Skip to Main Content
Extraction of simple, effective and efficient decision rules for fault diagnosis is one of the most important problems in mechanical fault diagnosis because available information is often inconsistent and redundant. Using inconsistent and redundant information, an extraction method of decision rules based on the rough set theory is developed in this paper. First, a decision table for fault diagnosis is obtained by discretization of continuous symptom attributes from original data; second, the fault symptom attributes discretized are reduced using a genetic algorithm; finally, a set of maximal generalized decision rules with certainty factor and coverage factor is generated by using a proposed algorithm. A turbine-generator unit vibration fault is used to illustrate the application process of the method.