1. Introduction
The wheel–rail rolling contact problem is a crucial issue in railway systems. The calculation of rolling contact characteristics is important for solving many contact problems, including wear, plastic deformation, RCF, vehicle–track interaction, and traction control. In many previous studies, the analysis of these contact problems used traditional half-space-based methods such as Vermeulen and Johnson’s analytical method [
1] and Kalker’s BEM [
2], which assume that the contact bodies are half-spaced and have isotropic elasticity. However, in reality, the material properties are complicated and the structures could be layered. Therefore, to accurately evaluate if it is necessary to consider the layer in further applications, it is crucial to study the layer effects on contact characteristics, including stresses at the interface and stresses in the subsurface.
In railway systems, layered structures present in the wheel–rail contact interfaces are unavoidable. For instance, head-hardened rails are widely installed in railway networks to increase wear resistance and fatigue strength [
3]. The tensile strength and hardness of head-hardened rails can be significantly increased compared with standard carbon rails, from 880 MPa and 260 HB to 1130 MPa and 340 HB for head layer material, respectively [
4]. In addition, while the rail is in service, the rail surface is affected by mechanical and thermal excursions under repeated wheel–rail rolling contact loads. As a result, the topmost surface layer of the rail is subjected to cyclic plastic deformation and phase transformation, where the microstructure evolves and its property changes to form a harder layer or a white etching layer (WEL) [
5]. The plastic deformation layer can reach a thickness of 1 mm. A WEL is brittle and has a hardness of over 700 HV [
3]. The thickness of the WEL can be up to 2 mm [
6]. Meanwhile, the elastic modulus increases with the hardness [
7]. Moreover, when repairing the rail with laser cladding, a cladding material is added to the rail surface, producing a bonded clad layer on the rail surface [
8]. This means that a layered structure is formed on the rail’s top surface. Also, third bodies in railway engineering can be considered as layered structures [
9]. In these cases, the rails become layered structures or gradient structures. Besides rails, layered structures are also observed on the surfaces of wheels. For instance, through the metallurgical observation of wheel tread damages in the field, Zhang et al. [
10] found that a microstructure change could occur in the wheel surface and become large amounts of upper bainite. This upper bainite had higher hardness and elasticity than the matrix microstructure. This layered structure was associated with stress concentration and RCF crack initiation. Zhou et al. [
11] also found that inhomogeneities in materials can affect rolling contact fatigue life in rotating mechanical components. Additionally, Du et al. [
12] conducted experimental and numerical studies and found that multi-layered materials can induce tensile failure in rolling contact systems, indicating that layered structures exist in rails and wheels. How they influence the wheel–rail rolling contact should be evaluated.
Some effort has been made to solve rolling contact problems associated with layers. Since the 1960s, analytical and numerical models have been developed to study the effect of elastic and viscoelastic layers bonded to an elastic half-space in rolling contact. For instance, Burton [
13] proposed an analytical method for solving visco-elastic effects in the lubrication of rolling contact. Margetson [
14] analyzed the rolling contact of a rigid cylinder over a smooth elastic or viscoelastic layer. The bonded and non-bonded layer effects on contact pressure were also studied. Londhe et al. [
15] proposed an extended Hertz theory for the contact of case-hardened steels. In these above studies, only normal contact problems were considered, and tangential problems were not included. To solve the tangential problems associated with layers, Kalker [
16] proposed a numerical method based on the BEM for two-dimensional visco-elastic multilayered cylinders rolling contact. To study the influence of a third body layer, Meierhofer et al. [
17] proposed a model for calculating the traction characteristic for the wheel–rail contact with a third body layer. However, in these studies, only two-dimensional problems were considered. Analytical and numerical methods were also developed to solve three-dimensional rolling contact problems with layers. Zhang and Yan [
18] proposed a semi-analytical model for the effects of subsurface inhomogeneous on frictional rolling contact. Xi et al. [
19] recently proposed a numerical model for solving three-dimensional rolling contact problems with elastic coating layers. Goryacheva and Miftakhova [
20] developed a model for an elastic sphere rolling over a thin viscoelastic layer bonded to an elastic half-space. Guler et al. [
21] used a numerical method for the tractive rolling contact problem between a rigid cylinder and a graded coating. Although some of these studies considered both normal and tangential problems associated with layers, they assumed the substrate to be half-spaces. This means that the layers are bonded or not bonded to the half-spaces. The contact geometries were also simplified.
