Automatic anomaly detection algorithm
Centre National d'Etudes Spatiales CNES - 10/17/2016
Description
The method uses two operating modes: - A learning mode during which a nominal model for each parameter measured is built up using recorded measurements. - A detection mode in which new measurements are compared with models obtained during the learning phase to determine atypical performance time periods
Benefits
Simple installation Failure prediction Fields for which measurements are acquired continuously over a long period : Health, nance, automotive industry, rail, aeronautics, chemistry, home automation.
Innovations
System supervision method based on recorded measurements - Synthetic modelling of a system’s nominal performance - Can detect early warning signs of a malfunction Rapid and simple anomaly detection - Enhanced responsiveness in the event of an anomaly Reduction in system downtime