基于故障可诊性的模拟电路多音正弦测试信号进化生成

摘要

为增加模拟电路故障诊断中故障模式之间的可辨识性,提高诊断精度,提出一种基于故障可诊性的模拟电路多音正弦测试信号进化设计方法.该方法采用多音正弦信号作为模拟电路故障诊断的激励信号,根据遗传算法的进化生成过程,用混合编程法直接修改待测电路仿真程序中测试信号参数值,实现电路的动态响应.同时动态采集电路可测节点的故障样本集,采用核模糊聚类算法(KFCM)对故障样本进行预处理,计算样本在核空间的类间距离,以类间距离大小作为衡量故障模式可诊性的依据,建立遗传算法寻优的目标函数.实验结果表明优化得到的多音正弦测试信号有效提高了待测电路的故障诊断率.%In order to improve the distinguish ability of fault samples in analog circuits and enhance the diagnostic performance of fault diagnosis, a novel evolutionary generation method of multi-tone sine stimulus based on fault distinguishable analysis is proposed in this paper. In this method, a multi-tone sine signal is used as the test stimulus. According to the evolutionary generation process of genetic algorithm ( GA) dynamic response of the circuit under test (CUT) is implemented by directly modifying stimulated parameters by means of a mixed programming approach. The fault samples are collected dynamically from the test nodes, and then they are preprocessed by using the kernel fuzzy c-means clustering ( KFCM) algorithm. The among-class distance between samples in kernel space is calculated and is taken as a standard for the fault distinguish ability. Therefore, the max among-class distance is designed as the optimization goal in the GA algorithm. Experimental results reveal that the optimal multi-tone sine test signal succeeds in improving the diagnosis results.