Directional fault isolation procedure for diagnosis systems based on parameter estimation techniques

Abstract

This paper focuses on the design of a Fault Isolation procedure suitable for integration within Fault Detection and Diagnosis schemes which are based on parameter estimation techniques. A brief discussion regarding the structure and functionality of model-based technical diagnosis systems is provided within this context, and a summary of previously obtained results in the field of fault detection and fault identification using the Recursive Least Squares method with exponential weighting and constant forgetting factor is presented. A fault localization procedure is proposed based on these results and on the principles of the directional fault isolation approach. Also, an enhancement is made to the classical model-based fault isolation approach, consisting of using history data in order to increase the accuracy of the associated algorithm. An application example from the field of aircraft control is considered in order to illustrate the performances of the proposed method.