In order to improve the computational efficiency of load distribution of large three-row roller slewing bearing, a new computational model using nonlinear springs is presented in this paper. In this model, each roller is simulated by a group of nonlinear springs, which has the same load-deformation performance with solid roller–raceway contacts. Thus, equivalent finite element roller–raceway contact models were built to acquire the load-deformation performance and the solid roller experiments were carried out to validate the computational results. Then, the computational accuracy and efficiency using the nonlinear spring-based model were analyzed from three aspects, the maximal contact load, the root mean square errors, and computation time. The results show that a group of parallel springs can be used to replace the solid roller and simulate the line contact performance between the roller and raceway, but an equivalent nonlinear spring is not recommended because of the low computational accuracy, the greater the spring number, the higher the computational accuracy, and the computation time of nonlinear spring-based models is just about 3–4% of time of solid roller-based model, but the model complexity and establishment time of a group of nonlinear springs will increase with the number of springs increasing.

During the operation, a slewing bearing is always subjected to a set of combined loads. It is the source of deformation of ballraceway contacts, rings, and even supporting structures. In practice, deformation of rings and supporting structures is often neglected for simplification, that is, they are supposed to be ideally stiff. To take elasticity of rings and supporting (fixed) structures into consideration, the finite-element method (FEM) is applied. Due to hundreds of contact pairs and the difference in the scale of contact area and rings or supporting structures, it is difficult to simultaneously model both local ballraceway contacts and the global slewing rings in a slewing bearing. The article developed a simple method to solve the problem, where the contacts are replaced by non-linear springs.

The axle box in the bogie system of subway trains is a key component connecting primary damper and the axle. In order to extract deep features and large-scale fault features for rapid diagnosis, a novel fault reconstruction characteristics classification method based on deep residual network with a multi-scale stacked receptive field for rolling bearings of a subway train axle box is proposed. Firstly, multi-layer stacked convolutional kernels and methods to insert them into ultra-deep residual networks are developed. Then, the original vibration signals of four fault characteristics acquired are reconstructed with a Gramian angular summation field and trainable large-scale 2D time-series images are obtained. In the end, the experimental results show that ResNet-152-MSRF has a low complexity of network structure, less trainable parameters than general convolutional neural networks, and no significant increase in network parameters and calculation time after embedding multi-layer stacked convolutional kernels. Moreover, there is a significant improvement in accuracy compared to lower depths, and a slight improvement in accuracy compared to networks than unembedded multi-layer stacked convolutional kernels.