BEHAVIOUR OF FRP-TO-STEEL BONDED JOINTS

Transcription

1 ABSTRACT Proceedings o the International Symposium on Bond Behaviour o FRP in Structures (BBFS ) Chen and Teng (eds) International Institute or FRP in Construction BEHAVIOUR OF FRP-TO-STEEL BONDED JOINTS S.H. Xia 1 and J.G. Teng 2 1 School o Civil and Environmental Engineering, Adelaide University, Adelaide 55, Australia 2 Department o Civil and Structural Engineering The Hong Kong Polytechnic University, Hong Kong, China Externally bonded iber-reinorced polymer (FRP) reinorcement oers an attractive method or the strengthening o structures constructed o various materials. In such strengthened structures, the characteristics o FRP-to-parent material bonded joints play an important role. While extensive research has been carried out on the characteristics o FRP-to-concrete bonded joints, existing work on FRP-to-steel bonded joints is much more limited. In an FRP-to-steel bonded joint, the weak link is the epoxy adhesive, while in an FRP-to-concrete bonded joint, the concrete is the weak link. This paper examines, through a series o pull-o tests in which the FRP-to-steel interace is subjected to direct shear, the parameters that aect the behaviour o FRP-to-steel bonded joints. The test results are irst presented and discussed. Based on these test results, the bond-slip relationship relating the interacial shear stress to the interacial slip is then investigated, leading to the development o the irst ever bond-slip model or FRP-to-steel interaces. INTRODUCTION While extensive research has been conducted on the strengthening o concrete and masonry structures using FRP composites (Teng et al. 22), the potential o externally bonded FRP composites in strengthening steel structures has been explored only to a very limited extent (Hollaway and Cadei 23; Xia and Teng ). The limited existing work has mainly been concerned with the demonstration o the eectiveness o the FRP strengthening technique or steel structures. Nevertheless, this limited existing work has provided a useul understanding o the overall behaviour and identiied possible ailure modes o FRP-strengthened steel members, particularly beams. FRP-strengthened steel beams can ail in a number o dierent modes. Apart rom the conventional ailure modes o steel (or steel-concrete composite) beams, FRP-strengthened beams may ail by the tensile rupture o the FRP laminate or by the debonding o the FRP laminate along the FRP-to-steel interace, depending on the beam and strengthening parameters. For convenience o description, the term interace is used in two dierent ways in this paper: (a) it is used to reer to the adhesive layer between the FRP plate and the steel substrate; and (b) it is used to reer to a physical interace such as the FRP-to-adhesive interace and the adhesive-to-steel interace. The precise meaning o the term would be clear within its context. To be able to understand and model debonding ailures in FRP-strengthened steel beams, it is irst necessary to understand the interacial behaviour between FRP and steel, usually through experimental studies on simple FRP-to-steel bonded joints. Indeed, a number o researchers have recently investigated the behaviour o FRP-tosteel bonded joints (Xia and Teng ). However, no research has been reported on the nonlinear behaviour o FRP-to-steel bonded joints covering the ull range o behaviour and the identiication o bond-slip relationships. The present paper presents the results o a study that represents an initial step to ill this gap in existing knowledge. EXPERIMENTAL PROGRAM The single-shear pull-o test (Figure 1) was adopted in the present study. This test set-up allows easy monitoring and inspection o the ailure process as only one path or debonding is possible. The same test set-up has been widely used in studies on FRP-to-concrete bonded joints (Teng et al. 22). A test rig (Figure 1a) was careully abricated to carry out all the tests reported in this paper. A tensile orce was applied to the FRP plate, and the steel block was supported at the loaded end. Appropriate restraints were provided to the steel block to prevent the steel block rom upliting and to minimize any bending in the FRP plate. Each pull-o test specimen was composed o a steel block bonded with a CFRP plate (Figure 1). The steel block was ormed by welding two mm thick steel plates to two 7 mm x 5 mm rectangular hollow sections o 3 411

