Research on Indicator Extraction Method for Fault Diagnosis and Prognosis

Abstract

The key problems to be faced in the prediction of Remaining Useful Life (RUL) include the extraction of Condition Indicator (CI) that can accurately reflect the early fault characteristics of the research object and the Health Indicator (HI) that can stably track the irreversible degradation process. To solve this problem, the CI that can be used for early fault diagnosis and the HI that can be used for degradation trend tracking have been extracted based on the traditional indicator extraction method, deep learning method, Phase Space Warping (PSW) method, and Improved Phase Space Warping (IPSW) method respectively. Based on the different requirements of the feature indicator for early fault diagnosis and RUL prediction, the extracted CIs and HIs were evaluated, and the fusion strategy was adopted to screen the feature indicators, and the extraction of the CI and HI was realized scientifically. To verify the effectiveness of the proposed method, Both CI extraction for early fault diagnosis and HI extraction for RUL prediction were carried out based on the bearing lifetime accelerated degradation public data set. The results show that the proposed method can provide a reference for the development of RUL prediction in engineering practice.