摘要
提出了一种利用遗传算法优化的SVM多分类决策树(GADT-SVM)实现模拟电路故障诊断的新方法。介绍了GADT-SVM的设计思想和算法原理;利用传递函数对模拟电路进行建模,并用小波分解提取电路冲激响应的能量分布作为故障特征;使用GADT-SVM对故障特征样本进行分类实现故障诊断。仿真结果表明,与未经优化的DAG-SVM和DT-SVM故障诊断方法相比,该方法可以减小诊断"误差积累"的影响,具有更好的误差控制能力。
A new method for analog circuit fault diagnosis is presented based on genetic algorithm optimized support vector machine multi-class decision tree （GADT-SVM）. The design idea and algorithm principle of GADT-SVM is introduced firstly; then model of analog circuit is built by transfer function, and fault characteristic is picked-up by wavelet energy distribution of impulse response. Finally, fault samples are recognized by GADT-SVM. Experiment results show that our method can depress error accumulation phenomena of diagnosis and have stronger error control ability compared with the traditional DAG-SVM and DT-SVM.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2009年第4期553-558,共6页
Journal of University of Electronic Science and Technology of China
基金
部级基础科研项目(A1420061264)
部级预研基金(9140A17030308DZ02)
关键词
模拟电路
故障诊断
遗传算法
支持向量机
analog circuit
fault diagnosis
genetic algorithm
support vector machine
