1. Introduction
Along with the strikingly rapid development of composite materials in the field of civil engineering, the applications of fiber-reinforced polymers (FRP) in strengthening concrete structures have attracted more and more attention. CFRP has higher mechanical properties, excellent fatigue resistance, corrosion resistance, and creep resistance. BFRP and GFRP have relatively low prices and wide sources. However, their long-term performances in service environments may degrade [
1,
2,
3]. The applications of FRP can be divided into the form of FRP tendons as internal reinforcements or FRP laminates as externally bonded reinforcements. In the previous decades, extensive analytical and experimental studies have been conducted on the local bond–slip relationship of concrete flexural members reinforced by FRP tendons [
4,
5,
6,
7,
8,
9,
10,
11] or concrete members strengthened by externally bonded FRP laminates [
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22]. It was found that a sound understanding of the bond behavior between FRP reinforcement, and the concrete substrate played a major role in the development of design guidelines and the performance evaluation of FRP-strengthening concrete members. Therefore, a reliable and rigorous analytical model based on the innovative partial composite action, taking into account the corresponding bond characteristics, is essential to accurately assess the mechanical properties of strengthening or retrofitting concrete structures using FRP. A new method [
23] was previously developed for determining the transfer length of pretensioned concrete with prestressing FRP tendons, which was solved with closed-form solutions using composite beam theory associated with considering the local bond–slip relationship between FRP tendons and concrete. In the presented paper, finite element modeling of pretensioned concrete members with prestressing FRP tendons is proposed in order to provide further verification of the developed analytical methodology.
Composite beam theory, proposed by Granholm [
24] in 1949 and Newmark et al. [
25] in 1951, was initially used to solve the cases of nailed timber structures and T-beams consisting of a rolled steel I-beam and a concrete slab, respectively. In terms of partially composite action, the theoretical analysis was developed for the member consisting of two separate elements connected by discrete connectors. Furthermore, the influence of relative displacement between the two elements, i.e., the effect of slip, was fully considered. From this perspective, therefore, composite beam theory is not limited to the types of structures mentioned above but is instead devoted to a wide range of structures comprised of two or more interconnected elements under reasonable assumptions. For example, Bai and Davidson [
26] implemented a rigorous analysis of foam-insulated concrete sandwich panels in which structural deflection was discomposed into two components, shear and flexural. The structural behavior was taken into account as partially composite in terms of composite beam theory. Sha and Davidson [
23,
27] provided closed-form solutions using composite beam theory for determining the transfer length of pretensioned concrete members strengthened by FRP tendons as well as predicted the interfacial stress in concrete beams with externally bonded FRP laminates. Through the research on the developmental course of composite beam theory [
23,
25,
26], it has been observed that theoretical methods were mainly verified against the existing experimental data from the literature. However, in order to comprehensively evaluate the accuracy and reliability of the developed method, as a supplementary verification, finite element analyses (FEA) are an effective methodology that can be employed to compare with theoretical solutions.
As an important numerical technique, FEA has been widely used to study the behavior of prestressed concrete beams [
28,
29,
30]. Most research has focused on pretensioned concrete members with prestressing steel strands; only a very limited number of FE models are specifically available for pretensioned concrete with prestressing FRP tendons. The main reason for the lack of in-depth FE research in this field is the challenging nature of the interaction between FRP tendons and the concrete matrix. Hence, this paper establishes a three-dimensional FE model that simulates prestressing FRP tendons in which the transfer length is determined. In addition, the different friction coefficients between FRP tendons and the surrounding concrete obtained by experimental studies [
31] are fully considered to improve the accuracy of FE modeling approaches.
One of the most compelling advantages of FEA over other analytical solutions is that a simulation associated with fewer assumptions may be closer to the corresponding experimental outcomes. In addition, visualizations of the pre- and postprocessing of FEA can help engineers easily find vulnerabilities in the design. Despite some obvious advantages, mesh convergence is a critical issue that must be taken into account in the process of developing FE models. In this work, a comparative study is conducted between numerical simulation with fine and coarse meshes to illustrate the effect of mesh density on convergence. Another noteworthy point is that the concrete model used for numerical simulation is based on the linear elastic assumption. Although the concrete damaged plasticity (CDP) model from Abaqus [
32] has often been used to simulate the nonlinear behavior of concrete in other studies, the strains associated with the present paper are assumed to be in a range that essentially has a linear and brittle stress–strain relationship in compression. Furthermore, pretensioned members are designed for zero tension in the concrete under service load conditions through Rabbat et al.’s [
33] tests. The main focus of this study is to determine the transfer length at the serviceability state level in which concrete has not yet cracked, and, therefore, it is reasonable to assume that the concrete is within the linear-elastic range.
