论文部分内容阅读
针对机车主变流器故障,借助Matlab中的Simulink工具箱,仿真得到了主变流器发生不同故障时相应的电路输出电压;采用小波分析对输出电压进行信号处理,提取故障特征,构造特征向量,获得蕴涵电路故障信息的样本;利用SVM工具箱构建多故障分类器,并对故障样本进行诊断。由于SVM在无限空间建立分类超平面,容易导致误分类,针对此问题,提出了改进方法,使SVM在确定的有限空间进行分类。最后用仿真样本进行了训练和测试。实验结果表明诊断效果良好,实现了预期目标。
Aimed at locomotive main converter fault, with the aid of Simulink toolbox in Matlab, the corresponding circuit output voltage is simulated when different fault occurs in the main converter. The output voltage is processed by wavelet analysis to extract fault features and construct eigenvectors , Get the sample of implication circuit fault information; use SVM toolbox to build multi-fault classifier, and diagnose fault samples. Because SVM establishes classification hyperplane in an infinite space, it is easy to cause misclassification. In view of this problem, an improved method is proposed to classify SVM in a limited space. Finally, the simulation samples were used for training and testing. The experimental results show that the diagnosis is effective and achieve the expected goal.