An approach to optimally robust fault detection and diagnosis

Abstract

In this paper, an original approach for robust fault detection and diagnosis (RFDD) is presented. We propose a method that makes use of H/sub 2/ and H/sub /spl infin// polynomial estimators to elaborate a RFDD procedure. The system considered is linear, stochastic and subject to noise, unknown inputs and uncertainties. The residual is obtained via factorization approach and then parameterized by H/sub 2/ and H/sub /spl infin// polynomial estimators to generate optimal robust fault detectors (ORFD). Adaptive thresholds are also derived with respect to an evaluation index. The results are illustrated by simulations on an actual rolling system.