FEATURE EXTRACTION FOR BEARING PROGNOSTICS AND HEALTH MANAGEMENT (PHM) - A SURVEY

摘要

Feature extraction in bearing PHM involves extracting characteristic signatures from the original sensor measurements, which are sensitive to bearing condition and thus most useful in determining bearing faults. The quality of extracted features directly affects the performance and the effectiveness of bearing PHM. Feature extraction is therefore a critical component in bearing PHM. To optimally improve PHM effectiveness and minimize maintenance costs of bearings, a large amount research has been conducted in extracting salient features for PHM, which leads to a considerable number of feature extraction techniques. Our main effort in this paper is to survey some major techniques explored so far, focusing on more recent advancements. Our endeavor also includes pointing out the advantages and disadvantages of each of those techniques. This paper attempts to serve as a general reference for bearing PHM practitioners and as a general guide for choosing proper feature extraction methods for bearing PHM systems.