1. Introduction
Many factors can reduce the durability of a reinforced concrete (RC) structure. Among them, the corrosion of internal reinforcement materials, which is known as reinforcement corrosion, is the primary cause. The volume expansion caused by the corrosion of reinforcing bars in concrete leads to cracking or spalling in the concrete protective layer as corrosion develops, followed by durability failure of the concrete structure [
1,
2]. Corrosion of steel bars leads to a decrease in the bonding force with concrete [
3,
4,
5,
6]. Steel corrosion has caused incalculable damage to several RC structures, necessitating their repair or removal. Thus, the structural damage caused by steel corrosion has become a global disaster risk concern. Failure due to steel corrosion is a complex process, and several problems remain to be solved.
After the reinforcement has corroded, an extrusion force on the concrete around the reinforcement is induced by the corrosion products. Under this reinforcement rust expansion force, the concrete cracks further. As the degree of rust corrosion increases, the cracks gradually expand toward the surface of the protective concrete layer. Several theoretical studies have been conducted on the reinforcement rust expansion force. According to an elastic analysis in the cracking process of corroded products, Bazant et al. [
7] proposed three types of concrete failure. Weyers [
8,
9] performed a mechanical analysis of the cracking process by considering the diffusion of rust products into the concrete pores around the steel bars. Additionally, a circumferential stress formula based on elastic mechanics for the uncracked part of a cylinder with a unit length radius was proposed. Zhao et al. [
10,
11] analyzed the mechanical behavior of both the concrete protective layer and the rust layer, using elastic theory to obtain the deformation coordination relationship between the solidified soil and the rust layer. According to the theory of damage mechanics and mechanics of elastic solids, they introduced a damage variable to establish a model for the rust swell cracking of a partially cracked protective concrete layer. Li et al. [
12] employed fracture mechanics to analyze the stress and strain in concrete. Their model determined the relationship between the width of the cracks in the protective layer and the corrosion depth of the reinforcement. The model of the main theoretical basis is the mechanics of elastic solids and fracture mechanics. The calculation process was complicated and the field measured data were needed, so the engineering application is relatively difficult.
Experimental studies on the rust-driven cracking of a concrete protective layer focus mainly on two topics: (1) the time from the rusting of the steel bar to the point when cracks appear on the concrete surface or the corrosion loss at the time of cracking on the concrete surface, and (2) the development of a law to predict the width of cracks on the concrete surface resulting from reinforcement corrosion. Numerous scholars have performed experiments to establish empirical models for the concrete cracking time and crack width. Andrade et al. [
13] performed accelerated corrosion tests on four test blocks with different rebar positions, rebar diameters, and protective-layer thicknesses under the influence of different electric fields. Assuming the uniform corrosion of the steel bars, the corrosion depth of steel at a given time was calculated according to Faraday’s law, using the energizing current and the energization time. Alonso et al. [
14] adopted similar test equipment at different current intensities to establish the relationship between the corrosion depth and the current intensity. Oh et al. [
15] used the strain value to judge concrete cracking, determining that the initial crack appears on the concrete surface of the protective layer when the monitored strain reaches the value of the concrete’s cracking strain. Thus, the relationship between the corrosion rate of the steel reinforcement and the cover thickness was established by determining the point at which the concrete surface of the protective layer cracks. Song et al. [
16] investigated the corrosion rate of rust expansion cracking on admixture concrete using accelerated corrosion tests under an electric field. They established a mathematical model to describe the steel corrosion rate in the case of a concrete protective layer under different environmental conditions, with consideration of the similarity between rapid electrochemical corrosion and the natural corrosion process.
Various factors contribute to the cracking of concrete structures due to corrosion expansion, but these factors cannot be comprehensively analyzed using empirical formulas. Previous research findings cannot be applied to other rust crack tests based on a small subset of factors. Therefore, in this study, a steel corrosion rate model was established for the situation in which a protective concrete layer is cracked, with consideration of various factors, such as the concrete cover thickness, water–binder ratio, and diameter of the steel bar.
