Applied Study Of Artificial Immune System In Unit Fault Detection Technology

The essence of the fault diagnosis technology is to know the state of the equipment in the process of running, to forecast its reliability, to determine its normality or exceptionality, to discovery fault earlier and estimate and identify its reason, position and criticality. It is also to predict the development trend of fault and make decision through its concrete status.Immune system is the complex distributional information processing learning system. Its have kinds of function such as immune protection, immunological tolerance, immune memory, immune surveillance, especially have kinds of characteristic such as stronger adaptability, multiplicity, study, recognition and memory. The combination between functions and characteristics gained various artificial intelligence methods based on immune mechanism, which solve the massive non-linear scientific problem. The information processing mechanism of immune system is of important theoretical significance and practical value in fault diagnosis.The primary subject of this paper is to(1) The negative-selection algorithm that can be detected abnormity is improved by the mutation mechanism of genetic algorithm. The result of simulation show that the computing complexity of the improved algorithm which can basically cover the self-space declines in evidence. Meanwhile its efficiency greatly improves to fault detection.(2) In the mixture of the idea of Mind Evolutionary Computation (MEC) and the clonal selection principle of artificial immune system, a model of artificial immune evolutionary computing is designed. The models learn and recognize to fault mode in clone selection theory, and the immune similartaxis operator to expend antibody and the immune dissimilation operator to restrain antibody are defined using the thought of mine evolutionary computing. The algorithm are used in state recognize of simulation machine unit. The experiment result shows that the algorithm has a better validation to state recognize.(3 ) Based on conal selection principle and K-Nearest Neighbor method, a fault detection algorithm that can produce the valid detector to diagnose fault is designed by the three parts of immune vaccine, immune learning and immune response. The simulation examples show that the detector produced by the algorithm has a better accuracy to fault detection.