Rolling bearing fault diagnosis by Markov transition field and multi-dimension convolutional neural network

Abstract

Safe and reliable operation of mechanical equipment depends on timely and accurate fault diagnosis. When the actual working conditions are complex and variable and the available sample data set is small, recognition accuracy of the rolling bearing fault diagnosis model is low. To solve this problem, a novel method based on Markov transition field (MTF) and multi-dimension convolutional neural network (MDCNN) is proposed in this paper. Firstly, the original vibration signals are converted into two-dimensional images containing temporal correlation by MTF. Then, a neural network model is constructed by using multi-dimension attention (MDA) and E-Relu activation function to fully extract fault feature information. Finally, the MTF images are input into the model and the data is normalized using the group normalization method. The MDCNN model is validated on two different data sets, and the results show that compared with other intelligent fault diagnosis methods, the MDCNN has higher fault diagnosis accuracy and stronger robustness under both variable working conditions and small sample data sets conditions.

Fine-grained transfer learning based on deep feature decomposition for rotating equipment fault diagnosis

“…Model-based diagnosis algorithms can be divided into deterministic fault diagnosis methods [8], stochastic fault diagnosis methods [9], fault diagnosis for discrete-events and hybrid systems [10], and fault knowledge of health system symptoms, including the wavelet transform [16], empirical mode decomposition [17], and Hilbert-Huang transform [18]. Machine learning methods analyze faults by manually extracting fault features and then using machine learning algorithms such as support vector machine algorithm [19], K-nearest neighbor algorithm [20], and Markov model [21]. Deep learning algorithms, on the other hand, make fault diagnosis algorithms more intelligent and less dependent on humans by automatically extracting deep features of the data and by their powerful fitting capabilities [22][23][24][25].…”

Small sample fault diagnosis for wind turbine gearbox based on lightweight multiscale convolutional neural network

“…Zhang et al [24] proposed a two-dimensional multiscale cascaded CNN-based method to obtain sensitive wavebands for fault identification by reconstructing MC images from multiscale information. Lei et al [25] proposed a fault diagnosis method based on Markov transfer fields (MTFs) and multidimensional convolutional neural networks. MTF images contain temporal correlations and features at different time scales can be extracted through multidimensional feature extraction.…”

Bearing fault diagnosis method using the joint feature extraction of Transformer and ResNet

“…In a previous study [10], a CNN with a wide first-layer kernel structure (WDCNN) was presented to directly identify and process the collected bearing vibration signals [10]. Lei et al used the Markov transition field method to convert the onedimensional (1D) vibration signal into a two-dimensional (2D) image, and then mined features and classified fault types from the image via multi-dimension CNN [11]. In another study [12], a joint-loss CNN architecture, which can simultaneously realize fault diagnosis and remaining life prediction for rotating machinery, was proposed to improve the generalization ability of the model.…”