Study of Basalt Fibers and Graphene Enriched Polymers on Bond Behavior of FRP Bars in Concrete

Abstract

In this work is investigated the bond behavior of FRP bars considering basalt as a fiber and also cases where graphene is introduced as a polymer filler. Standard FRP bars where glass is used as fiber is tested under the same conditions in order to have a benchmark. Three different temperatures are considered in the tests, including room temperature and a temperature higher than the glass transition of the polymer. At room temperature the effect of concrete strength and the bar diameter on bond is also investigated. The study developed is experimental, through pull-out tests, as well as numerical, through simulations by the Finite Element method. Cases where basalt fibers are associated to helically-wounded surface treatment showed at least 41.7% superior bond behavior among all cases considered with this fiber at room temperature. Studies with graphene were exploratory and limited to only one percentage of addition. However, results suggest that the addition is able to improve behavior of the polyester resin of the matrix at room temperatures by 49% and at higher temperatures bond degradation was 44% smaller when compared with the literature. It was not observed a statistically significant influence of the concrete strength on bond. Diameter influence on bond was mostly linked to surface treatment.

1. Introduction

Fiber Reinforced Polymer (FRP) bars were introduced in reinforced concrete structures replacing steel bars primarily to mitigate problems related to corrosion [1,2,3,4,5,6]. Although the initial cost of the FRP use is higher than steel, the total life cycle cost of structural components reinforced with the former may be lower due to reduced maintenance expenses. In addition, D’antino et al. [7] highlight as advantages the high tensile strength and lower density compared to steel. However, unlike steel reinforcements, there is still no standardization for the characteristics of these bars. In addition to the existence of different types of fibers and polymers, there is also a wide variety of surface finishes (smooth, sand coated, grooved, ribbed, helically wrapped, helically wounded, among others) that directly impact bond mechanisms [2,8]. Concrete compressive strength and bar diameter are other usual factors that influence bond. FRP bar diameter influence is in general controversial. For instance in Tighiouart et al. [9], Baena et al. [2] strength decreases with larger diameters and Gotad et al. [10] observed an inverse effect. This happens because in FRP bars is difficult to isolate the diameter effect due to many variables involved. It should be noticed that rebar diameter influence on bond behavior in general is not well established, even in the case of steel rebars, see for instance discussion in Miranda et al. [11]. The same can be said regarding concrete compressive strength effect on bond, see for instance Cosenza et al. [12] and Shen et al. [13]. However, as pointed out by Achillides and Pilakoutas [1], the general trend is that for high strength concretes (compressive strength > 30 MPa) the rupture of the bond tends to occur on the surface of the FRP bar, so bond strength would not benefit much from an increase in concrete strength beyond this point.

FRP is a composite material formed by a polymeric matrix reinforced with fibers. The matrix holds fibers together and distributes forces among them, in addition to protect fibers against abrasion, impact and other environmental hazards. Thermosetting resins such as epoxi, polyester and vinyl-ester are the most common material that compose the matrix [14]. Epoxi has higher mechanical strength, thermal stability and lower thermal contraction/expansion when compared to the others, but the cost is higher. Polyester is the cheapest in term of costs, but it is also the poorest in terms of mechanical properties [14,15]. Thermal properties also tend to be poorer in the latter case. For instance its glass transition temperature, where mechanical properties degrade substantially, is in general 10 °C lower than corresponding vinyl-ester polymer [16].

Thermo-mechanical behavior of polymeric resins, in general, can be improved by the addition of fillers. In this regard, studies have been done considering Graphene Nano Platelets (GNPs), see e.g., Prolongo et al. [17] and Saleem et al. [18]. GNPs present exceptional mechanical strength, low weight and can be easily incorporated into polymeric resins through solvents. The cost is lower when compared to other nanostructured materials such as carbon nanotubes with an equivalent performance [19]. GNPs have been also used as fire retardant in polymers due to its high temperature resistance, as reported in Dittrich et al. [20] and Inuwa et al. [21]. However, the present authors are unaware of bond studies at different temperatures considering FRP bar polymers enriched by GNPs, which is a motivational aspect for the present work.

Fibers give the mechanical strength and the stiffness of the bars. The more traditional fibers are glass (GFRP), carbon (CFRP) and aramid (AFRP). More recently, basalt (BFRP) was considered and it is becoming a very competitive alternative to other types of fibers [3,22,23,24,25,26]. BFRP bars present excellent physical and mechanical properties and good cost-benefit ratio. Among the studies already done, it was found that they have higher tensile strength and elasticity modulus and better chemical stability than GFRP bars, besides being cheaper than CFRP bars [27,28]. Another important advantage considering environmental and ecological issues is the fact that basalt fibers are produced by natural volcanic rocks. Regarding bond studies, helically wrapped surface were tested by Wang et al. [29] and Shen et al. [13], while grooved surfaces were studied by El Refai et al. [3] and Liu et al. [30]. Finally, Henin et al. [31] and Wei et al. [32] considered sand coated surfaces. In all cases concrete ranged from C30 to C50 strength grade. Not all types of surfaces, bar diameters, concrete strengths have been tested and bond behavior of cases such as helically wounded, smooth, among other types have not been addressed, as far as these authors know. Therefore there is still an insufficient knowledge about BFRP bars, which have prevented them to be included in design standards for reinforced concrete structures so far, which is another motivational aspect for the present work.

