类间距节点优化DDAG-SVM算法在多故障诊断中的应用

Application of Class Separation Distance Node Optimization DDAG-SVM in Multi-fault Diagnosis

针对决策导向无环图支持向量机(DDAG-SVM)方法根节点的选择会影响分类结果的不同及影响故障诊断的准确性的问题,文中将DDAG-SVM多分类方法中的节点进行优化,得到了一种通过计算类间距确定分类树根节点的改进算法.实验结果表明:类间距节点优化的DDAG-SVM方法较传统DDAG-SVM分类方法准确率提高了4%,且分类效率提高了26.1%.

The study aims to solve the problem that the DDAG multi-classification method affects different classification results and the accuracy of fault diagnosis .The node of DDAG-SVM classification method was optimized ,with an improved algorithm obtained for determining classification tree root node by calculating the distance between classes .Experiment results show that ,compared with the traditional DDAG-SVM classification method , the DDAG-SVM method for node optimization between classes improves the accuracy by 4% and the classification efficiency by 26 .1% .