Failure of concrete under the action of the steel rust expansion force is simulated using finite element software or a custom computer program developed for this purpose. Val et al. [
17] studied the rust filling process in the free expansion stage of corrosion using a finite element model of a two-dimensional (2D) plane with holes. The difference between the experimental corrosion value at the time of cracking and the corresponding numerical simulation results reflects the quantity of rust filling. Jang et al. [
18] performed a numerical simulation to examine the concrete cracking caused by non-uniform corrosion. The diameter of the steel bar, the cover thickness, and the rebar diameter at the time of concrete cracking were determined by adjusting the loads, material parameters, and geometrical dimensions. Zhao et al. [
19] analyzed the distribution model of a rust layer of varying thickness around a steel bar. As the basis of the displacement load in the finite element model, corroded RC specimens were prepared to facilitate the study of the non-uniform rust-driven cracking process. The stress field distribution of the concrete caused by the non-uniform corrosion of the internal steel bar, the development and distribution of the rust expansion crack, and the corrosion expansive force were obtained by a finite element model. Ožbolt et al. [
20] used a three-dimensional (3D) numerical model for transient analysis of the processes after depassivation of reinforcement in concrete, which is relevant to the calculation of the corrosion rate. The model predicted that cracks do not influence the corrosion rate for the case where the only influence of the crack is on the rate at which oxygen can reach the steel. Du et al. [
21] examined the cracking of the concrete protective layer caused by non-uniform reinforcement corrosion using a microscale numerical simulation. The effects of the reinforcement diameter, cover thickness, and reinforcement-position on the cracking of the concrete protective layer under non-uniform corrosion conditions were investigated. Zhang et al. [
22] numerically simulated a non-uniform corrosion of reinforcement that causes concrete cracking in chloride-contaminated RC structures. The effects of the concrete cover thickness, rebar diameter, and rebar spacing on the failure patterns of the concrete cover and the crack propagation were examined, and limiting criteria for the cracking modes were established. Fahy et al. [
23] considered the transport of corrosion products into pores and cracks in concrete when predicting corrosion-induced cracking in RC structures. The pressure-driven transport was studied using an axisymmetric thick-walled cylinder model and a network method. Zhen et al. [
24] studied the effects of different factors on the crack initiation time and crack propagation by using the thermal simulation method with a 3D nonlinear finite element model. The results indicated that the types of corrosion products, thickness of the interfacial transition zone, and corrosion rate are the most significant parameters that affect the crack initiation time. Xi et al. [
25,
26] established a mesoscale mixed-mode fracture model for concrete structure cracking. The effect of aggregate randomness on the crack width development of the difference between the uniformity and non-uniformity of concrete structures is that comprehensive parameters of corrosion have been investigated and proposed. Yang et al. [
27] performed a numerical prediction of the concrete crack width and developed a numerical method for predicting the concrete crack width for corrosion-affected concrete structures. Accurate prediction of the crack width and timely maintenance are important for the service life of RC structures.
At present, the standard approach for the simulation of the process of concrete cracking caused by corrosion expansion is the use of a 2D model with a circular hole inside the concrete. However, such a 2D model cannot be used to analyze the interaction between sections, fracture development, or concrete spalling after rust-driven cracking. Therefore, in this study, focusing on the non-uniform corrosion of reinforcement in natural environments, an unequal radial displacement distribution function for concrete around angular and non-angular steel bars was deduced using the characteristics of the rust layer. A mathematical model for the critical radial displacement at the point when the concrete protective layer cracks was obtained in accordance with the results of a numerical regression analysis. The relationship between the radial displacement and the steel corrosion rate was used to obtain a calculation model for the corrosion rate of steel at the time of cracking. The prediction model was used to estimate the cracking time of the protective layer. Finally, the non-uniform rust calculation model was validated by comparing the results of the numerical analysis with experimental results.
3. Numerical Analysis Model
At present, there are two approaches for the analysis of the concrete protective layer cracking: the theoretical analytical approach and the experimental approach. The theoretical analysis mainly involves mechanics analysis and finite element analysis. The elastic mechanics analysis approach adopted by scholars involves the consideration of the elastic mechanics of a thick-walled cylinder model in which the expansion force of the corrosion product is assumed to be evenly distributed on the inner wall of the concrete cylinder. However, this assumption is unrealistic. An additional drawback of elastic mechanics analysis is the difficulty of accurate quantitative determination of the damage caused by concrete cracking. Although finite element analysis is a recommended and an accepted method for the analysis of concrete crack damage caused by steel corrosion, an alternative approach—experimentation—must also be considered. The experimental techniques include the impressed-current accelerated test method and the simulation test method. The former is generally widely used by researchers, as it has the advantage of relatively short test time. However, the corrosion results of the impressed-current accelerated test method correlate poorly with natural corrosion conditions. Meanwhile, the simulation test method is poor for simulating the uneven nature of reinforcement corrosion. Compared with the experimental method, the theoretical analysis method’s advantages of a short analysis period and low cost ensure that it plays an important role in the field of concrete crack damage research, including the establishment of a reinforcement corrosion model both before and at the time of concrete cracking [
31,
32,
33] and the consideration of the quantitative relationship between the crack width and the corrosion rate. In the present study, the finite element software was used to investigate the cracking process of the concrete protective layer, caused by non-uniform corrosion and expansion of the steel reinforcement, and to identify the influencing factors.
3.1. Constitutive Relation and Failure Criterion of Concrete
The concrete constitutive relation refers mainly to the stress–strain relation of concrete subjected to uniaxial and multiaxial stresses. A concrete dispersion crack model is used to describe the nonlinear behavior of concrete caused by isotropic hardening under compression elastoplastic and elastic cracking tension. The compression yield surface and crack detection surface function on the p–q plane (compression damage) is shown in
Figure 3, and the compressive yield surface and crack detection surface function under biaxial stress (tensile damage) is shown in
Figure 4.
