Experimental and theoretical study of bond behaviour between FRP bar and high-volume fly ash-self-compacting concrete

The combination of fibre-reinforced polymer (FRP) and high-volume fly ash self-compacting concrete (HVFA-SCC) reinforced is expected to solve the problem of steel corrosion in traditional structures and develop sustainable infrastructures. Bond behaviour has a strong effect on serviceability of FRP reinforced concrete structures. To achieve the acceptance of this novel composite structures in practical construction and design, it is significantly important to investigate the bond behaviour of FRP reinforced HVFA-SCC. In this study, a series of pull-out tests were carried out to investigate the bond behaviour of HVFA-SCC reinforced by FRP bar, which included bond strength, bond-slip response, and failure mode. The investigated experimental variables were reinforcing materials, surface treatment of FRP bars, bar diameters and concrete materials. The test results revealed that using HVFA-SCC resulted in higher average bond strength compared to those in the normal concrete test specimens. The reinforcing materials (steel vs. GFRP) had strong effect on bond behaviour, including bond strength, bond stiffness and failure mode. The failure mechanism of bond interaction between FRP and HVFA-SCC is dependent on friction and chemical adhesion. Subsequently, theoretical models for bond strength and development length were proposed. Finally, the parameters of analytical models of bond-slip curve are calibrated for GFRP reinforced HVFA-SCC by using the experimental data.


Introduction
The service life of steel reinforced concrete structures is reduced by the corrosion of steel bars, especially in marine or near aggressive environments [1]. In addition, a large amount of energy is consumed induced by the cement production. In fact, 5-7% of global carbon dioxide emissions comes from cement production [2]. Therefore, structural performance degradation of concrete infrastructure and the development of high-performance concrete materials with low carbon footprint are the major challenges. Many engineers and researchers have proposed some viable alternative construction materials. For the steel-corrosion problems, fibre reinforced polymer (FRP) can be used as internal reinforcing materials instead of steel reinforcement, which can also improves the durability of those structures [3]. Additionally, by using post-waste materials, such as fly ash ground, limestone powder and granulated blast furnace slag, self-compacting concrete (SCC) has the characteristics of high-performance, low-energy consumption, high flowability and convenient construction [4,5]. It has been reported that the SCC mixed with highvolume fly ash (50% or more) can produce an environmentally friendly concrete with high flowability and durability [6]. Therefore, these two engineering materials, high-volume fly ash-SCC (HVFA-SCC) and FRP, have attracted wide attention in the structural constructions [7]. Due to those advantages of FRP reinforcement and HVFA-SCC, combining the two materials can offer a promising technology to solve two sustainability problems (steel corrosion and high carbon footprint) [8]. FRP reinforced concrete structures have larger deflection and crack width than steel reinforced concrete structures [8,9]. The service behaviour, including deflection and crack width, is affected by the bond behaviour of FRP and concrete. Also, the research on the structural behaviour of HVFA-SCC reinforced by FRP bars is rather limited. To achieve widespread acceptance of the combination of the two materials in practical constructions, it is necessary to investigate the bond behaviour between FRP and HVFA-SCC for ensuring a good loadtransferring mechanism.
The bond behaviour of FRP reinforced concrete is inferior to that of steel reinforced concrete structures [10][11][12]. This is attributed to different surface treatments and the low axial stiffness of FRP bars. The influence of the embedment length, rebar type and rebar diameter on the bond behaviour of FRP reinforced normal concrete (NC) has been studied widely [13]. The compositions of HVFA-SCC are different from that of NC, which could result in the various internal microstructure and mechanical properties between FRP reinforced HVFA-SCC specimen and FRP reinforced NC specimen [14,15]. It has been reported that the concrete compositions, including powder types and content, aggregate types and contents, and admixture, have a significant effect on the bond properties [16,17]. This means that the FRP/ steel reinforced HVFA-SCC, SCC and NC have different bond behaviour. Existing studies shows that using SCC to replace normal concrete (NC) can result in higher or equal bond behaviour, which is due to its higher compactness and filling properties [18]. In addition, the lower variability of bond stress is presented in SCC rather than NC [19]. Arezoumandi et al. [20] found that the similar bond strength is presented in the steel reinforced HVFA-SCC and the steel reinforced NC. Martí-Vargas et al. [21] indicated that the transmission length is affected by the compositions of concrete. Currently, bond behaviour of GFRP reinforced HVFA-SCC is rarely discussed in the existing literature. Therefore, it is of significantly important to fully study the bond-slip behaviour and bond failure process between GFRP bars and HVFA-SCC for establishing the prediction model of bond strength and development length. This is beneficial for developing further study on the serviceability behaviour and establishing the design guidelines of FRP reinforced HVFA-SCC structures.
