Least order fault and model detection using multi-models

Abstract

Multi-model based fault detection is often a viable alternative to various multi-model based state estimation techniques using banks of Kalman filters. A main advantage of the fault detection techniques based approach is the possibility to use detectors having low order dynamics with disturbance decoupling capabilities. The proposed synthesis algorithm of detectors relies on numerically reliable rational nullspace techniques enhanced with optimal tuning of detection sensitivities. The applicability of the multi-model based approach to solve fault identification problems is illustrated by solving a flight actuator fault detection problem with simultaneous faults.