Strength Degradation in Curved FRP bars as Concrete Reinforcement

Steel reinforcement in concrete has the tendency to corrode and this process can lead to structural damage. FRP reinforcement represents a viable alternative for structures exposed to aggressive environments and has many possible applications where superior corrosion resistance properties are required. The use of FRP rebars as internal reinforcements for concrete, however, is limited to specific structural elements and does not yet extend to the whole structure. The reasons for this relate to the limited availability of curved or shaped reinforcing elements on the market and their reduced structural performance. Various studies, in fact, have shown that the mechanical performance of bent portions of composite bars is reduced significantly under a multiaxial combination of stresses and that the tensile strength can be as low as 25% of the maximum tensile strength that can be developed in the straight part. In a significant number of cases, the current design recommendations for concrete structures reinforced with FRP, however, were found to overestimate the bend capacity of FRP rebar. This paper presents the state-of-the art review of the research works on the strength degradation in curved FRP composites and highlighted the performance of exiting predictive models for the bend capacity of FRP reinforcement. Recent practical predictive model based on the Tsai-Hill failure criteria by considering the material at marcromechanical level is also discussed and highlighted. The review also identifies the challenges and highlights the future directions of research to explore the use of shaped FRP composites in civil engineering applications and the trends for future research in this area.


Introduction
Since the late 1980s, Fibre Reinforced Polymer (FRP) reinforcement has emerged as an alternative to replace conventional steel bars in reinforced concrete (RC) structures [1]. Since FRP reinforcement does not corrode and is more durable, it can extend the structures' service life and reduce maintenance/repair costs in concrete structures [2]. To date, internal FRP reinforcement for concrete is mainly limited to specific structural applications such as bridge decks, road barriers, marine structures, and tunnel and underground infrastructure. The limited use of internal FRP reinforcement can be partly due to the lack of commercially available curved or shaped reinforcing elements needed for complex structural connections [3] [4].
In current construction practice, most of the curved/shaped steel bars are pre-bent and cut offsite. Unlike FRP reinforcement, steel bars have an elastoplastic behavior and therefore they can be easily shaped by cold bending. Existing guidelines for the cold bending of steel bars (e.g. BS 8666 [5]) specify a bend radius to diameter ratio (r/d) of 2 for mild steel, which would induce a maximum strain value of 20% in the steel (see Figure 1). In the case of cold-bent FRP reinforcement, however, there are problems associated with the potential buckling of the fibres located on the compression side. Moreover, the typical ultimate strain value of FRP products varies from 1% to 2.5%, hence, the induced strain in the fibres needs to be controlled to avoid premature failure of the reinforcement. As a result, cold bending of FRP reinforcement requires larger r/d ratios than those currently specified for steel reinforcement. To date, only a few of the FRP bars commercially available are supplied in bent configurations, and all of them are pre-bent during manufacturing. Bends are usually created while the material is partially cured. Typical bent shapes available include: closed stirrup of thermoplastic FRP strip (Figure 2a), J-hook thermoplastic FRP strip, pre-bent GFRP thermoset composite (Figures 2b and 2c), and U-shaped thermosetting FRP bar (Figure 2d). Whilst Carbon (CFRP), Glass (GFRP), Aramid (AFRP) and basalt (BFRP) bars exist in the market, CFRP and GFRP seem to be much more widely used in actual RC applications and research. CFRP has better properties than all the other composites, whereas GFRP is significantly cheaper than other composites.  [3] degradation occurs at the bent portion of an FRP bar can be quantified using empirical equations such as the one initially proposed by the Japanese Society of Civil Engineering (JSCE) [10] which is currently adopted in many design guidelines for the use of FRP reinforced concrete structures. To account for this potential failure, several design guidelines ( [19] [10], [20]- [22]) limit the design strain values in the case of curved FRP reinforcement in reinforcement concrete structures. However, equations included in the current design guidelines to predict the bend strength degradation at the bent portion of an FRP bar is an empirically derived equation which is mainly a function of bent geometry and does not seem to yield consistent results when different types of composite are used [23].
This article provides an overview of existing and ongoing research on the strength of curved FRP reinforcement in RC structures. Extensive experimental works to investigate the strength degradation of curved FRP composites are systematically presented and test data available from literature is also included in Appendix as an additional source. Modern techniques used to fabricate customized/ complex shaped FRP composites are also discussed as emerging challenges.

