摘要:介绍了EMD分解的条件、原理及步骤,提出了特征向量构造的方法和步骤,阐述了LS-SVM的原理,说明了LS-SVM用于多类分类问题的算法。以旋转整流器一个和两个功率二极管断路故障为例,通过对励磁电流信号的EMD分解,得到了以能量为基础的特征向量,建立了高斯径向基核函数的LSSVM故障模式分类器,对分类结果进行了测试、分析和比较,结果表明,所建分类器故障识别率高、用时短,比神经网络分类器更适合用于在线诊断系统。 关键词:EMD;LS-SVM;旋转整流器;特征提取;模式识别
Abstract: The condition, truth and approach of EMD were introduced in the work, and the method and approach of building up the characteristic vector, the truth of LS-SVM and the arithmetic in the classification were also included. taking the faults of one and two diodes turning off for example, the characteristic vector based on energy can be got after the EMD of inspiring current, and the classifying method based on Gauss RBF LS-SVM can be reached, after testing, analysis and comparing, the results show that the classified method owns a higher exactness, takes less time and has more application on the on-line diagnosing than neural network. Key words: EMD LS-SVM rotating rectifier the distilling of character the identification of modes
|