in precast concrete walls. Failure occurred in all specimens due to one, or a combination, of the following modes: CFRP rupture, delamination of the CFRP from the concrete, and concrete surface shear failure. The stress distribution along the composite appeared to be approximately linear, and to go asymptotically to zero after a certain distance, the effective bond length of the connection. Although a considerable amount of research has been devoted to bond of FRP laminates to concrete, to date there is no acceptable prediction model of the bond failure load and the effective bond length. In this project, flexural test specimens were prepared to address some of the factors expected to affect bond, namely, bonded length, concrete strength, number of De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 5 plies (stiffness), sheet width and, to a limited extent, surface preparation. Results are presented and discussed in this paper. A shear lag approach, along with a simple shear model for the evaluation of the slip modulus, is then used to model the strain distribution at moderate load levels. Finally, expressions of the peeling load and the effective bond length are presented. A design equation is proposed to calculate the effective FRP ultimate strain to be used in design to account for bond-controlled failure. EXPERIMENTS Specimens The specimen was a plain concrete beam with an inverted-T shape (Figure 1). The beam was simply supported, with a span of 42 in (1067 mm) and a total length of 48 in (1219 mm). A steel hinge at the top and a saw cut at the bottom, both located at mid-span, were used to control the distribution of the internal forces. During loading, the saw cut caused a crack to develop at the center of the beam and extend up to the hinge. Therefore, the compressive force in the beam at mid-span was located at the center of the hinge and the internal moment arm was known and constant for any given load level above the cracking load. This allowed to compute with accuracy the tensile stress in the FRP. A 2-in. (51-mm) wide CFRP strip was bonded to the t ension face of the beam. A transverse sheet was placed on one side to force failure to occur at the other end. Also, the sheet was left unbonded approximately 2 in (51 mm) on each side of midspan. The design choices were made to ensure that no cracking would occur within the bonded area. Three series of specimens were tested. Each series consisted of six specimens with three different bonded lengths. Either concrete strength or number of plies of CFRP was varied between each series. Description of the specimens is reported in Table 1. De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 6 After the beams had cured for 10-14 days, the concrete surface on which the CFRP had to be applied was sandblasted to remove the top layer of mortar, just until the aggregate was visible. The approximate depth of sandblasting was 0.06 in. (1.5 mm). A composite system including a unidirectional fiber tow CFRP sheet, primer and saturant was used for this experimental program. The sheet had an ultimate tensile strength of 620 ksi (4360 MPa), a design tensile strength of 550 ksi (3870 MPa), a modulus of elasticity of 33000 ksi (232 GPa) and a fiber thickness of 0.0065 in (0.165 mm), as indicated by the manufacturer (MBrace, 1998). The tensile properties of primer and saturant are reported in Table 2. The resin was allowed to cure for at least 7 days prior to testing of the beams. The thickness of each layer of the composite system was determined in previous work (Tumialan, 1998) by using a Scanning Electron Microscope (SEM). The resulting thicknesses were: 0.017 in. (0.432 mm) primer, 0.037 in. (0.940 mm) first resin layer, 0.0065 in. (0.165 mm) CFRP sheet. After testing of the three series of specimens was complete, three more specimens were tested. Two of them were made with 4-in. (102-mm) wide CFRP sheets instead of 2-in. (51mm) sheets. The purpose was to determine whether the width of the sheet had an effect on the strain distribution. The bonded lengths of the specimens were 8 and 12 in. (203 and 305 mm), while the nominal concrete strength was 6000 psi (42.2 MPa). These specimens were instrumented similarly to those of the first three series. Finally, one specimen was tested in which the surface preparation was changed. The surface was roughened by adding notches using a hammer and chisel. This specimen had 12-in. (305 mm) bonded length and 6000-psi (42.2 MPa) nominal concrete strength. The purpose of this test was to determine if surface preparation affected the average bond strength. No strain gauges were used in this test because the main interest was to see if the ultimate load was increased by the different surface preparation. De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 7 Data collection was accomplished by a series of strain gauges located along the length of CFRP sheet. One strain gauge was placed at the center of the unbonded region. For the 4-in. (102-mm), 8-in. (203-mm) and 12-in. (305-mm) bonded lengths, there were 3, 4 and 6 strain gauges placed along the centerline of the CFRP sheet in the bonded region, respectively. An LVDT was located at mid-span to monitor the beam deflection. Testing was performed on a universal testing machine. The beam was first loaded with 1500 lbs (6.67 kN) and then unloaded to 500 lbs (2.22 kN) to ensure that all data was being recorded properly. Next, the beam was loaded until a crack formed at midspan of the beam, and then unloaded to 500 lbs (2.22 kN). Load was then applied until failure. Results Test results in terms of ultimate load of the specimens are reported in Table 1. The data collected from the strain gauges was used to develop strain-location curves. These curves illustrate the strain vs. the distance the strain gauge is located from mid-span of the beam. Each curve is plotted for a given load level. A typical strain-location graph for the 4-in. (102mm) bonded length can be seen in Figure 2, while Figure 3 shows a typical graph for the longer bonded lengths. By comparing Figures 2 and 3, it can be seen that, at early stages of loading, the curves show the same behavior. They both have a non linear shape, and the strain gauges far from the center do not read strain. Also, as the load increases, the curves tend to attain a linear shape. It can be assumed that joint failure begins immediately after the point when the curve becomes linear. This corresponds to the attainment of a uniform bond stress along the portion of laminate that is taking the load. The length of this portion is what has been previously indicated as effective bond length. Once failure begins, the behavior in the two cases is different. The specimens with the 4-in. (102-mm) bonded length fail suddenly. For the 8 and 12-in. (203 and 305-mm) bonded lengths, the longer bonded length De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 8 causes failure to occur in stages. This progressive failure is indicated by the strain distribution. The strain becomes horizontal at the beginning of the bonded length, which means that no load is transferred into the concrete in that portion of the bond because the joint has begun to fail. In other words, the effective length of the CFRP sheet takes the entire load to a certain point at which localized joint failure occurs causing the effective bond length to shift. This shifting continues until the CFRP sheet has completely peeled from the concrete. Visual inspection of the specimens after the test revealed that failure occurred in the concreteadhesive interface, with very little or no sign of damage in the concrete surface. When comparing results, it was found that the bonded length did not affect the bond failure load. In fact, it was found that the bond strength decreases as the bonded length increases for all three series. It was concluded that an effective length exists beyond which no stress is transferred until peeling occurs. As already outlined in the introduction, other researchers have also reported the existence of a measurable bonded length beyond which no further increase in the transferred load can be achieved. While it was expected that the concrete strength would have an effect on the bond strength, there was no evidence from this investigation. Since failure occurred at the concrete-epoxy interface, the concrete strength did not affect the ultimate load. The number of plies used to make the CFRP laminate affects the bond failure load. In order for two plies of CFRP sheet to be as efficient as one ply, the ultimate load would have to double. As expected, this does not occur. The average of the ultimate loads of Series II is only 1.5 times that of Series I. Two specimens with 4-in. (102-mm) sheet width were tested in order to verify that the width of the CFRP sheet did not change the bond strength. The two specimens, with 8-in. and 12in. bonded length, failed at a load of 7890 lbs (35.09 kN) and 7990 lbs (35.54 kN), respectively. These ultimate loads are approximately twice those of the specimens with same bonded length and 2-in. wide sheets. Also, failure mode and strain distribution resulted the De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 9 same. It was concluded that the width of the sheet did not influence the bond strength. Finally, the performance of the specimen with roughened surface was much better than that of the specimen with sandblasted surface. The former failed at a load level of 5590 lbs (24.86 kN), as opposed to 3830 lbs (17.04 kN) and 3390 lbs (15.08 kN) of specimens 6-1-12-1 and 6-1-12-2. CFRP rupture was attained with the roughened surface. The sheet began peeling until it reached the location of the first set of notches. The notches seemed to anchor the sheet to the concrete. It was concluded that the surface preparation of the concrete can significantly affect the average bond strength. ANALYSIS OF RESULTS Linear Analysis of the Bonded Joint The analysis of the bonded joint in the linear elastic range can be conducted by means of a simple shear lag approach. The differential equation governing bond is as follows: 0 )) ( ( 1 2 2 = ⋅ ⋅ − x s E t dx s d τ (1) where s is the slip, τ the bond stress, x the coordinate along the bonded length of the laminate, t the thickness and E the elastic modulus of the FRP. Equation (1) comes from equilibrium and compatibility relations on a finite element of sheet of length dx, along with the assumptions that: (1) the FRP sheet is linear elastic, (2) the concrete strain is negligible if compared to that of the FRP, (3) the adhesive is only exposed to shear forces. At moderate load levels, a linear bond stress-slip behavior can be adopted: s K ⋅ = τ (2) Solving (1) with τ(s) given by (2) yields the following functions: x cosh C x sinh C ) x ( s α α ⋅ + ⋅ = 2 1 (3a) x sinh C x cosh C ) x ( α α α α ε ⋅ ⋅ + ⋅ ⋅ = 2 1 (3b) De Lorenzis, L., B. Miller, And A. Nanni, "Bond of FRP Laminates to Concrete", ACI Materials Journal, Vol. 98, No. 3, May -June 2001, pp. 256-264 10 ) ( ) ( x s K x ⋅ = τ (3c)