基于小波分析的最优故障特征提取研究

Research of Optimal Fault Feature Extraction Based on Wavelet Analysis

针对因模拟电路的故障模型复杂、有容差、非线性等导致的模拟电路故障特征提取难度大、严重依赖于专家的经验的现状,对基于小波分析的模拟电路最优故障特征提取技术进行了研究;以四运放电路为实验基础,采用Morlet和Haar两种小波基分别从不同的维度上做数据预处理,能量化、归一化后组成故障特征,而后通过克隆选择算法的诊断结果分析对比特征提取的效果;实验结果为通过两种小波基提取的故障特征在不同的情况下达到最高故障诊断率均接近89%,表明基于两种小波基的故障特征提取技术都是优秀可用的,以及单点采样数据的有效性;同时实验结果还反映了模拟电路故障特征的详细程度与诊断正确率成正比例关系;这对实际复杂模拟电路的故障特征提取具有指导性的意义。

Due to the complexity, nonlinearity and tolerance of analog circuit fault model, the feature extraction of analog circuit is difficult and rely heavily on the expert＇ s experience. In order to solve this situation, this paper tried to find a method of the optimal analog circuit fault feature extraction based on the wavelet analysis. The Experimental circuit is four op--amp biquad high--pass filter circuit. The fault feature was extracted from the voltage data hy using both Morlet and Haar wavelet with multiple perspectives, then comparing the effect of feature extraction with the diagnostic results of the Clonal Selection Algorithm. Results shows that the best fault diagnostic rate is closed to 89% in different circumstances of two kinds of wavelet, which prove that both two method are available and useful. And results shows the effectiveness of the single point sampling data, at the same time, the level of details of fault feature is positively related to the accuracy of diagnostic. These have the guiding significance for the fault feature extraction in the practical large-scale analog circuit.