Remaining Life Prediction Method for Long-Range Correlation Degradation Process Depending on Time and State

Abstract

The invention discloses a method for predicting the remaining life of a long-range correlation degradation process dependent on time and state, belonging to the technical field of health management, comprising the following steps: first, collecting sensor data sampled at equal intervals; The degradation model of the model; use the method based on quadratic variation to estimate the Hurst exponent in the model; maximize the likelihood ratio function constructed by Radon‑Nikodym derivatives to estimate the unknown parameters of the drift term; use the maximum likelihood method to estimate the unknown parameters of the diffusion term ; Then the original degradation process is approximated as a stochastic process based on Brownian motion with a time-varying diffusion term coefficient by the weak convergence theory; the degradation process is further simplified by a set of transformations; finally the analytical remaining life distribution is obtained. The present invention can predict the remaining life distribution more accurately.

Description

technical field [0001] The invention belongs to the technical field of health management, in particular to a method for predicting the remaining life of a long-range correlation degradation process dependent on time and state. Background technique [0002] The normal operation of industrial equipment is the fundamental premise to ensure production safety and enterprise benefits. Once industrial equipment fails, it will not only bring huge economic losses to production, but may also cause serious accidents. Therefore, health management is of great significance to industrial equipment. Remaining life prediction is an important part of health management. Accurate remaining life prediction results can effectively guide equipment maintenance strategies and spare parts supply, thereby avoiding waste caused by excessive maintenance and losses caused by failures. [0003] In order to obtain accurate life prediction results, it is necessary to establish a model that can describe e...

AI Technical Summary

Problems solved by technology

Although there are currently a small number of works focusing on modeling both time- and state-dependent degradation processes and long-range dependent degradation processes, none of them can simultaneously consider time- and state-dependent long-range dependent degradation process modeling and residual Life Prediction Methods

[0004] The remaining lifetime prediction of time- and state-dependent long-range correlation degradation processes mainly has the following difficulties: First, since the degradation state itself is included in the degradation model, it is impossible to write an analytical likelihood function, and it is difficult to use traditional estimation methods for parameter estimation

Second, since the degenerate process is neither a Markov process nor a semi-martingale, it is difficult to obtain the analytical first-arrival time distribution

Images