The Application Research Of Rough Set Based On Information Entropy In Fault Diagnosis

Rolling bearing is one of the rotating parts that are used most widely and one of the parts that are damaged most easily in mechanical equipments. Its state of running will influence the performances of the whole machine. So the research on the inspection of the working conditions and the technology of fault diagnosis is paid more and more attention by people and becomes an important measure that ensures rotating machines to run well.For a long time, many researchers have proposed some intelligent methods that apply to the fault diagnosis of rolling bearing. These methods include the method based on model, expert system, fuzzy logic and ANN. They set up the intelligent diagnosis of rolling bearing, and improve the efficiency of the running of system and the decision-making of maintenance personnel. However, these diagnosis methods are essentially based on pattern-recognition and pattern classification. When there are many patterns, there exit these problems, such as rule explode and the difficulty to deal with incomplete and inconsistent information. Considering this, we introduce rough set theory.The main contribution of the dissertation can be summarized as follows:In the course of applying rough set method, we discuss how to deal with the value of absent data and how to discretize the data. The general algorithm of attribute reduction and value reduction is presented, from which the attribute reduction algorithm based on information is proposed.The conception, development and effect of entropy are expounded in detail. In the selection of character of the fault of rolling bearing, entropy is introduced for the first time and time domain entropy and frequency domain entropy are selected.The application of discretizing data and reducing information table in rough set is studied, and the fault diagnosis system of rough set theory based on information entropy is set up.In this thesis, a method that applies to the fault diagnosis of rolling bearing is described. In this method, training samples whose decision attribute values are equal are selected to form information table for the first time, and the information table is reduced by the attribute reduction algorithm based on information, and the information table is analyzed from the point of information entropy for the first time.Through experiment and calculation, the method is proved to be feasible.Finally, some future research directions on fault diagnosis based on rough set are highlighted.