Fault Gear Categorization: A Comparative Study on Feature Classification using Rough Set Theory and ID3

Abstract

Fault diagnosis on a gear box is a difficult problem due to the non-stationary type of vibration signals it generates. Usually, one method of fault diagnosis can only inspect one corresponding fault category. Vibration based condition monitoring using machine learning methods is gaining momentum. In this paper, rough sets theory, is used to diagnose the fault gears in a gear box. Through the analysis of the final reducts generated using rough sets theory, it is shown that this method is effective for diagnosing more than one type of fault in a gear. The performance of rough set method are compared with those of the ID3 decision tree algorithm and the results prove that the rough set method has greater capability to bring out the different fault conditions of the gear box under investigation. The study reveals that the overall classification efficiency of the decision tree is to some extent better than the classification efficiency of rough sets method.