基于偏最小二乘法回归的工序质量建模.pdfVIP

第35 卷第5 期 东南大学学报（自然科学版 ） Vol. 35 No. 5 2005 年9 月 JOURNAL OF SOUTHEAST UNIVERSITY （Natural Science Edition ） Sept. 2005 基于偏最小二乘法回归的工序质量建模 张赤斌 史金飞 易 红 （东南大学机械工程系，南京210096 ） 摘要：针对制造工序质量控制问题，应用多元统计分析中的偏最小二乘回归法建立了质量模型. 利用该模型可以定量分析加工工序与最终成品率之间的关系，进而通过将大量的工序影响因子 约简得到主要影响因子子集. 根据在线生产的相关质量数据，采用非线性迭代偏最小二乘法获得 影响因子的权重. 得到偏最小二乘因子权重可以在线预测成品质量变化，避免离线测试. 在半导 体制造实例研究中，以工序质量水平为自变量，成品质量水平为因变量，建立了质量水平传递模 型，应用该方法可实现多工序质量异常的在线诊断和预测，为质量控制提供了定量依据. 关键词：偏最小二乘回归；质量模型；质量控制 中图分类号：TH166 文献标识码：A 文章编号：100 1 - 0505 （2005 ）05-0702-04 Manufacture process guality modeling based on partial least sguare regression Zhang Chibin Shi Jinfei Yi Hong （Department of Mechanical Engineering ，Southeast University ，Nanjing 210096 ，China ） Abstract ：Partial least sguare （PLS ）regression ，a multivariate statistical approach ，is applied to the manufacture process guality control ，so that guality modeling can be established. The guality model- ing can be used to analyze the relationship between finished product ratio and individual manufacture process ，and reduce the large set of process guality variables to a much smaller set of PLS compo- nents. The PLS component scores can be freguently computed by the nonlinear iterative partial least sguare method from the on-line process data. With PLS component scores the guality variables can be predicted on-line before conducting off-line tests. In terms of the semiconductor manufacturing re- search ，guality level transfer model was established on the basis of taking process guality levels as in- dependent variables and finished guality