Shear Interface Capacity of GFRP-Reinforced Concrete Joints

Highlights

What are the main findings?

• Existing design provisions and analytical models for interface shear transfer (IST) in glass fiber-reinforced polymer reinforced concrete (GFRP-RC) systems showed inconsistent predictions, either underestimating or overestimating the experimental shear transfer capacity across different interface conditions.
• A modified strain-based IST model incorporating a concrete-strength-dependent cohesion parameter and a GFRP strain limit of 0.003 provided accurate, yet conservative, predictions for 107 push-off specimens with different interface conditions.

What are the implications of these findings?

• The proposed model establishes a more reliable and unified framework for evaluating IST resistance in GFRP-RC composite systems, improving consistency and IST predictions across various interface conditions.
• The findings provide a basis for the future development and refinement of rational GFRP-RC IST design provisions, reducing uncertainty in current design approaches and supporting the broader adoption of GFRP reinforcement in durable, corrosion-resistant, and sustainable concrete infrastructure.

Abstract

Interface shear transfer (IST) is a critical mechanism governing composite action in reinforced concrete (RC) structures. While the IST behavior in steel-RC is well established, its application to glass fiber-reinforced polymer (GFRP)-RC remains uncertain due to the scatter of experimental data and the absence of a unified design model. This study assesses the accuracy of current IST design provisions and analytical models for GFRP-RC using a database of 107 push-off tests from the literature, including 56 specimens with an as-cast interface, 20 specimens with an intentionally roughened interface, 26 specimens with a monolithic interface, and five specimens with a smooth interface. Predictions of available models were compared with experimental peak loads. The results show that current provisions in design codes and standards either significantly underestimate or overestimate the IST capacity. The proposed analytical strain-based models in the literature improved predictions but exhibited inconsistencies across different interface conditions. Accordingly, a modified IST model is proposed based on regression analysis, incorporating a cohesion parameter as a function of the concrete strength with a GFRP strain limit of 0.003. The proposed model provides accurate, yet conservative, predictions across different interface conditions.

1. Introduction

Composite concrete construction is an approach that integrates precast and cast-in-place concrete, combining the continuity and performance of monolithic systems with the speed, efficiency, and cost-effectiveness of accelerated construction. This method is widely adopted in reinforced concrete (RC) bridges, where the structural behavior depends on the effective transfer of external loads across concrete-to-concrete joints through a mechanism known as interface shear transfer (IST). A prominent example of IST is the connection between precast concrete girders and cast-in-place bridge decks, which facilitates composite action between the girder and the deck, thereby improving the overall stiffness and strength of the structure [1]. Cold joints commonly occur in composite structures, where fresh concrete is poured on hardened concrete, forming a critical interface at the junction between the two components. These interfaces are inherently vulnerable to interface shear failure when the applied shear stresses exceed the IST capacity of the joint. To achieve a unified response and prevent interface separation under loading, IST reinforcement is commonly provided, ensuring the composite elements act as a single, efficient structural system [2].

Current design provisions for IST in steel-RC members are based on two models: the shear friction model and the cohesion-plus-friction model. The shear friction model, introduced by Birkeland and Birkeland [3] and Mast [4], operates on the principle that IST is entirely due to friction resulting from clamping force, ignoring the aspect of cohesion. This model considers that slip, when it occurs, generates tension in the perpendicular reinforcement, which in turn creates a clamping force across the joining surfaces. This clamping force is passive and requires crack opening to be activated. The model considers that the influences of both adhesion bonding and dowel action are minimal. According to this model, the interface shear strength primarily depends on the tensile strength of the reinforcement; therefore, without reinforcement, the interface cannot resist shear forces. Any compressive stress applied normal to the interface contributes an additional active clamping force that does not depend on crack opening. Differing from the shear friction model, the cohesion-plus-friction model proposed by Hofbeck et al. [5] and Mattock et al. [6] accounts for the cohesive bond between concrete surfaces. In this approach, shear stress is partially transferred through cohesion along the sliding plane, which includes both the adhesive bond between the surfaces and friction generated by their relative sliding, and it depends on the condition of the interface. This model also considers the clamping force from reinforcement crossing the interface and any external compressive force.

The IST across concrete interfaces is governed by the combined action of adhesive bonding, friction from aggregate interlock induced by reinforcement clamping, in addition to any generated friction by the external applied compressive stress normal to the interface, and the dowel action of reinforcement crossing the shear plane [7]. The first mechanism in the IST system is the adhesion bonding between the old and new concrete. Once this bond reaches its capacity, debonding occurs and a crack opens at the interface, shifting load transfer to friction generated by the aggregate interlock. As the shear load is further increased, relative sliding movement between two surfaces is induced, causing crack widening, and tension stresses in the reinforcement crossing the shear plane, which exerts compression (clamping) stresses normal to the interface, reducing the crack opening and providing frictional resistance along the interface to the slip movement. After the development of the full crack along the interface and complete degradation of the aggregate interlock, the IST is governed primarily by the dowel action of the reinforcement until failure. Figure 1 illustrates the contribution of each mechanism to the overall IST strength.

Harries et al. [2] provided a detailed explanation of the IST mechanism, dividing it into three distinct phases based on experimental observations, i.e., the pre-cracking phase, the post-cracking phases, and the post-peak phase, as shown in Figure 2. Initially, in the pre-cracking phase, the load is resisted by the shear strength of the concrete and the adhesion bond, with minimal displacement and negligible reinforcement contribution. As cracks develop in the post-cracking phase, the IST capacity depends on the friction from the aggregate interlock induced by the clamping force of the IST reinforcement crossing the interface. During this phase, both slip and crack width exhibit a relatively linear relationship with the applied shear load. The crack width significantly affects IST by increasing the clamping force in the reinforcement and reducing the cohesion of the shear plane. In the post-peak phase, the displacement, crack width, and reinforcement strain increase after the peak load, with no additional load-carrying capacity.

The utilization of steel reinforcement as IST connectors in composite concrete structures is well established and standardized in design codes and standards such as ACI CODE-318-25 [8], CSA S6-25 [9], CSA A23.3-24 [10], and AASHTO LFRD [11] where the IST strength is determined based on the yield strength of the IST reinforcement crossing the shear interface. However, steel-RC structures are susceptible to corrosion in aggressive environments, undermining both the service life and structural integrity of these structures. Over the past two decades, glass fiber-reinforced polymer (GFRP) reinforcement has been utilized as a viable, non-corrosive alternative to traditional steel reinforcement in concrete structures [12,13,14]. Although GFRP reinforcement provides high tensile strength and enhanced durability, it possesses a lower modulus of elasticity along with anisotropic properties compared to steel reinforcement. Consequently, GFRP-RC interfaces may experience wider shear cracks due to the reduced aggregate interlock, weaker dowel action, and lower clamping stresses compared to those reinforced with steel reinforcement.

The present study aims to review the existing experimental studies on the IST mechanism with GFRP reinforcement. The study evaluates the accuracy of existing GFRP-RC IST design provisions and analytical models through direct comparison with experimental results. Additionally, the study identifies the key limitations in the existing formulations, and their impact on IST predictions. Based on a comprehensive push-off test database, a modified strain-limited regression-based IST model accounting for concrete strength and GFRP strain is proposed to provide consistent and conservative predictions across different interface conditions.

2. Review of Experimental Studies on FRP-RC IST Systems

In the last few years, several studies have explored the use of GFRP as IST reinforcement. Alkatan [15] and Alruwaili [16] explored the use of GFRP as IST reinforcement across interfaces in cold joint IST systems incorporating three different shapes of GFRP reinforcement, i.e., stirrups, headed-end bars, and bent bars (angles). The study emphasized that the stiffness of the reinforcement is a key factor affecting the IST mechanism. A minimum reinforcement stiffness parameter (K_c), calculated as the product of the reinforcement modulus of elasticity (E_f) and the reinforcement ratio (ρ_v) of 203 MPa, was recommended to activate the role of GFRP reinforcement in providing shear frictional resistance after interface cracking. Below this value, the specimens failed suddenly after cracking, with only 9–14% higher interface shear capacity than those without connectors. Alkatan [15] reported that interfaces with a similar K_c of different types of GFRP connectors, i.e., headed bars and stirrups, achieved similar ultimate capacities within a difference of only 8–13%. However, the specimens reinforced with headed-end bars showed a remarkable post-peak performance, compared to those reinforced with stirrups. Additionally, it was observed that the impact of concrete strength on the shear capacity is more pronounced with a higher K_c. A recent study by El Ragaby et al. [17] introduced an IST model for predicting the capacity of IST cold joints reinforced with GFRP, based on the push-off tests conducted by Alkatan [15] and Alruwaili [16].

