The Study Of Fault Diagnosis Expert System Of Rotating Machinery Based On Artificial Neural Networks

Rotating machinery is widely used in various fields of heavy industry. Usually they ate key equipments in every production field. Since they mostly work in harsh environment with complicated working condition, rotating machinery failure frequently happens which can not only impact on the entire production line, but also cause significant losses even severe industrial injury. In order to ensure these eqipments operate safely, reduce maintenance costs and improve the equipment utilization rate as well, it has become an important research area to develop fault diagnosis expert system that can automatically obtain knowledge and reason in a high speed in the study of rotating machinery fault diagnosis.In view of rotor faults, bearing failure, floating seal failure, fluid vibration fault in vane machine, gear box failure and other types of rotating mechanical failure, their fault symptoms and vibration mechanism are discussed in the dissertation. Then according to the intrinsic characteristics of rotating machinery, it has put forward vector fusion energy spectrum theory to make fault analysis of rotating parts after comparing different signal analysis technologies. However, because the rotating machinery fault diagnosis belongs to a fuzzy zone, the fuzzy cluster analysis is studied. Fuzzy theory can make the uncertain knowledege or qualitative konwledge quantitative. But when using the membership degree, rules and expert's experience are difficult to be accurately described. Further more, it needs large amounts of operating data to construct fuzzy matrix. Therefore, a more effective fault diagnosis method—Artificial Neural Network has been put forward in the diagnosis process.Taking rotor unbalance, rotor misalignment, rotor rub-impact and base flexible into account, BP neural network model is designed and corresponding parameters such as learning rate are set through Matlab toolkit. Then it has obtained the relevant diagnosis results. Based on the analysis of Artificial Neural Network, the dissertation has introduced the basic concept, basic structure, knowledge representation method and reasoning process of expert system. Then the basic syntax of expert system tool CLIPS and Rete pattern matching algorithm are stated. Taking a rotating machine as an example, the fault diagnosis expert system has been developed based on CLIPS.Because the interface of CLIPS-based expert system is too simple, its application and promotion will be limited. Then it attempts to realize the development of expert system combining CLIPS and VC++ programming based on dynamic link library method. In addition to this, JAVA and CLIPS mixed programming methods is also discussed. In the process, the program has implemented internationalization in JAVA. Ultimately the hybrid programming rotating machinery fault diagnosis expert has been developed by rewriting the corresponding code in the appropriate knowledge structure and reasoning mechanism in CLIPS.Finally, taking multilevel planetary gear box as research objective, the thesis takes diagnosis vector broken tooth, rift gear root, tooth surface abrasion, bearing outter race fault, inner race fault and rolling element fault as the output of the artificial neural network. The obtained results are about 33.3% inconsistent with that from expert system. So in the case of incomplete data and inaccurate diagnosis, it can provide a conservative but more reliable diagnostic result for users. On the basis, rotating machinery fault diagnosis expert system is developed with the help of VC++6.0 software. It has realized the development and design of diagnostic reasoning module, knowledge base management module, neural network training module, and verified the results of multilevel planetary gear box fault diagnosis.