Concerned with the problem of lacking fault samples of complex equipments, it studies the principle and application of negative selection algorithm of artificial immune system. The detectors generation mechanism of real — valued negative selection algorithm is introduced. Intuitively, to maximize the covering produced by a set of detectors, it is necessary to reduce their overlapping and not covering the self set. This paper presents an optimization strategy based on re—heating simulated annealing algorithm to modify the location of detectors, not changing their number. This method can improve the covering effect of non — self space. Two synthetic 2— dimension datasets are used to demonstrate the properties of optimized VRNS. Detection rate is improved and false alarming rate is decreased. It is used for 12 soft faults detection in the filter circuit, the whole detection is 95%. The result demonstrates that the proposed algorithm is better than artificial neural network in fault detection of this circuit.%针对复杂装备故障样本少的难题,研究了人工免疫系统中的阴性选择算法原理及应用；介绍了变尺寸实值阴性选择算法检测器产生机制,以减小检测器交叠和对自体的覆盖为目标,提出一种基于重升温模拟退火算法的检测器分布优化策略；该优化方法不改变原有检测器数量,提高了对非己空间的覆盖效果；对两种不同几何形状的二维数据集进行仿真,结果表明,优化方法提高了算法的检测率,降低了虚警率；将优化后的算法应用于滤波电路12种软故障的检测,总体检测率达95％,结果优于基于人工神经网络的故障检测方法.
展开▼
