扩散映射K近邻在工业过程故障检测中的应用  

K-Nearest-Neighbor rule based on diffusion maps used in semiconductor industry process fault detection

作　　者：李元[1] 刘亚东[1] 张成[1,2] 

机构地区：[1]沈阳化工大学技术过程故障诊断与安全性研究中心,辽宁沈阳110142 [2]东北大学信息科学与技术学院,辽宁沈阳110819

出　　处：《辽宁工程技术大学学报（自然科学版）》2015年第12期1370-1376,共7页Journal of Liaoning Technical University (Natural Science)

基　　金：国家自然科学基金项目(60774070;61174119);国家自然科学基金重点课题基金项目(61034006)

摘　　要：针对半导体工业过程多工序、变量非线性、非高斯分布等特征,提出一种基于扩散映射的K近邻(DMKNN)故障检测方法.充分利用扩散映射(DM)降维,提取低维流行特性,保留数据集内在非线性结构特性,应用改进的KNN故障诊断方法在低维流行特征空间进行检测.研究结果表明:与传统K近邻技术的统计方法相比,DMKNN的故障检测率高于其他算法,提升了对数据样本关联性信息的有效提取能力,保持了K近邻处理非线性、多模态检测问题的性能,验证了该方法的有效性.In the semiconductor industry process, because of the features of multi-mode, nonlinear and non-Gaussian, this paper proposed a fault detection method diffusion maps based K-Nearest-Neighbor （DMKNN） used in batch process is proposed. First, the proposed method takes full advantage of the dimensionality reduction and information preserving properties of to extract the low dimensional manifold feature that optimally preserves the intrinsic nonlinear structure of the data set. Then the method of fault detection based the adapted KNN is applied to the low dimensional manifold feature space to detect potential faults. The research results of an industrial semiconductor example show that the fault detection rate of DMKNN is higher than that of the traditional statistical algorithms based KNN, and it improves the ability of fetching relevance information in data samples, keeps the performance of KNN rule to deal with nonlinear and multi-mode characteristics, and verifies the effectiveness of the proposed method.

关 键 词：扩散映射 K近邻 故障检测 低维流行特性 半导体工业过程 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置]