Machine Health Indicator Construction Framework for Failure Diagnostics and Prognostics

Condition monitoring (CM) data should undergo through preprocessing to extract health indicators (HIs) for proper system health assessment. Machine health indicators provide vital information about health state of subcomponents(s) or overall system. There are many techniques in the literature used to construct HIs from CM data either for failure diagnostics or prognostics purposes. The majority of proposed HI extraction methods are mostly application specific (e.g. gearbox, shafts, and bearings etc.). This paper provides an overview of the used techniques and proposes an HI extraction, evaluation, and selection framework for monitoring of different applications. The extracted HIs are evaluated through a compatibility test where they can be used in either failure diagnostics or prognostics. An HI selection is carried out by a new hybrid feature goodness ranking metric in feature evaluation. The selected features are then used in fusion to get the representative component HI. Several case study CM data are used to demonstrate the essentiality of the proposed framework in component monitoring.