Research Of Attribute Reduction And Engine Fault Diagnosis Based On Rough Set

Because of the complex structure and the bad working environment, the fault data show incompleteness and uncertainty, it make great trouble in car engine fault diagnose. Rough set theory in which the priori knowledge and prerequisite knowledge except measurement data are not necessary can find out implicit information and reveal latent rules by directly analyzing and reasoning all the incompleteness and uncertainty data effectively. The theory is increasingly mature and is becoming a powerful tool for engine fault diagnosis and rules knowledge acquisition.The method of fault diagnose based on rough set is mainly obtaining the useful rules according to calculating the fault data of incompleteness and uncertainty. The rules will directly affect the result of fault diagnose, which are obtained by attribute reduction of the decision tables firstly. Thus, the algorithm of attribute reduction will be mainly researched. The main work of this paper are as follows:(1)Improved the traditional algorithm of attribute reduction based on rough set discernibility matrix by researched rough set theory deeply, which is a algorithm of attribute reduction use core and pseudo-core as heuristics information. The method of attribute reduction based on improved algorithm is by calculate the discernibility function and change it into minimum disjunctive, then, add the core properties to every conjunction normal form of the minimum disjunctive to obtain the result of reduction. This algorithm can improving the computational efficiency significantly when the decision tables not have core or a few condition attributes;(2)Combined the advantages of the HORAFA-AFVDM algorithm to further improve the algorithm and make it has a universal applicability. The improved algorithm use the frequency of the condition attributes in the simplified discernibility matrix as the heuristics information to make the attribute reduction, and judge whether the result of the attribute reduction is correct or not by according to the positive region of the simplified and original decision tables;(3)Constructed the engine fault diagnose system based on the rough set, which can diagnose according to the fault symptoms and fault data. Some functions of this sysytem are programming. The software development used the VS2010integrated development environment. The construction of the example data and the stored rules are based on SQL SERVER2008, and use the ADO as the database access interface. The algorithm of the car engine fault diagnose is programming, whose correctness is proved by the examples.The improved algorithm of attribute reduction is tested and verified through some fault data, the verification results show that the improved algorithm is feasible and superiority, by case analysis and testing of each module of the software, achieves a better diagnostic results.