In this paper, a series of direct pull-out tests were conducted to investigate the bond-slip behaviour of FRP and HVFA-SCC. The structural variables adopted in this study include reinforcing materials, geometry and surface treatments of GFRP bars, GFRP bar diameters and concrete materials. Also, the test results are used to discuss the influence of those parameters on the bond strength, failure mechanism and bond-slip relationship of FRP reinforced HVFA-SCC specimens. The results of interface damage can be verified by the test results of piezoelectric ceramic sensors given in Reference [22]. Considering the influence of filling performance of HVFA-SCC on bond properties, the theoretical models of bond strength and development length of GFRP reinforced HVFA-SCC are proposed. Finally, the parameters of analytical models of FRP reinforced HVFA-SCC bond slip curves are calibrated by using test data. The Ttst results indicate that the average bond strength of GFRP reinforced HVFA-SCC specimens is higher than that of GFRP reinforced NC specimens, especially in specimen of GFRP with a lower spiral height.

Experimental program
The properties of the adopted raw materials and the test procedures conducted to investigate the bond-slip behaviour of FRP reinforced HVFA-SCC are discussed in this section.

Materials
Three reinforcing bars (GFRP, BFRP and steel) provided by three manufacturer companies were used in this test. The surface treatments of these bars are presented in Fig. 1. The GFPR and BFRP bars had spiral wound surfaces. In addition, the surface of the GFRP bars is coated with silica sand, which helps improve the bond behaviour by enhancing friction and interlock forces [23]. To study the effects of spiral height on bond behaviour, two types of GFRP bars with different spiral heights were adopted in this test. Additionally, the ribbed steel bars were employed for comparison purposes. The test results of the mechanical properties of reinforcing bars are shown in Table 1.
The binder materials for the preparation of HVFA-SCC were composite Portland cement, fly ash and limestone powder. Crushed granite with a particle size of 4.75-16 mm as coarse aggregate and river sand with a fineness modulus of 2.39 as fine aggregate were used in this test, respectively. BASF-F10 superplasticizer was used to improve the fluidity of HVFA-SCC. The compositions of HVFA-SCC and NC are summarized in Table 2. In the production of the HVFA-SCC mixtures, controlling the mixing sequence and time is essential for the workability of HVFA-SCC.
The one-time mixing method was adopted to produce HVFA-SCC. The batching sequence consisted of homogenising all the mixture except water for 120 s in a single horizontal-axis forced mixer. Then, 50% of the water was added into the mixer and continue to mix for 60 s at the same speed. Finally, the remaining water was added and the concrete was mixed for a further 120 s. The slump flow diameter (the slump for NC), the J-ring diameter, the U-Box height difference and the V-funnel time were tested to determine the workability of fresh HVFA-SCC. The test procedure is shown in Fig. 2 and the test results are listed in Table 3.
In this test, three cubes (150 9 150 9 150 mm) and three cylinders (300 9 150 mm) were respectively poured for both HVFA-SCC and NC to determine its compressive strength and elasticity modulus. The test specimens were demoulded after 24 h of concrete pouring and then were cured for another 55 days under the condition of 20 ± 2°C and 95% humidity. The mechanical properties of hardened concrete are listed in Table 3.

