欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在基于支持向量数据描述（suppoflvectordomaindescription，SVDD）的模拟电路故障诊断中，故障样本易陷入多个球体的交叉区域产生误诊。为了改进标准SVDD松弛的球体描述边界以提高故障诊断性能，提出一种基于图谱空间映射SVDD（graphspectrummappingSVDD，GSM—SVDD）的模拟电路故障诊断新方法。采用高斯核函数构造Laplace矩阵，然后进行特征值分解，由特征值对应的Laplace特征向量描述SVDD球体的边界，最后采用SVDD的最小相对距离法则诊断故障样本。实验结果表明，通过Laplace谱映射改变原始特征样本的空间分布，GSM—SVDD方法能有效提高模拟电路的故障诊断性能。
英文摘要：
For analog circuits fault diagnosis based on support vector domain description （ SVDD ）, as faulty samples usually distribute in the overlap region of several hyperspheres, they are misdiagnosed eas- ily. In order to overcome the slack description boundary of SVDD classifier and improve the diagnosis performance of SVDD, an improved method of GSM _ SVDD is proposed to fault diagnosis of analog cir- cuits in this paper. Firstly, the Gauss kernel function was used to construct the Laplace matrix. After im- plementing eigenvalue decomposition, the Laplace eigenvectors were applied to describe the hyperspheres boundary of SVDD classifier. Finally, the fault samples were diagnosed by the smallest relative distance rule of SVDD. The results show that cuits as the space distribution change the new method improves the diagnosis performance for analog cir- of original feature samples by Laplace spectrum mapping.
位置：
同期刊论文项目
