A Method for Rule Extraction Based on Granular Computing: Application in the Fault Diagnosis of a Helicopter Transmission System

Abstract

How to extract decision rules from incomplete decision table is of importance in fault diagnosis of helicopter transmission system. This paper introduces a knowledge acquisition method based on Granular Computing (GrC) for fault diagnosis of helicopter transmission system. First, following semantic analysis of missing attribute values in decision table, the basic idea of construction and interpretation of granules based on characteristic relation is studied. Then, the definition of GrC model based on characteristic relation as well as its construction algorithm is developed. Thus, a set of granules can be obtained completely and its implied information is consistent with the original decision table. Subsequently, the algorithm of attribute reduction in GrC is proposed. According to the definition of generalized decision rule, the way of extracting optimal decision rule from granules is studied. At last, Combined with an incomplete decision table for fault diagnosis of transmission system, this method has been achieved, and the analysis result shows its validity.