2 mm in thickness as illustrated in Figure 1b. Two pull-o tests were conducted on each steel block, one on each o the two thick steel plates. The two test suraces o the steel block were sandblasted and cleaned with Acetone to remove any rust, residues and grease to enhance their bonding capability. Ball bearings were used as spacers to achieve the desired adhesive thickness. The ball bearings were adhesive-bonded on the steel plate at six dierent locations within the bond area beore bonding the FRP plate. The bonding o an FRP plate involved the application o an adhesive layer, the placement o the FRP plate, and the pressing-down o the FRP plate to the steel block with a moderate orce until the adhesive was suiciently cured. The adhesive was cured or seven days during which strain gauges were installed on the FRP plate. 355 Strain gauge P CFRP plate t a Adhesive t p 3 1 Steel tube Steel plate Test specimen Figure 1 Pull-o test specimen and set-up Table 1: Specimen details, test results and predictions Intended/measured adhesive thickness (mm) Test ailure load (kn) Predictions o the proposed Debonding ailure theoretical model mode L e (mm) P ult P ult /Test (kn) A-1 1/ Adhesive A-2a 2/ Adhesive A-2b 2/ Delamination* A-4 4/ Delamination A-6 6/ Delamination B-1 1/ Adhesive B-2a 2/ Adhesive B-2b 2/ Adhesive B-4 4/ Adhesive/delamination B-6 6/ Delamination C-1 1/ Adhesive/delamination C-2a 2/ Adhesive/delamination C-2b 2/ Adhesive/delamination Mean 1.6 Standard Deviation Adhesive Tensile strength t,a (MPa) Table 2: Material properties o adhesives Young s Modulus E a (MPa) Poisson s ratio ν a Ultimate tensile strain (%) A B C The eects o the adhesive properties are the ocus o the study. The variables considered in the present tests relect this ocus. Three dierent adhesives were used in the test program. The type o adhesive used or a particular specimen is indicated by the irst letter, while the thickness o the adhesive layer is indicated by the.57 4

3 number that ollows a hyphen. Two nominally identical specimens are distinguished rom each other by letters a and b. To achieve a wide range o values or the adhesive stiness, the thickness o the adhesive layer was varied, as the elastic modulus o a commercially available adhesive cannot be readily modiied. Four thicknesses were used: 1 mm, 2 mm, 4 mm, 6 mm. It should be noted that the irst two thicknesses are realistic, but the last two thicknesses were used to achieve a wide range o the adhesive layer thickness. The bond length (Lp) and the plate axial rigidity (Eptp) also have a signiicant eect on the bond behaviour but they were not varied in the test program because their eects are believed to be deducible rom existing work on FRP-to-concrete bonded joints (e.g. Cheng and Teng 21; Yuan et al. 24; Lu et al. ). Details o the parameters varied in the 13 test specimens are listed in Table 1. The CFRP plate used in the test had a bond length o 35mm, a width (bp) o 5mm, a thickness (tp) o 1.2mm and an elastic modulus o 165GPa based on strain measurements on the unbonded part o the CFRP plate in pull-o tests. For each adhesive, three coupons were tested in tension up to ailure. Both longitudinal and transverse strains were measured to determine the elastic modulus and Poisson s ratio o each adhesive. All three adhesives behaved linearly initially, became slightly nonlinear gradually and ailed suddenly by rupture. Their properties are given in Table 2, in which the elastic modulus is the secant modulus at rupture ailure. Along the length o the plate, strain gauges were installed (Figure 1) to determine the axial strains in the FRP plate and to deduce the interacial shear stresses. Both the adhesive coupon tensile tests and the bond tests were conducted with load control. TEST RESULTS AND DISCUSSIONS Failure Modes All specimens ailed along the FRP-to-steel interace with a wedge o adhesive attached to the FRP plate near the loaded end, as shown in Figure 2. Following the ormation o the adhesive wedge, cracks propagated in all specimens along the weakest components o