Purpose – The purpose of this paper is to present research in detecting and identifying abrupt faults in controlled auto‐regressive (CAR) processes. Design/methodology/approach – Model‐based approach is adopted in this paper. Two series of fault‐tolerant iterative estimators are set up to estimate online the coefficients in a CAR process. Based on these fault‐tolerant estimators, the detailed detecting and identifying algorithms are obtained for not only the pulse‐type faults but also the step‐type faults in CAR process. Findings – This paper illustrates the useful information that can be obtained from residuals and that can be used to detect pulse‐type faults as well as step‐type faults. A fault‐tolerant recursive estimator for the coefficients of the CAR process is put forward. Using a simple transformation from step‐ to pulse‐type faults, all kinds of diagnosis methods to detect and identify step‐type faults can be used. Research limitations/implications – Fault‐tolerant estimators and fault detection and identification algorithms are aimed at abrupt faults in CAR processes. Practical implications – Most of the algorithms given in this paper can be used in many different fields, such as process monitoring, safety control and change detection, etc. Originality/value – This paper contributes to research of abrupt faults and abrupt changes in a CAR process and emphasizes identification of magnitudes of abrupt faults. The fault‐tolerant estimators are effective not only to detect faults but also to identify safely the coefficients CAR model.
International Journal of Intelligent Computing and Cybernetics – Emerald Publishing
Published: Jun 6, 2008
Keywords: Programming and algorithm theory; Autoregressive processes; Fault tolerance