Detection and isolation of actuator faults for a class of non-linear systems with application to electric motors drives

In this study, the diagnosis of actuator faults is addressed for a class of non-linear systems and applied to electric motors drives. The residuals are synthesised using a model-based strategy by applying a differential geometry perspective. According to the dynamic structure of the systems considered in this work, it is deduced that the observability co-distribution built for this class of systems can be independent of the state (constant) under certain conditions, so that the resulting coordinates transformations in the state and output spaces are linear mappings. Consequently, the dynamic structure of the resulting subsystems is not altered after the coordinates transformation, and departing from these subsystems, dedicated observers can be designed for residual generation. One important property of the analysis presented in the paper is that the studied dynamical structure can be fitted to several electrical motors. Hence simulation results are illustrated for a three-phase induction motor drive application that verify the ideas presented in this work.