Neurofuzzy Modelling and Pattern Matching for Online Fault Detection and Isolation of Nonlinear DC Motors

An online fault detection and isolation scheme for nonlinear systems based on neurofuzzy modelling and pattern matching is developed in this paper. The system is first modelled offline by a neurofuzzy network using data obtained under normal operating conditions. Another neurofuzzy network is then used to model the residual, which is the difference between the output of the system and that from the neurofuzzy network. For online fault monitoring, it is necessary to construct first a fault database that contains fuzzy rules for all possible faults in the system. Recursive least squares algorithm is used to train the network online, from which the IF-THEN rules are extracted. Faults are isolated online by comparing these fuzzy rules with those in the fault database using a nearest neighbour classifier. A simulation example involving a nonlinear DC motor control system is used to demonstrate the implementation and performance of the proposed FDI scheme.