Al-Kaimakchi and Rambo-Roddenberry [18] investigated the shear performance of two full-scale AASHTO Type II girders reinforced with either GFRP or duplex stainless-steel stirrups. Both girders had similar transverse reinforcement crossing the girder-deck smooth interface. The GFRP-RC girder failed at the girder-deck interface due to the low transverse shear capacity of the GFRP bars, whereas the stainless-steel girder failed in flexural shear. Using GFRP stirrups resulted in approximately an 8% reduction in shear capacity and altered the failure mode, emphasizing the need for specialized design guidelines in GFRP-RC IST systems. Mahmoud et al. [19] examined the IST behavior of six full-scale T-beams with monolithic and intentionally roughened interface conditions, considering different shear interface reinforcement ratios (0.00, 0.32, 0.35 and 0.48%), reinforcement types (steel and GFRP), and GFRP interface reinforcement configurations (stirrups and bent bars). It was found that GFRP reinforcement significantly enhanced load transfer between the web and flange, improving composite action compared to the specimen without shear connectors. For comparable reinforcement ratios, the GFRP-RC specimen achieved about 75% of the shear capacity of its steel-RC counterpart.

Aljada et al. [20,21] studied the performance of GFRP closed stirrups as IST reinforcement in concrete joints under different interface conditions (i.e., as-cast, intentionally roughened, and monolithic) and varying reinforcement ratios (ranging from 0.24 to 0.47%) using two concrete strengths (35 and 70 MPa). The GFRP-RC specimens demonstrated a more brittle failure due to their lower stiffness compared to steel-RC specimens experiencing a sudden load drop with large slips and reinforcement strains after reaching the peak load, whereas steel-RC specimens displayed a gradual load reduction, maintaining structural integrity for longer due to their ductile yielding behavior. The study identified a critical K_c threshold of 176 MPa, below which the GFRP-RC specimens were prone to brittle failure. Vega et al. [22,23] conducted push-off tests on monolithic joints and roughened cold joints reinforced with steel and GFRP stirrups. The results revealed that, although the GFRP stirrups improved the interface shear capacity compared to specimens with no connectors, the steel-RC specimens showed 30% greater shear resistance than the GFRP-RC specimens, even with a 50% increase in GFRP quantity. This was attributed to the higher stiffness of the steel reinforcement, which provided superior resistance to crack opening. Montaser et al. [24] experimentally investigated the IST behavior in GFRP-RC cold joints that were either roughened or smooth. It was reported that, at similar reinforcement ratios, roughened interfaces achieved about 40–70% higher IST capacity than smooth interfaces.

In a recent study [25], the IST behavior of GFRP-RC systems reinforced with headed-end connectors was investigated, considering three interface conditions (i.e., as-cast, intentionally roughened, and monolithic) and two concrete strengths (35 and 55 MPa) at varying reinforcement ratios. The results showed that GFRP headed-end connectors provided a significant contribution to the IST mechanism, with specimens exhibiting the three characteristic phases identified in Figure 2 (i.e., cracking, post-cracking, and post-peak) when adequate reinforcement ratios were used. In contrast, when reinforcement ratios significantly below the CSA S6-25 [9] minimum of 0.45% were employed, the connectors offered little improvement in capacity, and the specimens failed immediately after cracking without any post-cracking load recovery. Interface roughening increased the IST capacity by up to 23% regardless of the reinforcement ratio, while monolithic specimens achieved the highest peak loads due to enhanced adhesion and aggregate interlock. Despite these improvements, the GFRP-RC specimens generally exhibited lower shear capacities and more brittle behavior than their steel-RC counterparts. In particular, specimens with higher concrete strength or roughened interfaces exhibited sharper post-peak load drops due to rapid crack propagation along the irregular interface. However, increasing the reinforcement ratio effectively mitigated the severity of this load drop.

3. Available GFRP-RC Models

Despite recent research efforts on utilizing FRP as IST reinforcement, the IST provisions in codes and standards for FRP-RC structures remain limited, mainly due to a lack and wide scatter of experimental data. North American codes and standards, such as CSA S806-12 [26] and ACI 440.11-22 [27] do not incorporate any provisions or guidelines for IST. On the other hand, while CSA S6-25 [9] and AASHTO LRFD [28] allow the use of FRP as IST reinforcement, they adapt two different approaches in estimating the IST capacity, resulting in extremely different estimations that will be discussed in this section.

3.1. Code Provisions

3.1.1. AASHTO LFRD Model [28]

This model is derived from the IST provisions outlined in the AASHTO LRFD 2024 [11], originally developed for steel-RC IST systems. The model considers the interface cohesion and the friction resistance induced by the reinforcement crossing the shear interface (Equation (1)). However, the steel yield stress (f_y) is replaced by the FRP tensile strength (f_fu), neglecting the significant difference in reinforcement stiffness between both materials:

$$v_r=c+\mu ({\rho }_v{f}_{fu}+\frac{p_c}{A_{cv}}),$$

(1)

$$v_r\le K_1{f}_c^{\prime },$$

(2)

$$v_r\le K_2,$$

(3)

where v_r is the nominal interface shear resistance (MPa), c is the cohesion stress (MPa); A_cv is the shear plane area (mm²); μ is the friction coefficient; P_c is the unfactored permanent load normal to the shear plane (zero in this study); K₁ is the fraction of concrete strength available to resist interface shear, and K₂ is a limiting interface shear resistance factor. Values of c and µ K₁, and K₂, which depend on the interface condition, are given in

Table 1. Different values of c and µ.

Surface Condition	CSA S6-25 [9] a		AASHTO LFRD 2018 [28]
	c (MPa)	µ	c (MPa)	µ	K1	K2 (MPa)
Concrete placed against hardened concrete with the surface clean but not intentionally roughened (as-cast)	0.25 [0.00]	0.60 [0.00]	0.52	0.60	0.20	5.52
Concrete placed against hardened concrete with the surface clean and intentionally roughened to a full amplitude of at least 5 mm (6.4 mm for AASHTO LRFD 2018) and a spacing of about 15 mm	0.50	1.00	1.66	1.00	0.25	10.35
Concrete placed monolithically	1.00	1.40	2.76	1.40	0.25	10.35

Note: 1 MPa = 0.145 ksi. ^a Values between brackets are for the new Case 1 [9].

.

3.1.2. CSA S6-25 Model [9]

The model in Chapter 16 of CSA S6-25 [9] is based on the formulation for steel connectors in steel-RC IST systems. The model has been adapted for FRP reinforcement by replacing the steel yield strength (f_y) with an equivalent FRP stress term of 0.004 E_f, where E_f is the elastic modulus of FRP and 0.004 is a limiting strain value. Exceeding this limiting strain in FRP shear connectors is considered to initiate excessive crack widening at the interface, undermining aggregate interlock.

In the latest edition of the standard, the model was further updated by introducing and evaluating two design cases to account for the possibility of lower reinforcement clamping in the case of not intentionally roughened (as-cast) surfaces, with the lower value governing the design. The first case (Case 1) accounts only for the transverse shear capacity of FRP connectors, disregarding the interface cohesion and friction contribution considering the possibility of completely smooth surfaces. The second case (Case 2) adopts the conventional approach that combines cohesion and clamping friction. The revised model is outlined in Equations (4)–(9). It should be noted that the cohesion factors in this model are much lower than those adopted in the AASHTO LRFD model [28] as listed in Table 1.

$$v_r={\varphi }_c(c+\sigma )\le 0.25{\varphi }_c{f}_c^{\prime }\mathrm{o}\mathrm{r}6.5\mathrm{M}\mathrm{P}\mathrm{a}$$

(4)

where

$$\sigma ={\alpha }_v{\rho }_v{f}_{d}sin{\alpha }_f+{\displaystyle \frac{\mu N}{A_{cv}}}.$$

(5)

$${\rho }_v={\displaystyle \frac{A_{vf}}{A_{cv}}}\ge 0.45\%,$$

(6)

$${\rho }_v{E}_f\ge 225\mathrm{MPa}.$$

(7)

$$\begin{gathered} f_d=\mathit{transverse}\mathit{shear}\mathit{strength}\mathit{of}\mathit{FRP}\mathit{dowel}\mathit{bar}(\mathit{Case}1) \\ {f}_d=0.004E_f\left(Case2\right); \end{gathered}$$

(8)

$${\alpha }_v={\displaystyle \frac{{\varphi }_c}{{\varphi }_f}}=1(Case1),{\alpha }_v=\mu (Case2),$$

(9)

where v_r is the factored shear resistance of the plane (MPa); ϕ_c and ϕ_f are the material resistance factors (taken as 1 in this study); σ is the effective normal stress (MPa); f_d is a factor that accounts for the contribution of the IST reinforcement to the shear plane (taken as the transverse shear strength of the bars in Case 1 and 0.004 E_f in Case 2), MPa; N is the unfactored permanent load normal to the shear plane (zero in this study), and α_f is the angle between the IST reinforcement and the shear interface. Values of c and µ, which depend on the concrete interface condition, that are incorporated in the CSA S6-25 [9] model are given in Table 1.

