Experimental Research on Bond Behavior between GFRP Bars and Stirrups-Conﬁned Concrete

Featured Application: In this paper, the bond performance between Glass-ﬁber-reinforced polymer (GFRP) bars and stirrups-conﬁned concrete was studied for the ﬁrst time, which is very important to promote the engineering application of FRP bars. Abstract: This paper presents the results of a series of pullout tests that were performed on Glass-ﬁber-reinforced polymer (GFRP) bars embedded in concrete, while providing a detailed report on the inﬂuence of various variables that impinge upon bond behavior, such as the surface characteristics and diameter of the bars, concrete strength, as well as the conﬁned effect of stirrups. The Bertero-Popov-Eligehausen (BPE) and Cosenza-Manfredi-Realfonzo (CMR) models analyzed the bond stress ( τ )–slip ( s ) relationship between GFRP bar and stirrups-conﬁned concrete. The tests results indicate that when the bond failure interface only occurs on the surface of a GFRP bar, the bond strength is not dependent upon the concrete strength. Moreover, the results indicate that in comparison to specimens without stirrups, their stirrup-containing counterparts are more prone to pullout failure with greater ductility and higher bond strength and corresponding slip. The BPE and CMR models are able to investigate the τ - s relationship between GFRP bars and the stirrups-conﬁned concrete with accuracy. With the experimental data, the speciﬁc parameters in the models classiﬁed by surface characteristics have been suggested.


Introduction
The use of fibre reinforced polymers (FRP) as reinforcement in concrete structures is considered to be a possible alternative to steel in situations where corrosion is present. It is imperative that the bond behavior between FRP bars and concrete be given top priority prior to studying the performance of FRP-reinforced concrete (RC) structures. This is due to the crucial role that bond plays in terms of the coaction between FRP bars and concrete. An adequate level of bond is required between an FRP bar and concrete in order to ensure the successful transmission of forces from one to another. Unlike steel bars, the FRP bars behave anisotropic, non-homogeneous, and linear elastic properties, which may result in different force transfer mechanisms between bars and concrete. Some research on the performance of FRP-RC members [1,2] or structures [3,4] purport that the bond stress (τ)-slip (s) relationship differs greatly between that of FRP bars and that of steel bars, notably influencing the usability of the concrete structure. Contrarily, unlike steel bars, no uniform manufacturing standards have been established for FRP bars, resulting in divergences between the performance of FPR bars across different countries and manufacturers. Research into the bond behavior of different FRP indispensable that the influence of the confined effect of stirrups on the bond behavior between GFRP bars and concrete is studied.
In this paper, 100 pullout tests for GFRP bars and 20 pullout tests for steel bars have been carried out according to CSA S806-02 [29] and ACI 440.3R-04 [41] standards. These tests investigated the effects of various parameters on bond behavior (e.g., bond strength, slip capacity, and failure mode) of GFRP bars that are embedded in concrete. Furthermore, the analytical model was also used in order to predict the τ-s relationship of GFRP bars in stirrups-confined concrete, and the appropriate parameters were derived from the present experimental data.

Material Properties
The GFRP bars in this study were made of unidirectional E-glass fibres and vynilester resin. There are three types of external surfaces of the GFRP bars: helical wrapping, helical wrapping with sand coating, and ribbed (deformations by resin). Pullout tests have also been performed upon steel bars to form a comparison. Figure 1 shows the surface deformation and the characteristics of the bars. The nominal diameters of these bars provided by manufacturers were 8,12,16, and 20 mm. Normalised tests were conducted to determine the cross-sectional areas (experimental bar diameter) and mechanical properties (tensile strength and elastic modulus) of the bars, according to ACI 440.3R-04 [41] and CSA S806-02 [29]. Table 1 displays the geometric and mechanical information of the bars.
Appl. Sci. 2019, 9,1340 3 of 20 In this paper, 100 pullout tests for GFRP bars and 20 pullout tests for steel bars have been carried out according to CSA S806-02 [29] and ACI 440.3R-04 [41] standards. These tests investigated the effects of various parameters on bond behavior (e.g., bond strength, slip capacity, and failure mode) of GFRP bars that are embedded in concrete. Furthermore, the analytical model was also used in order to predict the τ-s relationship of GFRP bars in stirrups-confined concrete, and the appropriate parameters were derived from the present experimental data.

