Physics-Informed Deep Neural Network for Bearing Prognosis with Multisensory Signals

Xuefeng Chen; Meng Ma; Zhibin Zhao; Zhi  Zhai; Zhu  Mao

doi:10.37965/jdmd.2022.54

Authors

Xuefeng Chen School of Mechanical Engineering, Xi’an Jiatong University, China
Meng Ma School of Mechanical Engineering, Xi’an Jiatong University, China
Zhibin Zhao School of Mechanical Engineering, Xi’an Jiatong University, China
Zhi Zhai School of Mechanical Engineering, Xi’an Jiatong University, China
Zhu Mao Department of Mechanical and Materials Engineering, Worcester Polytechnic Institute, MA, USA

DOI:

https://doi.org/10.37965/jdmd.2022.54

Keywords:

deep learning;physics-informed neural network (PiNN);Prognostics and Health Management (PHM); remaining useful life

Abstract

Prognosis of bearing is critical to improve the safety, reliability and availability of machinery systems, which provides the health condition assessment and determines how long the machine would work before failure occurs by predicting the remaining useful life (RUL). In order to overcome the drawback of pure data-driven methods and predict RUL accurately, a novel physics-informed deep neural network, named degradation consistency recurrent neural network, is proposed for RUL prediction by integrating the natural degradation knowledge of mechanical components. The degradation is monotonic over the whole-life of bearings, which is characterized by temperature signals. To incorporate this knowledge of monotonic degradation, a positive increment recurrence relationship is introduced to keep the monotonicity. Thus, the proposed model is relatively well-understood and capable to keep the learning process consistent with physical degradation. The effectiveness and merit of the RUL prediction using the proposed method are demonstrated through vibration signals collected from a set of run-to-failure tests.

Conflict of Interest Statement
The authors declare no conflicts of interest.