Title: Real-time sensor fault diagnosis method for closed-loop system based on dynamic trend

Authors: W.B. Na; Y. Gao; T.H. Zhu; X.T. Zheng

Addresses: College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou, Zhejiang-310000, China ' College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou, Zhejiang-310000, China ' College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou, Zhejiang-310000, China ' College of Mechanical and Electrical Engineering, China Jiliang University, Hangzhou, Zhejiang-310000, China

Abstract: To solve the common problems of sensor fault detection, fault setting and fault location in the first-order closed-loop control system, a static fault diagnosis model based on the real-time trend of dynamic data flow is constructed. Aiming at the common fixed-value system and servo system in the first-order closed-loop control system, a data processing model based on sliding window is designed by collecting and analysing a large amount of real-time data, and the fault is isolated in the window based on the calibration parameters. Then, the adaptive threshold binarisation method is used to calculate the fault vector, and then the analytical model of fault location is obtained by the linear regression method. Finally, the feasibility and validity of closed-loop sensor fault diagnosis method based on dynamic data flow trend are verified by online simulation of “complex process system innovation experimental platform” based on OPC communication technology.

Keywords: dynamic trend; sliding window; sensor; fault diagnosis; closed-loop system; threshold; linear regression; fault detection; fault setting; fault location.

DOI: 10.1504/IJMIC.2019.104381

International Journal of Modelling, Identification and Control, 2019 Vol.33 No.2, pp.98 - 107

Received: 17 May 2019
Accepted: 29 Sep 2019
Published online: 06 Jan 2020 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article