In order to provide convenient use in engineering practice, the key technical challenge of this study is to develop a general form of the governing equations specifically for FRP-strengthening concrete members in terms of composite beam theory. Taking account of the empirical bond–slip relationship between FRP tendons and the concrete matrix, governing differential equations are derived in terms of the equilibrium of axial force acting on each element as well as the balance of the overall bending moment. Using the FEA commercial software Abaqus [
32], a comparison of the transfer length of prestressing FRP tendons in pretensioned concrete with those obtained by using composite beam theory is conducted. The present FE model has been established with consideration of the friction coefficient from the experimental study on the FRP tendons and prestressed concrete members. Additionally, different mesh densities are compared for the convergence analysis. As a result, a satisfactory agreement has been reached between the theoretical solutions and FEA responses, which further demonstrates the feasibility and effectiveness of the developed composite beam theory.
2. Background of the Bond Mechanism
Understanding the nature of bond behavior plays a critical role in assessing how the prestress force is transferred from the prestressing FRP tendons to the concrete. A large amount of research [
28,
30,
34,
35] indicates that the chemical adhesive, friction, and mechanical interlocking could explain the interaction between prestressing tendons and concrete. Chemical adhesives only affect the bond strength in the minimal slip range. With the increase of slip, friction and mechanical interlocking play roles in the bond strength when the adhesive bond gradually decreases. For prestressed tendons with a rough surface, such as seven-wire strands, ribbed bars, and deformed rebars, the mechanical action of the helical outer wire of a strand bearing against the surrounding concrete matrix is referred to as mechanical interlocking. It should be noted, however, that, although the contribution of mechanical interlocking to bond strength is important, it is still not the key factor. This is because the rough surface of surrounding concrete that is in contact with prestressing tendons will eventually be sheared off due to the mechanical interlock action if the pretensioned structure has sufficient confinement. However, that does not seem to be occurring [
35]. In other words, this would imply that friction known as the “wedge effect” dominates the interaction between prestressed tendons and concrete.
Friction can be defined as a relationship that is responsible for transmitting the shear and normal forces between contacting bodies, i.e., prestressing tendons and the surrounding concrete matrix. According to the commercial FE program Abaqus [
32], friction behavior is generally analyzed using the base form of the Coulomb friction model in which the critical shear stress is given by the following expression:
where
is the critical shear stress,
is the contact pressure, and
is the friction coefficient. In the Coulomb friction model shown in
Figure 1, the shear stresses between two contacting surfaces,
, is the case in which the two contacting bodies are in a state of sticking before sliding occurs, and
is when shear stresses exceed a certain magnitude defined as the critical shear stress
, which refers to the transition from sticking to slipping along the interface of contacting bodies. The slope of the function, the friction coefficient
, is in the range of 0.3 to 0.7 according to most research literature [
29,
35]. However, AASHTO [
36] reports that the value of the friction coefficient increases from approximately 0.6 to 1.4, depending on the concrete surface conditions and the shape of the reinforcement. Thus, it can be seen that some inconsistencies exist between the specification and the literature used to explain the bond behavior between concrete members and the reinforcement, which directly affects the reliability of the analysis results based on the value of the friction coefficient.
It is worth mentioning that the current work using FEM to estimate the transfer length is based on the friction coefficient specifically for FRP tendons in pretensioned prestressed concrete members. Previous studies on finite element analysis [
28,
29,
30] of pretensioned concrete members used the value of the friction coefficient recommended by the specification to address the bond behavior, which is suitable for steel reinforcement as a prestressed strand. However, when FRP tendons are considered, it is necessary to redefine the friction coefficient through the available experimental data. Khin et al. [
31] carried out pull-out tests of Vinylon and Carbon FRP tendons with cement mortar and confined by highly expansive material (HEM). During the test, the bond stress versus confining pressure for specimens was recorded using high-precision pressure transducers to determine the friction coefficient of FRP tendons from the slope of the curve. These values from Khin et al. [
31] are listed in
Table 1 and used as the friction coefficient in the presented FE model.
