Open AccessAn Improved Method for Constructing Health Factors of Rolling Bearing
Author(s) -
Tao Shu,
Yichi Zhang,
Yuye Xu,
Pengxiang Zhang
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2009/1/012028
Subject(s) - robustness (evolution) , bearing (navigation) , vibration , computer science , frequency domain , monotonic function , boltzmann machine , engineering , artificial intelligence , mathematics , deep learning , computer vision , chemistry , acoustics , mathematical analysis , physics , gene , biochemistry
Aiming at the problems of low performance index and redundant information of multi-class physical health factors, a method of building rolling bearing health factors based on improved restricted Boltzmann machine is proposed. By fusing multi-class physical health factors, the virtual health factor of rolling bearings is constructed. Extracting physical health factors in time domain and frequency domain of rolling bearing vibration monitoring signals as inputs; the performance degradation mechanism model is added to the regularization term of the Restricted Boltzmann Machine, the performance degradation information contained in the input data is mined, the virtual health factor construction model of rolling bearings is built, and the model parameters are adjusted by the health factor evaluation criteria to improve the performance of the virtual health factor. The life cycle test of rolling bearings shows that compared with the principal component analysis method, the monotonicity of virtual health factors based on the improved restricted Boltzmann mechanism is increased by 147.6% and 38.5%, the trend is increased by 113.8% and 16.1%, and the robustness is increased by 60% and 8.42%, respectively.