GENERATION OF SET MEMBERSHIP TESTS FOR FAULT DIAGNOSIS AND EVALUATION OF THEIR WORST CASE SENSITIVITY

Fault diagnosis based on the characterisation of operation ranges is proposed in this work. Set membership approaches are used to compute zonotopes used to implement detection and isolation tests. The proposed approach relies on the integration of residual generation and evaluation by testing the membership of the observations to some domains. An a priori evaluation of the worst case sensitivity of the tests is also proposed. The application to a current controlled DC motor illustrates the approach developed in the paper. Copyright © 2006 IFAC