摘要
为提高模拟电路故障诊断效率,克服依据单一信息进行诊断的不足,提出了一种支持向量机信息融合的模拟电路故障诊断方法;首先构建了基于支持向量机的信息融合诊断模型,其次给出了基于小波包变换的能量特征提取和基于主元分析特征压缩方法,分析了支持向量机一对一多分类方法,最后通过模拟电路的仿真实验,与未进行信息融合,以及BP、RBF和PNN等神经网络对比,结果显示,基于支持向量机信息融合方法的诊断精度最高,约为97.3%。
To improve the efficiency of analog circuit fault diagnosis according to single information, a method based on information fusion and SVM and is proposed. At first, information fusion diagnosis model based on SVM is introduced. Secondly, energy feature extraction based on wavelet packet transform and feature compression based on PCA are talked about. Multi--class one vs. one method of SVM is analyzed. At last, the fusion method, single information diagnosis, BP, RBF, and PNN are used in a simulation experiment of some analog circuit. The result shows that the method is much better than others, whose correct rate is 97. 3%.
出处
《计算机测量与控制》
CSCD
北大核心
2009年第11期2177-2180,共4页
Computer Measurement &Control