Dynamic Prognosis of Hybrid Systems Under Imperfect Maintenance and Intermittent Faults With Unknown Features

In this article, a dynamic prognosis method is proposed for hybrid systems with imperfect maintenance (IM) and intermittent faults of unknown features. First, two feature degradation models, i.e., fault magnitude feature degradation model and fault duration feature degradation model, are developed which incorporate the effects of mode change and IM. In the fault magnitude feature degradation model, the IM effect is characterized as negative jump and the modes where the fault magnitude feature is applicable are determined by judging whether the predefined conditions are satisfied. In the fault duration feature degradation model, the concept of virtual tumbling window (TW) age is proposed where the IM effect on fault duration feature is depicted by reducing the TW age with one TW delay and the cumulative effect of IM on TW age across different modes is involved. After that, a mode-triggered feature selection mechanism and a dynamic prognosis scheme are developed. On this basis, the prognosis module can function with unknown intermittent fault features where the predicted remaining useful life can be dynamically updated catering to the mode-varying intermittent fault feature as well as the mode-varying IM effect. Finally, the proposed method is validated by experimental results.