In reality, more complicated material properties and actual geometries have to be considered in the wheel–rail contact as they have significant effects on wheel–rail contact behavior; thus, the half-space assumption has to be abandoned. FE methods can meet the requirements. To this end, an explicit FE method has been developed to solve rolling contact problems. The Explicit FE method is suitable for solving wheel–rail transient rolling contact [
22]. With this method, all possible creepages, complex materials, realistic geometries, and dynamic effects can be included. It was first applied to the dynamic wheel–rail frictional rolling contact by Zhao et al. [
22]. Later, it was used for rolling contact with spin [
23] and compression-shift-rolling contact with partial slip [
24]. Its accuracy has also been validated against traditional methods by approximating it to quasi-static states. Furthermore, this method has been applied to more complex geometries and materials such as surface defects [
25], corrugation [
26], and material discontinuities [
27]. In the dynamic wheel–rail contact, it was compared with methods based on multibody dynamics [
28] and was shown to have better performance. All these studies focused on the wheel–rail rolling contact and wheel–rail dynamic interactions with isotropic materials. However, the effects on the wheel–rail rolling contact characteristics have not been particularly studied using the explicit FE method where the layered structures of the rail are involved.
In this study, the explicit FE method proposed in [
22] is employed to investigate the influence of rail surface layered structure on wheel–rail rolling contact characteristics such as contact stresses, contact patch, and stick–slip. The study considers a general case of the gradient layered structure instead of focusing on a particular engineering problem such as the head-hardened rail surface or WEL. The top layer is assumed to have a different elastic modulus from the matrix material. Different elastic moduli and layer thicknesses are considered. In total, five thicknesses, from 0.2 mm to 6 mm, are included; four different elastic moduli are employed for each thickness. The resulting stick–slip condition, surface contact stresses, and subsurface stresses in the rail are investigated.
5. Discussion
Layered structures exist in rails and wheels. The analysis of contact characteristics in previous studies often neglected the layered structures. To determine whether it is necessary to consider the layered effects in actual engineering applications, this study proposed an explicit FE method to study the layer effects. The proposed explicit FE method has high accuracy for wheel–rail rolling contact when an appropriate mesh is applied. The accuracy of frictional rolling has been validated by Zhao and Li [
22] for the 0.33, 0.63, and 1.3 mm mesh sizes in the contact surface when the material is homogeneous. Ref. [
22] also suggested that the accuracy may be acceptable when the size is 1/10 of the minor axis of the contact patch. To balance the computational cost and accuracy in the present study, the 1 mm mesh is used on the surfaces of the wheel and the rail. This mesh can also give reasonable accuracy compared with Kalker’s BEM, as shown in
Figure 4. The element size is about 1/10 of the minor axis of the contact patch, which agrees with the conclusion in [
22]. When the top layer structure is induced, the stresses at the interface change. Kalker’s BEM is not appropriate for analyzing cases with layered structures as the accuracy of the contact stresses is difficult to validate directly; however, based on the results in
Figure 4 and
Figure 5, the harder layer tends to produce larger stresses. This tendency is reasonable, because a harder layer produces a smaller contact patch and, therefore, larger stress.
It is worth noting that there is only one element across the thickness of the layer in the model. To evaluate if this mesh is reasonable for the analysis of the layer effects, two elements across the layer thickness is applied to the case of 1 mm thickness for comparison. The calculated stresses at the interface and the stresses in the subsurface are compared, as shown in
Figure 12. As can be seen, the difference is within 1%. Thus, the results of the two mesh types are in good agreement. Moreover, the stresses in the subsurface based on the two mesh types are compared in
Figure 13. As can be seen, the stress distribution and the magnitude are very close, and the difference in the maximum value is less than 1%. The small differences indicate that the mesh has a negligible influence on both the stresses at the interface and in the subsurface. Therefore, a mesh with only one element across the thickness is reasonable for studying the layer effects.
When the layer structure appears, the accuracy of the stresses in the subsurface is difficult to validate directly. The present study can be qualitatively validated by comparing studies on multi-layer materials under scratch contact [
12,
33]. In these studies, the contact in scratch is full sliding contact. Some contact characteristics are similar to those in the rolling contact; therefore, they can be used to verify the results in the present study. The authors of [
33] studied the stress in multi-layer polymetric systems under scratch contact. They found that a soft substrate was incorporated to induce larger stress fields at the interface of multi-layer materials under scratch contact. The finding is in line with the finding of the effects of elastic modulus of the top layer on the stress at the interface, as shown in
Figure 8 and
Figure 9. Ref. [
12] studied the delamination of multi-layer materials under scratch. They found that tensile maximum principal stress developed at the interface of the multi-layer polymeric structure under the scratch, and the magnitude and direction of the peak tensile maximum principal stress were affected by both the thickness and the material parameters of each layer. Therefore, the findings on the effects of the layer on the maximum principal stress (See
Figure 10 and
Figure 11) in the present study are in agreement with the observed phenomena in these references [
12,
33].