Preparation of pull-out test specimens
One end of the FRP bar is enclosed by a steel tube and the inside is filled with epoxy adhesive, which aims to prevent FRP bas from failing due the gripping forces applied by the clamps of the hydraulic machine during     testing. After curing the epoxy, the FRP and steel bars were embedded in the centre of the concrete sample, which had dimensions of 200 9 200 9 200 mm. The bond lengths were determined using five times of the diameters of the reinforcing bars (l = 5d) and can be achieved by using two PVC pipes in the unbonded areas of the reinforcing bars. To control the expansion of wood moulds in concrete casting, metal moulds were used for all the test specimens. All the test specimens were removed from the steel moulds a week after casting.

Test setup
The test method for a direct pull-out test reported in ACI 440.3 R-12 standard [24] was adopted in this test to investigate the bond behaviour of FRP reinforced HVFA-SCC. The pull-out loading test setup and data acquisition system are presented in Fig. 3. The pullout tests were carried out by using a microcomputercontrolled electro-hydraulic servo universal testing machine with a capacity of 1000 kN. The displacement loading control at a rate of 0.02 mm/s was adopted in all direct pull-out test specimens, which aims to capture the post-failure behaviour in the bondslip relationships. As shown in Fig. 3a, four linear variable differential transformers (LVDTs) were symmetrically placed at the loaded and unloaded end of the test specimens for measuring the slip values during the loading test. Those LVDTs were connected to TDS-530 data acquisition instrument for obtaining these slip values of loaded and unloaded end. The applied load was recorded by the computer connected to test machine. During loading, these slip values and applied load were continuously collected synchronously at a rate of one acquisition per second. It is worth noting that the slip values at the loading end are not discussed in the following section and the slip values of unloaded end were used as the overall slips. In addition, the configuration for the piezoelectric test is shown in Fig. 3c. The piezoelectric signals were recorded synchronously in the pull-out test and the piezoelectric test results had been discussed in detail in Reference [22]. In this paper, the piezoelectric test results are used to verify the results of interface damage. In this test, a total of 24 test specimens were tested up to failure.

Test result and discussion
In this test, the influence of concrete materials, reinforcing materials, FRP bar surface treatment and FRP bar diameter is discussed. In the pull-out test, it is assumed that the bond stress of reinforcing bars and concrete is evenly distributed along the bond length. The average bond stress, s, is defined as: where P is the ultimate tensile load, d is the bar diameter and l is the bond length. The relationship between the average bond stress determined by Eq. (1) The experimental results obtained from the pull-out tests are listed in Table 4. It is worth mentioning that the failure mode of pull-out was found in all test specimens since sufficient transverse restraint was provided by the relatively thick concrete cover.

Bond strength
The failure load and bond strength of all the test specimens are summarised in Table 4. The recorded average bond strength for the test specimens of steel reinforced HVFA-SCC (S-13-50%-series) is about two times higher than that of FRP reinforced HVFA-SCC test specimens (G-13-50%-series and B-12-50%series). This contributed to the higher axial stiffness of steel reinforcement compared to FRP bars and steelribbed surface treatment. It could introduce different failure modes in the steel and FRP reinforced HVFA-SCC specimens, which will be discussed in the next section. Fly ash and limestone powder, as the substitutes for cement cementitious material, were added into HVFA-SCC, which can improve the compactness property of HVFA-SCC [25,26]. As a result, using HVFA-SCC to replace NC resulted in higher bond strength in this test due to the denser internal structure and stronger chemical adhesion. Interestingly, spiral height has a slight effect on the bond strength of the test specimens of HVFA-SCC, while bond strength of the test specimens of NC is increased by about 80% when the spiral height of GFRP bars increases from 0.21 to 0.51 mm. This mean that the contribution of spiral height to bond strength is reduced in HVFA-SCC due to its high interfacial compactness. In addition, increasing fly ash volume in HVFA-SCC has barely effect on the peak bond strength. As reported in the literature [27][28][29], increasing the diameter of the GFRP bars results in a decrease in the average bond strength. This phenomenon can be explained for by the following reasons: (a) An uneven distribution of normal stresses through the cross section of the bar is presented in the pull-out of FRP bar, which results in a lower average bond strength; (b) Longer embedment length is required to develop the same normal bond stress for larger diameter bars, and the nonlinear stress distribution along the direction of the embedment length reduces the average bond strength; (c) The contact interface between concrete and FRP bar is increased by increasing the diameter of the reinforcing bar, and a large amount of water in concrete is absorbed and retained at the interface, reducing the chemical adhesion and frictional stresses. Thus, the average bond strength is decreased by increasing the diameter of the GFRP bars.

