Analysis of Decision Rules for Rolling Bearing Fault Diagnosis Based on Rough Set Theory

摘要

In order to extract simple and effective diagnostic rules from original datum, the method based on rough set theory is proposed. The fault diagnostic problems are described by defining information system and decision table from fault states. The decision rules are established based on indiscernible relation of attributes by producing the sets of attribute symptoms and selecting values of attributes. It is shown that every decision rule reveals some probabilistic properties. It satisfies the Total Probability Theorem and the Bayes' Theorem. These properties give a new method of drawing conclusion from datum without referring to prior and posterior probabilities. The efficiency of diagnostic rules is improved by pruning rules properly with certainty factor. The redundant information can be removed by the coverage factor of rules effectively. The availability of this method is proved by a fault diagnosis example of rolling bearing.