ROBUST FAULT DIAGNOSIS IN THE PRESENCE OF PROCESS UNCERTAINTIES Zhengang Han, Weihua Li and Sirish L. Shah 1 Department of Chemical and Materials Engineering University of Alberta Edmonton, AB, T6G 2G6, Canada Abstract: This paper proposes a novel scheme for the generation of primary residual vector (PRV) for sensor or actuator fault detection and isolation (FDI) in multivariate dynamic systems. The PRV, which is used for fault detection purpose, is designed to be insensitive to process uncertainties, including model–plant mismatch (MPM) and process disturbances. To generate the PRV, we do not need a precise system model. Instead, all we need is an estimate of the system model, which may be biased from the true model. Under the condition that the number of process uncertainties is less than the number of outputs, the generated PRV can be made perfectly insensitive to process uncertainties. Even when this condition does not hold, the most important elements in the process uncertainties can still be decorrelated from the PRV. A numerical example to demonstrate the theory is given. The newly proposed approach is compared with existing robust FDI schemes, e.g., the Chow–Willsky scheme. Copyright c 2002 IFAC Keywords: robust sensor or actuator fault detection and isolation, process uncertainties, primary residual vector, structured residual vector, multivariate systems 1. INTRODUCTION Since the 1970s, tremendous research efforts have been invested into model–based sensor or actu- ator fault detection and isolation (FDI). Survey papers in this area have been published by Willsky (1976), Gertler (1988), Frank (1990), and Patton et al. (2000). More recent advances have been reviewed by Qin and Li (2001), and Li and Shah (2002). Most model–based FDI approaches assume that an accurate plant model of the system under consideration is available, and at the same time, they also assume the process disturbances to be zero-mean Gaussian noise. Since model–plant 1 Author to whom correspondence should be addressed. Email: Sirish.Shah@ualberta.ca mismatch (MPM) is inevitable and process dis- turbances can be any functions of time. Their presnece will affect the residuals generated for fault detection. The need for robust FDI schemes which enable the decoupling of MPM and process disturbances from the residuals has been identi- fied. So far there have been very few published studies with respect to robust FDI. Patton and Chen (1992) have developed an observer–based approach toward the removal of process distur- bances from the residuals, but the approach does not consider the issue of MPM. Gertler and Kun- wer (1995) have proposed a modelling error de- coupling method, which decouples the modelling error at each time instant. This method is com- putationally intensive and not practical. Qin and Li (2001) and Li and Shah (2002) have proposed the subspace identification–based approaches for Copyright © 2002 IFAC 15th Triennial World Congress, Barcelona, Spain