Finite Memory Observers for linear time-varying systems: Theory and diagnosis applications

Abstract Fault detection and diagnosis are important issues in process engineering. Hence, a considerable interest exists in this field now from industrial practitioners as well as academic researchers, as opposed to 30 years ago. The literature on process fault diagnosis, ranging from analytical methods to artificial intelligence and statistical approaches, is largely widespread. In this paper, the modeling of the real process is known, and the state-space representation is used. The properties of the Finite Memory Observer (FMO) are studied from a global point of view for the class of linear time-varying (LTV) systems with stochastic noises. The FMO performances are framed by the study of their properties, and that of their influences on diagnosis results. Fundamentally, the generation of residuals is an essential procedure in diagnosis. So, the determination of the optimal window length of the observer is resolved, and the generation of residuals for diagnosis completed. In the first part, the design of the observer and the residual generation are shown. The second part is devoted to the study of the sensitivity and robustness of the observer and of residuals generated from the observer.