Research on strength degradation of curved FRPs
Although extensive work has examined the behaviour of RC structures internally reinforced with FRP, relatively little information is available on curved FRP reinforcing bars [24]- [26]. These studies focused on determining the "bend capacity" of curved FRP reinforcement using pullout test on bent FRP bars embedded in concrete specimens ( Figure 3). Different test configurations were used in examining the bend capacity of FRP bars, as summarised in the following paragraphs.
Ozawa et al. [24] tested studied the static and flexural fatigue behaviour of concrete beams reinforced with FRP flexural and shear reinforcement, consisted of continuous glass and carbon fibres impregnated with resin and formed by the filament winding method. A total of 10 beam specimens were tested under two-point bending; two of them were statically loaded and the other eight were fatigue loaded. The authors reported that, if the beams failed in shear, FRP stirrups could fail at the bent portion at a stress lower than the ultimate strength of the equivalent straight bar. Ozawa et al. concluded that the stress concentration that developed at the bent portion of the bar caused rupture, a failure which originated from the inside of the bend. Similar conclusions were also reported by Miyata et al. [25] after carrying out a series of pull-out tests that studied the effect of bends on hybrid FRP bars embedded in concrete blocks (see Figure 3a). Direct tensile tests were performed on the reinforcement, which consisted of a 10 mm-diameter hybrid FRP composite made of continuous glass and high strength carbon fibres impregnated with resin. The main parameter investigated was the variation of the tensile strength of bent FRP bars as a function of the internal bending radius (r). Five different bar diameters were used in the test and bending radius was set to three times the bar diameter (i.e. r/d=3). The authors reported that most of the bent specimens failed due to the rupture of the FRP bars at the bent section, and that the fibres started to break from the inside portion of the bend. They also concluded that the failure load increased as the internal bending radius increases. Although the studies by Ozawa et al. [24] and Miyata et al. [25] provided some insight into the strength degradation of bent FRP bars, the tests only considered a few test parameters and therefore their conclusions were not generalised. However these tests did not consider the bond contribution along the bent portion and the effect of tail anchorage. Other parameters that could affect the bond stress such as concrete strength and surface treatment of FRP bars were also excluded in this test. . Different pullout setup for examining the bend capacity of FRP reinforcement (illustration adopted from [25], [12], [8], [11]) To examine the factors that influence the shear capacity of concrete beams contributed by FRP stirrups, Nagasaka et al. [12] tested 35 half-scale beams internally reinforced with FRP bars. Parameters investigated were the type and reinforcement ratio of FRP stirrups, as well as the concrete strength. Nagasaka et al. also tested four panel specimens to investigate the bend capacity of the FRP stirrups with the main reinforcement to simulate the bond at the bent location around the main bar, using the pullout arrangement shown in Figure 3b. The FRP bars were aramid, carbon, glass, and hybrid of glass and carbon FRP. The vertical leg was left unbonded to the beginning of the bent portion and the bend radius was 2 times the bar diameter (r/d=2). They reported that the ultimate shear capacity of concrete beams reinforced with FRP stirrups was determined by the tensile rupture of stirrups at the curved sections, or by crushing of a concrete strut formed between diagonal cracks. They also found that the tensile strength of curved FRP bars was only 25%-80% of that of a straight counterpart. One of the main contributions of Nagasaka et al. study is that the degree of bend capacity reduction depended on the type of composites.
