The Research And Application Of Fault Diagnose Expert System Based On AIS

Modern mechanical industry's automation level became higher and higher, and its structure also became more and more complex. Existing fault diagnose techniques could not satisfy such complex equipments for fault diagnose. The main purpose of this paper was to research a fault diagnose technique which aimed at such complex equipments, and then overcame the problem of acquiring fault sample.This paper mainly researched the fault diagnose expert system. Based on the existing research results of fault diagnose expert system, this paper introduced artificial immune system, which was hot spot of artificial intelligence, to improve traditional fault diagnose expert system, and built a new method of fault diagnose with mixed artificial intelligence.This paper introduced negative selection to solve to the problem of acquiring fault data in complex structure equipments. According to lots of normal operation data and abnormal operation data, detector set was generated and saved to knowledge base. It could solve trouble bought from lack of abnormal operation data. But the detectors which were generated by valued negative selection algorithm overlapped between self and non-self detectors, so this paper used an improved negative selection algorithm which was chaotic-based negative selection algorithm. Because of the advantage of chaotic on the aspects of ergodicity and non-reproducibility, it solved the problem of overlapped between detectors. In the comparison simulation by Iris dataset, improved negative selection algorithm was better than valued negative selection algorithm in aspect of detection effect.In the process of detectors training, this paper proposed the method of knowledge base dynamic update based on K-Nearest neighbor algorithm and clonal selection. The method realized the self-study function of expert system. Knowledge base dynamic updating extended coverage of detector set and improved correctness and adaptability of fault diagnose expert system. The inference mechanism used antigen-antibody binding energy to match the sample. The lower binding energy, the more suited between sample and detector. The sample attributed to the type of this detector with the lowest binding energy.Finally, this paper applied the fault diagnose expert system based on artificial immune system to fault diagnose of refrigeration system and checked its application effect. The results showed that the correct rates achieved anticipative effect after using such expert system, and proved that this method of mixed artificial intelligence fault diagnose had its feasibility.