The low thermal resistance associated to FRP bars is another critical aspect. One of the pioneer studies in this regard was done by Katz et al. [33] considering GFRP bars. It was shown that, for temperatures of 250 °C, bond strength could drop up to 90% compared to room temperature strength. Surface treatment was a factor that changed thermal degradation. Another studies of thermal degradation can be found in Solyom et al. [34], Li et al. [35] and Hamad et al. [36]. The first focused on GFRP rebars while the others compared the behavior between GFRP and BFRP rebars. Thermal damage was slightly smaller for BFRP bars until certain temperature. For higher temperatures behavior was similar in both cases. Again surface treatment was considered an important factor for thermal degradation.

Due to significant number of variables involved in FRP bond-slip behavior, the use of machine-learning techniques can be appropriate in order to provide a reliable estimation of bond strength in these cases. It may reduce the extensive testing required for assessing variables influence, reducing time and costs significantly. The use of the technique will not be pursued in the present work, but applications to FRP can be found for instance in Salameh et al. [37], Kazemi et al. [38], among others.

This work proposes an experimental/numerical analysis of the bond behavior by pull-out tests for BRFP bars and bars where GNP is added as an additive. In the first case, helically wounded and smooth surfaces are considered, which are cases still not fully investigated in the literature. In the second case, an exploratory study of the GNP addition on GFRP rebars with polyester matrix is proposed. Only helically wrapped surfaces are considered in the latter cases. Standard GFRP/helically wrapped bars are also tested in this work as a benchmark. At room temperature the effect of bar diameter and concrete strength on bond are studied. Thermal degradation is also studied for one particular combination of diameter and concrete. The tests are also reproduced by Finite Element (FE) simulations in order to validate experiments, explore stresses in the concrete and obtain parameters for an analytical model.

In Section 2 the experimental study and materials are described. Numerical simulation details are shown in Section 3. Results and discussions are presented in Section 4. Finally, final remarks are stated.

2. Experimental Study

Experiments are based on pull-out tests done at three different temperatures. In this Section, first materials are characterized and then pull-out testes are described.

2.1. Materials

Three types of bars are considered in this work: BFRP and GFRP with vinyl-ester matrix and GFRP with polyester matrix where GNP was added. In the latter case the cheaper polyester polymer was enriched with a minimum weight of GNP in order to keep costs at the same level of the more traditional and more expensive vinyl-ester polymer. These cases will be referred as GFRPg. Two diameters are considered for the bars: 8.0 and 12.5 mm. In GFRP and GFRPg cases, surface treatment is helically wrapped while BFRP 8.0 mm bars have smooth surface and 12.5 mm bars have helically wounded surface treatment, see Figure 1 (nomenclature according to definition in Fahmy et al. [39]). FRP applications are relatively common in footing, slabs on grade applications, which in general requires lower concrete strength grades, such as C20. Another typical FRP applications are industrial pavements, residential slabs and beams which requires at least C30 concretes [40]. This is the main reason these concretes were chosen here and also by other relatively recent publications [30].

For the identification of the combinations tested, the nomenclature C xx −ϕN − yy was defined, where xx corresponds to the compressive strength class of the concrete; $\varphi$ is the nominal diameter of the reinforcement (8 = 8.0 mm diameter and 12 = 12.5 mm diameter); N represents the type of bar (G = GFRP, Gg = GFRPg and B = BFRP) and yy is the test temperature in degree Celcius, which is omitted in case of room temperature tests.

Table 1. Properties of the FRP bars.

Bar	rh (mm)	rs (mm)	rw (mm)	Ra (μm)	Fiber Fraction (%)	${\mathit{\sigma }}_{\mathit{t}}$ (MPa)	E (GPa)	Tg (°C)
Cxx-8G-yy	0.30	22.52	7.59	4.79	86.07	955.8	53.1	66.99
Cxx-12G-yy	0.71	16.73	6.42	6.94	79.06	917.1	50.4	66.99
Cxx-8Gg-yy	0.35	8.24	3.16	4.79	85.60	817.6	54.4	62.53
Cxx-12Gg-yy	0.67	10.42	2.82	6.94	85.29	707.9.1	51.2	62.53
Cxx-8B-yy	0.04	13.45	3.15	4.53	83.87	702.9	55.0	99.02
Cxx-12B-yy	0.61	10.46	2.12	5.65	79.20	1013.2	52.0	99.02

indicates geometrical and mechanical properties of the bars. Geometrical properties (rh, rs and rw) are defined according to Figure 1. The average roughness of the bars (Ra) was based on the ISO 21920-1 [41] Standard. Fiber fraction is measured by weight fraction through resin burn-off. Values obtained are higher than the minimum values established by ASTM D7957 [42]. The tensile strength (${\sigma }_t$) and longitudinal elasticity modulus (E) of the bars were determined according to ASTM D7205 [43] and are related only to axial direction. Poisson coefficient is 0.2 in all cases. $T_g$ is the glass transition temperature. In this temperature there is a substantial degradation of the polymer, so the bars loose most of their bond capabilities. Regarding test temperatures, the fundamental information is how the bars behave before and after glass temperature transition. So, besides room temperature, other two are considered based on $T_g$s. One below $T_g$ (60 °C) and another above $T_g$ (120 °C). As bond tends to drop significantly after $T_g$ and then keeps approximately constant, there is no reason to test different higher temperatures.For the cases above room temperature only concrete C30 and 12.5 mm bars are considered.

Concrete mix proportions are presented in

Table 2. Concrete mix proportion.