3.2. Models Proposed in the Literature

3.2.1. Alkatan [15]

Alkatan [15] proposed an IST model (Equations (10) and (11)) based on push-off tests performed on as-cast cold joint specimens. Similar to the previous two models, this one incorporates both interface cohesion and friction generated by the clamping action of the interface shear reinforcement, considering a rough interface with a friction coefficient of 1.0. Interestingly, this is the value typically assigned for intentionally roughened interfaces in design codes and standards and much higher than the value assigned for not intentionally roughened (as-cast) interfaces, i.e., 0.6 as listed in Table 1 [9,28]. The cohesion contribution is expressed as a function of concrete strength (0.04 f′_c). This model considers a higher interface shear reinforcement strain limit of 0.005:

$$v_r=0.04{f^{\prime }}_c+{\epsilon }_f{E}_f{\rho }_{v}sin{\alpha }_f+{\epsilon }_f{E}_f{\rho }_{v}cos{\alpha }_f,$$

(10)

and

$$\left(0.04{f^{\prime }}_c+{\epsilon }_f{E}_f{\rho }_{v}sin{\alpha }_f\right)<0.25{f^{\prime }}_c.$$

(11)

3.2.2. Vega et al. [23]

This model is adapted from the IST provisions in the AASHTO LFRD 2018 model [28] for GFRP-RC IST systems based on regression analysis of the available experimental tests, and with the same cohesion and friction coefficients listed in Table 1. However, this model (Equation (12)) adopts a more realistic approach by imposing a strain limit of 0.002 on the GFRP reinforcement strain, rather than considering the full strength of the GFRP reinforcement:

$$v_r=c+0.002\mu {\rho }_v{E}_f.$$

(12)

3.2.3. Montaser et al. [24]

This model entails two formulae for predicting the IST capacity of GFRP-RC cold joints. The first equation is for intentionally roughened interfaces (Equation (13)) based on the cohesion-plus-friction models proposed by Harries et al. [2] and Alkatan [15]. It adopts a limiting factor of 0.003 for the reinforcement contribution, which corresponds to the average reinforcement strain at peak load observed in the specimens tested in their study and other studies from the literature, multiplied by a friction coefficient of 1.0. In addition, the model adopts the same cohesion factor ($0.04{f^{\prime }}_c$) proposed by Harries et al. [2] for cold joint interfaces:

$$v_r=0.04{f^{\prime }}_c+0.003{\rho }_v{E}_f.$$

(13)

On the other hand, the second formula of the model (Equation (14)) is a friction-based model with no cohesion contribution, introduced for not intentionally roughened interfaces including both as-cast and smooth interfaces. It was determined that the capacity of not intentionally roughened interfaces would be governed by either the shear friction in as-cast specimens (i.e., $0.005{\rho }_v{E}_{FRP}$) or the dowel action in smooth interfaces (i.e., ${\rho }_v{f}_{ft}$), consistent with CSA S6-25 [9], where f_ft is the guaranteed transverse shear strength of the reinforcement. The formula adopts a friction coefficient (μ) of 1.0, not 0.6. Also, it adopts a GFRP strain limit of 0.005, which is higher than the 0.003 limit adopted for intentionally roughened interfaces:

$$v_r=0.005{\rho }_v{E}_f\le {\rho }_v{f}_{ft}.$$

(14)

3.2.4. El Ragaby et al. [17]

This model was proposed based on push-off tests on specimens with not intentionally roughened (as-cast) surface conditions. Unlike the model proposed earlier by Alkatan [15], this model adopts a friction coefficient of 0.48 based on the best-fitting linear analysis of test specimens by Alkatan [15] and Alruwaili [14]. Additionally, the model adopts a strain factor of 0.005 based on experimental GFRP strain measurements at the peak loads resulting in a limiting factor of 0.0024. Furthermore, this model neglects the concrete contribution for more conservative predictions:

$$v_r=0.0024{\rho }_v{E}_f.$$

(15)

4. Database of Experimental Tests on GFRP-RC IST Systems

As mentioned earlier, either push-off or T-beam tests have been used to investigate the IST mechanism in GFRP-RC members. The T-beam tests capture composite action under combined shear and flexural effects, closely simulating bridge deck behavior, whereas push-off tests isolate pure IST under controlled conditions, eliminating flexural effects. The push-off test is a compression-based experiment in which relative displacement develops between two L-shaped blocks. This movement is facilitated by gaps intentionally introduced during specimen fabrication. As relative slip occurs, the reinforcement crossing the interface becomes engaged and contributes to the IST resistance.

A comprehensive database of 107 push-off tests on GFRP-RC specimens reported in the literature was assembled to assess the accuracy of existing IST design provisions and analytical models and to develop a unified regression-based equation. These specimens covered a range of concrete strengths ranging from 30 to 76 MPa, GFRP reinforcement ratios ranging between 0.17 and 1.58% with varying shear plane dimensions, and GFRP modulus of elasticity ranging between 49 and 68 Gpa. The database incorporates 56 specimens with as-cast (AC), 20 specimens with intentionally roughened (IR), 26 specimens with monolithic (M), and five specimens with smooth (S) interfaces, as listed in Table 2, Table 3, Table 4 and Table 5, respectively. In all specimens, the GFRP IST reinforcement was oriented perpendicular to the interface, ensuring consistent activation of the IST mechanism.

Table 2. Database of push-off tests with as-cast (AC) interfaces.

