A class of observer-based fault diagnosis schemes under closed-loop control: performance evaluation and improvement

This study deals with a fundamental issue of evaluating the performance of widely used fault detection and diagnosis (FDD) schemes within a closed-loop framework. The focus is to examine how certain implemented controller would impact on the FDD performance and how such performance can be further improved. For this purpose, the authors consider the FDD problem for a class of linear discrete-time systems (with and without unknown disturbances) under typical proportional-integral control using observer-based methods. It is revealed that some existing observer-based FDD approaches are no longer applicable in the closed-loop situation due to the feedback control. To solve the problem, by appropriately modifying the structure and re-designing the parameters of the observers, it is proven that the dynamics of closed-loop residuals can be made identical with those of the residuals obtained with known control inputs at each time step. A numerical example is provided to show the effectiveness of the proposed algorithm.