WIT Press

FAULT DIAGNOSIS USING A DECISION TREE OF SIMPLE MODULAR NEURAL NETWORKS

Price

Free (open access)

Paper DOI

10.2495/AI980421

Volume

20

Pages

8

Published

1998

Size

40 kb

Author(s)

P.R. Drake & M.K. Kidwai

Abstract

FAULT DIAGNOSIS USING A DECISION TREE OF SIMPLE MODULAR NEURAL NETWORKS P. R. Drake & M. K. Kidwai Intelligent Systems Laboratory, School of Engineering, University of Wales Cardiff, P. O. Box 688, Newport Road, Cardiff CF2 3TE, UK. Email: Drake@cf.ac.uk, KidwaiMK@cf.ac.uk ABSTRACT A method of fault diagnosis using simple modular neural networks in a decision tree is proposed. The diagnostic accuracy of such a classifier is shown to be better than a single holistic neural network when applied to diagnosing faults in a seven component RC–network. 1 INTRODUCTION Neural networks have been used widely and effectively for fault diagnosis. Typically the approach is one of using a single holistic neural network with a large number of outputs, one for each fault class. Such a large network can be difficult to train and analyse when the number of fault classes is large. The approach presented here is to use simple modular neural networks – one for each fault class. Modular neural networks (MNN’s)

Keywords