Change Detection for Improved Maintenance Notification and Remaining Useful Life Calculation.

HUMS (Health and Usage Monitoring Systems) provide enhanced safety and availability by alerting maintenance personnel when components (shaft/gears/bearings) have degraded or become damaged. Proactive maintenance ensures that a degraded component will not fail in flight. Often, the damage is from high cycle fatigue or other relatively slow degradation processes. However, there are events, such as maintenance, that result in a step-change in the component's health. While high cycle fatigue degradation can be trended, a step-change in component health presents a challenge. For example, HUMS data from recent maintenance on a Thompson coupling resulted in a large 3/rev increase in vibration, which caused a step-change in the coupling health. Trending tools, which are, in effect, low-pass filters, would not immediately alert maintenance or the crew to the anomaly. Thus, a process to detect the change in health is needed which is both sensitive to change and provides protection against false alarms. Change detection in manufacturing process control has long been used for quality control. Several statistical tools have been developed to alert when the manufacturing status is out of specification. However, the research team is interested in determining when the high-cycle fatigue process is out of control. That is, when the difference between the component health trend, and measured component health, are statistically different enough, an alert should be sent to the maintainer, and the trend is reset to reflect the step-change. This paper describes the development of a CUSUM (cumulative sum) statistic between a component health trend and the measured component health. This allows the automated capture of maintenance or other events that would result in a component step-change in health. [ABSTRACT FROM AUTHOR]