Fault Diagnosis of Stochastic Systems

Solutions to the fault detection, estimation and isolation problems are presented when the model of the supervised system is a linear stochastic continuous-variable system. Faults are modelled as additive signals. The resulting diagnosis system is separated in two parts: a residual generator based on Kalman filters, and a decision system based on stochastic change detection/isolation algorithms. The link between these two parts is the object of particular attention.