Bearing fault vibration signals collected in real engineering cases often contain environmental noise which can easily mask the fault type characteristics of vibration signals, making it difficult to determine the corresponding fault type when traditional deep learning methods are used for fault diagnosis. To solve the above problem, a neural network model named multiscale CNN-LSTM (convolutional neural network-long short-term memory) and a deep residual learning model was designed, which combines a multiscale wide CNN-LSTM module and a deep residual module for rolling bearing fault diagnosis. In this model, a wide convolution kernel CNN-LSTM structure with different convolution scales is used to extract a variety of different types of frequency and sequential features from vibration signals. It is worth noting that the wide convolution kernel CNN-LSTM structure not only has stronger feature extraction performance compared with the common convolution layer but can also reduce the interference of high-frequency noise. Moreover, the deep residual module with a wide convolution kernel CNN-LSTM structure is used to further improve the feature expression ability of the proposed model. The above algorithm enables the proposed model to better extract the fault features hidden in the noise signal. When compared with some state-of-the-art methods, the experimental results showed that this model has better anti-noise performance and better generalization ability for rolling bearing fault diagnosis.

Axle box bearings are the most critical mechanical components of railway vehicles. Condition monitoring is of great benefit to ensure the healthy status of bearings in the railway train. In this paper, a novel fault diagnosis model for axle box bearing based on symmetric alpha-stable distribution feature extraction and least squares support vector machines (LS-SVM) using vibration signals is proposed which is conducted in three main steps. Firstly, fast nonlocal means is used for denoising and ensemble empirical mode decomposition is applied to extract fault feature information. Then a new statistical method of feature extraction, symmetric alpha-stable distribution, is employed to obtain representative features from intrinsic mode functions. Additionally, the hybrid fault feature sets are input into LS-SVM to identify the fault type. To enhance the performance of LS-SVM in the case of small-scale samples, Morlet wavelet kernel function is combined with LS-SVM for the classification of fault type and fault severity and the particle swarm optimization is used for the optimization of LS-WSVM parameters. Finally, the experimental results demonstrate that the proposed approach performs more effectively and robustly than the other methods in small-scale samples for fault detection and classification of railway vehicle bearings.

It is difficult to obtain sufficient data for some machines, in addition, different working conditions result in different distributions of training data and test data, which lead to the failure of traditional deep learning methods in engineering applications. To solve these problems, we propose a novel deep learning framework called 1D-WCLT for rolling bearing fault diagnosis that combines wide kernel deep convolutional neural network and long short-term memory (WDCNN-LSTM). In this approach, a wide convolution kernel is utilized for local convolution, because the local convolution receptive field is increased, the fault feature contained in the low-middle frequency component is extracted. It is worth mentioning that the network complexity has not increased. Then, the transfer learning is applied to solve related but different domain problems, the useful knowledge learned by WDCNN-LSTM model under sufficient data conditions is transferred to diagnosis tasks with small dataset. After that, some state-of-the-art methods are applied to compare with the proposed method. At the end, experimental results showed that the proposed approach is an excellent algorithm for fault feature extraction of machinery and has much better identification accuracy and applicability than the other existing techniques.

Degradation modeling using heterogeneous degradation data of a system population is critical for prognostics and health management. The heterogeneity may result from the inherent differences in the degrading systems and measuring instruments, as well as from the extrinsic environmental differences, making it difficult to predict the failure. Current studies have not considered the multiple sources of heterogeneity simultaneously, and consequently, have failed to capture the actual degradation process. To explain and quantify the combined effect of multi-source heterogeneity, we present a random effects Wiener process model with heteroscedastic measurement errors, where the parameters of the drift, diffusion, and error variance are assumed as random variables following a certain distribution. The Markov chain Monte Carlo method is adopted to estimate the posterior distributions of the actual degradation states and the model parameters under a hierarchical Bayesian framework. Based on the concept of first hitting time, the failure time distribution is estimated. To verify the presented approach, comparative studies are conducted with simulation and laser degradation data. The results show that considering multi-source heterogeneity can further eliminate the inter-individual variation in degradation data and improve the model fitting and prediction accuracy.

The fault information of rolling bearings is generally contained in vibration signals. How to efficiently unearth fault information from the raw signals is the key to detecting and evaluating the health condition of mechanical equipment. Therefore, a hierarchical dispersion Lempel–Ziv complexity (HDLZC) feature extraction method is developed in this paper to improve the accuracy of fault diagnosis. In this method, dispersion theory addresses the deficiency of Lempel–Ziv complexity, and can obtain more fault features from the raw signal. Second, the hierarchical extraction of high- and low-frequency components from time series can improve the ability to describe dynamic features. Simulations and experiments respectively demonstrate the predominance of HDLZC. The experimental results reveal that this method is significantly better than multiscale dispersive Lempel–Ziv complexity, hierarchical Lempel–Ziv complexity, multiscale dispersion entropy, and multiscale permutation entropy in extracting fault information.