Similar tests were carried out by Maruyama et al. [8] who tested 14 bent FRP samples embedded in concrete blocks with a 50 mm embedment length, ldb and the anchor at the end of tail to improve the bond stress ( Figure 3c). The main parameters studied were different types of composite materials, bending radii, and concrete strength. Curved pultruded CFRP rods, 7-strand CFRP rods, and braided AFRP rods were tested in direct tension and compared to steel bars with similar configurations. The bending radii (r) considered in this study were 5, 15, and 25 mm for each type of rod. Two different concrete strengths were used; (f'c = 50 MPa and 100 MPa) for each type of FRP rods. It was reported that all of the specimens failed due to rupture of the composite at the beginning of the bend on the loading side and all of the bend capacity of FRP bars were found to be lower than the tensile strengths of the straight portions, ranged from 48-82%. Moreover, the bend capacity trended to decrease hyperbolically as the bending radius decreased, and the bend capacity of FRP specimens were increased when the higher strength concrete was used and became more pronounced with 7-strand CFRP rods, and braided AFRP rods. This may be due to better bond developed by the stranded and braided composites, and the resulting lower amount of tensile stress transferred to the bend. In case of pultruded CFRP rods, the concrete strength had little effect on the bend capacity. This may be because the bond given by the roving wrapped around the rod was lost during the pullout tests, and thus adhesion at the bar-concrete interface becomes less significance. The authors also reported that the tensile strength at the bend varies with the type of fibre and the method of bending. The highest bend capacity-to-strength ratio was mobilized by braided AFRP rods, followed by strand CFRP and pultruded CFRP rods. These results indicated that the bend capacity depeneded on the type of FRP and the reinforcement surface (i.e. on bond properties). It should be mentioned that the test results of Maruyama et al. [8] were later used to validate the predictive equation for calculating the bend capacity of FRP reinforcement which is referred as the JSCE's equation [10] and now included in the current ACI guidelines [27] used to predict the bend strength of FRP bars. Ehsani et al. [11] investigated the bond behaviour of 90 o degree-hooked GFRP bars in concrete through thirty-six direct pullout tests as those shown in Figure 3d. The main parameter examined in Ehsani et al.'s study was the relationship between the strength capacity of curved FRP bars and the concrete compressive strength (f'c), which varied between 28 MPa and 56 MPa. Other examined parameters included bend radius to FRP bar diameter ratio (r/d= 0,3) (diameters, d= 9.5, 19.0, 28.6 mm), embedment length, and tail length (lc) beyond the hook. In these tests, the tensile load was horizontally applied through a gripping system (Figure 3d). Ehsani et al. found that the bend capacity was highly affected by the bend radius and bar diameter. When using r/d=3, the bend capacity ranged from 64% to 70% of the ultimate tensile strength and the bend capacity trends to increase when higher concrete strength was used. Based on their results, the authors recommended a minimum bend radius of 3d for GFRP hooks, as well as a tail length of 12d since the tail length beyond 12d had no beneficial effect on the strength of the bent bar. As the bend capacity increased with the embedment length, Ehsani et al. also recommended a minimum development length of 16d for a 90 o standard GFRP hook. The results from this study confirmed that concrete strength, embedment length and tail length are important parameters that influence the bent portion of FRP bars. Unfortunately, the studies by Ehsani et al. [11] did not consider types of composite used as well as different bending geometries that could affect the bend capacity of FRP bars.
The effectiveness of bent FRP reinforcement depends on the bond characteristics of the reinforcement itself, but also on the characteristics of the embedment and tail lengths. Accordingly, Vint and Sheikh [6] examined the bond performance of GFRP bars with different anchorage configurations (90 o degree-hooked bars and straight bars with mechanical anchor heads). 72 pullout GFRP specimens (as shown in Figure 3b) were tested using different anchorage configurations: straight anchorage, mechanical anchor heads or bends. Bent GFRP bars with different bending radius and surface coatings were used to examine the performance of this anchorage solution. Vint and Sheikh concluded that the full tensile strength in the fibre direction could be developed for bonded lengths of 5d in specimens with bent bars, and 10d for specimens with an anchorage head. However, the bend capacity of the GFRP bars was only 58-80% of the ultimate tensile strength of the straight portion. This indicates that, although mechanical anchor heads can potentially enhance the bond behavior of bent FRP bars, the theoretical ultimate tensile strength of the bars could not be achieved. The above mentioned studies examined the bend capacity of FRP bars using geometries typical of end anchorages (e.g. relatively large corner radius) However none of the previous studies tested FRP reinforcement with geometries similar to those used in steel stirrups.
