基于优选小波包和ELM的模拟电路故障诊断

Analog circuit fault diagnosis method based on preferred wavelet packet and ELM

为提高模拟电路故障特征提取的有效性以及实现对故障模式的准确分类,提出了一种优选小波包和极端学习机相结合的模拟电路故障诊断新方法。为获取最优故障特征,提出了特征偏离度的概念,可作为评价小波包变换在不同小波基函数下获取的故障特征的一种测度,可据此选择特征偏离度最大的小波基进行故障特征提取;在此基础上,引入极端学习机对故障进行分类识别,并将诊断结果与目前几种主要神经网络方法进行了比较。仿真实验结果表明:利用优选小波包提取的最优故障特征能够得到更高的诊断精度,而极端学习机在测试时间和诊断精度上都优于其他3种神经网络方法,能够在不到1 ms时间内实现94.44%的诊断精度,说明了所提方法在模拟电路故障诊断中的有效性。

In order to enhance effectiveness of feature extraction and classify fault patterns exactly in analog circuit fault diagnosis, a new analog circuit fault diagnosis method based on preferred wavelet packet and extreme learning machine （ELM） is proposed. In order to obtain the best fault feature,the concept of feature departure degree is de- fined, which can be used as the measure of the fault features extracted by wavelet packet transform using various wavelet basis functions, and the wavelet basis function with maximum feature departure degree is selected and used to extract the fault feature. Further, the ELM is introduced for fault classification and identification, and the diagnosis re- sult is compared with those using three popular neural networks. The simulation results show that the better diagnosis precision can be achieved using the preferred wavelet packet, and the test time and the classification precision of the ELM are all better than those using other methods; the proposed method can achieve the diagnosis precision of 94.44% in less than 1 ms,which verifies the effectiveness of the proposed method in analog circuit fault diagnosis.