Author	Plane Dimension (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-N	38.0	NA	NA	0.15	NA	308
		Headed-end	GH2-16-N	36.3	0.33	212.8	0.17	420	321
			GH4-13-N	37.3	0.43	294.3	0.46	3290	377
			GH2-19-N	36.1	0.47	295.7	0.42	2695	384
			GH4-16-N	36.8	0.66	425.6	0.43	2340	428
			GH6-16-N	37.4	1.00	644.8	0.56	2200	522
Ahmed [25]	300 × 400	NA	XX-0-H	57.8	NA	NA	0.31	NA	550
		Headed-end	GH2-16-H	54.6	0.33	212.8	0.19	290	530
			GH4-13-H	56.1	0.43	294.3	0.60	4250	652
			GH2-19-H	56.5	0.47	295.7	0.68	3530	715
			GH4-16-H	55.3	0.66	425.6	0.75	3840	769
			GH6-16-H	55.3	1.00	644.8	0.77	3770	885
Aljada et al. [20]	300 × 400	NA	X0-000-N	36.0	NA	NA	0.18	NA	293
		Z-shaped	G1-15Z-N	38.0	0.17	91.8	0.17	50	298
			G2-15Z-N	38.0	0.33	178.2	0.17	190	308
		Stirrups	G4-10C-N	33.0	0.24	117.6	0.20	410	399
			G4-10C-N	33.0	0.24	117.6	0.20	280	387
			G6-10C-N	36.0	0.36	176.4	0.21	440	439
			G4-13C-N	36.0	0.43	227.9	0.20	310	419
			G8-10C-N	33.0	0.47	249.1	0.22	500	451
			G4-10C-H	76.0	0.24	117.6	0.49	3900	668
			G6-10C-H	76.0	0.36	176.4	0.66	3600	890
			G8-10C-H	76.0	0.47	230.3	0.74	5000	897
Alkatan [15]	250 × 500	NA	C0-30	30.0	NA	NA	0.34	NA	332
		Stirrups	FS2-30	30.0	0.41	205.0	0.58	4466	385
			FS3-30	30.0	0.61	305.0	0.64	4847	384
		Headed-end	FH3-30	30.0	0.30	180.0	0.44	1472	362
			FH5-30	30.0	0.51	306.0	0.94	4973	433
		Angles	FA3-30	30.0	0.30	150.0	0.24	100	342
		NA	C0-50	50.0	NA	NA	0.66	NA	296
		Stirrups	FS1-50	50.0	0.20	100.0	0.34	22	334
			FS2-50	50.0	0.20	100.0	0.48	3881	402
			FS3-50	50.0	0.61	305.0	0.37	402	617
		Headed-end	FH2-50	50.0	0.20	120.0	0.31	389	336
			FH3-50	50.0	0.30	180.0	0.28	2260	323
			FH5-50	50.0	0.51	306.0	0.77	2953	569
		Angles	FA2-50	50.0	0.30	150.0	0.15	NA	255
			FA3-50	50.0	0.30	150.0	0.66	4525	540
Alruwaili [16]	250 × 400	NA	C0-A1	35.0	NA	NA	0.41	NA	343
		Stirrups	FS2-A1	35.0	0.51	255.0	0.34	2209	268
			FS3-A1	35.0	0.76	380.0	0.72	5962	490
			FS4-A1	35.0	1.01	505.0	0.11	3976	428
		Headed-end	FH3-A1	35.0	0.38	228.0	1.00	5993	328
			FH4-A1	35.0	0.51	306.0	0.37	2383	293
			FH5-A1	35.0	0.61	366.0	0.67	6988	497
			FH6-A1	35.0	0.76	456.0	0.80	5752	547
			FH8-A1	35.0	1.01	606.0	1.00	5889	550
	250 × 300	NA	C0-A2	35.0	NA	NA	0.28	NA	217
		Stirrups	FS2-A2	35.0	0.68	340.0	0.41	3113	172
			FS3-A2	35.0	1.01	505.0	0.92	8828	275
			FS4-A2	35.0	1.35	675.0	0.94	3619	484
		Headed-end	FH3-A2	35.0	0.51	306.0	0.97	6563	213
			FH4-A2	35.0	0.68	408.0	0.30	2308	218
			FH5-A2	35.0	0.84	504.0	1.00	9010	264
			FH6-A2	35.0	1.01	606.0	0.60	3543	304
			FH8-A2	35.0	1.35	810.0	0.97	6081	397

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v).

Table 2. Database of push-off tests with as-cast (AC) interfaces.

Author	Plane Dimension (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-N	38.0	NA	NA	0.15	NA	308
		Headed-end	GH2-16-N	36.3	0.33	212.8	0.17	420	321
			GH4-13-N	37.3	0.43	294.3	0.46	3290	377
			GH2-19-N	36.1	0.47	295.7	0.42	2695	384
			GH4-16-N	36.8	0.66	425.6	0.43	2340	428
			GH6-16-N	37.4	1.00	644.8	0.56	2200	522
Ahmed [25]	300 × 400	NA	XX-0-H	57.8	NA	NA	0.31	NA	550
		Headed-end	GH2-16-H	54.6	0.33	212.8	0.19	290	530
			GH4-13-H	56.1	0.43	294.3	0.60	4250	652
			GH2-19-H	56.5	0.47	295.7	0.68	3530	715
			GH4-16-H	55.3	0.66	425.6	0.75	3840	769
			GH6-16-H	55.3	1.00	644.8	0.77	3770	885
Aljada et al. [20]	300 × 400	NA	X0-000-N	36.0	NA	NA	0.18	NA	293
		Z-shaped	G1-15Z-N	38.0	0.17	91.8	0.17	50	298
			G2-15Z-N	38.0	0.33	178.2	0.17	190	308
		Stirrups	G4-10C-N	33.0	0.24	117.6	0.20	410	399
			G4-10C-N	33.0	0.24	117.6	0.20	280	387
			G6-10C-N	36.0	0.36	176.4	0.21	440	439
			G4-13C-N	36.0	0.43	227.9	0.20	310	419
			G8-10C-N	33.0	0.47	249.1	0.22	500	451
			G4-10C-H	76.0	0.24	117.6	0.49	3900	668
			G6-10C-H	76.0	0.36	176.4	0.66	3600	890
			G8-10C-H	76.0	0.47	230.3	0.74	5000	897
Alkatan [15]	250 × 500	NA	C0-30	30.0	NA	NA	0.34	NA	332
		Stirrups	FS2-30	30.0	0.41	205.0	0.58	4466	385
			FS3-30	30.0	0.61	305.0	0.64	4847	384
		Headed-end	FH3-30	30.0	0.30	180.0	0.44	1472	362
			FH5-30	30.0	0.51	306.0	0.94	4973	433
		Angles	FA3-30	30.0	0.30	150.0	0.24	100	342
		NA	C0-50	50.0	NA	NA	0.66	NA	296
		Stirrups	FS1-50	50.0	0.20	100.0	0.34	22	334
			FS2-50	50.0	0.20	100.0	0.48	3881	402
			FS3-50	50.0	0.61	305.0	0.37	402	617
		Headed-end	FH2-50	50.0	0.20	120.0	0.31	389	336
			FH3-50	50.0	0.30	180.0	0.28	2260	323
			FH5-50	50.0	0.51	306.0	0.77	2953	569
		Angles	FA2-50	50.0	0.30	150.0	0.15	NA	255
			FA3-50	50.0	0.30	150.0	0.66	4525	540
Alruwaili [16]	250 × 400	NA	C0-A1	35.0	NA	NA	0.41	NA	343
		Stirrups	FS2-A1	35.0	0.51	255.0	0.34	2209	268
			FS3-A1	35.0	0.76	380.0	0.72	5962	490
			FS4-A1	35.0	1.01	505.0	0.11	3976	428
		Headed-end	FH3-A1	35.0	0.38	228.0	1.00	5993	328
			FH4-A1	35.0	0.51	306.0	0.37	2383	293
			FH5-A1	35.0	0.61	366.0	0.67	6988	497
			FH6-A1	35.0	0.76	456.0	0.80	5752	547
			FH8-A1	35.0	1.01	606.0	1.00	5889	550
	250 × 300	NA	C0-A2	35.0	NA	NA	0.28	NA	217
		Stirrups	FS2-A2	35.0	0.68	340.0	0.41	3113	172
			FS3-A2	35.0	1.01	505.0	0.92	8828	275
			FS4-A2	35.0	1.35	675.0	0.94	3619	484
		Headed-end	FH3-A2	35.0	0.51	306.0	0.97	6563	213
			FH4-A2	35.0	0.68	408.0	0.30	2308	218
			FH5-A2	35.0	0.84	504.0	1.00	9010	264
			FH6-A2	35.0	1.01	606.0	0.60	3543	304
			FH8-A2	35.0	1.35	810.0	0.97	6081	397

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v).

Table 3. Database of push-off tests with intentionally roughened (IR) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-IR-N	37.8	NA	NA	0.23	NA	370
		Headed-end	GH2-16-IR	37.8	0.33	212.8	0.18	220	377
			GH4-13-IR	35.2	0.43	294.3	0.46	3380	406
			GH2-19-IR	35.2	0.47	295.7	0.42	2570	392
			GH4-16-IR	34.8	0.66	425.6	0.46	2495	528
			GH6-16-IR	36.4	1.00	644.8	0.64	2600	602
Aljada et al. [21]	300 × 400	Stirrups	G4-10-CR	40.0	0.24	117.6	0.24	410	452
			G4-10-CR	40.0	0.24	117.6	0.26	260	417
			G4-13-CR	35.0	0.43	227.9	0.26	230	438
			G8-10-CR	35.0	0.47	230.3	0.32	3300	423
Vega et al. [22]	165 × 330	NA	N-0-1-J	38.0	NA	NA	0.35	NA	146
			N-0-2-J d	38.0	NA	NA	0.47	NA	174
			N-0-3-J d	38.0	NA	NA	0.50	NA	141
		C-shaped	F-3-1-J	38.0	1.42	883.2	5.49	2772	228
			F-3-2-J d	38.0	1.42	883.2	5.79	3180	252
			F-3-3-J d	38.0	1.42	883.2	5.86	3656	245
Montaser et al. [24] e	150 × 180	NA	R-0-0	57.3	NA	NA	0.22	NA	92
		U-shaped	R-GU-1-10	59.3	0.26	135.2	0.15	184	85
		Stirrups	R-GS-4-10	59.3	1.05	546.0	0.65	2460	166
			R-GS-6-10	59.3	1.58	821.6	0.85	2975	181

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v). ^d Replicated specimen. ^e Only specimens with the first peak load reported by the authors are included.