the bonded joint, leading to the eventual ailure o the joint. Two distinct debonding ailure modes were observed: cohesive ailure within the adhesive layer (Figures 2a and 2b), and delamination o the FRP plate with the crack propagating within the FRP plate separating some carbon ibres rom the resin matrix (i.e. a thin layer o ibres was attached to the intact adhesive layer ater ailure) (Figure 2c). The occurrence o this plate delamination ailure mode indicates that in such FRP-to-steel bonded joints, the adhesive and the FRP-to-adhesive and the adhesive-to-steel interaces can be stronger than interaces between ibres and the resin matrix within the FRP plate. (a) Specimen A-1 (b) Specimen B-2a (c) Specimen A-4 (d) Specimen B-4 Figure 2 Failure modes o pull-o test specimens Specimens with 1 mm and 2mm thick adhesive layers ailed predominantly by debonding in the adhesive layer adjacent to the FRP-to-adhesive interace (e.g. Figures 2a and 2b). A thin layer o adhesive was attached to the FRP plate ater ailure. In some specimens, plate delamination occurred ater the cohesive debonding crack had propagated over a substantial part o the interace towards the ree end o the FRP plate. Since this delamination occurred late in the debonding ailure process, it is not taken to be the ailure mode o these specimens. That is, these specimens are taken to have ailed by cohesive ailure in the adhesive layer. In other cases, plate delamination occurred irst, ollowed by cohesive ailure in the adhesive layer. Such a ailure is taken to be a combination o the two distinct modes (Figure 2d) and denoted as adhesive/delamination in Table 1. Failure o the FRP-to-adhesive and adhesive-to-steel interaces (i.e. pure interacial debonding) were not observed, 413

4 testiying the strong bond capacities o the adhesives to the roughened steel and the FRP plate suraces. The ailure mode o each specimen is given in Table 1. Specimen A-2b should be noted as an exception. In Specimen A-2b, the CFRP plate was detached rom the steel block in the vicinity o the loaded end over a distance o 5mm on one side and 7 mm on the other side. This specimen ailed by plate delamination rapidly. In practical applications, adhesive thicknesses o 4 and 6 mm are unlikely to be used, so the plate delamination ailure mode appears to be unlikely based on the present tests, although this conclusion is believed to be dependent on the type o adhesive used. For the development o design methods, it is also desirable or ailure to occur within the adhesive layer, as there is much greater uncertainty i plate delamination or pure interacial debonding ailures control. Delamination o the FRP plate should be prevented using a strong resin or throughthickness ibres. Interacial debonding at the FRP-to-adhesive and the adhesive-to-steel interaces should be avoided by appropriate roughening/treatment o the suraces o the steel substrate and the FRP plate. For these reasons, the cohesive ailure mode in the adhesive layer is ocussed on in the remainder o the paper. For a detailed discussion o those joints which ailed by plate delamination, the reader is reerred to Xia and Teng (). Load-Displacement Behaviour Figure 3 shows the load displacement curves o our specimens (A-1, A-2a, B-1 and B-2a) which all ailed by cohesive ailure within the adhesive layer. The slips (or displacements) o the FRP plate were ound by integrating the measured strain distribution along the plate length (Xia and Teng ). Initially, the displacements o all specimens increase almost linearly with the load. The initial slopes o the loaddisplacement curves are apparently related to the type and thickness o the adhesive. The initial slopes are higher or the adhesive B specimens whose adhesive had a higher elastic modulus, and are lower or the adhesive A specimens whose adhesive had a lower elastic modulus. Overall, the initial slope deceases with the adhesive layer thickness, but the dierence between the two slopes or the two dierent adhesive layer thicknesses is much smaller than the degree o variation in the adhesive layer thickness. Indeed, assuming that only the adhesive layer is deormable, the predicted slope is ar greater than the corresponding experimental slope derived rom strain