Adaptive residual generation on a winding machine described by multiple models

An adaptation method for fault detection on systems described by multiple models is presented in this paper, based on the properties of the synthesis parameters Q and R of Kalman filters. A non linear winding machine described by multiple linearized models is used to highlight the principle of this adaptation method. The application to this winding machine is proposed, in the case where a fault occurs after a loss of the information used for the determination of the current valid model.