Table 3. Database of push-off tests with intentionally roughened (IR) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-IR-N	37.8	NA	NA	0.23	NA	370
		Headed-end	GH2-16-IR	37.8	0.33	212.8	0.18	220	377
			GH4-13-IR	35.2	0.43	294.3	0.46	3380	406
			GH2-19-IR	35.2	0.47	295.7	0.42	2570	392
			GH4-16-IR	34.8	0.66	425.6	0.46	2495	528
			GH6-16-IR	36.4	1.00	644.8	0.64	2600	602
Aljada et al. [21]	300 × 400	Stirrups	G4-10-CR	40.0	0.24	117.6	0.24	410	452
			G4-10-CR	40.0	0.24	117.6	0.26	260	417
			G4-13-CR	35.0	0.43	227.9	0.26	230	438
			G8-10-CR	35.0	0.47	230.3	0.32	3300	423
Vega et al. [22]	165 × 330	NA	N-0-1-J	38.0	NA	NA	0.35	NA	146
			N-0-2-J d	38.0	NA	NA	0.47	NA	174
			N-0-3-J d	38.0	NA	NA	0.50	NA	141
		C-shaped	F-3-1-J	38.0	1.42	883.2	5.49	2772	228
			F-3-2-J d	38.0	1.42	883.2	5.79	3180	252
			F-3-3-J d	38.0	1.42	883.2	5.86	3656	245
Montaser et al. [24] e	150 × 180	NA	R-0-0	57.3	NA	NA	0.22	NA	92
		U-shaped	R-GU-1-10	59.3	0.26	135.2	0.15	184	85
		Stirrups	R-GS-4-10	59.3	1.05	546.0	0.65	2460	166
			R-GS-6-10	59.3	1.58	821.6	0.85	2975	181

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v). ^d Replicated specimen. ^e Only specimens with the first peak load reported by the authors are included.

Table 4. Database of push-off tests with monolithic (M) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-M-N	39.9	NA	NA	0.46	NA	626
		Headed-end	GH2-16-M	39.7	0.33	212.8	0.71	4953	700
			GH4-13-M	39.0	0.43	294.3	0.92	4524	724
			GH2-19-M	42.9	0.47	295.7	0.84	3663	775
			GH4-16-M	37.4	0.66	425.6	0.96	3777	837
			GH6-16-M	36.4	1.00	644.8	1.32	5508	1018
Aljada et al. [21]	300 × 400	Stirrups	G4-10-M	46.0	0.24	117.6	0.78	5000	624
			G4-10-M	40.0	0.24	117.6	0.60	3300	611
			G6-10-M	37.0	0.36	176.4	0.82	3940	694
			G4-13-M	34.0	0.43	227.9	0.68	2950	671
			G8-10-M	35.0	0.47	230.3	0.84	4300	719
Vega et al. [22]	165 × 330	NA	N-0-1	38.0	NA	NA	0.48	NA	213
			N-0-2 d	38.0	NA	NA	0.57	NA	250
			N-0-3 d	38.0	NA	NA	0.61	NA	278
			N-0-4 d	38.0	NA	NA	0.52	NA	142
			N-0-5 d	38.0	NA	NA	0.64	NA	234
			N-0-6 d	38.0	NA	NA	0.31	NA	166
		C-shaped	F-1-1	38.0	0.47	292.3	0.81	1670	239
			F-1-2	38.0	0.47	292.3	0.93	1889	236
			F-1-3	38.0	0.47	292.3	1.00	1565	261
			F-3-1	38.0	1.42	883.2	1.63	1347	308
			F-3-2	38.0	1.42	883.2	1.34	1400	324
			F-3-3	38.0	1.42	883.2	1.43	1469	299
			F-3-4	38.0	1.42	883.2	1.59	1697	300
			F-3-5	38.0	1.42	883.2	1.74	1381	350
			F-3-6	38.0	1.42	883.2	1.12	1619	301

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v). ^d Replicated specimen.

Table 4. Database of push-off tests with monolithic (M) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Ahmed [25]	300 × 400	NA	XX-0-M-N	39.9	NA	NA	0.46	NA	626
		Headed-end	GH2-16-M	39.7	0.33	212.8	0.71	4953	700
			GH4-13-M	39.0	0.43	294.3	0.92	4524	724
			GH2-19-M	42.9	0.47	295.7	0.84	3663	775
			GH4-16-M	37.4	0.66	425.6	0.96	3777	837
			GH6-16-M	36.4	1.00	644.8	1.32	5508	1018
Aljada et al. [21]	300 × 400	Stirrups	G4-10-M	46.0	0.24	117.6	0.78	5000	624
			G4-10-M	40.0	0.24	117.6	0.60	3300	611
			G6-10-M	37.0	0.36	176.4	0.82	3940	694
			G4-13-M	34.0	0.43	227.9	0.68	2950	671
			G8-10-M	35.0	0.47	230.3	0.84	4300	719
Vega et al. [22]	165 × 330	NA	N-0-1	38.0	NA	NA	0.48	NA	213
			N-0-2 d	38.0	NA	NA	0.57	NA	250
			N-0-3 d	38.0	NA	NA	0.61	NA	278
			N-0-4 d	38.0	NA	NA	0.52	NA	142
			N-0-5 d	38.0	NA	NA	0.64	NA	234
			N-0-6 d	38.0	NA	NA	0.31	NA	166
		C-shaped	F-1-1	38.0	0.47	292.3	0.81	1670	239
			F-1-2	38.0	0.47	292.3	0.93	1889	236
			F-1-3	38.0	0.47	292.3	1.00	1565	261
			F-3-1	38.0	1.42	883.2	1.63	1347	308
			F-3-2	38.0	1.42	883.2	1.34	1400	324
			F-3-3	38.0	1.42	883.2	1.43	1469	299
			F-3-4	38.0	1.42	883.2	1.59	1697	300
			F-3-5	38.0	1.42	883.2	1.74	1381	350
			F-3-6	38.0	1.42	883.2	1.12	1619	301

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v). ^d Replicated specimen.

Table 5. Database of push-off tests with smooth (S) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Montaser et al. [24]	150 × 180	NA	S-0-0	57.3	NA	NA	0.03	NA	1.5
		U-shape	S-GU-1-10	59.2	0.26	135.2	9.50	8160	49
		Stirrups	S-GS-2-10	59.2	0.53	275.6	8.93	7870	53
			S-GS-4-10	59.2	1.05	546.0	6.04	8943	112
			S-GS-6-10	59.2	1.58	821.6	3.88	4589	127

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v).

Table 5. Database of push-off tests with smooth (S) interfaces.

Author	Plane Dimensions (mm)	Connector Type	Specimen ID	${\mathit{f}}_{\mathit{c}}^{\mathbf{\prime }}$ a (MPa)	IST Reinforcement		Results at Peak Load
					ρv b (%)	K c (MPa)	Slip (mm)	Connector Strain (με)	Load (kN)
Montaser et al. [24]	150 × 180	NA	S-0-0	57.3	NA	NA	0.03	NA	1.5
		U-shape	S-GU-1-10	59.2	0.26	135.2	9.50	8160	49
		Stirrups	S-GS-2-10	59.2	0.53	275.6	8.93	7870	53
			S-GS-4-10	59.2	1.05	546.0	6.04	8943	112
			S-GS-6-10	59.2	1.58	821.6	3.88	4589	127

Note: NA = Not applicable. 1 MPa = 0.145 ksi. ^a Concrete strength. ^b IST reinforcement ratio: ratio of IST reinforcement area to interface area. ^c Reinforcement stiffness factor: product of modulus of elasticity (E_f) of the reinforcement and reinforcement ratio (ρ_v).

5. Comparisons of Available GFRP-RC IST Models

Based on the compiled database in the previous section, this section encompasses detailed comparisons of the available analytical GFRP-RC IST models against the database.

Table 6. Predictions of different models for specimens with AC interface.

