Active diagnosis of hybrid systems — A model predictive approach

A method for active diagnosis of hybrid systems is proposed. The main idea is to predict the future output of both normal and faulty model of the system; then at each time step an optimization problem is solved with the objective of maximizing the difference between the predicted normal and faulty outputs constrained by tolerable performance requirements. As in standard model predictive control, the first element of the optimal input is applied to the system and the whole procedure is repeated until the fault is detected by a passive diagnoser. It is demonstrated how the generated excitation signal can be used as a test signal for sanity check at the commissioning or for detection of faults hidden by regulatory actions of the controller. The method is tested on the two tank benchmark example.