Intermittent fault detection for discrete- time linear stochastic systems with time delay

Abstract

This study investigates the problem of intermittent fault (IF) detection for a class of discrete-time linear stochastic systems (LSSs) with time delay. The IF has the characteristics of intermittency, randomness and uncertainty, and the time delay is constant. By using the lifting method and a reduced-order observer, the truncated residual over a sliding-time window is presented for the IF detection of discrete-time LSSs with time delay. Two hypothesis tests are proposed to set the thresholds for detecting the appearing (disappearing) time of IF. Moreover, the magnitude of IF is estimated and the detectability of IF is analysed based on the analysis of statistical characteristics of the truncated residual. The efficacy and practicality of the derived detection scheme are verified via a numerical simulation of the Williams–Otto process.