Previous research also studied the effect of bends in FRP stirrups, but using geometries similar to those used in conventional steel stirrups [28][13], [16], [23], [29]- [37]. In these conditions, the tight corner radius of FRP stirrups tend to limit the shear capacity of the concrete beams where premature failure was generally observed at the proximity of the bent portion. To study the failure behaviour of thermoplastic FRPs as shear reinforcement in concrete beam, bent tests on thermoplastic FRP stirrups were nylon/carbon and nylon/aramid FRP fibres formed using a thermoplastic matrix resin in the pultrusion process was investigated by Currier et al. [38]. Thermoplastic FRP strips were bent in the laboratory by the application of heat to create the closed shape of shear links, having the internal bending radius of 12.7 mm and the bend capacity of the thermoplastic FRP links were evaluated using the test setup similar to the ACI B.5 method. Based on their test, the bend capacity of the thermoplastic FRP bars was 25% of the ultimate tensile strength of the straight portion and the failure on all of the tested specimens were observed at the bend portion of the stirrup.
Initial research by Ueda at al. [30] investigated the performance of FRP stirrups partially embedded into a concrete block, which aimed to simulate a shear crack crossing the FRP stirrups. The 6 mm-diameter FRP rods used in Ueda at al.'s study were braided epoxy-impregnated aramid fibres. The main variables in the study were the embedment length and the distance from the artificial crack to the bend. Tensile forces were transferred through steel plates, and steel rods to the bearing plates. The test setup was adopted from the ACT B.5 method (e.g. Figure 5a), except the free distance between two concrete blocks was not 200 mm but using the artificial crack instead. An artificial crack initiated with a 0.5 mm gap began to open and the tensile forces were then induced in the bent portions of the FRP sample. Ueda at al. also conducted Finite Element Analyses (FEA) to assess the nature of the stress-strain fields developed in the bent region. Their results showed that the bend capacity varied between 40% to 100% of the ultimate capacity in the direction of the fibres. The FEA showed that high strains developed in the inner portion of the bend, which was assumed as the location of failure initiation. For an embedment length of 100 mm, the failure stress was higher than the nominal strength of the straight bar. Accordingly, the numerical analysis performed by Ueda at al. was perhaps the first who attempted to investigate the stress-strain field at bent portion of FRP bars and the results agreed with the previous research where premature failure mostly initiated at the proximity of the bends.
Morphy et al. [18] tested sixteen specially-designed specimens using different types of FRP stirrups using the ACI B.5 method [39] as shown in Figure 5a. Parameters investigated were the type of FRP material, bar diameter, stirrup anchorage and embedment length of the stirrup into the concrete and the configuration of the stirrup anchorage. Three types of FRP reinforcement were used: Carbon FRP Leadline bars, Carbon Fibre Composite Cables (CFCC), and GFRP bars (C-BAR). All of the bent stirrups were embedded in concrete blocks with f'c = 45 MPa. The embedment length within the block varied by debonding part of the stirrups. The authors found that a decrease in the embedment length increases the tendency of failure at the bent region of the stirrup, which resulted in a bend capacity of 40% of that developed in a straight bar. From the results, it was suggested that a 150 mm embedment length was sufficient to achieve the full strength in the direction of the fibres. Morphy et al. also found that when a large bending radius to bar diameter ratio (r/d) is used, a higher bend capacity was observed. Based on their test results, and using the stirrup spacing recommended by the ACI codes [40], they proposed to limit the strength of CFRP stirrups to 50% of the unidirectional tensile strength to account for the strength degradation due to bend effect.
More recently, Imjai et al. [6] studied of the bend capacity on bent FRP stirrups using the pullout test shown in Figure 4a. 47 bent thermoset and thermoplastic FRP bars with 19 different configurations were investigated. The parameters investigated included the ratio r/d, surface treatment, embedment length (lb) and concrete strength (f'c). It was found that the capacity of the curved FRP composites could be as low as 25% of the ultimate tensile strength of the material parallel to the fibres. Based on the results, it was recommended to use a minimum ratio r/d=4 to guarantee that the composite could resist 40% of its unidirectional tensile strength parallel to the fibres. Imjai et al. also conducted FEA to study the bond stress along the bent portion of FRP bar embedded in concrete. The bond mechanism between the bent bar and the concrete was explicitly modelled with identical non-linear spring elements with the stiffness determined from the load-slip characteristic from the pullout tests. The FEA results confirmed that high stress concentrations develop at the beginning of the bent portion, thus indicating that failure could be expected to occur at this location ( Figure 4c). However, by using a larger bending radius or providing sufficient bond along the bent portion, the stress concentration at the beginning of the bend can be significantly reduced, and a higher bend capacity can be achieved. Although all of the experimental works mentioned above have dealt with the study of the behaviour of curved FRP bars embedded in concrete structures such as stirrups or anchorages, externally bonded FRP reinforcement (EBR), which have gained widespread use as strengthening material for RC structures in order to provide confinement and/or shear capacity may also suffer the bent effect, especially when bending the fibres over the member corners. The need for bending the composites may deteriorate the performance of the FRP laminate and the efficiency of its confining/strengthening action. Yang et al. [41] studied the effects of corner radius on the strength of FRP lamina using the test setup similar to the ACI B.12 [39]. In their experimental programme, one and two plies CFRP lamina were applied by the manual lay-up procedure over interchangeable corner inserts. They concluded that the corner radius (r) affects the strength of CFRP laminates. The test results showed that only 67% of the ultimate laminate strength could be developed when a large radius of insert was used. As the corner radius was decreased, the strength capacity of the FRP lamina further reduced. A higher failure stress was achieved by increasing the number of layers used.
