The Application Of Data Mining Technology In The Fault Diagnosis

With the development of the producing technology, the comprehensive production capacity of the production equipment is improved, and the structure of the equipment is becoming more and more complicated, that the economic benefit of the enterprise is directly affected by the operation state of the equipment. As the throughput of the daily monitoring information of the equipment is too large, artificial data analysis is impractical, and data mining is the first choice to solve this problem.Data mining technology can be understood as a new type of “knowledge” acquisition tool, also a technology that its internal information is explored from the data itself and its results are displayed. The development of computer application technology has laid a solid foundation for the application of data mining technology in fault diagnosis,by which the intelligent diagnosis is done for the equipment fault. It helps equipment maintenance personnel and fault diagnosis experts diagnose the equipment timely,propose correct maintenance advices, speed up the maintenance, improve maintenance quality, and save maintenance fund. Whether is starting off from the perspectives of economic or safety, how to detect the equipment failure timely in the production is the most important in ensuring the normal production of the equipments,therefore, the application research of data mining technology in fault diagnosis is of great significance.In this paper, the technical route of application of data mining technology in equipment fault diagnosis is firstly determined, fault diagnosis related algorithms in data mining technology of the current stage is mainly studied. According to the condition of actual production data, rough sets are utilized to conduct data reduction for the fault characterization property, and the needed fault matching rules are digged out from the reduction results using decision tree and association rules, thus fault diagnosis is conducted for the equipment by the fault matching rules. The detailed work is as follows:1 、 The genetic algorithm is used to optimize the algorithm of fault attribute reduction in data preprocessing, and the rough approximation precision in the rough set theory is utilized to determine the importance of information attribute. From which the fault diagnosis decision table is constituted by selecting the attributes which have higher degree of attribute importance, and the attribute core of decision information isobtained by using the identification matrix. The initial population is constructed on the standard of the attribute core, the search area of genetic algorithm is reduced.Finally, the correction operator based on the rough approximation precision is introduced, and the algorithm is made to conduct in the correct solution space, thus the speed of fault attribute reduction is improved, furthermore, the optimal results of fault attribute reduction are obtained.2、Because the data is needed to be converted into executable data files when the original ID3 algorithm implements, and the process is comparatively complex. The high efficiency of the SQL language and the flexibility of C# language are combined together, that classification of fault data is operated directly in the fault database,which solves the deficiency that the ID3 algorithm can not operate on a large number of fault data. This has improved the enforceability and execution efficiency of the ID3 algorithm, and made the increase of data in the fault database not influence the execution and efficiency of the algorithm, that it is made to have a good scalability.3、When the Apriori algorithm is searching for the frequent itemsets, it usually scan the database repeatedly for many times and generate large quantities of useless candidate sets. For this problem, an improved Apriori algorithm based on matrix reduction is proposed. It only needs to scan the database once, and the database information is converted into a Boolean matrix. Then the process of pruning and connection in the original algorithm are reduced according to the conclusions deduced by the properties of K-frequent itemsets, which has reduced the production scale of invalid candidate itemsets effectively, and improved the efficiency of fault rule acquisition.In this paper, data of equipment fault in hydraulic system is the research object,from which data mining algorithm is used to dig out the required fault matching rules,and it’s compared with the original fault data, by which the correct fault matching rules are selected. Then the fault matching rules, fault attribute representation, and maintenance opinions are employed to compose the fault rule base, which provides technical support and theoretical basis for the equipment fault diagnosis of the hydraulic system. On the basis of the related research and under the Windows 7operating system, the development tool of Microsoft Visual Studio 2010, Oracle10 g database and its management system OEM, and the object-oriented development language C# are utilized to design a equipment fault diagnosis system in the structure of C/S. Finally, the system is tested using the actual production data, it’s shown thatthe system meets the needed requirements of the equipment fault diagnosis.