For another description of the bond-behavior model, the local bond stress–slip relationship
was used in the analytical model [
23] developed by using composite beam theory for determining the transfer length for FRP tendons in prestressed concrete. The results of pullout tests [
4,
5,
7,
37,
38,
39] show that the local bond stress as a function of slip depends on a variety of factors, including concrete strength, the roughness of the reinforcement surface, concrete cover, bar diameter, and epoxy resin properties. Over the years, numerous existing models of the bond stress
and slip
have been proposed to evaluate the bond performance that is established on the basis of the nonlinear local bond stress–slip relationship
between concrete and the reinforcement. Three well-known models have been developed for steel and FRP tendons, namely the Bertero–Eligehausen–Popov (BEP) model [
4], the modified Bertero–Eligehausen–Popov (mBEP) model, and the Cosenza–Manfredi–Realfonzo (CMR) model [
5]. The BEP model is defined by Equation (2), which is adopted in CEB-FIP Model Code 90 [
37]:
where
is the maximum shear stress,
is the slip corresponding to
, and
is the coefficient of 0.4 that is available for the case of steel [
37]. Considering different requirements in the engineering analysis process, the mBEP and CMR expressions were proposed as the bond stress–slip alternative analytical models given by the following Equations (3) and (4), respectively.
For the mBEP model,
is rewritten by
and, assuming
from the BEP expression,
is the slip related to
. The expression of CMR,
is the peak bond stress, and the unknown parameters
and
are determined by the curve fitting of the experimental data. More detailed reviews of these analytical models for the curve
can be found in the literature [
4,
5,
7,
23]. In previous work by Sha and Davidson [
23], the BEP expression with calibrated parameters of
and
from Focacci et al. [
7] was used as the constitutive bond–slip definition between the FRP tendons and concrete, as illustrated in
Figure 2. The latter two models, i.e., the mBEP and CMR expressions, are equivalent to the BEP expression in the case of structural analyses in which the slip is sufficiently small.
Thick-wall cylinder theory [
29] depends on the Coulomb friction model to perform the analysis on prestress transfer in pretensioned concrete members. The concrete is conceived as a hollow cylinder in which the inner diameter is equal to that of prestressed tendons and the outer diameter is the distance across the short side (diameter) of the component. Accordingly, the estimation of bond behavior relies on the radial compressive stress as well as deformation compatibility conditions of the interface between prestressed tendons and the surrounding concrete. Based on extensive experimental and analytical investigations [
5,
39,
40], many researchers nevertheless point out that the confinement pressure has a small effect on the bond strength between the reinforcement (steel or FRP) and the surrounding concrete for the situation in which the outer surface of the reinforcement bar is a spiral. However, the bond resistance strongly depends on the confined stress known as radial compressive stress in other cases such as smooth rods. Different from the thick-wall cylinder theory, the nonlinear bond stress–slip relationship is taken into account for deriving the governing differential equations using composite beam theory developed herein for analyzing the behavior of pretensioned concrete members with prestressing FRP tendons. Consequently, in addition to further verifying the previous work of predicting transfer length for prestressing FRP tendons by means of the developed FE model, the second aim of the current study is to prove the superiority of composite beam theory considering the slip effect through the comparative studies between the analytical and numerical results.
5. Comparison and Discussion
For further proving the performance of the application of composite beam theory on predicting the transfer length for prestressed FRP tendons strengthening pretensioned concrete members, the present FE model is used to compare with the previously developed analytical solutions. As can be seen from
Figure 15, it is worth noting that theoretical results for 50% and 100% release levels are in excellent agreement with those from the experiment compared to the FE model’s result. In particular, the slope of both curves related to the rate of strain change within the transfer zones can be accurately calculated by the analytical model using composite beam theory. In the corresponding zones, a small discrepancy exists between the FE model for 50% release force and strain profile measurements, which fully demonstrates that determining the transfer length is mainly dependent upon the understanding and definition of bond behavior during the simulation process. This is also the reason why the FE fine model with
significantly overestimates the values measured in the tests at 50% force release by 107%.
Since it is not possible to exactly match the values measured from testing, the local bond–slip relationship between the FRP tendons and the concrete is taken into account to predict the transfer length using closed-form solutions from Equation (26) in the analytical model. This results in an error between the predictions and the test values of 7% and 26% for high pretension (100% force release) and low pretension (50% force release), respectively. In this perspective, the accuracy of the transfer length obtained from the theoretical solution in terms of partially composite action is superior to that of the numerical simulation by using a Coulomb friction model based on FEA.
In addition, extensive studies [
26,
42] have shown that the influence of interface slip on the mechanical behavior of composite structures cannot be neglected. For this reason, the curves presented in
Figure 16 and
Figure 17 are used to conduct the comparison between the slip predicted by the closed-form solution given by Equation (25) and those from the FE models with various friction coefficients. Note that the value predicted by the analytical solution using composite beam theory is smaller than the FE model predictions; the main reason is attributed to different bond-behavior models that are adopted in theoretical and numerical solutions. The local bond stress–slip relationship
used in the analytical model is based on the experimental investigation from the available literature [
7], in which the effects of three factors on bond behavior are comprehensively considered, including chemical adhesive, friction, and mechanical interlocking, as mentioned in the previous section. Whereas only friction is modeled as a tangential behavior associated with the friction coefficient from the pullout test [
31] during the process of FE simulation, many studies have confirmed that friction plays a major role in the interaction between FRP tendons and concrete. However, the interface slip can be still reduced by the other two factors, i.e., adhesion and mechanical interlocking. This is also the reason why the prediction values for transfer length from FE models are larger than the test results summarized in
Table 3. To a certain extent, it is further proved that the analytical solution for FRP-strengthening concrete members, considering the empirical bond–slip relationship in terms of composite beam theory, is reasonable.