Concrete	Cement Consumption (kg/m3)		Unit Mix
		Cement	Fine Aggregate	Coarse Aggregate	Water/Cement Ratio
C20	292.78	1.0	2.9	3.6	0.66
C30	452.52	1.0	1.6	2.4	0.47

. High-early-strength Portland cement of 3.00 g/cm³ specific mass and Blaine finesses of 5001 cm²/g was used for the concrete mixes. Fine aggregates were quartz sand of natural origin, with a 4.8 mm maximum grain diameter and 2.49 g/cm³ specific mass, determined according to the ABNT NBR 17054 [44] and ABNT NBR 16916 [45], respectively. Basalt-based coarse aggregates with a 25 mm maximum size, 2.79 g/cm³ specific mass and 1.03% water absorption were utilized, determined according to the ABNT NBR 17054 [44] and ABNT NBR 16917 [46]. Throughout the research, supplied tap water was used for both mixing and curing purposes.The evaluation of the concrete compressive strength at 28 days of age was made according to ABNT NBR 5739 [47] recommendations. Based on the 28 day compressive strength of the concrete, its elasticity modulus was estimated using the FIB Model Code 2010 [48] and the results are presented in

Table 3. Concrete properties.

Combinations	Slump (mm)	Average Compressive Strength (MPa)	Standard Deviation	Elasticity Modulus (GPa)	Poisson’s Ratio
C20-8G	140	23.2	0.91	34.2	0.2
C30-8G	145	33.3	0.77	38.5	0.2
C20-12G	140	23.2	0.91	34.2	0.2
C30-12G	145	33.3	0.77	38.5	0.2
C20-8Gg	140	20.8	0.95	32.9	0.2
C30-8Gg	140	30.3	0.42	37.3	0.2
C20-12Gg	130	23.8	0.17	34.4	0.2
C30-12Gg	130	32.4	0.76	38.2	0.2
C20-8B	145	22.6	0.48	33.9	0.2
C30-8B	125	30.8	0.32	37.5	0.2
C20-12B	145	22.6	0.48	33.9	0.2
C30-12B	125	30.8	0.32	37.5	0.2
C30-12G-60	120	34.3	1.95	38.9	0.2
C30-12Gg-60	120	34.3	1.95	38.9	0.2
C30-12B-60	120	34.3	1.95	38.9	0.2
C30-12G-120	120	33.9	1.61	38.8	0.2
C30-12Gg-120	120	33.9	1.61	38.8	0.2
C30-12B-120	120	33.9	1.61	38.8	0.2

. In all cases, 3 samples were considered.

2.2. Pull-Out Tests

The molding and curing of the pull-out test specimens were done according to ASTM D7913 [49] in a controlled environment at 23 ± 2 °C and 50 ± 10% relative humidity. In order to avoid the crushing of the reinforcement during the test, it was necessary to place an anchoring device at the loaded end of the bars. For this purpose, a steel pipe was used and the space between the pipe and the bars was filled with an epoxy based structural adhesive, as recommended by ASTM D7205 [43]. After finishing the anchoring system on the bars, the specimens were molded for the pull-out tests. The molding procedures were performed according to the ASTM D7913 [49]. Cubic molds, with a 200 mm dimension on each edge, were used to manufacture the pull-out specimens. Figure 2 shows the dimensions of the specimens schematically. An embedment length $l_b=\varphi 5$ was considered. A PVC pipe was used as a bond breaker in the debonded area.

Bars were positioned horizontally in the middle of the formworks, then the concrete was poured in two layers. Each layer was compacted with 25 strokes of a metal rod. The surface was smoothed with a mason trowel. After 72 hours, the concrete cubes were demolded, marked and transferred to the curing room at 23 ± 2 °C and 95.0% humidity until the age of 28 days. Three specimens were molded for each combination.

In the cases subjected to heating, the methodology adopted by authors such as Li et al. [35] and Hamad et al. [36] was followed in the present work. The specimens were molded with the insertion of K-type thermocouples at the bar-concrete interface. After the 28-day curing, the specimens were heated in an electric oven. The part of the bar inside the oven were thermally insulated with ceramic fiber blankets in order to avoid thermal damage to the polymeric matrix. The heating rate was 27 °C/min up to the desired temperature, remaining at this temperature range for at least one hour until a uniform distribution of heat through concrete was achieved, according to measurements done. (The remaining time at the final temperature were 6 h in Li et al. [35] and 30 min in Hamad et al. [36]). In all cases, including the cases in the references, the specimens were then left in the oven for natural cooling. Afterwards, pull-out tests were done at room temperature.

To evaluate the compressive strength of the concrete exposed to higher temperatures, cylindrical specimens were placed in the oven, which remained under heating for the same period as the specimens for the pull-out test. After cooling, the compressive strength was determined according to ABNT NBR 5739 [47] and the results are shown in

Table 4. Concrete properties after thermal exposure.

Combinations	Exposure Temperature (°C)	Average Compressive Strength (MPa)	Standard Deviation	Elasticity Modulus (GPa)	Poisson’s Ratio
C30-12G-60	60	31.7	2.09	37.9	0.2
C30-12Gg-60	60	31.7	2.09	37.9	0.2
C30-12B-60	60	31.7	2.09	37.9	0.2
C30-12G-120	120	29.6	2.91	37.0	0.2
C30-12Gg-120	120	29.6	2.91	37.0	0.2
C30-12B-120	120	29.6	2.91	37.0	0.2

.

The pull-out tests were performed according to ASTM D7913 [49] in a Universal Testing Machine (Instron/Emic) with a 200 kN capacity. The concrete cubes were positioned at the top of the machine, remaining stationary during the test. A Linear Variable Differential Transformer (LVDT) was attached to the free end of the bar for slip measurements. The load was applied to the bars at 1.3 mm/min rate. Figure 3 shows a specimen positioned in the equipment during the test.