Author	Specimen ID	Experimental Peak Load	Experimental/Predicted
			CSA S6-25 [9]		AASHTO 2018 [28]	Alkatan [15]	Vega et al. [23]	Montaser et al. [24]	El Ragaby et al. [17]	Equation (17)
			Case 1	Case 2
Ahmed [25]	XX-0-N	308	NA	10.27	4.97	1.69	4.97	NA	NA	1.69
	GH2-16-N	321	3.34	3.53	0.73	1.06	3.45	3.34	5.26	1.28
	GH4-13-N	377	3.09	3.28	0.65	1.06	3.59	3.09	4.44	1.32
	GH2-19-N	384	3.05	3.34	0.68	1.09	3.66	3.05	4.52	1.37
	GH4-16-N	428	2.23	2.80	0.65	0.99	3.45	2.23	3.48	1.29
	GH6-16-N	522	1.79	2.42	0.79	0.92	3.39	1.79	2.81	1.27
	XX-0-H	550	NA	18.33	8.87	1.99	8.87	NA	NA	1.99
	GH2-16-H	530	5.52	5.82	1.21	1.36	5.70	5.52	8.69	1.56
	GH4-13-H	652	5.34	5.67	1.12	1.46	6.21	5.34	7.67	1.74
	GH2-19-H	715	5.67	6.22	1.27	1.60	6.81	5.67	8.41	1.89
	GH4-16-H	769	4.01	5.03	1.16	1.48	6.20	4.01	6.25	1.84
	GH6-16-H	885	3.03	4.10	1.34	1.36	5.75	3.03	4.76	1.79
Aljada et al. [20]	X0-000-N	293	NA	9.77	4.73	1.69	4.73	NA	NA	1.69
	G1-15Z-N	298	6.34	5.32	1.60	1.26	3.97	6.34	11.46	1.39
	G2-15Z-N	308	3.35	3.80	0.99	1.07	3.50	3.35	6.04	1.25
	G4-10C-N	399	7.98	6.23	1.57	1.74	5.05	7.98	11.74	2.00
	G4-10C-N	387	7.74	6.05	1.52	1.69	4.90	7.74	11.38	1.94
	G6-10C-N	439	5.78	5.42	1.25	1.57	5.05	5.78	8.61	1.87
	G4-13C-N	419	3.81	4.36	1.06	1.35	4.41	3.81	6.35	1.64
	G8-10C-N	451	4.56	4.42	1.01	1.46	4.56	4.56	6.26	1.81
	G4-10C-H	668	13.36	10.44	2.63	1.54	8.46	13.36	19.65	1.64
	G6-10C-H	890	11.71	10.99	2.54	1.89	10.23	11.71	17.45	2.08
	G8-10C-H	897	9.06	9.34	2.01	1.78	9.34	9.06	13.59	2.00
Alkatan [15] a	C0-30	332	NA	10.71	5.11	2.21	5.11	NA	NA	2.21
	FS2-30	385	NA	4.14	0.93	1.38	4.05	3.01	6.21	1.70
	FS3-30	384	NA	3.12	0.66	1.13	3.46	2.01	4.17	1.45
	FH3-30	362	NA	4.26	0.99	1.38	3.93	3.20	6.70	1.66
	FH5-30	433	NA	3.52	0.77	1.27	3.90	2.27	4.71	1.64
	FA3-30	342	NA	4.50	0.94	1.40	3.89	3.64	7.60	1.65
	C0-50	296	NA	9.55	4.55	1.18	4.55	NA	NA	1.18
	FS1-50	334	NA	5.48	1.40	1.07	4.18	5.30	11.13	1.16
	FS2-50	402	NA	6.59	0.98	1.28	4.23	6.38	13.40	1.23
	FS3-50	617	NA	5.02	1.05	1.40	5.56	3.23	6.71	1.70
	FH2-50	336	NA	5.01	1.27	1.03	4.05	4.48	9.33	1.14
	FH3-50	323	NA	3.80	0.89	0.89	3.51	2.86	5.98	1.02
	FH5-50	569	NA	4.63	1.01	1.29	5.13	2.98	6.18	1.56
	FA2-50	255	NA	3.36	1.07	0.74	3.19	2.71	5.67	0.89
	FA3-50	540	NA	7.11	1.66	1.57	6.14	5.74	12.00	1.76
Alruwaili [16] a	C0-A1	343	NA	13.72	6.60	2.45	6.60	NA	NA	2.45
	FS2-A1	268	NA	3.12	0.67	1.00	3.27	2.09	4.39	1.24
	FS3-A1	490	NA	4.22	0.89	1.48	5.00	2.58	5.38	1.93
	FS4-A1	428	NA	2.93	0.78	1.09	3.79	1.69	3.54	1.47
	FH3-A1	328	NA	4.10	0.93	1.29	4.15	2.88	5.96	1.58
	FH4-A1	293	NA	2.99	0.65	1.00	3.29	1.92	4.01	1.27
	FH5-A1	497	NA	4.40	0.90	1.54	5.07	2.72	5.65	1.96
	FH6-A1	547	NA	4.08	0.99	1.49	5.11	2.40	5.02	1.97
	FH8-A1	550	NA	3.24	1.00	1.24	4.40	1.82	3.79	1.70
	C0-A2	217	NA	11.42	5.56	2.07	5.56	NA	NA	2.07
	FS2-A2	172	NA	2.15	0.45	0.74	2.49	1.34	2.82	0.95
	FS3-A2	275	NA	2.50	0.66	0.94	3.24	1.46	3.02	1.26
	FS4-A2	484	NA	3.46	1.17	1.35	4.84	1.91	3.97	1.88
	FH3-A2	213	NA	2.88	0.63	0.97	3.23	1.85	3.87	1.23
	FH4-A2	218	NA	2.37	0.53	0.84	2.87	1.42	2.99	1.11
	FH5-A2	264	NA	2.42	0.64	0.90	3.11	1.40	2.90	1.21
	FH6-A2	304	NA	2.38	0.73	0.92	3.23	1.34	2.79	1.26
	FH8-A2	397	NA	2.41	0.96	0.97	3.54	1.31	2.72	1.38
Mean			5.54	5.40	1.65	1.33	4.71	3.85	6.76	1.58
SD			3.10	3.27	1.70	0.37	1.62	2.62	3.84	0.35
COV %			55.98	60.45	103.28	27.63	34.40	67.93	56.76	22.04
Mean Error %			81.94	81.49	39.37	24.93	78.78	74.03	85.22	36.52

^a The transverse shear strength of the GFRP connectors is not reported. Therefore, Case 1 of the CSA S6-25 model [9] and the upper limit of the Montaser et al. model [24] are not applicable. Note: SD: standard deviation and COV: coefficient of variance. Note: NA = Not applicable. Error % = [(Experimental − Predicted)/Experimental] × 100.

,

Table 7. Predictions of different models for specimens with IR interface.

Author	Specimen ID	Experimental Peak Load	Experimental/Predicted
			CSA S6-25 [9]	AASHTO 2018 [28]	Vega et al. [23]	Montaser et al. [24]	Equation (17)
Ahmed [25]	XX-0-IR	370	6.17	1.86	1.86	2.04	2.04
	GH2-16-IR	377	2.33	0.46	1.51	1.46	1.46
	GH4-13-IR	406	2.02	0.38	1.50	1.48	1.48
	GH2-19-IR	392	1.94	0.38	1.45	1.43	1.43
	GH4-16-IR	528	2.00	0.51	1.75	1.65	1.65
	GH6-16-IR	602	1.63	0.55	1.71	1.48	1.48
Aljada et al. [21]	G4-10-CR	452	3.90	0.87	1.99	1.93	1.93
	G4-10-CR	417	3.59	0.80	1.84	1.78	1.78
	G4-13-CR	438	2.59	0.58	1.72	1.75	1.75
	G8-10-CR	423	2.47	0.50	1.66	1.69	1.69
Vega et al. [22]	N-0-1-J	146	5.41	1.62	1.62	1.76	1.76
	N-0-2-J a	174	6.44	1.93	1.93	2.10	2.10
	N-0-3-J a	141	5.22	1.57	1.57	1.70	1.70
	F-3-1-J	228	1.04	0.44	1.22	1.00	1.00
	F-3-2-J a	252	1.15	0.49	1.35	1.11	1.11
	F-3-3-J a	245	1.11	0.47	1.31	1.08	1.08
Montaser et al. [24]	R-0-0	92	6.57	2.04	2.04	1.48	1.48
	R-GU-1-10	85	3.04	0.73	1.63	1.13	1.13
	R-GS-4-10	166	2.31	0.59	2.24	1.54	1.54
	R-GS-6-10	181	1.77	0.65	2.03	1.38	1.38
Mean			3.14	0.87	1.70	1.55	1.55
SD			1.85	0.57	0.27	0.31	0.31
COV %			58.93	65.94	15.85	20.07	20.07
Mean Error %			68.10	14.81	41.06	35.42	35.42

^a Replicate specimens. Note: SD: standard deviation and COV: coefficient of variance. Note: Error % = [(Experimental − Predicted)/Experimental] × 100.

