Research Of Fault Diagnosis Based On Artificial Immune Algorithm

Three fault diagnosis methods based on Artificial Immune Algorithm(AIA) are researchedin this paper, and use them to diagnosis the fault of the pump-jack. The detail work is asfollows:Fault diagnosis method based on Immune neural network is designed in this paper. theimmune mechanism is introduced when adjust the weight of neural network, this method makeuse of the immune algorithm's capability of keeping individual diversity to search the weight ofneural network globaly, it can avoid the BP algorithm from being trapped in the lacal valueefficiently, meanwhile it can improve the convergence speed of algorithm, the fault diagnosisresearch are done on the pump-jack, and two methods are compared and analysed.Fault diagnosis method based on the immune response is proposed. The real code isadopted in this paper in order to solve the problem that using binary code increase thecomputing. Clone selection and hypervariation is introduced besides genetic operator by theauthor, it can improve the speed of generating memory antibodies, and improve blind searchcapability of the genetic algorithm when genarate antibodies, meanwhile the diversity ofmemory antibodies is ensured. Inorder to make the memory antibodies express structures andcharacters for more antigens, this paper proposes the idea that using multi-antigens to generatememory antibodies and proposes the strategy that stimulate and suppress the antibodies whichare generated by the same antigen based on the living expection and that stimulate and suppressthe antibodies which are generated by different antigens belong to the same fault pattern basedon the concentration, This can avoid the phenomenon of immaturity convergence. Besides, knearest neighbor(KNN) method based on threshold is proposed in this paper. The faultdiagnosis research is done on the pump-jack, The result show that this method can get memoryantibodies which can identify antigens effectively, and the effect to identify antigens which arevariated is presented.Abnormity detection and fault diagnosis method based on negative selection is designed,The author introduce the idea of alterable detector radius in order to exert the detection functionof detectors and needn't to set detector radius. In order to solve the problem of thedetectors'number and detection area, this paper adopts the method that use anneal algorithm tooptimize detectors and make the cover space for nonself lager with limited detectors. Besides,inorder to detect and diagnose the unknown fault pattern, this paper put the detectors which arenot activated together to get unknown pattern fault detectors, and to detect new pattern fault.The fault diagnosis research is done on the pump-jack, the result show that this method can getrapid and correct result based on a lot of normal data, and can get good fault diagnosis effectusing less fault samples.