PATTERN RECOGNITION OF HEALTH - DATA DERIVED PROGNOSTIC HEALTH MANAGEMENT

摘要

Prognostic health management (PHM) provides system stakeholders the ability to anticipate problems and address them in a planned, rather than unscheduled, manner. PHM technology creates models of expected system operations and uses variances from predicted performance to indicate abnormalities and future failures. The models described in this paper are based on the concept of "Pattern Recognition of Health (PRoH)". PRoH is a data-derived solution combining pattern recognition algorithms with principles from statistics and signal-processing to create models of expected system performance using available sensors and data sources. PRoH applies these models to determine and report system health states. In the presence of degrading or unexpected conditions, PRoH creates fault mode signatures that identify failure modes and corrective action. PRoH has been designed to overcome obstacles inhibiting application of PHM technology. It is independent of detailed system knowledge, makes accurate decisions using data available from existing sensors and sources, and interfaces with current logistics systems. It has been tested in challenging environments including aircraft engines, rotorcraft, wheeled vehicles, navigation equipment, oil production and power plants. In all cases the algorithms have been able to detect problems earlier than current methodologies and in time to permit corrective action prior to actual system failure.