On the other hand, it is obvious to see that a remarkable increase in interface slip occurs when the value of the friction coefficient decreases as shown in
Figure 16 and
Figure 17. This is because the bond strength between concrete and FRP tendons is reduced as the decrease of the friction coefficient, resulting in the larger slip.
Through normalizing the parameters, the influence of the friction coefficient
tabulated in
Table 1 and the bond-stress coefficient
(
) given by the expression of
on the transfer length is compared. It has been found from the result, as represented in
Figure 18, that the transfer length of prestressed FRP tendons in pretensioned concrete members is exponentially proportional to the bond-stress coefficient
and inversely proportional to the friction coefficient
. Furthermore, from the perspective of the varying tendencies of the curves, the extent of the effect of the bond-stress coefficient is more distinctive than the friction coefficient. In essence, different forms of the function that describes the bond behavior are adopted in the theoretical solution and numerical simulation that lead to differing impacts.
In analytical solutions using composite beam theory, the BEP expression is chosen as the bond–slip relationship to estimate the transfer length of pretensioned concrete members prestressed with FRP tendons. An important difference from the linear equation modeling the interaction between contact bodies using FEM is that the function form of is a power function of the slip in the analytical solution. By understanding the concept of transfer length, the distance that the effective prestressing force is transferred by the bond stress from the prestressed tendons to the concrete in which the axial force of concrete increases from zero to a constant. In other words, there is no interactive shear stress related to the bond between FRP tendons and concrete outside of the transmission zone. This exactly fits the typical characteristics of power functions with , where the slope of the curve gradually becomes flat as the variable increases. With the use of the BEP relationship to explain the bond mechanism, another advantage is that the closed-form solution for transfer length by means of Dirichlet and Neumann boundary conditions avoids many approximations and computational effort compared to the 95% AMS method in the numerical simulation.
6. Conclusions
Testing is considered to be the best way to predict a phenomenon and obtain necessary information. However, large-scale testing is time-consuming, expensive, and has many limitations. With this consideration, along with the need to further verify the accuracy and feasibility of the previously developed method, a three-dimensional FE model of pretensioned concrete members with prestressing FRP tendons was developed. Despite the numerous numerical research on reinforced concrete beams strengthened with conventional FRP bars, none of the existing studies considered the effects of friction coefficients on the transfer length. To bridge this gap in the literature, the different friction coefficients between FRP tendons and the surrounding concrete obtained by experimental studies are fully considered in this study. Based on the data reported from the pullout test, fine and coarse FE models were implemented for convergent analysis, respectively.
In addition, the general approach of composite beam theory, as the key innovation of this research, is derived for providing convenient use in engineering practice, specifically for FRP-strengthening concrete members. Lastly, a comparison between the analytical solution and the FE simulation is carried out and discussed. The main accomplishments and conclusions are as follows:
A general form of the governing equations has been presented specifically for FRP-strengthening concrete members in terms of composite beam theory. Associating with the knowledge of the local bond stress–slip relationship between FRP and concrete, the closed-form solution can be solved under corresponding boundary conditions;
Comparisons with the experimental data demonstrate good agreement, which indicates that the proposed FE model with fine mesh is acceptable. The measured transfer length for high pretension agrees with the prediction from the fine FE model with the friction coefficient within a 10% range. The consistency between the FE model results and the previously developed analytical solutions demonstrate that theoretical results using composite beam theory are superior to that of the numerical simulation;
Although friction plays a key role in the interaction between concrete and prestressed tendons, the slip-prediction comparisons show that if the adhesion and mechanical interlocking are ignored, the bond behavior cannot be accurately evaluated;
The transfer-length prediction is strongly dependent on the adopted function form of the bond–slip relationship between the concrete and FRP tendons in the analytical model using composite beam theory. For the analytical solution of the mechanical behavior of concrete members strengthened with FRP in terms of partially composite action, the most critical issue is to have knowledge of the local bond–slip relationship in the interface region.
Therefore, it is necessary to adopt a new method to describe the bond behavior between concrete and FRP tendons in the future FE simulation. In this process, adhesion, friction, and mechanical interlocking must be fully considered in order to provide more accurate predictions and facilitate engineering applications.