Material Properties
The GFRP bars in this study were made of unidirectional E-glass fibres and vynilester resin. There are three types of external surfaces of the GFRP bars: helical wrapping, helical wrapping with sand coating, and ribbed (deformations by resin). Pullout tests have also been performed upon steel bars to form a comparison. Figure 1 shows the surface deformation and the characteristics of the bars. The nominal diameters of these bars provided by manufacturers were 8,12,16, and 20 mm. Normalised tests were conducted to determine the cross-sectional areas (experimental bar diameter) and mechanical properties (tensile strength and elastic modulus) of the bars, according to ACI 440.3R-04 [41] and CSA S806-02 [29]. Table 1 displays the geometric and mechanical information of the bars.
In order to analyze the influence of concrete strength on the bond behavior, two different concrete strengths, C1 and C2, were used. Three concrete cubes of 150 mm × 150 mm × 150 mm were reserved and then cured for 28 days with all specimens being subjected to the same conditions, obtaining an average compressive strength of 28.2 ± 0.5 MPa for C1 and 4 47.6 ± 0.7 MPa for C2. Table  2 shows the compositions of C1 and C2.    In order to analyze the influence of concrete strength on the bond behavior, two different concrete strengths, C1 and C2, were used. Three concrete cubes of 150 mm × 150 mm × 150 mm were reserved and then cured for 28 days with all specimens being subjected to the same conditions, obtaining an average compressive strength of 28.2 ± 0.5 MPa for C1 and 4 47.6 ± 0.7 MPa for C2. Table 2 shows the compositions of C1 and C2.

Specimens, Setup and Testing Equipment
According to CSA S806-02 [29], the bars were pulled from 150 mm concrete cubes. The embedment length was 5d (d is the bar diameter). The diameter of the stirrup was 8 mm and the spacing was 40 mm. In order to be closer to the actual situation in engineering, longitudinal steel bars were configured in concrete cube and then tied with stirrups. The assembled steel cages are shown in Figure 2 and the tested specimens are shown in Figure 3. Two nominally identical specimens without stirrups (N) and three nominally identical specimens with stirrups (Y) were tested for each specimen type to obtain some measure of uncertainty. Each specimen has been identified with a five-part code. For instance, R1-C1-8-N-1 indicates the surface characteristic of GFRP in the specimen is helical wrapping, concrete strength is C1, the bar diameter is 8 mm, no stirrup is configured in the concrete cube, and the specimen is numbered as 1 in the identical specimens.
A pullout load was applied using a servo-hydraulic testing machine (1000KN/SHT4106-G) with displacement control with the loading speed set at 1.0 mm/min, as shown in Figure 4. Contact between the concrete and the bar along the debonded length was broken using a Polyvinyl chloride (PVC) tube to minimize the stress concentration that is near the loaded end. A spherical seat was installed between the steel plate and the base plate at the bottom of the steel frame of the fixed concrete specimen. This enabled the specimen to spin freely within a narrow range during the loading process in order to prevent the GFRP bar or the steel bar from bending and torsion due to the load. The free (unload) end slips of the bar relative to concrete were measured with two linear variable differential transformers (LVDTs) that were fixed on the surface of the bar. Outputs from the servo-hydraulic testing machine and the LVDTs were recorded while using an automatic data acquisition system.

Experimental Result
In the pullout tests, the pullout load and the displacement values were used to calculate the corresponding average bond stress and slip. The average bond stress were calculated as the pullout force divided by the surface area of the bar that is embedded in the concrete with the assumption of uniform bond stress distribution along the embedment length: where τ is the average bond stress; P is the imposed force; d is bar diameter; and, l is the embedment length. Due to the low elastic modulus of the GFRP bar, a significant degree of elastic deformation occurs at its loading end, resulting in deviations in the calculation of slip at the loading end, which influence the accuracy of the τ-s relationship [24]. Nevertheless, measuring of the slip at the free end enables the accurate reflection of the actual slippage of the GFRP bar in the concrete.
Consequently, in this study, the slips corresponding to average bond stress are derived from the free of slip. Table 3 depicts the test results, in which τmax is representative of the maximum average bond stress (bond strength), and sm is the free end slip corresponding to the maximum average bond stress; τ* is the average bond strength that is obtained by calculating mean value of bond strength for identical specimens. s* is the average free end slip corresponding to τ*; P represents pullout failure;

