Multivariate Relevance Vector Regression Based Degradation Modeling and Remaining Useful Life Prediction

Relevance vector regression (RVR) is a useful tool for degradation modeling and remaining life (RUL) prediction. However, most RVR models are 1-D processes can only handle univariate observations. This article proposes path-based RUL prediction framework using dynamic multivariate relevance model. Specifically, multistep model established describing the dynamics extending classical into one with consideration of environment. The introduces matrix Gaussian distribution-based approach then estimates hyperparameters Nesterov’s accelerated gradient method to avoid exhausting re-estimation phenomenon in seeking analytical solutions. It further forecasts path monitoring status. Based on forecasted path, predicted by first hitting time method. Finally, proposed methods illustrated two case studies, presented this other supplement, which investigate capacitors’ bearings’ performance degradations.