基于核相似度支持向量数据描述的间歇过程监测

发布时间：2018-06-12 00:35

本文选题：核相似度 + 支持向量数据描述　；参考：《化工学报》2017年09期

【摘要】：基于支持向量数据描述的间歇过程监测方法选择历史过程数据中最大的核距离作为控制限,忽略了高维空间中超球体的不规则性,导致基于该方法的过程监测精度不高。针对上述问题,提出了一种基于核相似度支持向量数据描述的间歇过程监测方法,将间歇过程数据待监测样本与支持向量之间的核函数值作为相似度权重,利用该相似度对不同时刻的支持向量球心距加权求和,得到待监测间歇过程数据样本的动态控制限,通过判断待监测样本的球心距是否超过其动态控制限,实现间歇过程监测。所提方法综合考虑了超球体的不规则性和过程数据在高维空间分布的局部特性,以及间歇过程数据待监测样本的时变性,提高了间歇过程监测的准确性。利用数值仿真实验和半导体金属刻蚀实验验证了该方法的有效性。
[Abstract]:The intermittent process monitoring method based on support vector data description selects the maximum kernel distance in the historical process data as the control limit, neglecting the irregularity of hypersphere in the high-dimensional space, which leads to the low accuracy of the process monitoring based on this method. To solve the above problems, a method of batch process monitoring based on the description of kernel similarity support vector data is proposed. The kernel function value between the sample and support vector of batch process data to be monitored is taken as the similarity weight. By weighted summation of support vector spherical center distance at different times, the dynamic control limit of batch process data sample to be monitored is obtained. By judging whether the spherical center distance of the monitored sample exceeds its dynamic control limit, batch process monitoring is realized. The proposed method takes into account the irregularity of hypersphere and the local characteristics of process data distribution in high-dimensional space, as well as the time-varying of batch process data to be monitored, which improves the accuracy of batch process monitoring. The effectiveness of the method is verified by numerical simulation and semiconductor metal etching experiments.
【作者单位】：北京化工大学信息科学与技术学院;
【基金】：国家自然科学基金项目(61240047) 北京市自然科学基金项目(4152041)~~
【分类号】：TB497

【相似文献】