Remaining Useful Life Prediction for Deteriorating Equipment with Multivariate Degradation Processes

Remaining Useful Life (RUL) prediction with multivariate degradation processes is one of the research hotspots in the field of reliability. However, most existing methods primarily focus on static correlation models, and ignore the time-varying correlations between degradation indicators. In this paper, a novel approach of RUL prediction is proposed based on time-varying Copula function. Firstly, the nonlinear Wiener process is employed to characterize each individual degradation process, and the time-varying Copula function is utilized to construct the dynamic correlation between multivariate degradation processes. Secondly, the parameters of the degradation model and the time-varying Copula function are estimated based on the Bayesian Markov Chain Monte Carlo (MCMC) and maximum likelihood estimation (MLE), respectively. Thirdly, the distribution of predicted RUL is derived in combination with the marginal distribution by using time-varying Copula function. Finally, a numerical example is provided to validate the proposed method.