Based on the literature summarized in this section, it is evident that numerous factors affect the bend capacity of FRP reinforcement such as bent geometry, materials of which type of composite is made, concrete strength, and bond stress between concrete/FRP bar interface. Advanced FE technique were used to study stress-strain field along the bent portion of FRP bars and the results confirmed that premature failure always initiated at the proximity of the bends which confirmed the reports from companion works from literature. However, issues such as mechanics at macro-scale of the material composition of composite bent portion when subjected to external loads, irregular shape and cross-section and bond stress along the bent portion have not been investigated yet and are a matter of future research. The results from the tests discussed in this section have also been reflected in the development of the predictive equation to the current design guidelines, as discussed in the following section.

ACI test methods of determining bend capacity of FRP bars
Different tests have been proposed to calculate the strength reduction in bent bars. The test rigs applicable to FRP bars for use in nonprestressed concrete outlined in ACI 440.3R [39] includes the B.5 method (bent bar capacity) and the B.12 method (corner radius), and these are illustrated in Figure  5a and b, respectively. The B.5 method measures the ultimate capacity of the FRP by testing (in tension) the straight portion of a FRP C-shaped stirrup whose bent ends are embedded in two concrete blocks (Figure 5a). The bend capacity of bent FRP bars are measured and compared to the ultimate tensile strength of the bar to obtain the strength reduction factor due to bend effects. The B.12 method measures the effect of the corner radius on the tensile strength of the FRP bar using a the testing apparatus shown schematically in Figure 5b. The apparatus applies tension in the U-shaped FRP that reacts against the bent portion mounted on a yoke.  Chronological development of predictive model for bend capacity and code provisions is shown in Figure 6. In 1995, Nakamura and Higai [17] conducted a theoretical study on the bend capacity of FRP stirrups based on test results from Miyata et al. [25]. As a result of their study, the authors proposed an empirical model to calculate the bend capacity of FRP composites (fb) as shown in Table  3; Equation (1). The model depends primarily on the bend ratio r/d, and therefore neglects the variation of the composite cross-section, the type of composites and the influence of bond characteristic between FRP/concrete interface. Based on test results from Ueda et al. [30], Ishihara et al. [15] analysed the behaviour of bent FRP stirrups embedded in concrete using a 2D FEA. The results of their study showed that the strength of a bar at its bent portion increases directly with the radius of the bend. Based on a FEA parametric study, Equation (2) was proposed to assess the strength of the bent portion (fb). Note that Equation (2) is a special case of Equation (1) in which λ replaces d/r. The study by Ishihara et al. showed that the reduction in bend strength was also a function of the different types of FRP composites. Ishihara et al. suggested that bond characteristics and differential slippage of the FRP rod (which were not considered in their FEA) could play an important role in the strength reduction. Figure 7a and b compare, respectively, the predictions given by Equations (1) and (2) and test data from Miyata et al. [25] and Ishihara et al. [15]. The results show that the experimentally derived bend capacity increases with increasing r/d ratio. Figure 7a also shows that the predictions from Equation (1) agree better with the test results when compared to Equation (2). This is not surprising as Equation 1 was empirically derived using test data from Miyata et al. [25]. In Figure 7b, it can be observed that Equation (2), as proposed by Ishihara et al. [15] predicts the experimental results more accurately than Equation (1). This is because the equation proposed by Ishihara et al. [15] was empirically derived using their own test data.  