A machine learning approach to generate rules for process fault diagnosis

Shastri, S., Lam, C.P. and Werner, B. (2004) A machine learning approach to generate rules for process fault diagnosis. Journal of Chemical Engineering of Japan, 37 (6). pp. 691-697.

Abstract

Expert systems can play a very important role in manufacturing processes by locating problems as soon as they arise. The most important ingredient in any expert system is knowledge. The current knowledge acquisition method is slow and tedious and there exist substantial difficulties in acquiring the knowledge for complex processes. An approach is proposed that makes use of the machine learning technique, C4.5, to generate a decision tree. The decision tree is translated into rules that are implemented into the expert system shell, G2. The rules are tested using a sensitivity analysis of the system. The approach works well, but depends on both the quality and quantity of available training data.

Item Type:	Journal Article
Murdoch Affiliation(s):	School of Engineering
Publisher:	The Society of Chemical Engineers, Japan
Copyright:	© 2004 The Society of Chemical Engineers, Japan
URI:	http://researchrepository.murdoch.edu.au/id/eprint/33184

Item Control Page