Experimental Result
In the pullout tests, the pullout load and the displacement values were used to calculate the corresponding average bond stress and slip. The average bond stress were calculated as the pullout force divided by the surface area of the bar that is embedded in the concrete with the assumption of uniform bond stress distribution along the embedment length: where τ is the average bond stress; P is the imposed force; d is bar diameter; and, l is the embedment length. Due to the low elastic modulus of the GFRP bar, a significant degree of elastic deformation occurs at its loading end, resulting in deviations in the calculation of slip at the loading end, which influence the accuracy of the τ-s relationship [24]. Nevertheless, measuring of the slip at the free end enables the accurate reflection of the actual slippage of the GFRP bar in the concrete.
Consequently, in this study, the slips corresponding to average bond stress are derived from the free of slip. Table 3 depicts the test results, in which τmax is representative of the maximum average bond stress (bond strength), and sm is the free end slip corresponding to the maximum average bond stress; τ* is the average bond strength that is obtained by calculating mean value of bond strength for identical specimens. s* is the average free end slip corresponding to τ*; P represents pullout failure;

Experimental Result
In the pullout tests, the pullout load and the displacement values were used to calculate the corresponding average bond stress and slip. The average bond stress were calculated as the pullout force divided by the surface area of the bar that is embedded in the concrete with the assumption of uniform bond stress distribution along the embedment length: where τ is the average bond stress; P is the imposed force; d is bar diameter; and, l is the embedment length. Due to the low elastic modulus of the GFRP bar, a significant degree of elastic deformation occurs at its loading end, resulting in deviations in the calculation of slip at the loading end, which influence the accuracy of the τ-s relationship [24]. Nevertheless, measuring of the slip at the free end enables the accurate reflection of the actual slippage of the GFRP bar in the concrete. Consequently, in this study, the slips corresponding to average bond stress are derived from the free of slip. Table 3 depicts the test results, in which τ max is representative of the maximum average bond stress (bond strength), and s m is the free end slip corresponding to the maximum average bond stress; τ* is the average bond strength that is obtained by calculating mean value of bond strength for identical specimens. s* is the average free end slip corresponding to τ*; P represents pullout failure; and, S represents concrete splitting failure. Figures 5-9 shows the τ-s curves of the different specimens. Due to the results of the identical specimens being basically the same, one curve was chosen for each specimen type in Figures 5-9. It is apparent that the τ-s curves of the specimens exhibit distinctive features with different surface characteristics of the bar. Section 4 discusses this phenomenon.    Bond stress (MPa)             Bond stress (MPa)

Influence of Bar Surface Characteristics
In this paper, the tests compare the bond behavior of three types of surface characteristics that are observable in GFRP bars, R1/helical wrapping, R2/helical wrapping and sand coating, and R3/Ribbed. Figure 10 shows the typical failure modes of GFRP bars that are exhibited by specimens with pullout failure. Close examination of the surface of GFRP bars indicated that the bond stress collapse of the GFRP bar was primarily due to the detachment of the fiber spirals and resin layer, for R1 bars (Figure 10a), and sand grains for R2 (Figure 10b) bars, while the concrete surface remained damaged. It was evident that a friction-resistant type mechanism had been activated via this failure mode. The bond stress was predominantly dependent upon chemical adhesion and after slip upon friction. Furthermore, sanding provides many small anchoring points that are distributed over the surface, leading to an increase of friction between GFRP bar and concrete, thereby increasing bond strength. Table 3 shows that the average bond strength of R2 bars is 1.2-1.35 times that of R1 bars with the same diameter. This is also reflected in the experiments by Zhang et al. [42]. Subsequent to pullout failure, the R1 bars exhibited a constant residual bond stress that is caused by the friction between the bar and concrete, whereas the R2 bars exhibited a small amount of oscillating residual bond stress due to the existence of mechanical interlocks between the shallow sand coating and the concrete (Figures 5 and 6). Close examination the interface between the R3 and R4 bars and the concrete revealed that the bond stress collapse of R3 bar was a consequence of the crushing of concrete and the shearing off of ribs (Figure 10c), whilst the failure of R4 bars was solely due to the crushing of concrete (Figure 10d). This failure mode verified that a bearing type mechanism was activated. The

Influence of Bar Surface Characteristics
In this paper, the tests compare the bond behavior of three types of surface characteristics that are observable in GFRP bars, R1/helical wrapping, R2/helical wrapping and sand coating, and R3/Ribbed. Figure 10 shows the typical failure modes of GFRP bars that are exhibited by specimens with pullout failure. Close examination of the surface of GFRP bars indicated that the bond stress collapse of the GFRP bar was primarily due to the detachment of the fiber spirals and resin layer, for R1 bars (Figure 10a), and sand grains for R2 (Figure 10b) bars, while the concrete surface remained damaged. It was evident that a friction-resistant type mechanism had been activated via this failure mode. The bond stress was predominantly dependent upon chemical adhesion and after slip upon friction. Furthermore, sanding provides many small anchoring points that are distributed over the surface, leading to an increase of friction between GFRP bar and concrete, thereby increasing bond strength. Table 3 shows that the average bond strength of R2 bars is 1.2-1.35 times that of R1 bars with the same diameter. This is also reflected in the experiments by Zhang et al. [42]. Subsequent to pullout failure, the R1 bars exhibited a constant residual bond stress that is caused by the friction between the bar and concrete, whereas the R2 bars exhibited a small amount of oscillating residual bond stress due to the existence of mechanical interlocks between the shallow sand coating and the concrete (Figures 5  and 6). Close examination the interface between the R3 and R4 bars and the concrete revealed that the bond stress collapse of R3 bar was a consequence of the crushing of concrete and the shearing off of ribs (Figure 10c), whilst the failure of R4 bars was solely due to the crushing of concrete (Figure 10d). This failure mode verified that a bearing type mechanism was activated. The bond stress was primarily dependent on the interlocking interaction between ribs and concrete. Therefore, R3 and R4 bars exhibited a higher bond strength than friction-resistant type bars (R1, R2). The average bond strength of R3 bar manifested as 1.3-1.5 times that of R1 bar; that of R4 bar manifested as 1.75 times that of R1 bar with the same diameter (Table 3). After pullout failure, R3 exhibited a large oscillating residual bond stress due to the existence of interlocking between the residual ribs and the concrete (Figure 7).
When concrete splitting failure became apparent in the specimen, the bond stress-slip curves (Figures 5-8)

Influence of Bar Diameter
As has been indicated in prior literature [6,9,[11][12][13], the larger the bar diameter, the smaller its bond strength. Two widely recognized viewpoints can explain this phenomenon. The former reasoning stating that bars with a larger diameter require longer embedment length in order to develop the same normal bond stress; larger embedment lengths reduce the average bond strength, as confirmed by the references [6,13]. As the latter reasoning states, Poisson's effect [9,13] can lead to a slight reduction in bar diameter as a result of longitudinal stress. This bar reduction increases with bar diameter, which can lead to reduced frictional/mechanical locking stresses. This conclusion has been confirmed by experiments that were conducted previously; this tendency is not affected by the failure mode, concrete strength, the confined effect of stirrups, or the surface characteristics of GFRP bars (Figures 5-8). Additionally, with an increase of GFRP bar diameter, there was an increased likelihood of concrete splitting in specimens (Table 3).

Influence of Concrete Strength
Research on the bond behavior of steel bars has indicated that, when pullout failure occurs, there is an increase in bond strength of steel bars, along with an enlargement of concrete strength. However, this conclusion is not always applicable in the case of GFRP bars. Figure 11 shows a comparison between the average bond strength of R1 and R3 bars in C1 and C2 concrete without stirrups. With the occurrence of pullout failure, the bond failure interface only transpires at the surface of the R1 bar. When the concrete strength increases from C1 to C2, the average bond strength

Influence of Bar Diameter
As has been indicated in prior literature [6,9,[11][12][13], the larger the bar diameter, the smaller its bond strength. Two widely recognized viewpoints can explain this phenomenon. The former reasoning stating that bars with a larger diameter require longer embedment length in order to develop the same normal bond stress; larger embedment lengths reduce the average bond strength, as confirmed by the references [6,13]. As the latter reasoning states, Poisson's effect [9,13] can lead to a slight reduction in bar diameter as a result of longitudinal stress. This bar reduction increases with bar diameter, which can lead to reduced frictional/mechanical locking stresses. This conclusion has been confirmed by experiments that were conducted previously; this tendency is not affected by the failure mode, concrete strength, the confined effect of stirrups, or the surface characteristics of GFRP bars (Figures 5-8). Additionally, with an increase of GFRP bar diameter, there was an increased likelihood of concrete splitting in specimens (Table 3).