The effects of the layer sometimes fluctuate with the elastic modulus, such as the influence on the surface shear stress in
Figure 5 and the influence on
in
Figure 11. There are two factors that can induce this fluctuation. One is that the effects are determined by both the value of the elastic modulus and the thickness. Another one could be dynamic fluctuations. Although dynamic relaxation is applied to minimize the dynamic effects for comparison of different cases in the numerical simulations, some fluctuations are unavoidable due to the inherent high-frequency vibration of the continuum [
22], such as the fluctuations in the surface shear stress in
Figure 5. The fluctuations would cancel out or amplify the layer effects. In this case, it is difficult to see the trends of slight effects, such as the effects on surface shear stress in
Figure 5b,c. However, when the effects become more significant as the thickness increases, the trends in the effects become more apparent (see
Figure 5e). Therefore, a general trend still exists in the effects. The layer effects should be primarily determined by both the values of the elastic modulus and the thickness.
Based on the numerical results, the influence of the elastic modulus of the top layer is associated with its thickness, such as the influence on the contact pressure in
Figure 4 and the influence on
in
Figure 10 and
Figure 11. The critical factor behind this could be the ratio between the size of the contact patch and the thickness of the top layer. Therefore, if the size of the contact patch and the thickness are changed, the influence could change. Moreover, the traction coefficient significantly influences the surface shear stress and subsurface stresses. Thus, the layer effects could be also associated with the traction coefficient. In future studies, more traction coefficient values can be considered for comparison.
Although layer effects in rolling contact problems have been extensively studied, e.g., in references [
13,
14,
15,
16,
17]. In these studies, the substrates connected with the layers were assumed to be half-spaces. In reality, the geometries of the wheel and rail are complicated, and the realistic geometries influence the contact characteristics. A robust method that considers realistic geometry is vital in the study of layer effects in engineering applications. Therefore, the explicit FE method was proposed in this study. The realistic geometries of the contact and the substrate were considered. More realistic results were obtained. The results (
Figure 4,
Figure 5,
Figure 6,
Figure 7,
Figure 8,
Figure 9,
Figure 10 and
Figure 11) proved that the layer has significantly influence on the contact characteristics, including stresses at the interface, stick-slip behavior and stresses in the subsurface. This implies that the layer structure may need to be considered in more precise practical engineering applications, such as the analysis of wear, RCF and adhesion. For instance, as shown in
Figure 10, the layer can induce high tensile stress, which could be the driving force for RCF crack initiation in some conditions. RCF due to WEL could belong to this case, as reported in [
34], where crack initiation due to WEL is caused by a brittle fracture. This is why RCF cracks often occur in WEL on rails [
35]. The present study analyzed a general case of the layer. The precise RCF mechanism still needs detailed analysis to consider more realistic material parameters. Moreover, to study more specific conditions, future studies should include the layer effects of the wheel.
The implication of the results can be addressed from another point of view. At present, some traditional methods are still employed for some wheel–rail contact problems, such as wear prediction (e.g., [
36]) and analysis of RCF (e.g., [
37]). These methods cannot deal with the problems of layered structures. Since layered structures exist and have a great influence on stress distribution, these methods are not applicable to some practical railway problems. Instead, an FE method should be considered. Precise analyses with consideration of more realistic parameters are still needed to study specific engineering problems.
In the present study, two elastic layers were considered, and the thickness of the layer was assumed. In practice, the material may have multiple layers, and the thickness may vary due to different causes of the layer. Even the material properties gradually change across the thickness. These cases still need to be studied further. In future studies, elastic/plastic material properties should also be considered when dealing with engineering problems.
6. Conclusions
This study employed an explicit FE method to analyze the elastic layer effects on contact characteristics at the interface and in the subsurface of wheel–rail rolling contact. The top layer structure in rails is considered to have a different elastic modulus from the matrix material. Contact characteristics such as stick–slip behaviors, surface contact stresses, and subsurface stresses are investigated. The accuracy of the solutions with applied mesh was validated by comparing them with Kalker’s BEM and other existing studies. The following conclusions can be drawn:
The top layer of the rail may alter the contact stresses at the interface. A harder layer induces higher stresses at the interface while reducing the size of the contact patch. The ratios of the stick area and the slip area change as the layer becomes harder. These layer effects tend to become more significant when the thickness of the layer is larger. However, the degrees of these effects are determined by both the value of the elastic modulus and the thickness of the top layer.
The layer also affects the stresses in the subsurface of the rails. There is a tendency for a harder layer to induce larger v-m stress in the subsurface. However, the harder layer may increase or reduce the tensile stress, depending on the thickness. Simultaneously, the layer may change the location of the tensile regions from the surface to the subsurface when the elastic modulus is increased. In general, the degrees of these effects depend on both the elastic modulus and the thickness of the layer.
Layer structures exist in practice, and layer effects are important for wheel–rail rolling contact consequences. The analysis of rolling contact consequences such as wear and RCF may require the inclusion of layer effects by considering more realistic parameters. Moreover, from another point of view, traditional popular methods such as Kalker’s BEM and the Hertz theory without consideration of layered structures are not sufficiently accurate for analyzing contact consequences when layered structures exist in the wheel and rail. Instead, methods that consider layered structures should be used.