基于人工免疫模式识别的故障诊断方法研究

摘要

A fault diagnosis method based on artificial immune pattern recognition is proposed. The fault is regarded as the antigen.Memory antibodies used to figure and recognize antigens are produced based on the mechanism of clone selection, hyper-mutation, immune memory and diversity maintain. Then the antigens i. e. faults are recognized using the classification method based on KNN and threshold.The simulations are done through the experiments of the pump-jack, in which the threshold is 1, the number of the memory antibodies which are nearest to the input faults are 3 and 5, the result show that the accuracy is 100％. The good generalization of the algorithm is proved through recognizing the mutations of the antigens using the method proposed.%提出了一种基于人工免疫模式识别的故障诊断方法,将故障视为抗原,基于生物免疫系统的克隆选择、超变异、免疫记忆以及多样性保持等机制生成能够表示和识别抗原的记忆抗体,然后采用基于KNN的阈值分类法对抗原即故障进行识别;以抽油机井为对象进行仿真研究,阈值取1,与输入故障距离最近的记忆抗体个数取3和5两个值进行试验,仿真结果表明算法的诊断准确率均为100%.通过对变异抗原的识别,表明该方法具有较好的泛化能力.