Generating Adaptive Alarms For Condition Monitoring Data

Maintenance actions cannot be defined and undertaken in isolation from the business and production objectives of a typical company. They must be predicted and subsequently planned into the production schedule. Condition monitoring identifies changes within defined monitoring parameters to trigger maintenance. These changes when combined with engineering knowledge, historical data and other monitoring parameters enable the prediction and identification of these changes in state. The data carries with it uncertainties in measurement accuracy, running conditions, and regularity. This paper reviews methods for trending and prediction, and describes how an adaptive approach to the definition of alarm levels can be used to improve the prediction capabilities when analysing routine condition monitoring measurement data.