The average bond strength (${\tau }_{avg}$) as a function of slip was calculated based on the recorded force in the experimental test (F), bar diameter $\varphi$ and bond length ($l_b$) according to Equation (1).

$${\tau }_{avg}={\displaystyle \frac{F}{\pi \varphi l_b}}$$

(1)

3. Numerical Model

In order to simulate the bond process, a FE model was used. The model was defined on the bar scale (see [50,51,52]), i.e., the reinforcement and concrete are modeled as volumetric FEs and the discretization of the ribs or details of the reinforcement is not done. A bond stress-slip law is applied to the model considering cohesive interface elements between concrete and reinforcement. In this work the model developed by Rolland et al. [53] was used as a bond stress-slip law. This model was chosen in this work because it was specifically proposed for FRP applications and it is also a complete model in the sense that ascending, descending branch of the bond-slip law, as well as residual bond, are all included in the model. Classical models as in ref. [9,12] only consider the ascending branch. The model also presents a greater generality than others permitting a very good fit with experiments. Bond stress τ(s) is calculated as a function of the slip s, considering three different zones according to Equation (2) and Figure 4. The model is governed by seven parameters, $\alpha$, $\beta$, $s_0$, ${\tau }_0$, ${\tau }_{\infty }$, ${\tau }_1$ and $s_1$.

In the first part of the model, the parameters $s_0$ and ${\tau }_0$ define an ascending linear phase. Subsequently, when the slip is greater than $s_0$, there is an ascending hyperbolic curve governed by the $\alpha$ parameter, which goes up to the maximum bond stress ${\tau }_1$ (bond strength) for a corresponding slippage $s_1$. Then, in the residual phase for $s>s_1$, the curvature of the descending hyperbolic branch is controlled by the $\beta$ parameter and, finally, the horizontal asymptote is governed by the ${\tau }_{\infty }$ parameter. ${\tau }_1$ and $s_1$ are obtained experimentally through the pull-out test and the other parameters are here obtained through an optimization process, which is detailed below.

$$\begin{array}{c}\tau \left(s\right)={\tau }_0{\displaystyle \frac{s}{s_0}}:0\le s\le s_0\\ \tau \left(s\right)={\tau }_0+({\tau }_1-{\tau }_0)\left(1+{\displaystyle \frac{1}{\alpha }}\right)\left(1-{\displaystyle \frac{1}{1+\alpha \left(\frac{s-s_0}{s_1-s_0}\right)}}\right):s_0<s\le s_1\\ \tau \left(s\right)={\tau }_{\infty }+({\tau }_1-{\tau }_{\infty })\left({\displaystyle \frac{1}{1+\beta \left(\frac{s-s_1}{s_1}\right)}}\right):s>s_1\end{array}$$

(2)

3.1. Determination of the Parameters

To determine the parameters of the bond stress-slip model, an optimization process was used based on an algorithm with a metaheuristic approach called Whale Optimization Algorithm (WOA), proposed by Mirjalili and Lewis [54]. Parameters were determined by the minimization criterion of the objective function described in Equation (3) by the method of least squares, which seeks to minimize the sum of the differences between the values estimated by the model and the values from experimental data. 300 search agents and 100 iterations were defined in the WOA algorithm to reach the best solution for the parameters. When processing starts, the WOA algorithm provides a set of random solutions. In each iteration, the research agents update their positions in relation to the best solution obtained so far, thus reaching the model parameters. After obtaining the parameters, the constitutive law of the cohesive interface model was implemented in the FE code to perform numerical solutions of the experiments. The numerical results were compared with experimental results for each case analyzed, as will be shown in the Section 4.

$$Obj=\sum _{i=1}^n{\left(\tau {\left(i\right)}_{exp}-\tau {\left(i\right)}_{mod}\right)}^2$$

(3)

In the equation above, $\tau {\left(i\right)}_{exp}$ is the experimentally obtained bond stress, $\tau {\left(i\right)}_{mod}$ corresponds to the bond obtained by the model and n is the number of experimental points extracted from a curve.

3.2. Thermal Degradation Model

The model developed by Correia et al. [55] is applied in the present work to represent the decrease in bond strength with increasing temperatures. The model is based on extensive experimental evidences. In Equation (4) P represents normalized bond strength as a function of the temperature T.

$$P\left(T\right)=P_r+(1-P_r)(1-e^{B{e}^{CT}})$$

(4)

$P_r$ is the bond strength at a given temperature normalized by room temperature bond strength ${\tau }_1$ and B and C are parameters fitted to the experimental data. WOA metaheurist approach is once again used to obtain parameters.

3.3. FE Model

The specimens were modeled as 2D with axisymmetric geometry. This was possible because the size of the cubic specimens is much larger than the minimum cover to avoid splitting, so bond results are supposed to be insensitive to the specimen shape. Rebar was positioned along the axis of symmetry. Standard 4-node bi-linear quadrilateral FEs were used in the volume. Cohesive FEs are 4-node linear elements. The boundary conditions consisted of a displacement applied to the nodes at the loaded end of the bar, while vertical displacements for concrete nodes in contact with a support was blocked (see Figure 5). In addition, the nodes of the bar positioned at the symmetry axis were constrained in the horizontal direction. The meshes employed in the numerical analysis are shown in Figure 5. It is worth to mention that meshes with a double number of elements were tested in relation to the meshes shown in Figure 5 and no major differences were found in the numerical results, therefore results obtained can be considered mesh independent. Bars and concrete were considered elastic-linear according to properties in Table 1, Table 3 and Table 4. Therefore, all non-linearity and rupture process is confined at the interface.

