The Adaptive Anomaly Detector expands on constant false alarm rate techniques to calculate adaptive thresholds.
The processing behind the adaptive thresholds automatically sets and maintains a unique threshold for each component or asset being monitored and adapts to the “normal” behaviour of the component/asset. This means that no threshold management is required.
Where standards exist which define fixed safety thresholds for the safe monitoring of assets, these are incorporated by means of a rule based interface.
Any deviations from “normal” behaviour are identified as anomalies such as events, discontinuities and trends in the data. The earliest detection of these types of events are essential for the implementation of predictive/prescriptive maintenance in order to maximise the remaining useful life of the asset.
Additional processing optimises the false alert rate performance by subjecting the data to a binomial integration process to identify a sustained change that constitutes an anomaly.