Research Of The Knowledge Discovery Application Based On Rough Set In Mechanical Fault Diagnosis

It will cause serious economical lose if the faults would not be diagnosed when the machines go wrong. So it has great significance to research the intelligent system for mechanical equipment. Therefore, the obtaining and effective processing of knowledge has become the bottleneck for the diversity and uncertainty. In recent, manual neutral net has been widely used in machanical fault diagnosis as knowledge obtaining method, it requires a great deal of typical training samples or knowledge proved effective. But it is hard to obtain the typical samples of big type machines.In recent years, the data processing method based on Rough Set has been developed. It does not need the given attributes of some parameters, which can be obtained by means of predigestion. Rough Set has become a very powerful tool in soft computing. We use it to condense the redundant data in expert system, which will reduce the complexity of the fault diagnosis system and improve the whole efficiency.This paper first introduces the base knowledge base of knowledge discovery, the method Rough Set theory and the fault diagnosis method based on knowledge discovery. Second, the paper presents the algorithm discussion of equivalence relation, attribute core, data reduction, relatively positive region. Third , the structure and algorithms of KDP system are raised, many kinds of algorithms of decision rule extraction are researched. The KDP is realized with MATLAB, the thought and idea of interface design are mentioned and MATLAB program is given.The paper also introduces an ES frame named ESTA developed by the Prolog Development Center. The KDP is embedded in the ESTA by C++, which realizes data transfer based on text document. The paper introduces the method of chinesization, extending and database design of the ESTA. It gives the examples of the use of KDP, ESTA and the union of KDP and ESTA in fault diagnosis.