Failure mechanism
The common failure mode observed in all the test specimens is the pull-out failure of GFRP bars in HVFA-SCC. This is due to the fact that the short development length for the bonding test cannot result in wider longitudinal cracks along the concrete prism. Also, the test specimens were split to observe the interface damage and analyse the failure mechanisms of FRP reinforced HVFA-SCC, as shown in Fig. 4. It should be noted that the reinforcing materials had a strong effect on the failure modes of the bond test specimens, as shown in Fig. 4a-c. Although the bond strength of HVFA-SCC specimens reinforced with GFRP and BFRP is similar, the surface abrasion damage of BFRP bars is more significant than GFRP bars (see Fig. 4a, b). This occurred due to the surface treatment of the BFRP bar which is formed by winding a spiral basalt fibre bundle. On account of the low elasticity modulus and weak lateral shear resistance of the fibre bundles, they were abrasion by concrete during the pull-out test. Additionally, using the sandcoated surface in the GFRP bars resulted in delamination of the resin-rich outer layer from the fibre core and damage to the concrete attached to the bar surface, as shown in Fig. 4a. These test results indicate that debonding failure of the GFRP and BFRP reinforced test specimens is dependent on friction and chemical adhesion. On the contrary, it can be found from Fig. 4c that the surface of steel bars is almost intact and some concrete blocks are inserted between the steel ribs. This means that the concrete is tore between two adjacent steel ribs in the steel reinforced HVFA-SCC specimens due to the low rib spacing-to-height ratios and enough concrete cover surrounding the steel rebar. Similar phenomena have also been observed in the references [30,31]. It can be summarised that the steel reinforced HVFA-SCC specimens failed in the mechanical interlocking between the ribs and concrete paste and the bond strength is determined by shear strength of the surrounding concrete.
The effect of the surface geometry of GFRP bars on the bond performance has been introduced in Gu's study [32]. As the standards and quality of FRP bars produced by different manufacturers are not uniform, the surface treatment of various FRP bars is still not accurately quantified. This could make the specification of FRP bars difficult and hinder the practical application of FRP bars in new structures. The failure mode shown in Fig. 4a, d reveals that the surface damage of the GFRP bar with a spiral height of 0.51 mm is more obvious. This also indicates that increasing the spiral height of GFRP bars results in stronger mechanical interaction between reinforcing bars and HVFA-SCC.
The influence of varying the surrounding concrete matrix materials (HVFA-SCC vs. NC) on failure mechanism in bond tests is shown in Fig. 4a, e. The surface of GFRP reinforcing bar is damaged by the delamination of the resin-rich outer in the pull-out test for the HVFA-SCC test specimens, while the surface of the GFRP bar in NC test specimens is almost intact and some residual concrete is attached to the reinforcement. This is due to the fact that the compactness of HVFA-SCC is better than the NC due to the filling effect of fly ash and limestone powder in HVFA-SCC mixture [26,33]. It can be inferred from the interface damage of reinforcement that the stronger adhesive and friction are presented in GFRP reinforced HVFA-SCC specimens rather than NC specimens. This results in the improvement of bond strength of GFRP reinforced HVFA-SCC specimens, which can be proved by the test result in Table 4.
As reported in the literature [22], an active sensing technology using a pair of Smart Aggregates (SA) is adopted to identify the failure mechanism of steel/FRP reinforced concrete in pull-out process. It is found that the failure of steel reinforced HVFA-SCC specimen depends on the mechanical interlocking damage, and the failure of FRP reinforced HVFA-SCC is mainly the chemical adhesion and friction damage. This is consistent with the interface damage presented in Fig. 4a-c.