Integration of stresses $\tau \left(s\right)$ at the cohesive interface of the FEs gives nodal forces F related to bond. These forces are added to the finite element forces. Equilibrium is obtained using a Newton-Raphson scheme until a relative error of 0.1% is achieved. Tangent matrices were calculated by numerical derivation. Finally, F in equilibrium was used in Equation (1) to obtain the numerical average bond stress.

4. Results and Discussions

4.1. Room Temperature Results

Table 5. Experimental results at room temperature.

Combinations	Specimens	${\mathit{\tau }}_1$ (MPa)	${\mathit{s}}_1$ (mm)	Rupture	${\overline{\mathit{\tau }}}_1$ (MPa)	Std Dev. (MPa)
GFRP bars
	SP1	8.42	2.40	pull-out
C20-8G	SP2	7.26	1.75	pull-out	7.58	0.73
	SP3	7.10	2.94	pull-out
	SP1	10.16	2.59	pull-out
C20-12G	SP2	10.45	1.66	pull-out	9.88	0.74
	SP3	9.04	1.88	pull-out
	SP1	7.83	3.09	pull-out
C30-8G	SP2	7.12	1.87	pull-out	6.69	1.41
	SP3	5.111	1.83	pull-out
	SP1	12.76	2.59	pull-out
C30-12G	SP2	11.34	2.23	pull-out	12.71	1.35
	SP3	14.04	2.25	pull-out
GFRPg bars
	SP1	6.02	2.80	pull-out
C20-8Gg	SP2	7.75	3.77	pull-out	6.16	1.53
	SP3	4.70	3.07	pull-out
	SP1	12.81	2.88	pull-out
C20-12Gg	SP2	12.97	2.54	pull-out	12.91	0.09
	SP3	12.96	2.33	pull-out
	SP1	7.85	3.31	pull-out
C30-8Gg	SP2	8.71	2.83	pull-out	9.78	2.64
	SP3	12.78	3.27	pull-out
	SP1	12.89	2.65	pull-out
C30-12Gg	SP2	8.08	3.47	pull-out	9.82	2.67
	SP3	8.48	2.34	pull-out
BFRP bars
	SP1	3.50	4.14	pull-out
C20-8B	SP2	2.54	3.28	pull-out	3.02	0.68
	SP3 *	-	-	-
	SP1	25.79	3.64	pull-out
C20-12B	SP2	19.33	2.60	pull-out	23.26	3.45
	SP3	24.66	2.53	pull-out
	SP1	5.16	2.69	pull-out
C30-8B	SP2	6.41	2.39	pull-out	6.84	1.94
	SP3	8.95	3.96	pull-out
	SP1	25.99	2.96	pull-out
C30-12B	SP2	19.91	2.60	pull-out	22.39	3.19
	SP3	21.27	2.52	pull-out

* Specimen was discarded due to slippage of the anchoring system.

presents the values of the maximum bond stress ${\tau }_1$ (bond strength), corresponding slip $s_1$ and the rupture mode developed for each pull-out test. It is also presents the average maximum bond stress ${\overline{\tau }}_1$ of the specimens and the corresponding standard deviation. In the Appendix A model parameters used in the numerical simulations are given for the average of each case considered.

In all specimens a rupture by pull-out occurred, with the outer surface of the bars being damaged, as can be seen in Figure 6. No signs of cracking were observed in the concrete, indicating an adequate confinement of the bars. For smooth surfaces, partial delamination occurred (Figure 6a). For helically-wounded surfaces, it was noted an abrasion of the outer surface (light gray areas in Figure 6b). This type of rupture have been also reported by other authors, such as Solyom and Balázs [56] and Fahmy et al. [39]. In helically-wrapped surfaces the strings wrapping the bars tended to delaminate, as seen Figure 6c.

Figure 7a, shows all average bond stress vs slip plots for GRFP cases. Experimental and numerical results are shown. An excellent fit between them is obtained. Figure 7b shows the average of the three specimens used per case. Based on these averages it is possible to access the influence of concrete strength increase on bond strength. For 12.5-mm bars there was an increase in bond and for 8-mm bars the bond actually decreased. A new set of average values is presented in Figure 7c, for both diameters considered. It can be seen a clear increase in bond with the increase in diameter. Finally Figure 7d shows a comparison of the average 12.5 mm GFRP bars with some results available in the literature, considering GFRP bars in similar conditions. It is observed that the present experiments do not depart significantly from what is expected by the literature, except the case tested by Baena et al. [2], where a grooved surface was used.

Figure 8a, shows all average bond stress vs slip plots for GRFPg cases. Experimental and numerical results are shown. An excellent fit is obtained, except for SP2 and SP3 C30-12Cg cases where an oscillatory answer was obtained experimentally during post-peak. The model used was not able to capture such oscillations but it was able to capture the behavior on average. Figure 8b shows the relations for the average of the three specimens used per case. The observed effect of the increase in concrete strength on bond strength is the opposite of the effect seen in GFRP bars, i.e., For 8-mm bars there was an increase in bond and for 12.5-mm bars the bond decreased. The average values for both diameters considered is presented in Figure 8c. It can be seen again an increase of the bond with the increase of diameter. As no other test of bars with GNP addition was found in the literature, it is still not possible to compare the present results with others. However, in the figure is included results obtained by Katz et al. [33] diameter 12.7 mm, polyester polymer and other conditions fairly similar to the present experiments. The better behavior observed in the present cases in terms of ${\tau }_1$ and $s_1$ may indicate a positive effect of the GNP, with an increase of 49% in the bond strength. Further analysis of GFRPg cases is done in the discussion section below.

