论文部分内容阅读
在传统支持向量机(C-SVM)的基础上,通过集成模糊聚类技术和支持向量机算法,构造了一种适合于故障诊断的多级二叉树分类器,并首次应用于水轮机调速系统故障诊断,取得了良好效果。该方法首先利用模糊聚类技术求取每类样本聚类中心,再对各聚类中心逐次二分,从而确定了一棵二叉树,然后在二叉树的每个节点处,根据样本聚类中心把相应样本分成两类,构造出SVM 子分类器。实验结果表明,对于k 类别故障诊断问题,只需构造k-1 个SVM 子分类器,简化了分类器结构,避免了不可区分区域的出现,且节省了内存开销,故障诊断正确率高。
Based on traditional support vector machine (C-SVM), a kind of multi-level binary tree classifier suitable for fault diagnosis is constructed by integrating fuzzy clustering and support vector machine algorithm, and is firstly applied to fault diagnosis of turbine governor system Diagnosis, achieved good results. Firstly, the fuzzy clustering technique is used to find the cluster centers of each type of samples, and then the cluster centers are sequentially divided into two points to determine a binary tree. Then, at each node of the binary tree, Divided into two categories, to construct SVM sub-classifier. The experimental results show that for k-class fault diagnosis, only k-1 SVM sub-classifiers are constructed, which simplifies the classifier structure, avoids the emergence of indistinguishable regions and saves the memory overhead, and the fault diagnosis has a high correct rate.