,

Table 8. Predictions of different models for specimens with M interface.

Author	Specimen ID	Experimental Peak Load	Experimental/Predicted
			CSA S6-25 [9]	AASHTO 2018 [28]	Vega et al. [23]	Equation (17)
Ahmed [25]	XX-0-M	626	5.22	1.89	1.89	1.74
	GH2-16-M	700	2.66	0.59	1.74	1.51
	GH4-13-M	724	2.28	0.62	1.68	1.45
	GH2-19-M	775	2.43	0.62	1.80	1.45
	GH4-16-M	837	2.06	0.75	1.76	1.51
	GH6-16-M	1018	1.84	0.93	1.86	1.57
Aljada et al. [21]	G4-10-M	624	3.14	0.80	1.69	1.32
	G4-10-M	611	3.07	0.78	1.65	1.46
	G6-10-M	694	2.90	0.69	1.78	1.65
	G4-13-M	671	2.46	0.66	1.64	1.59
	G8-10-M	719	2.61	0.68	1.76	1.66
Vega et al. [22]	N-0-1	213	3.94	1.42	1.42	1.37
	N-0-2 a	250	4.63	1.67	1.67	1.61
	N-0-3 a	278	5.15	1.85	1.85	1.79
	N-0-4 a	142	2.63	0.95	0.95	0.92
	N-0-5 a	234	4.33	1.56	1.56	1.51
	N-0-6 a	166	3.07	1.11	1.11	1.07
	F-1-1	239	1.66	0.46	1.23	1.07
	F-1-2	236	1.64	0.46	1.21	1.06
	F-1-3	261	1.81	0.50	1.34	1.17
	F-3-1	308	0.95	0.60	1.08	0.86
	F-3-2	324	1.00	0.63	1.14	0.91
	F-3-3	299	0.92	0.58	1.05	0.84
	F-3-4	300	0.93	0.58	1.05	0.84
	F-3-5	350	1.08	0.68	1.23	0.98
	F-3-6	301	0.93	0.58	1.06	0.84
Mean			2.51	0.87	1.47	1.30
SD			1.30	0.43	0.32	0.32
COV %			51.79	49.91	21.54	24.63
Mean Error %			60.21	−14.84	31.94	22.96

^a Replicate specimens. Note: SD: standard deviation and COV: coefficient of variance. Note: Error % = [(Experimental − Predicted)/Experimental] × 100.

and

Table 9. Predictions of different models for specimens with S interface.

Author	Specimen ID	Experimental Peak Load	Experimental/Predicted
			CSA S6-25 [9]		AASHTO 2018 [28]	Alkatan [15]	Vega et al. [23]	Montaser et al. [24]	El Ragaby et al. [17]	Equation (18)
			Case 1	Case 2
Montaser et al. [24]	S-0-0	1.5	NA	0.21	0.11	0.02	0.11	NA	NA	NA
	S-GU-1-10	49	2.33	3.06	0.86	0.60	2.72	2.72	5.44	2.33
	S-GS-2-10	53	1.26	2.12	0.54	0.52	2.30	1.43	2.94	1.26
	S-GS-4-10	112	1.35	2.67	0.75	0.81	3.50	1.51	3.20	1.35
	S-GS-6-10	127	1.02	2.12	0.85	0.73	3.10	1.14	2.40	1.02
Mean			1.49	2.04	0.62	0.54	2.35	1.70	3.50	1.49
SD			0.58	1.10	0.31	0.31	1.33	0.70	1.34	0.58
COV %			38.73	53.80	50.45	57.75	56.56	41.08	38.31	38.73
Mean Error %			32.89	50.88	−60.77	−86.57	57.37	41.18	71.39	32.89

Note: SD: standard deviation, COV: coefficient of variance. Note: Error % = [(Experimental − Predicted)/Experimental] × 100.

compare experimental capacities with model predictions.

Figure 3, Figure 4 and Figure 5 compare the experimental IST capacities with the predictions of the aforementioned models. Normalization of concrete compressive strength was utilized in some instances to facilitate the direct comparison of nominal capacities and eliminate the effect of concrete strength variability among specimens in the models that consider the cohesion contribution as a function of concrete strength. In addition, all material and environmental reduction factors were set to unity for consistency of the predictive performance of the models.

As shown in Figure 3, the predictions for specimens with not intentionally roughened interfaces (both AC and S specimens) were calculated using both prescribed cases in CSA S6-25 [9] to illustrate their differences. In the AC specimens, the newly introduced Case 1, which accounts for the transverse shear strength of the bars, yielded an experimental-to-predicted ratio of 5.54, which is close to the ratio of 5.40 resulting from Case 2 based on clamping friction. On the other hand, in the S specimens, Case 1 yielded considerably higher estimates than those of Case 2 with ratios of 1.49 and 2.04, respectively. Nonetheless, at reinforcement ratios equal to or higher than the 0.45% threshold prescribed in CSA S6-25 [9], Case 2 (clamping friction) was the governing case for both AC and S specimens. On the other hand, at much lower reinforcement ratios, the dowel action of the connectors was the governing parameter. Overall, CSA S6-25 [9] significantly underestimated the IST capacity of push-off specimens with GFRP shear connectors, regardless of the interface condition, with experimental-to-predicted ratios of 5.54, 3.14, and 2.51 for AC, IR, and M conditions, respectively. The accuracy improved slightly for S specimens with an experimental-to-predicted ratio of 2.04. This variability reflects the model’s limited ability to capture GFRP contributions, its neglect of concrete compressive strength effects, and its use of lower cohesion coefficients compared to the AASHTO LFRD model [28]. It should also be noted that the capacity threshold of 6.5 MPa was the governing upper limit specified in all specimens rather than the limit of $0.25{f^{\prime }}_c$. For the range of specimens tested to date, this cap would not govern the design.

In contrast, the AASHTO LFRD model [28] provided widely scattered estimates, while markedly overestimating the IST capacity of several GFRP-RC specimens, as shown in Figure 4. This model yielded average experimental-to-predicted ratios of 1.65, 0.57, 0.87, and 0.62 for AC, IR, M, and S interface conditions, respectively, with significantly high COVs up to 103%. This overestimation and scatter become more pronounced at higher reinforcement ratios, and stems from the assumption that GFRP bars can develop their full tensile strength at the peak load. While this assumption may be applicable for steel reinforcement, given its high modulus of elasticity and low yield strain of 0.002, it is not applicable to GFRP reinforcement. Owing to its lower modulus of elasticity, GFRP bars require significantly larger strains to reach their tensile strength and would typically fail before reaching the ultimate tensile strength. Consequently, the effective contribution of GFRP reinforcement must be capped by an appropriate strain limit.

As shown in Figure 5, the equation proposed by Alkatan [15] with a strain limit of 0.005 showed improved predictive accuracy compared to the CSA S6-25 [9] and AASHTO LFRD [28] models, with an average experimental-to-predicted ratio of 1.33. However, this model overestimates the capacity of several AC specimens at reinforcement ratios of 0.6% and higher. By comparison, the model proposed by Vega et al. [23] provided conservative predictions, as it disregards the effect of the concrete strength and considers a lower strain limit of 0.002. This model provided conservative results for the AC specimens with an average ratio of 4.71, which suggests that refinement for the cohesion and friction factors for different shear interface conditions may be required. It is worth noting that neither model was originally developed for smooth IST surfaces, hence both models yielded either unconservative or very conservative predictions for that interface condition.

The model proposed by Montaser et al. [24] exhibited consistent predictions for IR specimens with experimental-to-predicted peak load values of 1.55, with a relatively low COV of 20%. However, not accounting for the cohesion contribution in the capacity prediction of the AC specimens led to an overconservative ratio of 3.85. Additionally, most of the tested AC specimens in the database did not record reinforcement strains of 0.005 at the peak load. On the other hand, the model yielded better predictions for the S specimens, with a ratio of 1.70 and a COV of 41%. The model proposed by El Ragaby et al. [17] resulted in overconservative predictions for AC specimens due to neglecting the cohesion parameter and adopting a lower limiting factor of frictional resistance of 0.0024, with a ratio of 6.76 and a very high COV at approximately 57%; this limit severely underestimates the capacity of specimens, with a low reinforcement ratio. This conservatism was less pronounced in the S specimens, with a ratio of 3.50 and a COV of 38%.