measurements on the top surace o the FRP plate. This is believed to be due to shear deormation within the FRP plate whose shear rigidity is derived rom the resin matrix, which is expected to have an elastic modulus o the same order as those o the adhesives used in the present study. For those specimens with a small adhesive layer thickness, the load displacement curve becomes nonlinear as the ultimate load o the FRP-to-steel bonded joint is approached due to micro-cracking that initiates at the loaded end. Ater the initiation o debonding ailure when the ultimate load is reached, debonding propagates towards the unloaded end with very limited variations in the load level but substantial displacements, resulting in a plateau in the load-displacement curve. This is similar to the behaviour o FRP-to-concrete bonded joints (Yuan et. al. 24) Load P (kn) A-1 A-2a B-1 B-2a Displacement (mm) Figure 3 Load-displacement curves Shear Stress Distributions along FRP-to-Steel Interaces Figure 4 shows typical distributions o shear stresses along the FRP-to-steel interace at dierent load levels or specimens A1 and B1 in the order o displacement. These shear stresses were calculated rom the readings o strain gauges mounted on the top surace o the FRP plate (Xia and Teng ), so they represent the average shear stresses over strain gauge intervals and are thus smaller than the actual values in the specimen. At a low 414

5 load level, the shear stress is the largest at the loaded plate end and then gradually reduces to zero towards the unloaded plate end. As the load increases, the shear stress at the loaded end approaches the local bond strength (i.e. the maximum interacial shear stress the interace is able to resist). When this local bond strength is reached at the loaded end, the linear stage o the load-displacement curve ends, and the FRP-to-steel interace enters its sotening stage, during which the shear stress at the loaded end decreases. When the shear stress at the loaded end reduces to zero, the ultimate load o the specimen is reached. Debonding then propagates towards the ree end as the peak shear stress moves away rom the loaded end with only small luctuations in the load level (Figure 4). These stages o development are the same as those described or FRP-to-concreted bonded joints (Yuan et al. 24). Shear stress (MPa) Shear stress (MPa) kn 51.6kN 58.3kN 58.4kN 62kN 63.2kN 61.4kN kN Distance rom loaded end (mm) (a) Specimen A-1 14kN 3.4kN 38.9kN 39kN 39.4kN 38.5kN 37kN 37.8kN 37.3kN 37.2kN Distance rom loaded end (mm) (b) Specimen B-1 Figure 4 Shear stress distributions Overall, the peak shear stress (i.e. the local bond strength) captured by strain measurements does not vary much along the bond length. The higher peak shear stresses near the loaded end seen in Figure 4 are believed to have been due to the eect o local bending o the plate near the loaded end. The lower peak shear stresses observed or the interace beyond 2 mm rom the loaded end (Figure 4) are believed to be due mainly to the larger strain gauge intervals in this region. Figure 4 shows the local bond strengths or specimen A-1 are 4-5 MPa higher than those or B-1. This dierence in the local bond strength is more than can be expected based on the tensile strengths o the two adhesives. Visual inspections o the shear stress distributions (Figure 4) show that the eective bond lengths or the two joints (and hence the two adhesives) are about 1 mm or A-1 and 8 mm or B-1 due to a larger strain capacity o adhesive A. These values are only approximate as the strain gauges had a spacing o mm. The higher local bond strengths and the eective bond lengths o adhesive A specimens explain why they have higher ultimate loads than the adhesive B specimens. This also means that the ultimate load o an FRP-to-steel bonded joint does not depend only on the tensile strength o the adhesive but also the strain capacity o the adhesive. That is, the interacial racture energy G determines the ultimate load, as has been well established or FRP-to-concrete bonded joints (Yuan et al. 24). 415

