A Novel Scheme for Remaining Useful Life Prediction and Safety Assessment Based on Hybrid Method

Abstract

The prediction of remaining useful life (RUL) and safety assessment are the key of prognostics and health management (PHM) that provide decision support for it. A hybrid approach for the prediction of RUL which combines partial least squares (PLS) with support vector regression (SVR) and similarity based prediction (SBP) is proposed firstly. The SVR model, trained in a supervised manner, is employed to learn features extracted by PLS to capture the health indicator (HI) degenerate trajectory. Then the RUL prediction is implemented by calculating the similarity between the HI degenerate trajectories. Furthermore, on the basis of the prediction results, we construct a fuzzy comprehensive evaluation model to evaluate the safety level. To validate the proposed approach, a case study is performed on benchmark simulated aircraft engine datasets. The results show the superiority of the hybrid approach compared with other methods reported in the literature and indicate the effectiveness of the fuzzy comprehensive evaluation method in safety assessment.