Health Index-Based Prognostics for Remaining Useful Life Predictions in Electrical Machines

• Published in: IEEE transactions on industrial electronics (1982) Vol. 63; no. 4; pp. 2633 - 2644
• Main Authors: Feng Yang, Habibullah, Mohamed Salahuddin, Tianyou Zhang, Zhao Xu, Pin Lim, Nadarajan, Sivakumar
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.04.2016; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Aging; Data collection; Data models; Electrical Machine; Feature extraction; Health Index; Hidden Markov models; Induction motors; Predictive models; Prognostics; Remaining Useful Life; health index (HI); remaining useful life (RUL); Electrical machine; prognostics
• ISSN: 0278-0046; 1557-9948
• DOI: 10.1109/TIE.2016.2515054

Many industries have a growing awareness in utilizing new technologies to improve the reliability and availability of their systems. Prognostics, a subject concerned with the prediction of the remaining useful life (RUL), has been increasingly studied and applied to practical systems, such as electrical systems, over the past few years. Here, with the adoption of a data-driven prognostics framework, this paper proposed a health index (HI)-based prognostics method to predict the RUL of electrical machines. By assuming a linearly degrading HI over time, the proposed method predicts the RUL in two steps: 1) from input signals to HI; and then 2) mapping HI to RUL. The novelty of this method lies in the proposed dynamic HI smoothing approach where three characteristics of HI, namely monotonicity, gradualness, and consistency, are incorporated to smooth the current HI values with the previously predicted ones. Real data collected from eight electrical motors, subjected to accelerated thermal aging process, were used in the experimental studies, with the results showing the superiority of the proposed HI-based RUL prediction over the traditional direct RUL prediction (i.e., without HI).