Finally, in Figure 9a, the average bond stress vs slip relations for BRFP cases are shown, considering all experimental and numerical results. Again an excellent fit is obtained. The average for each case considered is presented in Figure 9b. The observed effect of the increase in concrete strength on bond strength is on the same direction as observed for GFRP bars: for 8-mm bars bond increased substantially and for 12.5-mm bars the bond presented a slight decreased. Figure 9c shows the average bond-slip relation for the two diameters tested. As in previous cases, the greater the diameter the greater the bond. However, in BRFP cases the substantial diameter effect on bond is believed to be related mostly to surface treatment. In the case of 12.5 mm bars, the presence of deformations increased the efficiency of the bar-concrete bond due to the better mechanical interlock they promote. In the 8.0 mm bar case, because of the smooth surface (see values of rib height, rh, Table 1), the bond behavior depends basically on friction and adhesion. Also 12.5 mm bars have greater roughness (Table 1), which increases friction. As Achillides and Pilakoutas [1] and others emphasize, mechanical interlock associated to surface treatment is a key factor to bond strength.

It is already possible to find other bond studies in the literature regarding BFRP bars, but not exactly the same combinations tested in the present study. Figure 9d shows a comparison of C30-XXB results with such cases. Wei et al. [32] analyzed the bond behavior for 8.0 mm diameter bars with sand coated surface treatment, obtaining a bond strength of 10.3 MPa or 33.6% higher value than current experiments. Liu et al. [30], on the other hand, investigated the bond between 12.0 mm bars with grooved surface treatment and C30 concrete, finding a bond strength of 15.8 MPa, which is 41.7% smaller than current experiments. Such comparisons suggest again that the surface conformation of the bars has a fundamental role governing bond behavior. The sand coated case of Wei et al. [32] also presented a similar peak slip $s_1$ as the present cases. However, for the grooved case investigated by Liu et al. [30], this slip was smaller.

Discussion

Despite the changes in bond strength with concrete strength in all cases reported in this work, a statistical analysis of variance (ANOVA) indicates these variations are not significant. This conclusion is in agreement with the findings of Achillides and Pilakoutas [1] and others, that concrete strength has a limited influence on bond because the bar surface will be damaged after certain level of bond, regardless the concrete.

GFRP bars will be considered here benchmarks to compare with the two new materials tested in this work. In the former cases there is an increase of the bond strength with greater bar diameter, but the main reason for the increase seems linked to the height of ribs (rh). Figure 10 then compares GFRP with GFRPg and BFRP. First, in Figure 10a, a comparison of the average values for 8 and 12.5 mm bars between GFRPg and GFRP cases is shown. As already discussed, the main difference between the two is the polymer used. It is observed that there is an equivalence in terms of bond strength ${\tau }_1$ for both cases, despite the tendency for GFRPg cases to have a slightly larger slip peak $s_1$. Therefore, results indicate that the addition of GNP was able to overcome the trend of more unfavorable behavior usually associated to polyester polymers, keeping the cost of the addition competitive with standard GFRP. This preliminary study encourages future studies to further access the effect of GNP additions in FRP bars. Finally, in Figure 10b a comparison of the average values for 8 and 12.5 mm bars between BFRP and GRFP cases is shown. For the 12.5 mm case, combination BFRP/helically-wounded presented a bond strength almost twice as much as the bond for combination GFRP/helically-wrapped. Considering that both surfaces have similar characteristics (e.g., rh and roughness), results indicate a superior behavior of the first combination when compared to the second. However, for the 8 mm cases the behavior of the combination BFRP/smooth is inferior to GFRP/helically-wrapped. The comparison emphasizes again the importance of the surface treatment of FRP bars in general, i.e., the type of fiber alone cannot be considered the determinant factor defining bond.

The parameters of the curves shown in Figure 10 are presented in

Table 6. Parameters of the numerical model obtained from the average experimental curves of same diameter and bar type.

Average Combinations	Parameters
	${\mathit{\tau }}_1$ (MPa)	${\mathit{s}}_1$(mm)	${\mathit{s}}_0$(mm)	${\mathit{\tau }}_0$(MPa)	${\mathit{\tau }}_{\infty }$(MPa)	$\mathit{\alpha }$	$\mathit{\beta }$
Cxx-8G	7.0224	2.26	0.0857	1.0096	−7.1129	7.7062	0.066633
Cxx-12G	11.1963	2.24	0.10931	3.0007	−4.8165	3.431	0.13681
Cxx-8Gg	7.8026	3.01	1.49521	4.91998	−89.6102	0.978	0.015289
Cxx-12Gg	11.0204	2.6	1.4489	7.94637	−10.8328	1.50158	0.181161
Cxx-8B	4.8111	3.16	0.188973	1.411561	−170.099	3.1396	0.006711
Cxx-12B	22.6302	2.69	0.83574	12.20998	−100.1095	1.87643	0.030432

. In general, $s_0$ is smaller and $\alpha$ is bigger for the GFRP bars, which showed higher initial stiffnesses in comparison to the other bars. Conversely, due to lower initial stiffness, GFRPg bars showed larger values of $s_0$ and smaller $\alpha$ values. The addition of GNP was unable to improve this typical characteristic of the polyester. Higher values of ${\tau }_0$ are associated to 12.5 mm bars when compared to 8 mm bars of the same type, which can be linked to higher values of rh in former cases (see Table 1).

In general, BFRP bars present more negative values for the parameter ${\tau }_{\infty }$, which is associated to a greater drop in bond after peak. Less negative values of ${\tau }_{\infty }$ are observed for GFRP bars, which presented a smoother drop after peak, when compared to other bars. The parameter $\beta$ showed little variation among the bars tested, with most of the values close to zero.