Influence of Concrete Strength
Research on the bond behavior of steel bars has indicated that, when pullout failure occurs, there is an increase in bond strength of steel bars, along with an enlargement of concrete strength. However, this conclusion is not always applicable in the case of GFRP bars. Figure 11 shows a comparison between the average bond strength of R1 and R3 bars in C1 and C2 concrete without stirrups. With the occurrence of pullout failure, the bond failure interface only transpires at the surface of the R1 bar. When the concrete strength increases from C1 to C2, the average bond strength of specimens R1-C1-8-N, R1-C1-12-N, and R1-C1-16-N becomes essentially akin to that of specimens R1-C2-8-N, R1-C2-12-N, and R1-C2-16-N ( Figure 11); the shape of the τ-s curves also appear to be similar (Figures 5a and 9a), which is concordant with the research that was noted in the references [7,9,42,43]. However, the bond failure interface occurs at the surfaces of both the R3 bar and concrete. When the concrete strength increases from C1 to C2, the τ-s curves of specimens R3-C1-8-N and R3-C1-12-N (Figure 7a) appear similar in shape to those that are evident in specimens R3-C2-8-N and R3-C2-12-N (Figure 9b). However, the average bond strength (τ*) increases by 37.3% and 23.6%, respectively ( Figure 11). It can be concluded that the bond failure interface takes place in the concrete, and the bond strength of the GFRP bar is directly related to concrete strength. The conclusion is unaffected by stirrups ( Figures 5, 7 and 9).

Influence of the Confined Effect of Stirrups
ACI 408R-03 [35] specifies that stirrups provide confinement to concrete, which not only helps to curtail the occurrence of splitting cracks, but also changes the failure modes and the τ-s relationship, culminating in a relatively higher ductile performance. This thereby increases the bond strength and corresponding slip of steel bars to concrete. This viewpoint is also applicable to the GFRP bars. Figure 12 (Table 3). At the failure point, an abrupt explosive noise was heard, and a penetrating crack appeared on concrete surface at the loading end of these specimens, e.g., R4-C1-20-N-1 (Figure 12a). The average bond strength of these specimens is only related to the diameter, as shown in Table 3. However, the failure modes of their stirrup-containing counterparts changed into pullout failure with many micro-cracks appears on the concrete surface at the loading end e.g., R3-C1-20-Y-2 (Figure 12b). This is with the exception of specimen R4-C1-20-Y. Despite the concrete splitting failure on specimen R4-C1-20-Y, unlike specimen R4-C1-20-N, many penetrating