The strength degradation at the bend portion of FRP composite has been quantified using the predictive model (Equation (3)) included in current design recommendations for concrete structures reinforced with FRP composite materials (see Table 1) [44][20], [22], [43], [45] which is based on the original work by the JSCE guidelines [10]. In Equation (3)  ) using previous ACI B.5 bent test data from the literature [14], [16], [29]. It should be noted that Equations (1)(2)(3)(4) are empirical and only depend on the geometry of the bend, whilst the bond characteristic between FRP bar/concrete interface, type of FRP and material composition are neglected. Recent research by Imjai et al. [23] demonstrated that the predictions of Equations 1-4 do not match the experimental data available in the existing literature. As a result, Imjai et al. proposed a new macromechanical-based equation (Equation 5) that calculates more accurately the bend capacity of bent FRP reinforcement. Equation 5 adopts the Tsai-Hill failure criterion [46] for a unidirectional orthotropic lamina with fibres in the one-direction and subjected to plane stress in the 1-2 plane. The bend capacity (fb) was expressed as a function of the strength reduction factor (k) multiplied by the ultimate strength parallel to the fibres (fu). The strength reduction factor (k) is less than unity and ranged from 0.25 -0.70, depending on the value of β. The factor β (ratio of the longitudinal tensile strength and transverse compressive strength of the FRP material). In their model, the factor β was explicitly derived from the Tsai-Hill failure criterion which represents the physical meaning of materials at the macro-scale and type of composite/resin composition is being considered when determining the bend capacity of unidirectional FRP composites. Table 1. Summary of equations to predict the strength degradation of curved FRP reinforcement Nakamura and Higai [17] = ln (1 + ) .
where ln = 0.90 + 0.73ln ( / ) Derived using test data from Ishihara et al. [15] and further compared to the numerical results obtained from a 2D FE analysis Empirical model based on test results performed by the Japanese researchers.
Unfortunately, information on these tests is not available for all of specimens and only selected test data from JSCE extracted from Ishihara et al. [15] are presented in the = bend capacity, = ultimate strength parallel to the fibres, r = bend radius, α =0.05 corresponds to a 95% confidence limit, and α =0.092 corresponds to a 50% confidence limit, d = nominal diameter of the bars, dfi = diameter of the equivalent circular section, = is the safety factor, = 4 or t for circular or rectangular cross-sections, respectively, = strength ratio, and k = strength reduction factor for bent FRP bars This section compares the accuracy of the equations shown in Table 1 Table 1. The comparative results presented in Figure 9 shows clearly that the JSCE's equation ( =0.05) 12 is conservative, with a mean prediction/experiment ratio P/E = 1.02, and a standard deviation SD 13 =0. 27. It can be also seen that the five equations yield quite different ranges of results. For instance,

16
Equation (4) predicts better the test results and has less scatter (P/E=1.08, SD=0.28). Equation (5) 17 shows the best agreement with the test results and has a low scatter (P/E=1.00, SD=0.25). The  Shear reinforcement is often produced from pultruded bars prior to resin polymerization in form 43 of circular, rectangular, and others such as spiral shaped stirrup [49]. A recent study [32] reported 44 that prefabricated 3D FRP reinforcement cages produced using filament winding were successfully 45 used in concrete elements. The manufacturing process of the 3D reinforcement FRP cage included 46 wet and dry winding process. In the wet-winding process, each layer of fiber was impregnated with a two-component epoxy resin, squeezed with a polytetrafluorethylene tool to remove any excess, and 48 wound around the mold. The stirrups were cured at room temperature for 72 hrs, prior to being 49 demolded. In the dry winding process, the pre-preg tow was wound around the mold, before being    Eq. (1) Eq. ( Eq. (3) Eq. Eq. (1) Eq. ( Eq. (3) Eq. Eq. (1) Eq. ( Eq. (3) Eq. Eq. (1) Eq. ( Eq. (3) Eq. Note: r is the internal bending radius, d is the nominal diameter (diameter for circular section and thickness for strip), dfi is the transformed diameter, fb is the experimental average failure stress, and fu is the ultimate strength of the FRP bar.