Title: SVM-based fault detection for double layered tank system by considering ChangeFinder's characteristics
Authors: Yosuke Furukawa; Mingcong Deng
Addresses: Department of Electrical and Electronic Engineering, The Graduate School of Engineering, Tokyo University of Agriculture and Technology, 2-24-16 Nakacho, Koganei, Tokyo, Japan ' Department of Electrical and Electronic Engineering, Tokyo University of Agriculture and Technology, 2-24-16 Nakacho, Koganei, Tokyo, Japan
Abstract: In this paper, a fault detection scheme for a tank system using support vector machine (SVM) combined with ChangeFinder, which are both machine learning methods, is studied. SVM can detect faults on a nonlinear system, but it can be late because SVM cannot recognise the nature of the time series. Combination with ChangeFinder enables SVM to recognise the nature of time series, and early detection using SVM becomes possible. Simulations and experiments assuming the temperature sensor fault case for the temperature control system of the tank system have been done and 5 s reduction of detection time was confirmed. In addition to the situation of the tank system, the sine wave input showed the effectiveness of the proposed method for general input. In addition, the superiority of the proposed method over ChangeFinder in an experimental environment was confirmed.
Keywords: nonlinear control; fault detection; fault tolerance; ChangeFinder; support vector machine; SVM.
DOI: 10.1504/IJAMECHS.2022.10044457
International Journal of Advanced Mechatronic Systems, 2022 Vol.9 No.4, pp.185 - 192
Received: 18 May 2021
Accepted: 01 Sep 2021
Published online: 31 May 2022 *