基于优化AdaBoost-SVM的模拟电路故障诊断

Fault Diagnosis of Analog Circuit Based on Optimized AdaBoost-SVM

为提高含容参元件模拟电路软故障的诊断率,并考虑到单分类器分类精度的提升已达到了一个瓶颈,提出一种优化AdaBoost-SVM算法并将其应用于模拟电路故障诊断中。以OrCAD/PSpice软件中对电路进行Monte-Carlo分析的数据为基础,选取特征时,采用对时频信号中易直接测量的物理量归一化后组合的方式。实验结果表明,通过选取的组合特征向量,利用优化的AdaBoost-SVM算法,构造出具有差异度的SVM分类器并集成后,能够自适应地提升单SVM分类器性能,表现出更好的分类精度与泛化性能,能较好地满足容差模拟电路软故障诊断要求。

In order to improve the diagnostic rate of soft faults in analog circuits with capacitive parameters,and considering that the improvement of classification accuracy of single classifier has reached a bottleneck,this paper proposes an optimized AdaBoost-SVM algorithm and applies it to analog circuit fault diagnosis.Based on the Monte-Carlo analysis data of the circuit in OrCAD/PSpice software,the normalized combination of physical quantities which are easy to measure directly in time-frequency signal is adopted when selecting features.The experimental results show that the optimized AdaBoost-SVM algorithm which is used to construct and integrate SVM classifiers with different combinations can adaptively improve the performance of single SVM classifiers and shows better classification accuracy and generalization performance.This method can well meet the requirements of soft fault diagnosis for tolerance analog circuits.