Linear Optimization of Parameters in Dynamic Threshold Generators

Authors:	Bask Michael, Lulea University of Technology, Sweden Johansson Andreas, Lulea University of Technology, Sweden
Topic:	6.4 Safeprocess
Session:	Active fault diagnosis/ Decision system design
Keywords:	Fault detection, Threshold selection, Parameter optimization,Linear programming, Uncertain linear systems

Abstract

A model-based fault detection algorithm for linear systems withuncertain parameters is treated. An error system, bilinear in theuncertainties, generates the residual. The residual is compared toa threshold, which is generated by a linear system with theunknown uncertainty upper bounds as parameters. These unknownuncertainty upper bounds can be substituted by design parametersand this article suggests an algorithm to choose design parametervalues such that the threshold is larger than the residual when nofault is present. This parameter design algorithm is applied to asensor fault detection algorithm for a jet engine.