Research On Fault Pattern Recognition Method Based On Rough Set Theory

It is one of immediate hot research areas to use rough set theory to solve fault pattern recognition problem for mechanical equipment. The scale and complexity of mechanical equipment have become faster and faster, in practical wielding, data contain varied uncertainty, tradition fault pattern recognition has difficulties to make correct decision while facing problem above. Rough set theory needs no priori knowledge and additional knowledge, and it has superiority while dealing with imperfect, uncertain, incompatible data. Therefore, it is possible to apply rough set theory in fault pattern recognition. Attribute reduction is an emphasis research direction, to obtain the whole reduction set or best reduction set is proved to be an NP problem, the target of researching is to find a relative ideal reduction while maintaining good efficiency and storage space. In this paper, works below aim at rough set theory and fault pattern recognition have been done:This paper analyzed current main attribute reduction methods, emphasis on discernibility matrix algorithm. Compare the efficiency and accuracy of basic, Hu and Ye algorithm. Binary discernibility matrix has good time and space complexity during processing of operation as a transform of discernibility matrix, and is appropriate for further research. Proved equivalence of Ye discernibility matrix and modified binary matrix, the effectiveness of algorithm for computing core as a theory basis, and approached a modified attribute reduction algorithm based on binary matrix. In this algorithm, amount of logical operation has been saved, instead we only need simple math operation to complete the process of attribute reduction, and it is also easy for computer to accomplish, corresponding value reduction algorithm is are also approached.Combined with traditional fault pattern recognition principle, fault pattern recognition system based on rough set theory has been developed applying the modified algorithm approached by this paper. UCI machine learning dataset has been used to test the effectiveness of the system.