| The kernel work in this thesis is the research and design of fault- rule automatic extraction based on rough set, which may improve the ability to reason and solve the decision question with incomplete information and to acquire rule-knowledge automatically in the fault diagnosis expert system. It is a useful exploration of building a fire-new fault diagnosis expert system. The application of the fault diagnosis expert system is so wide in the industry field that the research of our work is worth and important.At the beginning of this thesis, the necessity and feasibility of applying rough set method to the fault diagnosis expert system are presented, at the same time, a model of fault diagnosis expert system based on rough set is proposed. Then on the basis of discussing the representation of fault-sample information and the several key techniques of fault-rule automatic acquisition, the author presents and designs an approach of automatic extraction of fault-rule based on rough set. Some practical problems involved in modeling and practical techniques can be concluded as follows: (a) A model based on rough set of RDB, which represents the fault-sample information as a decision table is defined, (b) Get fault decision-table from database by ODBC, (c) A number of sub-tables are randomly sampled from the input fault decision table by the method of dynamic-reduction, (d) An algorithm of attribute reducing is developed, at the same time, the reduction sets, as the result of attribute reducing, is filtered in order to get the most "stable" reduction in some sense, (e) A structure of fault-rule is defined and a method of extracting fault-rule based on rough set is presented on basis of the rule structure, (f) Some parameters are calculated to value the fault-rule.In the process of system development, a UML model of automatic acquisition of fault-rule based on rough set is designed by OO technique and UML in Rational Rose, and some useful classes such as decision-table class, reduction class rule class and key functions are developed in the platform VC++. Then a fault sample of the noise-fault about automobile engine is given to analyze and prove the effectiveness of the system.At last, a summary of the thesis work is give, and further research interests are also prospected. |
