Neurofuzzy Approach to Fault Detection of Nonlinear Systems
Jinglu Hu*, Kotaro Hirasawa* and Kousuke Kumamaru**
*Department of Electrical and Electronic Systems Engineering Kyushu University, 6-10-1 Hakozaki, Higashi-ku, Fukuoka 812-8581, Japan
**Department of Control Engineering and Science Kyushu Institute of Technology, 680-4 Kawatsu, Iizuka 820-8502, Japan
Received:April 2, 1999Accepted:June 30, 1999Published:December 20, 1999
Keywords:Nonlinear system, Fault detection, Kullback discrimination information, Neurofuzzy model
This paper proposes a neurofuzzy approach to fault detection in linear systems. The system diagnosed is described by using a neurofuzzy model called LimNet that consists of a linear model and multiple local linear models with interpolation of a "fuzzy basis function". Fault detection is considered in two cases: when faults occur in the linear model part, a KDI-based robust fault detection is applied, where a multi-local-model part is treated as error due to nonlinear undermodeling; when faults occur in the multi-local-model part, a multi-model based fault detection method is developed, in which the identified LimNet is interpreted as several local ARMAX models, and KDI is used as an index to discriminate between each local model and its reference. This paper mainly concentrates discussions on multi-model based fault detection.
Cite this article as:J. Hu, K. Hirasawa, and K. Kumamaru, “Neurofuzzy Approach to Fault Detection of Nonlinear Systems,” J. Adv. Comput. Intell. Intell. Inform., Vol.3 No.6, pp. 524-531, 1999.Data files: