Health indicator based on signal probability distribution measures for machinery condition monitoring

Abstract

Health indicator (HI), which aims to make quantitative measures for machinery operating state at different degradation stages, is very critical in machinery condition monitoring. Some HIs from different aspects have been developed and reported in recent years. However, a preferable HI which is more robust to transient interferences, free of complicated model training and also sensitive to incipient defects in machinery condition monitoring still remains to be further investigated. To address these issues, a novel HI based on signal probability distribution measures is proposed in this paper. Firstly, characteristic parameters of the alpha stable distribution are preliminarily estimated based on the machinery degradation data, the consistency of which is quantitatively evaluated and optimized through the hypothesis test with a parameter calibration strategy. Afterwards, signal distribution models are accordingly constructed to describe the statistical characteristics of the machinery degradation data. On this basis, the deviation of the established signal distribution models between the current degradation state and the initial fault-free state is accordingly analyzed and quantified for machinery degradation assessment. Experimental validations by using simulated and industrial run-to-failure datasets demonstrate that the proposed HI can effectively recognize the state shift of the machinery during the degradation process and can be therefore applied for machinery condition monitoring.