4.2. Thermal Effects

Table 7. Experimental results for specimens heated at 60 and 120 °C.

Combinations	Specimens	${\mathit{\tau }}_1$ (MPa)	${\mathit{s}}_1$ (mm)	Rupture	${\overline{\mathit{\tau }}}_1$ (MPa)	Standard Deviation (MPa)
GFRP bars
	SP1	10.05	2.97	pull-out
C30-12G-60	SP2	11.35	1.93	pull-out	10.65	0.65
	SP3	10.55	2.05	pull-out
	SP1	8.59	2.24	pull-out
C30-12G-120	SP2	7.98	2.10	pull-out	7.87	0.79
	SP3	7.03	2.28	pull-out
GFRPg bars
	SP1	8.46	2.41	pull-out
C30-12Gg-60	SP2	8.32	3.54	pull-out	8.30	0.18
	SP3	8.11	2.67	pull-out
	SP1	6.48	2.95	pull-out
C30-12Gg-120	SP2	5.64	3.26	pull-out	6.32	0.62
	SP3	6.85	2.74	pull-out
BFRP bars
	SP1	23.70	2.53	pull-out
C30-12B-60	SP2	17.45	2.65	pull-out	20.11	3.23
	SP3	19.19	3.04	pull-out
	SP1	17.72	2.55	pull-out
C30-12B-120	SP2	13.87	2.73	pull-out	15.90	1.94
	SP3	16.12	2.63	pull-out

presents the values of ${\tau }_1$, $s_1$ and the rupture mode developed in the pull-out tests after exposing specimens to 60 and 120 °C. The average maximum stress value ${\overline{\tau }}_1$ and the corresponding standard deviation are also presented. Only C30 concrete and 12.5 mm bars are considered. Rupture always occurred by pullout without significant differences regarding room temperature ruptures. However, for 120-degree cases, in general, the polymer color changed and the degradation of the outer surfaces was more intense than lower temperature cases. Adjusted numerical parameters used in the numerical simulations are presented in the Appendix A.

Results of average bond stress versus slip for all GRFP cases are shown in Figure 11a and average values per temperature are shown in Figure 11b (the figures also include room temperature results). From Figure 11b is observed that bond strength after heating process at 60 and 120 °C decays by 16.21% and 38.08% respectively, when compared to room temperature results. Also in Figure 11b, temperature effects on GFRP bars are presented as tested by Katz et al. [33] and Solyom et al. [34]. In the former case, test conditions were fairly similar to the present work. However, tested temperatures were different (130 and 205 °C) but interpolating data to temperatures tested in this work, a bond strength decayed by 13.89% and 31.62% is observed, respectively. In Solyom et al. [34] grooved surface, 8 mm bars and test temperatures ranging from 80 and 165 °C were considered. Interpolating to temperatures tested in this work, these cases showed a decrease in bond strength of 23% for 60 °C case and 40% for 120 °C case. Therefore, thermal effects here presented for GRFP bars are on par with results of the literature. Interesting to note that all heated cases present a bond strength smaller than 12 MPa (minimal required by GOST 31938-2012 Standard [57]). Finally, Figure 12 shows temperature degradation curve according to the model by Correia et al. [55] for the analysis done in the present work, as well as the interpolated results obtained from Katz et al. [33] and Solyom et al. [34]. It is possible to see that results of the present work shows an intermediate behavior. For the present results the following parameters were obtained by optimization analysis (see Equation (4)): $P_r=0.62$, $B=-1238.95$ and $C=-0.1214$.

Results of average bond stress versus slip in the cases of GRFPg bars are shown in Figure 13a. Average values per temperature are shown in Figure 13b, including also room temperature results. From Figure 13b is possible to observe that bond strength after heating process at 60 and 120 °C decays by 15.48% and 35.64% respectively. As pointed out earlier, the existence of other GRFPg studies are unknown. However, as a matter of comparison, the Figure 13b shows again results presented by Katz et al. [33]. As discussed in the room temperature results, polymer used is also polyester and other conditions of the experiments in this case are similar to present GRFPg experiments, except for the addition of GNP. Temperature tested in Katz et al. [33] were not the same used in the present work, but through interpolation it was possible to estimate a decay in bond strength of 32.93% for 60 °C and a decay of 63.19% for 120 °C. Hamad et al. [36] also considered thermal effects in bars made with polyester but the bond strength at room temperature presented was far below 120 °C bond strength of the present experiments and for this reason they were not considered here. Visually the degradation with temperature can be seen in Figure 14 for the analysis done in the present work, as well as the interpolated results obtained by Katz et al. [33], considering the model by Correia et al. [55]. For the present results parameter obtained by optimization analysis are $P_r=0.64$, $B=-1124.95$ and $C=-0.1201$. It is observed that Katz et al. [33] results presented a degradation with temperature around 44% greater when compared to present results, which may indicate a positive effect of the GNP addition.

Finally, in Figure 15a thermal effects on average bond vs slip relations are shown for BRFP bars. The average per temperature obtained is shown in Figure 15b. Bond strength in this case deteriorated by 10.18% and 28.99% after 60 and 120 °C heating, respectively. Thermal effects for BRFP bars are also shown in Figure 15b for a case analyzed in the literature Li et al. [35] (Hamad et al. [36] also tested BRFP bar at high temperature. Results are not shown here because room temperature bond strength was already far below present results at 120 °C). Bond strength degradation for this case are 7.92% and 20.77% for 60 and 120 °C, respectively. Figure 16 shows the thermal degradation for both cases using the model by Correia et al. [55]. In the present simulations the parameters obtained by optimization are $P_r=0.71$, $B=-858.81$ and $C=-0.1119$. It is observed a greater degradation in the present results when compared to Li et al. [35], but room temperature bond strength of the latter is already relatively low when compared to present results.

