Research And Application Of Granular Computing Based On Rough Sets In Data Mining

With the rapid development of modern industry mechanization, there is a direct relationship between production security and reliability of equipment operation. Currently, the gear transmission is the most widely used form in mechanical system, so gear failures is one of the most important factor of equipment failure, therefore, how to prevent and reduce the occurrence of gear failures effectively, detect equipment malfunction in time, and estimate the type and location of gear failures accurately is a very meaningful work. At present, there are several common methods for gear fault diagnosis, such as times domain analysis, frequency domain analysis, DEA, wavelet analysis method and so on. But traditional methods require experienced professionals, it is a high demand for human; they identify the type of gear failure by determining waveform characteristics of gear vibration signals. With the development of intelligent technology, intelligent approaches have been widely introduced to gear fault diagnosis, it makes fault diagnosis systematic, intelligent, and improve diagnostic accuracy greatly.The messages conveyed by vibration signals of different types of gear fault are different; in order to discover potential rules, granular computing method is very effective. As a new concept of artificial intelligence, the nature of granular computing method is that by choosing a reasonable granulate method to get a similar, better solution. As one of the three theories of granular computing, rough set can effectively deal with imprecise, inconsistent, incomplete information and knowledge. Therefore, with the use of granular computing in data mining field, it is more effectively to translate mass data to potential useful knowledge, and reduce the difficulty of problem.Based on the idea of granular computing, this article mainly has a research on attribute reduction and rule extraction method of decision table information system. Proposed the reduction algorithm based on granular computing of information system, the basic idea is:remove the attribute which has no effect on the division one by one, thus get the reduction. Compared with traditional attribute reduction algorithms, this algorithm need not to seek nuclear attributes, the structure is clear. In this thesis, the concept of relative distribution function and relative distribution granularity are introduced in decision table information system, we use the difference between relative distribution as the heuristic information, then propose the attribute reduction algorithm based on relative distribution granularity of decision table information system, and use UCI data sets to validate the algorithm, the experimental result shows that this algorithm is effective and efficient. Then we give the representation of knowledge granule, study granular computing and the representation and conversion of granular layer. On this basis, we propose a rule extraction algorithm based on granular computing. Combining with data mining, a data mining model based on granular computing is given, and we use it to discover potential rules for four common types of gear fault, the result shows that the extract rules and actual fault signal characteristics are consistent, could diagnose faults effectively. And on this basis, the thesis realizes the gear fault diagnosis simulation system.