Instrument fault detection in systems with uncertainties

Abstract

We demonstrate how to detect instrument faults in non-linear time-varying processes that include uncertainties such as modelling error, parameter ambiguity, and input and output noise. The design of state estimation filters with minimum sensitivity to the uncertainties and maximum sensitivity to the instrument faults is described together with existence conditions for such filters. Simulations based on a non-linear chemical reactor with heat exchange and feedback control illustrate the validity of the proposed method.