Remaining useful life prediction using ranking mutual information based monotonic health indicator

In prognostic approaches, if the health indicators tracking damage level show obvious monotonic trend through the life cycle, good RUL prediction results can be expected. This paper proposes a method to generate such an indicator. Ranking mutual information is employed to measure monotonicity relevance between features and damage level. Furthermore, a case study based on similarity prognostic method is carried out to identify the effectiveness of the proposed prognostic indicator generation method.