Rolling Bearing Fault Diagnosis Based on Single Gated Unite Recurrent Neural Networks. (July 2020)

• Record Type: Journal Article
• Authors: Tan, Wenwen; Sun, Yuansheng; Qiu, Dawei; An, Yapeng; Ren, Ping
• Is Part Of: Journal of physics. Volume 1601:Number 4(2020)
• Publication Date: 2020-07
• Subjects: Physics -- Congresses; 530.5
• DOI: 10.1088/1742-6596/1601/4/042017
• Languages: English
• ISSNs: 1742-6588

Abstract

Abstract: In order to meet the real-time requirement of bearing fault diagnosis, a simplified strategy of the traditional long short term memory (LSTM) neural network structure is proposed. We designed a new single gated unite (SGU) recurrent neural network. In view of the non-stationary and non-linear characteristics of bearing fault vibration data, we use wavelet packet decomposition to extract the features as input signals of bi-directional single gated unite (Bi-SGU) to complete the diagnosis of 10 types of bearing data. Simulation results show that the proposed method can ensure the accuracy of bearing fault diagnosis, reduce the number of network parameters by 36%, and improve time efficiency by 41%.