A deep condition feature learning approach for rotating machinery based on MMSDE and optimized SAEs. (2nd December 2020)

• Record Type: Journal Article
• Authors: Ge, Ming-Feng; Ge, Ziyue; Pan, Hao; Liu, Yiben; Xu, Yanhe; Liu, Jie
• Is Part Of: Measurement science & technology. Volume 32:Number 3(2021)
• Publication Date: 2020-12-02
• Subjects: rotating machinery -- modified multi-scale symbolic dynamic entropy -- stacked auto-encoders -- fault diagnosis -- multiple working conditions; Physical measurements -- Periodicals; Scientific apparatus and instruments -- Periodicals; Equipment and Supplies -- Periodicals; Science -- instrumentation -- Periodicals; Technology -- instrumentation -- Periodicals; Mesures physiques -- Périodiques; Physical measurements; Scientific apparatus and instruments; Periodicals; 502.87
• DOI: 10.1088/1361-6501/ab89e3
• Languages: English
• ISSNs: 0957-0233

Abstract

Abstract: The failure of rotating machinery affects the quality of the product and the entire production process. However, it usually suffers the subsequent deficiency that the hyperparameters of the fault diagnosis model require constant debugging. This paper proposes a deep condition feature learning approach for rotating machinery based on modified multi-scale symbolic dynamic entropy (MMSDE) and optimized stacked auto-encoders (SAEs). Firstly, MMSDE has been used to extract fault characteristics of the original vibration signal, because such methods do not rely on prior knowledge and experience. MMSDE conducts multi-scale analysis on the original vibration signal and calculates the entropy of the multi-scale signal. The multi-scale fault characteristics are obtained. Then, Bayesian optimization-based SAEs are applied to select feature samples and classify the fault status in mechanical fault diagnosis without debugging. The effectiveness of the proposed method is verified by using open-source data and experimental data. Multiple working conditions are also considered and investigated.