Intelligent Fault Diagnosis Based On Rough Set Theory

With the development of modern industry and technology, the equipments become more and more large, complicated and intelligent, and the faults are fuzzy, random and uncertain. Most faults are caused by many factors and these factors are interactional, which causes the traditional fault diagnosis and detection technology(FDD) more and more difficult to exert its action in system fault. Therefore, with the development of computer engineering and artificial intelligence, especially the applications of knowledge engineering, expert system and artificial neural network in diagnosis fields, the FDD are more and more intelligent. The intelligent fault diagnosis and detection technology(IFDD), token on the applications of expert knowledge and artificial intelligence, is the combining of artificial intelligence and fault diagnosis. The IFDD is based on information processing and knowledge engineering of human thinking and makes decision to the circuiting state and faults of objects through obtainning, transferring and disposing diagnosis information effectively. The IFDD based on rough sets is one of them. Rough sets theory was proposed by Professor Pawlak Z, a Polish mathematician, in 1982, which was a new mathematic tool to deal with incomplete or inconsistent problems. In virtue of equivalence relation and approximate concept it reducts the data to obtain the knowledge information. The rough knowledge system is a system based on rules, which does not need accurate mathematic description, but a summary experience. It provides the theoretic base and research idea to implement the intelligent fault diagnosis, because it satisfies the demands of intuitionistic, simple, human comprehensible and intelligent.Singular Rough Sets, which was proposed by Professor Kaiquan Shi in 2002, was an improvement of Pawlak rough sets. S-rough sets is based on element transfer and its extended form is function S-rough sets, which was also put forward by Professor Shi. Function S-rough sets provides us a new research field and theorical guarantee in the research of system laws.The main contributions of this dissertation are as follows:1. Reviewed the developments and research situation of rough sets theory. Introduced the concepts of equivalence relation, approximate space, attribute core, attribute reduction and knowledge discovery of rough sets theory. Discussed and compared the attribute reduction approaches based on discernibility matrix and presented a new improved discernibility matrix model to reduct the attributes. At last, given the algorithm of rules obtainning and optimazing based on this model.2. Summarized the develpoments and research situation of fault diagnosis and detection technology, the classification and approaches of fault diagnosis. Especially introduced the processes of intelligent fault diagnosis based on rough sets theory. And give some examples of fault diagnosis to prove the validity of the approach of fault diagnosis presented in this dissertation.3. Through introducing the concepts of S-rough sets and its characteristics, it discussed the knowledge heredity mining of S-rough sets and presented the relation theorem of heredity-variation and algorithm of heredity mining. based on this, it introduced the concepts of function S-rough sets and its characteristics, put forward the F- recognition rule of system law and the F- diagnosis rule of fault law.