A Bank of Finite Memory Filters for Fault Detection and Adjusting Detection Latency

Abstract

This paper proposes a new fault detection scheme using a bank of finite memory filters for discrete-time dynamic systems with multiple sensors. In the proposed scheme, fault detection is carried out by testing the consistency of two filtered estimates, which are obtained from the primary estimation filter and the auxiliary estimation filter using a bank of finite memory filters, respectively. Detection latency is considered as one of important performance criteria and focus on the improvement of detection latency even for high threshold value. Through extensive computer simulations for the F-404 engine system, it is shown that detection latency can be adjusted by the window length. Simulation results show that the trade-off between the fast detection performance and the noise-suppressing estimation performance should be needed for the proposed fault detection scheme in real applications.