Bond-slip relationship
The bond-slip behaviours of the HVFA-SCC test specimens reinforced with different bars are shown in Fig. 5a. It can be seen that the stiffness of bond-slip behaviour in steel reinforced HVFA-SCC test specimens is higher than that in GFRP and BFRP reinforced HVFA-SCC test specimens. This is owing to the high elastic modulus of steel bar and the high mechanical interlock resistance between steel bar and concrete. The similar bond behaviour is presented in GFRP and BFRP reinforced HVFA-SCC test specimens, which corresponds to the similar stiffness of the two reinforcing materials. It can be found that the slip value of the test specimens reinforced with GFRP is smaller than that of BFRP reinforced specimens. This phenomenon is due to the various surface treatments of the FRP bars. As shown Fig. 1, the surface of the GFRP bar is spiral wound with a sand coating, and the BFRP bar has a spiral wound and a large spiral spacing. In addition, the bond-slip curve of the BFRP reinforced test specimen has obvious periodic fluctuation, which is consistent with the spiral spacing of the BFRP bar. When the slip value is equal to the spiral spacing of the BFRP bar, a new ''spiral wound'' enters the bond length and re-provides the mechanical interaction and friction in the pull-out test. A similar periodic fluctuation behaviour has also been reported in CEB-FIP Model Code 90 [34] and Baena's study [27]. The spiral spacing of the GFRP bar is larger than the measuring range of the pull-out test, therefore similar periodic fluctuation behaviour cannot be exhibited in the bond-slip curve. For the purpose of comparison, the pull-out test of steel bars is also presented. Figure 5a shows that the corresponding slip value is relatively small when the bond strength is reached. The ribs on the surface of the steel bars provide a reliable mechanical interaction between steel bars and the HVFA-SCC, so that the high initial stiffness is presented in bond-slip curve. A fast bond degradation behaviour is exhibited in the descending branch of the steel bar reinforced test specimen, which  is caused by the shear damage of the concrete between the ribs during the pull-out process.
The influence of spiral height of GFRP bars on bond performance is shown in Fig. 5b. In this study, GFRP bars of the same diameter (13 mm) were used to study the effect of the surface treatments of reinforcing bars on the bond performance. It was found that the bond strength of the two spiral heights is similar but the spiral height has a significant influence on the unloaded end slip value when the bond strength is reached. The specimen reinforced by GFRP bar with large spiral leads to the increase of pores at the interface, which means that the chemical bond and friction effect are decreased at the interface of FRP bars and concrete. In the early stages of the pull-out process, the chemical bond force is the key factor to resist pull-out. Therefore, the large unloaded end slip is presented in the specimen reinforced by GFRP bar with higher spiral. The specimen using GFRP bar with larger spiral height can provide higher mechanical interaction during the pull-out process. However, the chemical bond and friction play an important role to resist pull-out force when using GFRP bars with smaller spiral height, which is due to the denser interface of FRP bars and concrete. Therefore, the lower spiral height of the GFRP bar in specimen leads to a small unloaded end slip value. Figure 5c shows the effect of different concrete materials on the bond-slip behaviours. The variation of concrete materials does not have a significant effect on the unloaded end slip corresponding to bond failure. However, the effect of different concrete materials on the bonding degradation process is significant. The post-peak bond stress of the NC and SCC-50% series is maintained at a high level, but the post-peak bond stress of the HVFA-SCC-65% series decreases rapidly. The mechanical interlocking is negligible due to the small spiral height (0.21 mm) of GFRP bars, and the chemical adhesion is failed after the bond strength is reached. This implies that the bond stress after bond strength mainly depends on the friction between concrete and FRP reinforcement. Therefore, the compressive strengths of the concrete materials (52.1 MPa for NC, 45.4 MPa for HVFA-SCC-50%, 21.5 MPa for HVFA-SCC-65%) and the interface porosity caused by different concrete compositions are the key factors of affecting the degradation of bond slip curve after post-peak bond stress [16]. As shown in Fig. 5c, larger post-peak bond stress can be provided by the concrete with higher compressive strength, which indicates that the high compressive strength could result in the slow degradation of bond behaviour.