Discussion

In Figure 17a all thermal results are grouped for the three types of bars considered. It is evident the superior behavior overall of the combination BRFP/helically-wounded bars when compared to the others regarding bond strength ${\tau }_1$ in all temperatures considered, being this case the only where residual strength after the 120 °C heating cycle is above 12 MPa. Interestingly peak slip $s_1$ is very similar in all tests and ranges between 2–2.5 mm. Comparing curve parameter in all cases, see

Table A4. Numerical model parameters obtained from the average experimental curve for each combination tested as a function of the temperature.

Combinations	Interpolated Data		Optimized Parameters
	${\mathit{\tau }}_{\mathbf{1}}$  (MPa)	${\mathit{s}}_{\mathbf{1}}$  (mm)	${\mathit{s}}_{\mathbf{0}}$  (mm)	${\mathit{\tau }}_{\mathbf{0}}$  (MPa)	${\mathit{\tau }}_{\infty }$  (MPa)	$\mathit{\alpha }$	$\mathit{\beta }$
C30-12G-60C	10.2925	2.05	0.26389	3.498	5.1516	4.3894	0.3236
C30-12G-120C	7.8488	2.10	0.21999	2.5257	−5.5794	4.1643	0.17586
C30-12Gg-60C	8.0323	2.89	0.266385	1.88402	-75.5048	3.24813	0.0306622
C30-12Gg-120C	6.2872	2.74	0.26998	1.53913	−49.7424	3.31262	0.0413618
C30-12B-60C	19.9539	2.64	0.54549	4.0614	3.2486	3.588	0.20887
C30-12B-120C	15.8601	2.63	0.614717	4.1399	−31.1235	2.68845	0.13722

in Appendix A, there is a consistent drop in the initial slope (${\tau }_0/s_0$) of the bond curve with temperature, driven by a consistent drop in ${\tau }_0$. Combination GFRP/helically-wrapped remained the case with greater initial bond stiffness among all, in all temperatures.

GFRP cases showed the greatest initial slope (${\tau }_0/s_0$) ranging from 37 to 12 with temperature, while GFRPg had always the smallest initial slope ranging from 6.7 to 5.9 with temperature. Considering the same conditions, ${\tau }_0$ is always prone to be greater for greater diameters. Finally, BFRP cases tended to have the most negative ${\tau }_{\infty }$ values, or the drop of bond after peak was more severe.

Figure 17b compares thermal degradation of the three cases tested using the model by Correia et al. [55]. It is observed again a superior behavior of BRFP cases when comparing with glass fiber cases. This is a conclusion also obtained in comparative studies done by Hammad et al. [36] and Li et al. [35], despite considering other types of polymers, concrete and diameter. The particular comparison between GRFP and GRFPg shows the latter with an equivalent thermal degradation or a slightly superior behavior, in particular for 120 °C case. Considering that polyester in general has a greater degradation than vinyl-ester with temperature, it is possible that the enhanced behavior of GRFPg is due to the addition of GNP.

The better behavior overall at all temperatures associated to BRFP seems to be linked to combination basalt+helically wounded surface. The far better behavior of this combination at room temperature is the distinguishing factor. Regarding GRFPg results, they are equivalent to GRFP+vinyl-ester but far superior than GRFP+polyester. Therefore results are encouraging but more studies are necessary in this regard.

Finally, Figure 18 shows normal Cauchy stress in y direction for case C30-12B (a) and case C30-12B120 (b) at the peak of the bond for both cases. These are the cases with maximal bond at room temperature and at 120 °C, respectively. It is observed that concrete strength is not attained in any case, assuring that concrete cover was sufficient and splitting rupture was not an issue. Also, the hypothesis of considering concrete as elastic in the simulations is correct. It is observed that concrete stresses near the adherent zone was mostly uniform, except near the ends of this zone. This is an expected result because in these regions rebars stop sharing stresses with concrete. Other cases showed a similar stress distribution and for this reason are not depicted. Finally, as expected, stresses in concrete are much smaller for the 120 °C case due to bond and elastic properties degradation.

5. Conclusions

In this work, FRP bars using basalt as fiber (BFRP) and bars with the addition of GNP in the matrix (GFRPg) were tested against a more traditional type of FRP bar where glass fiber and vinyl-ester matrix (GFRP) were used. The study with GNP was exploratory and only a fixed amount was added in a polyester matrix. Temperature effects up to 120 °C were also considered as well as numerical simulations. Different diameters and concrete classes were taken into consideration at room temperature. The main conclusions are:

• Combination BRFP/helically-wounded presented a superior behavior in terms of strength than other cases studied in this work and other BRFP cases considered in the literature at room and higher temperatures. The former was the only case to have a bond strength above minimum levels defined in project standards such as GOST 31938-2012 [57] and CAN/CSA S807-19 [58], in all situations tested in this work.
• Cases were GNP was introduced in a polyester matrix (GFRPg) presented a similar behavior of bars were vinyl-ester matrix was used (GFRP). However, behavior of the former was far superior than similar cases taken from the literature (polyester without GNP), at room and higher temperatures. Therefore results are encouraging, but more studies are necessary regarding GNP use.
• Numerical simulations combining Rolland et al. model and an optimization method to search for parameters showed to be an excellent tool to fit experimental results. Stresses in the concrete remained always below concrete strength. At the adherent zone, these stresses stayed mostly uniform, except near the end of the zone where stress concentrations were observed.