热镀锌带钢表面粗糙度的多变量统计过程质量监控  

Application of multivariable statistical process control based on PLS to surface roughness quality monitoring

作　　者：姚林[1] 黎敏[1] 阳建宏[1] 徐金梧[1] 

机构地区：[1]北京科技大学机械工程学院

出　　处：《冶金自动化》2007年第4期23-27,49,共6页Metallurgical Industry Automation

摘　　要：以热镀锌带钢的表面粗糙度为研究对象,建立了基于偏最小二乘法(PLS)的多变量统计过程控制模型。该模型首先利用偏最小二乘回归对生产数据进行信息提取,然后根据T2统计量对提取出的信息进行过程监控,最后利用贡献图分析生产过程中出现异常的原因。新的方法能够有效克服多变量间的相关性,同时充分保留过程参数对质量参数最大的解释力。用鞍山钢铁集团新轧钢股份有限公司冷轧厂带钢热镀锌线的生产运行实绩进行验证,结果表明:基于PLS的多变量统计过程控制模型与基于主元分析(PCA)的统计过程控制模型相比,能够更准确地找出生产过程中的异常点及其原因。Based on partial least square method, a muhivariable statistical process control model of quality monitoring for hot-dip galvanizing strip surface roughness is proposed. To overcome correlation among process variables, partial least square regression was employed to obtain optimal components which carry the most significant information of all the production data set. Then the Hotelling T2 statistic of the components was calculated for quality monitoring, and causes of faults in production process can be tracked by analyzing the c0mponents＇contribution plots. Being different from the process control model based on PCA （Principle Component Analysis）, the new model maintains the maximum explanation of process variables to quality variables when extracting comprehensive components from production data set. Hot-dip galvanizing strip surface roughness production data from Anshan Iron and Steel Group Corporation are used for model validation, and the proposed model has more satisfactory performance in fault diagnosis of production process, compared with the PCA process control model.

关 键 词：多变量统计过程控制 偏最小二乘法 热镀锌带钢 表面粗糙度 质量监控 

分 类 号：TP277[自动化与计算机技术—检测技术与自动化装置] TG335.5[自动化与计算机技术—控制科学与工程]