Influence of the Confined Effect of Stirrups
ACI 408R-03 [35] specifies that stirrups provide confinement to concrete, which not only helps to curtail the occurrence of splitting cracks, but also changes the failure modes and the τ-s relationship, culminating in a relatively higher ductile performance. This thereby increases the bond strength and corresponding slip of steel bars to concrete. This viewpoint is also applicable to the GFRP bars. Figure 12 shows the typical cracking behavior of the specimens.  (Table 3). At the failure point, an abrupt explosive noise was heard, and a penetrating crack appeared on concrete surface at the loading end of these specimens, e.g., R4-C1-20-N-1 (Figure 12a). The average bond strength of these specimens is only related to the diameter, as shown in Table 3. However, the failure modes of their stirrup-containing counterparts changed into pullout failure with many micro-cracks appears on the concrete surface at the loading end e.g., R3-C1-20-Y-2 (Figure 12b). This is with the exception of specimen R4-C1-20-Y. Despite the concrete splitting failure on specimen R4-C1-20-Y, unlike specimen R4-C1-20-N, many penetrating and non-penetrating cracks appeared on the concrete surface at its loaded end, e.g., R4-C1-20-Y-2 ( Figure 12c). This phenomenon is also reflected in the research by Darwin and Graham [32] on the bond behavior of steel bars. Largely due to the existence of stirrups, the circumferential pressure of the steel bar to concrete becomes more uniform, resulting in the appearance of more cracks on the concrete surface.  Figure 13 shows the ratio of the average bond strength (τ* (Y) / τ* (N)) and the corresponding average slip (s* (Y) / s* (N)) of the specimens with and without stirrups. It is evident that the τ* and s* of specimens that contain stirrups will be increased if their stirrup-lacking counterparts undergo concrete splitting failure. The larger the diameter of the similar bars, the more evident was the improvement in its performance. Different types of bars ranged in the amount of ostensible improvement, from R1, R2, R3, and R4 bars with same diameter. The τ* (Y) / τ*(N) and s* (Y) / s* (N) values of R2, R3, and R4 bar of the 16mm diameter were 1.25, 1.32, 1.44 and 1.38, 1.57, 1.79, respectively; while the values of R1, R2, R3, and R4 bar of 20mm diameter were 1.12, 1.41, 1.58, 1.69 and 1.31, 1.68, 2.03, 2.18, respectively. This may be because the larger the diameter and bond stress between the bar and concrete, the higher the likelihood of the occurrence of concrete splitting failure, which in turn renders an increase in the confinement of stirrups on concrete.
In addition, no splitting cracks were apparent in specimens that underwent pullout failure. The τ-s curves of specimens with and without stirrups were basically identical ( Figures 5-8), and the values of τ* (Y) / τ*(N) and s* (Y) / s* (N) are close to 1 (Figure 13). This observation can be reckoned to the work by Malvar [18], who imposed different circumferential constraining forces onto the concrete surface of the pullout specimens. If there was no apparent cracking to the concrete with the GFRP bar pulled out, the different circumferential constraint force did not bear influence upon the bond stress-slip process of the GFRP bar. The same phenomenon was observable in the pullout testing of the R1 and R3 bar in concrete C2, hence Figure 9 only lists τ-s curves of R1 and R3 bars in concrete C2 without stirrups   Figure 13 shows the ratio of the average bond strength (τ* (Y) / τ* (N)) and the corresponding average slip (s* (Y) / s* (N)) of the specimens with and without stirrups. It is evident that the τ* and s* of specimens that contain stirrups will be increased if their stirrup-lacking counterparts undergo concrete splitting failure. The larger the diameter of the similar bars, the more evident was the improvement in its performance. Different types of bars ranged in the amount of ostensible improvement, from R1, R2, R3, and R4 bars with same diameter. The τ* (Y) / τ*(N) and s* (Y) / s* (N) values of R2, R3, and R4 bar of the 16mm diameter were 1.25, 1.32, 1.44 and 1.38, 1.57, 1.79, respectively; while the values of R1, R2, R3, and R4 bar of 20mm diameter were 1.12, 1.41, 1.58, 1.69 and 1.31, 1.68, 2.03, 2.18, respectively. This may be because the larger the diameter and bond stress between the bar and concrete, the higher the likelihood of the occurrence of concrete splitting failure, which in turn renders an increase in the confinement of stirrups on concrete.
In addition, no splitting cracks were apparent in specimens that underwent pullout failure. The τ-s curves of specimens with and without stirrups were basically identical ( Figures 5-8), and the values of τ* (Y) / τ*(N) and s* (Y) / s* (N) are close to 1 (Figure 13). This observation can be reckoned to the work by Malvar [18], who imposed different circumferential constraining forces onto the concrete surface of the pullout specimens. If there was no apparent cracking to the concrete with the GFRP bar pulled out, the different circumferential constraint force did not bear influence upon the bond stress-slip process of the GFRP bar. The same phenomenon was observable in the pullout testing of the R1 and R3 bar in concrete C2, hence Figure 9 only lists τ-s curves of R1 and R3 bars in concrete C2 without stirrups to the work by Malvar [18], who imposed different circumferential constraining forces onto the concrete surface of the pullout specimens. If there was no apparent cracking to the concrete with the GFRP bar pulled out, the different circumferential constraint force did not bear influence upon the bond stress-slip process of the GFRP bar. The same phenomenon was observable in the pullout testing of the R1 and R3 bar in concrete C2, hence Figure 9 only lists τ-s curves of R1 and R3 bars in concrete C2 without stirrups