6. Proposed Model

Based on the previous discussions, the existing models and design provisions for GFRP-RC IST systems do not accurately represent the capacity of the push-off specimens, either fully or partially. This is attributed to adopting inaccurate GFRP reinforcement strain limits, assuming full utilization of the tensile strength of GFRP connectors, or by neglecting or underestimating the concrete contribution in the IST mechanisms. Additionally, the models available in the literature do not provide a unified equation capable of consistently capturing all investigated IST surface conditions. Therefore, a proposed regression-based model is formulated using a cohesion–friction framework. In this model, the concrete compressive strength (f′_c) is used to characterize the cohesion component, while the stiffness of the GFRP reinforcement (E_f ρ_v) governs the reinforcement clamping contribution. The general form of the model is expressed as shown in Equation (16):

$$v_r=\alpha {f^{\prime }}_c+\mu {\epsilon }_f{E}_f{\rho }_v.$$

(16)

Unlike the approach followed by CSA S6-25 [9] and AASHTO LFRD [28], the AC and IR surface specimens were grouped into a single regression analysis, supported by the marginal differences in the IST capacity between both conditions at the same reinforcement ratio and concrete strength, as reported by Aljada et al. [21] and Ahmed [25] (2–23% difference). Therefore, the friction coefficient (μ) for IR surface conditions (i.e., 1.0) was also used for not intentionally roughened (AC) surface conditions, instead of 0.6. Also, the analysis did not include the M and S specimens due to the scarcity of experimental data.

The regression analysis was performed using a normalized formulation with respect to the concrete compressive strength to improve correlation by eliminating the influence of concrete strength variability and to calibrate the concrete strength parameter (α) in addition to the GFRP reinforcement strain limit (${\epsilon }_f$). Additionally, the conservative 5th percentile lower-bound model corresponding to a 95% confidence interval was used in the regression analysis to ensure reliability in the proposed formulation, as shown in Figure 6. This figure illustrates the relationship between the normalized IST capacity (v_u/f′_c) and the normalized GFRP reinforcement stiffness (E_f ρ_v/f′_c). The model for AC and IR specimens yielded a cohesion parameter of 0.042, closely matching the value of 0.04 proposed by Harries et al. [2], with a slope of 0.003.

Based on the comprehensive review of existing models and the regression-based analysis, a new model is proposed. For AC, IR, and M interfaces, the proposed model (Equation (17)) adopts a cohesion coefficient of 0.04 (

Table 10. Values of α and µ in the proposed model.

Surface Condition	α	µ
Cold joint as-cast and intentionally roughened interface (AC and IR)	0.04	1.0
Monolithic interface (M)	0.075	1.4

) for both IR and AC interfaces and a unified strain limit of 0.003. The latter value represents the average strains measured in GFRP shear connectors at peak load in the database in Table 2, Table 3, Table 4 and Table 5. This strain limit is lower than the limit proposed by Alkatan [15] (0.005) and CSA S6-25 [9] (0.004) as several specimens with sufficient reinforcement (ρ_v ≥ 0.45%) did not reach these values. Additionally, this value is higher than the limits of 0.002 and 0.0024 proposed by Vega et al. [23] and El Ragaby et al. [17], respectively. The lower limits would restrict the stress in GFRP bars to well below their effective capacity, evidenced by the higher recorded strain values at the peak load in most of the specimens. Even though regression analysis was not performed for M interfaces due to the scarcity of experimental data, the proposed model was extended to the M specimens based on the analogy in existing code provisions, with adjustments to the cohesion and friction coefficients. The higher friction coefficient of 1.4 utilized in CSA S6-25 [9] was adopted in the proposed model to reflect the enhanced reinforcement contribution and increased IST capacity in monolithic specimens, evidenced by the higher strains recorded at peak load in the M specimens. Additionally, the concrete strength coefficient of 0.075 was adopted as proposed by Harries et al. [2] to reflect the higher cohesion contribution in the monolithic interface compared to AC and IR specimens.

$$v_r=\alpha {f^{\prime }}_c+0.003{\mu \rho }_v{E}_f\le 0.20{f^{\prime }}_c.$$

(17)

For S interfaces, only five specimens have been tested [24]. Therefore, the conservative approach of the CSA S6-25 [9] model (Case 1) is adopted by disregarding the contribution of the interface cohesion and friction and relying on the transverse shear strength of the IST connectors (Equation (18)). The same approach is followed by Montaser et al. [24] (Equation (14)). Additional experimental data are required to further assess the IST performance at smooth interfaces.

$$v_r={\rho }_v{f}_{ft}\le 0.20{f^{\prime }}_c.$$

(18)

The model adopts the 0.20f′_c upper limit from ACI 318-25 [8], which is more conservative than the 0.25f′_c limit in CSA S6-25 [9]. This limit reflects the behavior of over-reinforced interfaces, where increases in reinforcement result in diminishing gains in shear capacity [29]. Additionally, the model only considered GFRP IST reinforcement oriented perpendicular to the interface. Hence, more experimental results are required to study the effect of IST reinforcement orientation on GFRP-RC IST behavior.

The predictions of the proposed model (Equations (17) and (18)) are listed in Table 6, Table 7, Table 8 and Table 9 and shown in Figure 7. The proposed model yielded more accurate, yet conservative, predictions for all surface conditions tested, with experimental-to-predicted ratios of 1.58, 1.55, 1.30 and 1.49 for AC, IR, M and S conditions, respectively.

7. Conclusions

This study assembled a comprehensive database of 107 push-off tests from the literature covering various parameters to evaluate the accuracy of existing IST design provisions and analytical models for GFRP-RC systems. The assessment revealed significant limitations in current formulations, leading to the development of a new regression-based model (Equation (17)). The new model more accurately captures the IST behavior by incorporating the effects of FRP reinforcement stiffness and concrete compressive strength. The following key conclusions can be drawn.

• The CSA S6-25 model [9] was found to be highly conservative, underestimating experimental capacities with mean experimental-to-predicted ratios of 5.54, 3.14, 2.51, and 2.04 for specimens with as-cast, intentionally roughened, monolithic, and smooth interfaces, respectively. This model disregards the contribution of concrete strength and utilizes relatively low cohesion factors.
• The AASHTO LRFD model [28] yielded mean experimental-to-predicted ratios of 1.65, 0.57, 0.87, and 0.62 for specimens with as-cast, intentionally roughened, monolithic, and smooth interfaces, respectively. The model consistently overestimated the IST capacity at higher reinforcement ratios across all interface conditions. This is attributed to the unrealistic assumption that FRP bars can develop their full tensile strength at peak load.
• The model by Alkatan [15], which is applicable to as-cast and smooth interfaces, recorded a reasonable mean experimental-to-predicted ratio of 1.33 for as-cast specimens. However, this model tended to overestimate several AC specimens at higher reinforcement ratios due to the relatively high reinforcement strain limit of 0.005. On the other hand, the model by Vega et al. [23], which incorporates a lower reinforcement strain limit of 0.002 and disregards the effect of concrete strength, provided conservative estimates across the different interfaces, with ratios of 4.71, 1.70, 1.47, and 2.35 for specimens with as-cast, intentionally roughened, monolithic, and smooth interfaces, respectively.
• The model by Montaser et al. [24] provided adequate predictions for specimens with intentionally roughened interfaces, with a test-to-predicted ratio of 1.55. However, it underestimated the capacities of specimens with as-cast interfaces, with a ratio of 3.85. The model by El Ragaby et al. [17] resulted in overconservative predictions, due to neglecting the cohesion parameter, with ratios of 6.76 and 3.50 for specimens with as-cast and smooth interfaces, respectively.
• The proposed model incorporates a GFRP strain limit of 0.003 and a cohesion coefficient of 0.04 (for both as-cast and intentionally roughened interfaces) and 0.075 (for monolithic interfaces). In addition, it adopts the transverse shear capacity approach for smooth interfaces. The model achieved accurate and consistent predictions across all interface conditions, with mean ratios of 1.58, 1.55, 1.30 and 1.49 for specimens with as-cast, intentionally roughened, monolithic, and smooth interfaces, respectively.
• For practical design provisions, Equation (17) is recommended for AC, IR and M interfaces, while Equation (18) is recommended for S interfaces, considering only the transverse shear strength of the IST connectors. However, additional experimental testing is recommended to further validate the applicability of the model to monolithic and smooth interfaces.