6 Shear stress (MPa) Slip (mm) Bond-Slip Curves Shear stress (MPa) Slip (mm) (a) Specimen A-1 (b) Specimen B-1 Figure 5 Bond-slip curves o the interace at dierent distances rom the loaded end Figure 5 shows the shear bond stress-slip curves at dierent distances rom the loaded end or specimens A-1 and B-1. The bond-slip curves ound rom specimens A-1 and B-1 are similar in shape, and the curves or dierent locations o the same interace are consistent, except near the loaded end where the strain readings are expected to have been aected by local bending o the plate. A bilinear bond-slip model can approximate these experimental curves closely. The slope o the ascending part is very dierent rom the theoretical shear stiness o the adhesive layer (G a /t a ) due to reasons given earlier in the paper. Elastic Sotening Region Debonding Shear stress (MPa) (δ 1, τ ) Area under the curve = G (,) (δ, ) Slip (mm) Figure 6 A bi-linear bond-slip model BOND-SLIP MODEL Using the experimental data obtained in the present study, a simple bi-linear bond-slip model was developed as described in this section. The present authors are not aware o any existing bond-slip model or FRP-to-steel interaces. The proposed bi-linear model (Figure 6) are deined by three key points: the origin (,), the peak shear stress point (δ 1, τ ), and the ultimate point (δ, ), with the area under the curve being the interacial racture energy (G ). The coordinates o the peak and ultimate points can be derived rom the experimental data. Based on existing analytical models (e.g. Yuan et al. 24), the ultimate load o a bonded joint between a thin plate and a stocky substrate block (i.e. the axial rigidity o the substrate block is much greater than that o the bonded plate) is given by Pult = bp 2G E pt p (1) i the bond length is greater than the eective bond length which is normally the case in practice. For a bilinear bond-slip model, the interacial racture energy is given by G =.5τ δ (2) so Eq. 1 becomes 416

7 P ult = b E t τ δ (3) where, E p, b p, and t p are the elastic modulus, width and thickness o the FRP plate. p p p The experimental data indicated that the value o the local bond strength τ varies with the type o adhesive but does not vary signiicantly with the adhesive thickness or the same adhesive and the same ailure mode. Based on the experimental local bond strengths rom those joints which experienced cohesive ailure in the adhesive layer at least over part o the interace (i.e. no results rom specimens A-4 A-6, and B-6 were included), this local bond strength can be reasonably closely approximated by (Xia and Teng ) τ =.8 t, a (4) where t, a is the tensile strength o the adhesive. In deriving this equation, only a single average value was used or each bonded joint. Due to the limited test data available, the dependence o τ on other parameters, such as the ultimate strain o the adhesive and the width o the FRP plate, is not yet clear and requires urther investigations. The interacial racture energy was ound to be related to the tensile strength t, a, the shear modulus G a and the thickness o the adhesive t adh. A nonlinear unction relating the interacial racture energy (and hence the product o τ and δ ) and the adhesive properties was chosen to approximate the test data, with the unknown coeicients varied to minimise the errors between the theoretical predictions and the test results. This process led to the ollowing expression:.56, adh t tadh G adh τ δ = (N mm/mm 2 ) (5) The above expression provides close predictions o the present experimental ultimate loads (Figure 7). To complete the deinition o the bond-slip model, it is proposed that the slope o the ascending part be equated to the shear stiness o the adhesive layer: K a = Ga / ta (6) Thereore, the slip δ 1 at the peak shear stress be deined as δ = / (7) 1 τ K a 7 R 2 =.85 6 Prediction (kn) 5 4 A B C Test (kn) Figure 7 Comparison between test and predicted ultimate loads The deinition o the initial slope o the bond-slip model excludes shear deormation o the resin matrix o the FRP plate, which is believed to be a rational approach as the shear modulus o the resin and the plate thickness varies rom one FRP material to another and the deormation o the resin matrix should be accounted or explicitly or each FRP product using its speciic properties. This aspect is very important and should be noted when the proposed bond-slip model is used in analysis. This situation contracts with that or FRP-to-concrete bonded joints. For the latter, the deormability o the adhesive layer can generally be ignored as ailure usually occurs within the concrete at much lower interacial shear stresses, and wet lay-up FRP sheets with a thin 417

