REDUCING THE SIZE OF DETECTABLE FAULTS IN NONLINEAR SYSTEMS

AbstractThe problem of threshold selection for fault detection in nonlinear systems is considered. For the linear case the existing result requires of factorization of transfer function from perturbations to the output. In the nonlinear case a different approach is proposed in this work. Using ideas based on the concept of uniform ultimate boundedness a procedure to construct a residual generation is considered. The residual obtained is defined in such a way that a corresponding threshold can be defined in a natural form, in fact, as term in the residual equation. The proposed threshold can reduce the size of detectable faults and in this way the result is comparable to the existing for linear systems. The benefits of the approach are discussed.