A Knowledge Acquisition Method Based on Rough Set Theory and Neural Networks to Electrostatic Monitoring Systems for the gas path in an Aeroengine

摘要

Electrostatic monitoring technology is a tool to improve the health management capability for aero-engines. This paper presents an application of intelligent information processing methods for extracting rules from the sparse experimental data which is available from an engine test-bench., A knowledge acquisition method is proposed which is based on a neural network, rough set theory and a genetic algorithm. Firstly, the continuous data set is discretized using a self-organizing map neural network, then the threshold for each attribute is obtained. Attribute reduction is performed and the sensitive features are selected using a reduction method based on rough set theory. An architecture-adaptive neural network with a better generalization is then constructed using a genetic algorithm. The trained neural network is applied to generate new data sets which can then be used for extracting “if-then” rules. The experimental results show that this method can effectively extract rules for fault identification and is feasible for use in an electrostatic monitoring system.