Bond Stress-Slip Relationship Model Between GFRP Bar and Stirrup-Confined Concrete
The cracking behavior of RC structures under tension is mainly governed by the relative slip relationship between the bar and the concrete. Moreover, most of structural problems are dealt with at the serviceability limit state. Thus, a refined modeling of the τ-s curve is only needed for the ascending branch (i.e., for slip less than s m ) [21]. In previous research, primarily the Malvar model [18], BPE, and CMR models [19,20] are used for the analysis of the bond behavior of FRP bars. When compared with the Malvar model, the BPE model and CMR models are simple and convenient for application [44], thus these two models will be the primary focus in this paper. Equation (2) describes the ascending branch of of the BPE model and the same branch of the CMR model by Equation (3).
where α is the parameter for the BPE model and β s r are the parameters for CMR model, which can be determined from curve fitting of test results.
Many studies indicate that the parameters that were outlined in the BPE and CMR models, as obtained from curve fitting based on experimental data, are affected by the surface characteristics of GFRP bars. Table 4 shows the values of BPE and CMR models parameters that are based on surface characteristics. The values reported in Table 4 also confirm that the trend of the τ-s constitutive law is strongly dependent on the surface characteristics of GFRP bars.
At present, no scholar in this field is yet to study the τ-s model between the GFRP bar and the stirrup-confined concrete. With the results of the pullout tests that were obtained from specimens with C1 strength, curve fitting has been performed utilizing the BPE and CMR models. Figure 14 shows the distribution of fitting parameters α β s r based on different surface characteristics. The high variation in Figure 14 Figure 16. The analytical curves that were obtained by using the parameters fitted to its experimental counterpart are represented by BPE and CMR models, and the analytical curves obtained by using the medians of parameter are represented by BPE* and CMR* Models. It is clear that the curves of the BPE and CMR models fitted to their experimental counterpart are in good agreement with the experimental curve, whereas the curve that was obtained via the obtained medians indicates slightly lower bond stress than the experimental results for the BPE model, and a slightly higher bond stress for CMR model, as shown Figure 16. This is indicative that the BPE model and CMR model can be used to investigated the τ-s process of GFRP bar and stirrups-confined concrete, and furthermore that the obtained medians α β sr are able to serve as the proposed values for engineering design. Certainly, without experimental data, it is necessary to have methods of theoretical derivation for the calculation of bond strength and the corresponding slip, in order to obtain the complete τ-s relationship curve between the GFRP bar and stirrups-confined concrete.   Figure 15 shows the box plots for parameters α β s r fitted by stirrup-containing specimens test results with regard to specific surface characteristics. Medians of α β s r were obtained as 0.2205, 0.3468, and 0.775 for R1bars, 0.2501, 0.445, 0.715 for R2 bars, 0.18, 0.635, 0.72 for R3 bars. Comparisons between the experimental and analytical τ-s curves are shown in Figure 16. The analytical curves that were obtained by using the parameters fitted to its experimental counterpart are represented by BPE and CMR models, and the analytical curves obtained by using the medians of parameter are represented by BPE* and CMR* Models. It is clear that the curves of the BPE and CMR models fitted to their experimental counterpart are in good agreement with the experimental curve, whereas the curve that was obtained via the obtained medians indicates slightly lower bond stress than the experimental results for the BPE model, and a slightly higher bond stress for CMR model, as shown Figure 16. This is indicative that the BPE model and CMR model can be used to investigated the τ-s process of GFRP bar and stirrups-confined concrete, and furthermore that the obtained medians α β s r are able to serve as the proposed values for engineering design. Certainly, without experimental data, it is necessary to have methods of theoretical derivation for the calculation of bond strength and the corresponding slip, in order to obtain the complete τ-s relationship curve between the GFRP bar and stirrups-confined concrete.

Conclusions
A total of 100 pullout tests were carried out in this paper in order to examine the bond behavior of the GFRP bars in concrete. Based on the results of this experimental and analytical study, the following conclusions may be drawn: (1) The surface characteristics of the bar influence the bond behavior of GFRP bars. In the tests, the bond stress collapse of GFRP bars with surface of helical wrapping (R1) and helical wrapping and sand coating (R2) was due to the detachment of fiber spirals, the resin layer, and sand grains; the bond strength of these GFRP bars was governed by the fiction between the GFRP bar and the concrete. However, the bond stress collapse that is evident in the GFRP bars with a ribbed surface (R3) was due to the crushing of concrete and the shearing off of ribs. The bond strength of these GFRP bars was governed by the interlocking interaction between the GFRP bar and the concrete. Therefore, the R3 bars exhibit higher bond strength than that of both the R1 and R3 bars. The average bond strength of R3 bars is 1.3-1.5 times that of R1 bars and 1.2-1.36 times that of R2 bars. However, it is notable that the bond strength of the specimens that displayed concrete splitting failure is not affected by the surface characteristics of the GFRP bar.
(2) The test results confirmed the notion that larger bar diameters develop lower bond strength. This tendency is not influenced by the surface characteristics of the GFRP bar, concrete strength, or the confined effect of stirrups.
(3) The improvement of concrete strength does not necessarily render an increase in the bond strength of the GFRP bars. When bond failure interface only occurs on the surface of a GFRP bar, the bond strength of a GFRP bar is not dependent upon the concrete strength. However, when a bond failure interface coexists on the surface of a GFRP bar and the concrete, an enhancement of concrete strength is able to effectively increase the bond strength of a GFRP bar. As a result, in the design of a GFRP-RC structure, an appropriate concrete strength should be selected according to the surface of the GFRP bar.
(4) Stirrups provide confinement to concrete, which in turn changes the failure modes and the τ-s relationship of the pullout specimens. If concrete splitting failure occurs in specimens without stirrups, their stirrup-containing counterparts have a tendency to exhibit pullout failure, leading to an increase in the bond strength and the corresponding slip observable in the specimens. The level of improvement ranges depending on the amount of increase of the diameter of the GFRP bar, and the bond stress between the GFRP bar and the concrete. As opposed to the beam test, the concrete was not subjected to bending stress during the pullout testing. No sign of splitting cracks was evident on these specimens due to the relatively small diameter of the bars in comparison to the dimensions of the specimens or the relatively high concrete strength. The bond behavior of these specimens was almost completely unaffected by the presence of stirrups. As a result, further research into the subject of the beam testing of GFRP bars should be taken into consideration in the presence of stirrups.

