A remaining life prediction method based on semi-random filter considering model uncertainty

Abstract

The remaining life prediction is at the core of the application of health management technology. Semi-random filtering method is one of widely used prognosis method in this field. Existing studies on semirandom filtering methods have solved the problem of insufficient initial distribution determination by considering the prior knowledge of degraded equipment. However, for a class of degraded equipment that has unknown distribution characteristics and even does not have prior knowledge, the initial distribution required by this kind of methods cannot be easily determined. To solve this problem, this paper proposes a method for predicting the remaining life of degraded equipment based on the semi-random filtering method considering the model uncertainty. The method is based on the Bayesian model averaging method to fuse different models with different initial distributions, and the fused distribution based on the Bayesian model averaging method is used instead of the initial distribution to predict the remaining life, and the parameter estimation is obtained by the maximum likelihood method based on the historical data. Finally, the proposed method is verified based on the fatigue crack growth data, and the results show that the Bayesian model averaging method can improve the remaining life prediction accuracy when considering the model uncertainty.