fault detection toolbox.pdf

A Fault Detection Toolbox for MATLAB A. Varga Abstract—The recently developed FAULT DETECTION Tool- box for MATLAB is described. The new toolbox provides a com- prehensive set of high level m-functions to support the design of residual generation filters using reliable numerical algorithms recently developed by the author. The basic computational layer is formed by the DESCRIPTOR SYSTEMS Toolbox which contains all necessary tools to solve the underlying numerical problems. The m-functions based user interfaces ensure user-friendliness in operating with the functions of this toolbox via an object oriented approach. I. INTRODUCTION The design of fault detection (FD) filters able to detect discrepancies between normal and erroneous plant operations has been an active area of research since decades (see for example the recent monographs [1], [2], [3] and hundreds of references cited therein). The need to address properly the numerical issues encountered in designing fault detection filters has been already recognized by Chen and Patton [2, p.219]. Among methods traditionally used to design fault detection filters we mention techniques based on unknown- input observers, parity equations, eigenstructure assignment, polynomial approaches, H2- or H∞-optimal filter synthesis techniques. However, in spite of many computational ap- proaches proposed in the literature, until recently there were no general and numerically reliable algorithms available able to address the underlying computational aspects for large dimensional systems. The existing lack of dedicated robust software tools to design FD-filters is a consequence of the existing delicate situation in algorithmic field. A systematic research in developing numerically reliable algorithms for the design of linear FD-filters has been conducted by the author in the recent years leading to a fairly complete collection of methods able to solve various FD-problems in the most general setting [4], [5], [6]. The proposed methods