A Deep Learning Approach based on MLP-mixer Models for Bearing Fault Diagnosis

Van-Quang Nguyen, Manh-Hung Vu, Van-Truong Pham, Thi-Thao Tran

Abstract: Bearing fault diagnosis plays an important role in monitoring activities of rotating machinery, especially electrical machine failures. Identifying bearing failure, therefore, is getting more and more attention. With the successes of deep learning techniques, various approaches based on convolutional neural networks (CNNs) have been developed for diagnosing bearing faults. Deep learning models can be effective for accurate diagnosis; nevertheless, in order to achieve this, a substantial amount of training data is typically required. To overcome this, we propose a new deep neural network approach based on the MLP-mixer model and CNN model that can dramatically improve training performance even with a small amount of data set. Various experiments and ablation studies have been conducted on the public Case Western Reserve University (CWRU) database. Evaluation results show that the proposed approach still obtains high accuracy when reducing the number of training samples. The promising performance of the proposed model when compared with another state of arts proved the advantages of the proposed approach.