Fault detection of uncertain systems based on probabilistic robustness theory

This paper studies the fault detection problem of uncertain linear systems with arbitrary uncertainty structure. With the aid of probabilistic robustness technique, an approach is developed to determine parameters of observer-based residual generators by incorporating probability distribution of model uncertainty in the design. One advantage of the approach is that it needs no assumptions on the structure of model uncertainty. Examples are given to illustrate the proposed design procedure.

I. Introduction

With the increasing requirement of modern complex control systems on safety and reliability, model-based fault detection and isolation (FDI) technique has attracted much attention during the last three decades [1]–[4]. The basic idea of model-based FDI is to generate analytical redundancy with the help of a mathematical model of the supervised system. The fault indicating signal, called usually residual, is generated by comparing measured outputs with their estimations. A number of model-based approaches, for instance, observer-based approach, parity space scheme, parameter estimation method, etc., have been proposed to the FDI of linear and nonlinear systems. Applications have been found in automobile industry, process industry, transportation systems, aerospace and aeronautics, etc.