Improved similarity-based residual life prediction method based on grey Markov model

Remaining useful life (RUL) prediction is a significant prognostic activity in many industrial applications. As an emerging data-driven method, the similarity-based residual life prediction (SbRLP) method is vital to RUL prediction. However, its prediction performance is unsatisfactory in the early and middle terms, which limits its application. Therefore, an improved SbRLP method based on the grey Markov model is proposed. First, monitoring variables are evaluated and selected based on five aspects, namely monotonicity, correlation, robustness, difference, and sensitivity, to construct a one-dimensional health index. Next, the grey Markov model is employed to predict the similarity measurement sequence of an operating sample, and a similarity measurement sequence is reconstructed based on the predicted information. Furthermore, the RUL of an operating sample is predicted based on the new similarity measurement sequence. Subsequently, a commercial modular aero-propulsion system simulation dataset is used to verify the effectiveness and superiority of the proposed SbRLP method. Implementation results show that the prediction performance of the proposed SbRLP method improves, particularly in the early and middle stages. Moreover, reasonable values of the prediction step F and control coefficient a can further improve its prediction accuracy.