Rolling bearing fault diagnosis method based on MTF and PC-MDCNN

A rolling bearing fault diagnosis method based on Markov transition field (MTF) and the pyramid cascade multidimensional convolutional neural network (PC-MDCNN) model is proposed to address the problems of poor fault diagnostic performance and generalization performance. These problems are caused by the complex and variable working conditions of rolling bearings and the difficulty in collecting fault samples. First, the one-dimensional vibration signal is transformed into two-dimensional images using MTF. Then, a pyramid cascade multidimensional feature extraction module (PC-MDFEM) is introduced to reduce model parameters and extract fault information of the feature maps comprehensively by focusing on different dimensions of the feature maps and combining convolutional layers at different levels. Afterward, PC-MDFEM is applied to the convolutional neural network to construct a PC-MDCNN model. Finally, the MTF feature maps are input into the proposed model for training, and the proposed model is compared with other fault diagnosis models. Experimental results demonstrate that the proposed method has strong classification performance, generalization performance and robustness.