8 adhesive layer between the ibre sheet and the concrete substrate are much more commonly used instead o pultruded FRP plates with a well-deined adhesive layer (Lu et al. ). Since the values rom Eq. 7 or δ 1 is generally very small compared to values o δ, the eective bond length can be approximated by the analytical expression or a bond-slip model with a rigid ascending branch ollowed by a linearly ascending branch, which is given by (Yuan et al. 24) π le = (8) 2 τ E pt pδ A comparison o the eective bond lengths and the ultimate loads rom the present tests with those predicted using the present bond model is given in Table 1. Very close agreement is seen. CONCLUSIONS This paper has presented a study into the interacial behaviour o a pultruded FRP plate bonded to a steel member, which is the basis or understanding debonding ailure mechanisms in FRP-plated steel members. Results rom a series o pull-o tests have been presented and discussed to understand the eects o the properties and the thickness o the adhesive layer on bond behaviour. Based on detailed strain measurements, a bond-slip model has been proposed or FRP-to-steel interaces. The results and discussions presented in the paper allow the ollowing conclusions to be made: The thickness o the adhesive layer has a signiicant eect on the ailure mode. When an adhesive layer o realistic thickness (< 2mm) is used, debonding is likely to occur within the adhesive layer with a ductile ailure process, but when a thick adhesive layer is used, debonding is likely to occur by plate delamination. Plate delamination is a brittle ailure mode and should be avoided in practice. For joints that ail by debonding in the adhesive layer, the local bond strength o the FRP-to-concrete interace is closely related to the tensile strength o the adhesive and does not depend on the adhesive layer thickness, but the interacial racture energy depends on both the ultimate tensile strain o the adhesive (i.e. the elastic/shear modulus or the same tensile strength) and the adhesive layer thickness. The slips ound rom strain measurements on the top surace o the FRP plate includes shear deormation o the resin matrix o the FRP plate, as the shear modulus o the resin matrix is o the same order as that o the adhesive layer. For joints that ail by debonding in the adhesive layer, the bond-slip curves can be very closely approximated by a bi-linear model. Based on this observation, a simple bi-linear bond-slip model has been proposed, which leads to accurate predictions o the ultimate loads and eective bond lengths o the present pull-o test specimens. Since the present tests covered only a limited range o variables, urther research is needed to assess and improve the accuracy and applicability o the proposed bond-slip model. ACKNOWLEDGEMENTS The authors grateully acknowledge the inancial support provided by The University o Adelaide through the small grant scheme and The Hong Kong Polytechnic University (project code: BBZH). They would also like to thank A/Pro. Deric Oehlers or his invaluable discussions during the preparation o the test specimens. REFERENCES Chen, J.F. and Teng, J.G. (21) Anchorage strength models or FRP and steel plates bonded to concrete, Journal o Structural Engineering, ASCE, 7(7), Hollaway, L.C. and Cadei, J. (23). Progress in the technique o upgrading metallic structures with advanced polymer composites, Progress in Structural Engineering and Materials, 4(2), Lu, X.Z., Teng, J.G., Ye, L.P and Jiang, J.J. (). Bond-slip models or FRP sheets/plates bonded to concrete, Engineering Structures, 27(6), Teng, J.G., Chen, J.F., Smith, S.T. and Lam, L. (22). FRP-Strengthened RC Structures, John Wiley and Sons Ltd, UK, 245pp. Xia, S.H. and Teng. J.G. (). Interacial behaviour between FRP and steel, in preparation. Yuan, H., Teng, J.G., Seracino, R., Wu, Z.S. and Yao, J. (24). Full-range behavior o FRP-to-concrete bonded joints, Engineering Structures, 26(5),