This paper proposed a novel fault diagnosis method based on SVM of learning using privileged information (LUPI-SVM),aiming at solving the problem of correctly identifying fault classes in analog circuit fault diagnosis. Firstly, the fault feature vectors were extracted by PCA (principal component analysis) feature extraction method. Then, after training the LUPI-SVM by faulty feature vectors, the LUPI-SVM model of the circuit fault diagnosis system was built. Finally, input the lest samples' feature vectors into the trained LUPI-SVM model to identify the different fault cases. The simulation results for analog and mixed-signal lest benchmark Sallen-Key filter circuits demonstrate that the proposed method improves classification ability. It correctly classifies not only the single hard fault classes with a highly average classification success rate more than 99% , but also the multiple fault classes. The method develops a new direction for the fault diagnosis of analog circuit.%针对模拟电路故障诊断复杂多样难于辨识的问题,提出了基于融合特权信息支持向量机的模拟电路故障诊断新方法.首先对采集的信号进行主成分分析( PCA)——特征提取；然后将训练集输入融合特权信息支持向量机进行训练获得故障诊断模型；最后将测试集输入训练好的支持向量机分类模型,实现对不同故障类型的识别.Sallen-Key滤波电路故障诊断仿真实验结果表明,该方法有效提高了分类的性能,不仅能够正确分类单故障而且能够有效分类多故障,其中单硬故障情况下平均故障诊断率达到了99％以上,为模拟电路故障诊断提供了新的途径.
展开▼
