Remaining useful life prediction for multiple degradation indicators systems considering random correlation

Abstract

With the increasing complexity of system structure, using only a single degradation indicator often fails to fully reflect the potential degradation state and failure type of the system. Meanwhile, there may have correlation between multiple degradation indicators. Therefore, multiple degradation indicators are significant to predict the remaining useful life of complex systems. An online RUL prediction model under multiple fault modes is constructed for multiple degradation indicator systems in the presence of random correlation among the degradation indicators. First, a degradation model based on multi-indicator stochastic correlation is established. On the basis of expectation maximization algorithm, forward Kalman filtering algorithm and reverse Rauch-Tung-Striebel smoothing algorithm are jointly estimated unknown parameters and hidden state. Then, the fault modes of the system with multiple indicators are analyzed, and the RUL distribution of the system with different fault modes is deduced and calculated. Finally, the correctness and validity of the RUL model with multi-degradation indicator correlation are verified by numerical experiments. Moreover, the C-MAPSS dataset is adopted as an example to validate the adaptability and superiority of the proposed approach compared with other prediction methods.