Prediction models for bond strength and development length
The stress transfer and collaborative deformation of FRP reinforced concrete structures depend on the bond stress. The bond strength is a key factor affecting the cracking behaviour, tension stiffening, plastic hinge rotation capacity and failure mechanism of concrete members reinforced with FRP bars. Also, it is responsible for the ultimate strength of the end anchorage joint components. To determine the appropriate development length to ensure the safety of FRP reinforced concrete structures, it is important to accurately predict bond strength by using an appropriate model. Table 5 summarises some of the models to evaluate the bond strength of FRP bars and ordinary concrete, which mainly consider factors such as compressive strength of the concrete, thickness of the concrete protective layer, diameter of the reinforcing bar and bond length, etc. The predicted values of bond strength are calculated and compared with the test results obtained from the pull-out test, and the comparison results are listed in Table 6. The bond strength of the test specimens GFRP with a diameter of 19 mm reinforced SCC and GFRP with a spiral height of 0.21 mm reinforced NC is significantly overestimated by all formulas in Table 5. This suggests that the effect of the diameter of the FRP bar on bond strength has not been accurately reflected in these models. Also, the effect of spiral height of FRP bars on the bond strength of FRP reinforced NC should be considered. For the other specimens, the bond strength is overestimated by Eqs. (5) and (6), and underestimated by Eqs. (3), (7) and (8). The calculated results of Eqs. (2) and (4) mostly overestimate bond strength except test specimens of G-13-65%-21.5. This means that the effect of compressive strength on bond strength of FRP reinforced HVFA-SCC is magnified in these formulas. Therefore, it is particularly important to put forward the prediction formula of bond strength for FRP reinforced HVFA-SCC due to the variance in damage mechanism between FRP bars reinforced HVFA-SCC and reinforced NC in the process of pulling-out.
As previously reported in the literature [41] that the SCC can significantly improve the bond strength of FRP reinforced SCC members due to high cementitious materials and high density. Mousavi et al. [41] proposed a prediction formula, as shown in Eq. (9), for bond strength of GFRP reinforced SCC, which considered the ratio of fine to all amount of aggregate (Fine/All) to reflect the filling ability.
The calculated results of bond strength from formula (9) are given in Table 7. It can be observed that the bond strength calculated by Eq. (9) significantly overestimated the test values since a high sandto-aggregate ratio was used to prepare HVFA-SCC in this test. Besides, the authors think that the ratio of fine-to-total aggregate cannot effectively reflect the filling ability and compactness of interface. However, the slump-flow can quantify the filling ability, and it is an important factor affecting the interfacial bonding of FRP bars and HVFA-SCC. Therefore, a new prediction formula of FRP reinforced HVFA-SCC considering the slump-flow is proposed for the pull-out test specimens with an embedment length of 5 times diameters of FRP bars in this paper, as shown below: where the coefficients a, b and c can be obtained based on curve-fitting the test values, and the SF represents the slump-flow in mm. Based on the least square method, the coefficients can be determined and the proposed prediction formula of FRP reinforced HVFA-SCC can be expressed as follows: The results of the bond strength from the test values and from Eq. (11) are presented in Table 7. It shows that the bond strength values predicted by Eq. (11) are consistent with the test values and the average value of the COV is 0.99. This implies that the proposed predicted model for bond strength of FRP reinforced HVFA-SCC is accurate and applicable.
