Hankel matrix-based Mahalanobis distance for fault detection robust towards changes in process noise covariance

Szymon Gres (1) , Michael Döhler (1) , Laurent Mevel (1)

1 I4S - Statistical Inference for Structural Health Monitoring

Statistical subspace-based change detection residuals have been developed to infer a change in the eigenstructure of linear systems. Their statistical properties have been properly evaluated in the case of a known reference and constant noise properties. Previous residuals have favored the family of null space-based approaches, whereas the possibility of using other metrics such as the Mahalanobis distance has been omitted. This paper investigates the development and study of such a norm under the premise of a varying noise covariance. Its statistical properties have been studied and tested on a numerical example of a mechanical system.

Mots clés: Change detection; Local approach; Mahalanobis distance; Linear systems

Domaines: Dynamique, vibrations; Applications [stat.AP]

Szymon Gres, Michael Döhler, Laurent Mevel. Hankel matrix-based Mahalanobis distance for fault detection robust towards changes in process noise covariance. SYSID 2021 - 19th IFAC Symposium on System Identification, Jul 2021, Padua / Virtual, Italy. pp.1-6, ⟨10.1016/j.ifacol.2021.08.337⟩. ⟨hal-03292515⟩