An improved remaining useful life prediction method for system with volatile degradation path

Remaining useful life (RUL) prediction is a key link in prognostics and health management. More accurate RUL prediction results will lead to more reasonable decision making on sequential management activities including maintenance, replacement, spare parts ordering, etc. In engineering practice, many products exhibit more volatile degradation paths when they have high degradation rates. By this observation, this paper concerns the problem of RUL prediction for a class of critical products which possess positive correlation between degradation rate and volatility. A Wiener-process-based degradation model with a special volatility parameter form is developed to achieve the aim. In this model, the volatility parameter has been set dependent on the drift parameter reflecting the degradation rate to describe the concerned problem. Both Bayesian updating and expectation maximization (EM) algorithm are used to estimate the unknown parameters in the model depend on the historically-observed degradation data. An exact and closed-form RUL distribution, which incorporates the random-effect capturing the unit-to-unit variability, is derived under the concept on the first passage time. Finally, a practical case study is used to illustrate and demonstrate the effectiveness of the presented method. The results show that the proposed approach can provide a more accurate RUL estimation for degradation system with volatile degradation path.