The development length is required in actual engineering structures to prevent the reinforcing bar from being pulled out. And the minimum development length (l ab ) in the pull-out test indicates that the fracture failure of the reinforcing bar and the pulledout failure occurs simultaneously. As a result, the equilibrium equation can be established as follows: where the f max and A represent the tensile strength and the cross section area of FRP bar, respectively. The minimum development length can be derived from Eq. (12), as shown below: Lee et al. [40] Okelo and Yuan [12] s  Table 6 The comparison of the test values and predicted values of bond strength, using the formulas of FRP reinforced ordinary concrete (Eqs. 2-8) The bond strength in Eq. (11) is substituted into Eq. (13) to obtain the minimum development length of the FRP reinforce HVFA-SCC component, and it can be expressed as follows: ffiffiffi ffi f c p À9:3508 þ 0:4778 c d þ 0:0108 Â SF To ensure a positive bond strength in Eq. (11) and avoid negative development length in Eq. (14), the ratio of the thickness of the concrete protective layer to the diameter of FRP bar should be set within a reasonable range, which can be obtained as follows: where s and s represent the bond stress and the slip, s b and s b s b are bond strength and its corresponding slip obtained from test results; F and G are parameters to be determined by fitting test data. It is noted that the using Malvar model results in a low prediction accuracy for the ascending segment due to the single expression. Also, the effect of bar diameter on the bond strength is ignored in the Malvar model. The modified BPE (mBPE) model was described by Cosenza et al. [45] and is presented in Eq. (17), as follows: where a and p are the curve-fitting parameters based on test data. The effects of type and diameter of FRP bars are not considered in this model. The CMR model suggested by Cosenza et al. [42] can be expressed as follows: where s r and b are the parameters that can be obtained by fitting test data. The slope of the initial stage of the CMR model is infinite, which better matches the physical meaning of chemical adhesion. However, the descending branches of the bond-slip curve are not represented in the CMR model. Equations 16-18 are calibrated based on the test results of FRP reinforced HVFA-SCC test specimens in this study. Considering that the bond behaviour of the reinforcing bar with a low spiral height is easily affected by interfacial pores of concrete materials, those specimens of GFRP with a spiral height of 0.51 mm reinforced concrete are selected to calibrate the parameters of the analytical models. These parameters are determined by using the least square method are listed in Table 8 and the predicted bondslip curves are shown in Fig. 6. It can be observed that the initial stiffness of the bond-slip curves is underestimated, and the tested results of the ascending branches cannot be accurately evaluated by the Malvar model. However, the test results of bond-slip curves are consistent with the predicted results by using the mBPE model. It is worth mentioning that the third equation (Eq. (17c)) of the mBPE model should be used based on the characteristics of the bond-slip curves of FRP reinforced concrete obtained from the experiment. If a stable residual segment, i.e. the phenomenon that the bond stress remains constant as the unloaded end slip continues to increase, was not found in the test results, the third equation (Eq. (17c)) of the mBPE model can be removed to predict the bond-slip behaviour.
In general, the bond-slip behaviour after peak bond stress cannot be predicted by using CMR model. Compared with the Malvar model, the mBPE model composed of piecewise expressions is more accurate in predicting the bond-slip behaviour of FRP reinforced HVFA-SCC and FRP reinforced NC. Furthermore, suggested values of the parameters of the mBPE model are determined and calibrated for FRP reinforced HVFA-SCC specimens based on the obtained experimental data, which is of great significance for the numerical and theoretical studies of FRP reinforced HVFA-SCC members considering the bond-slip constitutive model in the future research.

Conclusions
In this paper, the bond-slip behaviour of FRP reinforced HVFA-SCC has been investigated through a series of pull-out tests. The influence of parameters, such as the type of reinforcing bars, the geometry of the surface of GFRP bars, the bar diameters and the concrete material type, on the bond-slip behaviour is discussed. Based on the test and analysis results, the following conclusions can be drawn: 1. The average bond strength of steel bar reinforced HVFA-SCC is about two times higher than that of GFRP bar and BFRP bar reinforced HVFA-SCC. The use of large diameter FRR bars results in a significant reduction in bond strength. The bond strength can be improved using HVFA-SCC instead of NC reinforced with GFRP bars, especially in GFRP with a lower spiral height. The average bond strength of HVFA-SCC specimens is improved slightly, while the average bond strength of NC specimens is increased by about 80% when the spiral height of GFRP bars is from 0.21 to 0.51 mm. 2. The bond failure of the GFRP and BFRP reinforced test specimens depends on friction and chemical adhesion. While the bonding failure of the steel reinforced test specimens is controlled by mechanical interlocking. The surface damage on the GFRP bar of the HVFA-SCC series is more pronounced than in the NC series. This indicates that stronger chemical adhesion is presented in FRP reinforced HVFA-SCC. In addition, the surface damage of the GFRP bar with a spiral height of 0.51 mm is more severe than that of the 0.21 mm. 3. The bond behaviours, including initial bond stiffness, bond strength and bonding degradation, are significantly affected by the type of reinforcement. Rapid degradation after peak bond stress occurs in GFRP reinforced HVFA-SCC with 65%