Research On The Train Fault Diagnosis Expert System Intelligent Technologies

With the strong development of infrastructure construction in China, the development of high-speed rail and high-speed trains are also in full swing. In recent years, the number and speed of high-speed trains have been greatly improved, and the security requirements are also increasing. We developed the fault diagnosis expert system for high-speed trains, and studied the fault diagnosis expert system intelligence technology and the reasoning algorithm of fault diagnosis. This thesis’s key intelligence technologies are fuzzy reasoning, image recognition and association rules mining.In order to implement inference engine in the train fault diagnosis expert system, we proposed a new fuzzy inference method based on Lexical Analysis. Firstly, fault antecedents were processed by lexical analysis including removing redundant vocabulary and word segmentation. Secondly, for the purpose of improving the precision of the inference, we used the membership degree, and defined the confidence level of antecedent and the confidence level of conclusion and the closeness degree of the knowledge base rules. Finally we gave the reasoning processes and algorithms. To work through the above procedure, we could get accurate fault diagnosis reasoning results which were consistent with the reasoning results of experts in the field. Theoretical analysis and application showed that this method could achieve hybrid fuzzy inference of Chinese characters and English words and could easily extend to the language related to Chinese, which is universality and high efficiency in the field of fault diagnosis.Nowadays, fault diagnosis based on image recognition generally has the problem of low accuracy and instability. We proposed a new fuzzy inference method based on fuzzy image recognition set to implement fuzzy image inference engine in the train fault diagnosis expert system. Firstly, the concept of fuzzy image recognition set was introduced which is fault pictures with different angles and distances for a failure, and the membership degree was defined. Secondly, for the purpose of improving the precision of the inference, we defined the confidence level of antecedent and the confidence level of conclusion and the closeness degree of the knowledge base rules. Thirdly, we gave the reasoning processes and algorithms. Finally, this method is applied in the train fault diagnosis expert system.Theoretical analysis and application showed that this method is accurate and high efficient in the field of fault diagnosis.In the history library for fault diagnosis, we used the association rule mining method to find the relevant diagnostic history, got new knowledge through self-learning method, and expanded the knowledge base.