基于小波与LS-SVM集成的模拟电路故障检测

Analog circuits fault diagnosis based on wavelet ensemble LS-SVM

由于模拟电路的多样性、非线性和离散性等特点,模拟电路的故障诊断呈现复杂、难以辨识等问题。针对已有方法的数据不平衡,提出了一种支持向量机集成的故障诊断方法。使用小波变换方法提取特征向量,在多类别支持向量机的基础上,设计了模拟电路的最小二乘支持向量机预测模型,实现了对模拟电路的状态的故障预测。将该方法应用于Sallen-Key带通电路进行故障预测试验,结果表明,该方法比单一支持向量机、径向基神经网络、BP神经网络和APSVM有更好的分类和泛化性能,故障诊断准确率更高。

Due to characteristics such as variety, non-linear and discrete, fault diagnosis over analog circuits exhibits complicated and hard to recognize. To improve data imbalance of fault diagnosis methods over analog circuits, an ensemble SVM is proposed. This paper utilizes wavelet to extract feature vector from analog circuits signals and proposes analog circuits fault diagnosis model based on multi-class least square support vector machines. The proposed method is tested over Sallen-Key band-pass. Experimental results show that our method has better accuracy and generalization ability than popular fault diagnosis methods such as o-v-o SVM, RBFNN, BPNN and APSVM.