The nonlinear behavior of concrete under a load is expressed by the model partitions. The isotropic hardening elastoplastic model is used in the pressure–pressure model area and the elastic cracking model is used in other areas. The linear-elastic constitutive relation is adopted before concrete cracking, the crack detection surface (tensile damage) is determined by the stress state and fracture direction at the time of cracking, and the elastic constitutive relation is adopted after cracking. The equivalent hydrostatic pressure and equivalent deviatoric stress are used to express the compressive yield surface and crack detection (tensile damage), respectively. The “Failure Ratios” option defines the shape of the failure surfaces. Four values are necessary to express the model:
- (1)
The ratio of the biaxial ultimate compressive stress to the uniaxial ultimate compressive stress;
- (2)
The ratio of the uniaxial tensile ultimate failure stress to the uniaxial ultimate compressive stress;
- (3)
The ratio of the principal plastic strain component under the biaxial ultimate compressive stress to the plastic strain component corresponding to the uniaxial ultimate compressive stress; and
- (4)
The ratio of the principal tensile stress at cracking (the other two principal stresses combine to form the ultimate compressive stress) to the stress at the time of uniaxial tensile cracking.
The uniaxial compressive stress–strain relationship of the concrete was defined by the “concrete” option. The uniaxial compressive stress–strain relationship of GB50010-2010 [
34] was adopted in this study. The relational model is given as follows:
where
fc represents the maximum uniaxial compressive stress,
εc represents the maximum uniaxial compression strain, and
αa is the relevant parameter of the uniaxial compression stress–strain curve.
Cracking is one of the most important mechanical behaviors of concrete materials. The crack expression and the shape simulation after cracking are key parts of the model. In the numerical analysis of concrete, considering nonlinear behavior, the key problem is how to simulate concrete cracking. In this study, a dispersion cracking model with independent crack detection surfaces was adopted. Concrete cracks develop when the stress reaches the crack detection surface. The fracture directions are then stored and used for subsequent analysis and calculations. Because the damage theory is used, the fracture direction can affect the subsequent calculation once the crack is generated (crack may be open or closed). The subsequent failure behavior of the crack zone was simulated via “tensile hardening,” which must be defined in the concrete cracking model. There are two tension stiffening modes: the post-failure stress–strain relation and the fracture energy crack criterion. The post-failure stress–strain relation mode was used for plain concrete, but its calculation results exhibited mesh sensitivity. To solve this problem, the fracture energy crack criterion mode was adopted. The displacement of the crack surface was 0.05 mm when the stress was 0 in the model. The setting “TYPE=DISPLACEMENT” under the “TENSION STIFFENING” option in the model parameters was used to apply the initial displacement.
3.2. Finite Element Model
An RC specimen with a section size of 200 mm × 200 mm was simulated using the finite element analysis model, in which the reinforcement was replaced with an equal diameter hole. C3D8R elements were used to simulate the concrete. C3D8R is a 3D eight-node linear hexahedral solid heat conduction unit that adopts the linear reduction integral of fine mesh division. Structural grid generation technology was adopted, and the grid size was 0.005 m. The meshes around the holes are more refined, in which the grid size was 2 mm. The density and shape of the mesh division must be specified for finite element calculations. A relatively convergent solution can be obtained via trial-and-error calculations with different meshing densities. The steel and concrete were assumed to be perfectly bonded. An “embedded” command was utilized to simulate the interaction between the concrete and steel. Moreover, a fixed connection at the end of the concrete cover was used as a boundary condition. In the cracking analysis, the main tasks included adding the mechanical property parameters of the materials, modifying the analysis types, adding displacement boundary conditions, applying loads, and outputting calculation results.
The reinforcement corrosion was assumed to be uniform along the axial direction, such that the reinforcement corrosion expansion could be treated as a plane strain problem. The concrete angle bar model and concrete side bar model were simulated (
Figure 5). The surrounding grid division was encrypted because of the stress concentration around the hole.
3.3. Model Parameters
The main factors influencing concrete cracking include the concrete strength, concrete cover thickness, rebar diameter, and steel position. The influencing factors considered in this study are presented in
Table 1. These factors were combined in different ways to establish 54 models for the concrete cracking process.
3.4. Loading Method
The elliptical model described above was adopted to analyze the radial displacement field, and radial displacement was applied to simulate the uneven corrosion expansion of the reinforcement. Hinged bearings were set at the model boundary, such that the models were under tension in both the radial and axial directions. To facilitate analysis, initial defects and micro-cracks were ignored in the study. For the reinforcements in the angular and non-angular areas, the radial displacement of the concrete was applied according to Equations (13) and (14).