Deep neural networks-based rolling bearing fault diagnosis

• Published in: Microelectronics and reliability Vol. 75; pp. 327 - 333
• Main Authors: Chen, Zhiqiang, Deng, Shengcai, Chen, Xudong, Li, Chuan, Sanchez, René-Vinicio, Qin, Huafeng
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.08.2017
• Subjects: Deep Belief Networks; Deep Boltzmann Machines; Fault diagnosis; Rolling bearing; Stacked Auto-Encoders; Fault diagnosis; Stacked Auto-Encoders; Deep Belief Networks; Rolling bearing; Deep Boltzmann Machines
• DOI: 10.1016/j.microrel.2017.03.006
• ISSN: 0026-2714

Rolling bearing is one of the most commonly used components in rotating machinery. It's so easy to be damaged that it can cause mechanical fault. Thus, it is significant to study fault diagnosis technology on rolling bearing. In this paper, three deep neural network models (Deep Boltzmann Machines, Deep Belief Networks and Stacked Auto-Encoders) are employed to identify the fault condition of rolling bearing. Four preprocessing schemes including feature of time domain, frequency domain and time-frequency domain are discussed. One data set with seven fault patterns is collected to evaluate the performance of deep learning models for rolling bearing fault diagnosis, which is based on the health condition of a rotating mechanical system. The results proved that the accuracy achieved by Deep Boltzmann Machines, Deep Belief Networks and Stacked Auto-Encoders are highly reliable and applicable in fault diagnosis of rolling bearing. •Develop DNN-based rolling bearing fault diagnosis system.•Evaluate the performance of DBM, SAE and DBN for rolling bearing fault diagnosis.•Discuss four preprocessing schemes for DNN-based fault diagnosis.•Deeply analyze the parameter setting of DBM, SAE and DBN-based diagnosis system.