Conclusions
A total of 100 pullout tests were carried out in this paper in order to examine the bond behavior of the GFRP bars in concrete. Based on the results of this experimental and analytical study, the following conclusions may be drawn: (1) The surface characteristics of the bar influence the bond behavior of GFRP bars. In the tests, the bond stress collapse of GFRP bars with surface of helical wrapping (R1) and helical wrapping and sand coating (R2) was due to the detachment of fiber spirals, the resin layer, and sand grains; the bond strength of these GFRP bars was governed by the fiction between the GFRP bar and the concrete. However, the bond stress collapse that is evident in the GFRP bars with a ribbed surface (R3) was due to the crushing of concrete and the shearing off of ribs. The bond strength of these GFRP bars was governed by the interlocking interaction between the GFRP bar and the concrete. Therefore, the R3 bars exhibit higher bond strength than that of both the R1 and R3 bars. The average bond strength of R3 bars is 1.3-1.5 times that of R1 bars and 1.2-1.36 times that of R2 bars. However, it is notable that the bond strength of the specimens that displayed concrete splitting failure is not affected by the surface characteristics of the GFRP bar.
(2) The test results confirmed the notion that larger bar diameters develop lower bond strength. This tendency is not influenced by the surface characteristics of the GFRP bar, concrete strength, or the confined effect of stirrups.
(3) The improvement of concrete strength does not necessarily render an increase in the bond strength of the GFRP bars. When bond failure interface only occurs on the surface of a GFRP bar, the bond strength of a GFRP bar is not dependent upon the concrete strength. However, when a bond failure interface coexists on the surface of a GFRP bar and the concrete, an enhancement of concrete strength is able to effectively increase the bond strength of a GFRP bar. As a result, in the design of a GFRP-RC structure, an appropriate concrete strength should be selected according to the surface of the GFRP bar.
(4) Stirrups provide confinement to concrete, which in turn changes the failure modes and the τ-s relationship of the pullout specimens. If concrete splitting failure occurs in specimens without stirrups, their stirrup-containing counterparts have a tendency to exhibit pullout failure, leading to an increase in the bond strength and the corresponding slip observable in the specimens. The level of improvement ranges depending on the amount of increase of the diameter of the GFRP bar, and the bond stress between the GFRP bar and the concrete. As opposed to the beam test, the concrete was not subjected to bending stress during the pullout testing. No sign of splitting cracks was evident on these specimens due to the relatively small diameter of the bars in comparison to the dimensions of the specimens or the relatively high concrete strength. The bond behavior of these specimens was almost completely unaffected by the presence of stirrups. As a result, further research into the subject of the beam testing of GFRP bars should be taken into consideration in the presence of stirrups.
(5) The BPE model and the CMR model have relatively simple form and reliable results that can be applied to investigate τ-s relationship of GFRP bars to stirrups-confined concrete. The fitting parameters, α β s r , specified in these two models, were generated based on the test results of this paper, and the suggested parameter values that were classified based on surface characteristics were proposed. However, there are no universal analytical models that can be applicable to the general τ-s relationship of GFRP bar to stirrups-confined concrete. As a result, the universal analytical τ-s model of GFRP bar to stirrups-confined concrete should be further studied based on more test data.
Author Contributions: J.T. and K.G. conceived and designed the experiments; K.G., Z.L., J.Z. performed the experiments; X.L. is in charge of project administration; All authors participated in data analysis and paper writing.