Robust adaptive fault isolation observers for nonlinear systems

In this note we present a robust fault diagnosis scheme for parametric faults in nonlinear systems, and derive a parametric fault isolability condition under suitable "persistency of excitation" hypotheses. We show that the adaptive observer proposed is robust with respect to modelling errors, as well as external disturbances, and find an upper bound on the time required by the observer to successfully isolate the fault. We describe how by choosing properly the observer error gain it is possible to match asymptotically the isolation condition. The dimension of the observer scheme that we propose has the same order of magnitude of the parameter vector, rather then a dimension growing exponentially with respect to the number of unknown parameters, thus potentially allowing more complex faults to be isolated. Copyright (C) 2001 IFAC.