基于物联感知的卷烟工艺数据融合与统计分析研究  

Research on Fusion and Statistical Analysis of Cigarette Process Data Based on IoT Perception

作　　者：蒋一翔[1] 张青松[1] 高雪峰[1] 徐琦[1] JIANG Yixiang;ZHANG Qingsong;GAO Xuefeng;XU Qi(China Tobacco Zhejiang Industrial Co.,Ltd.,Ningbo 315504,China)

机构地区：[1]浙江中烟工业有限责任公司,浙江杭州315504

出　　处：《自动化仪表》2020年第3期107-110,共4页Process Automation Instrumentation

摘　　要：卷烟工艺数据是一组包含烟支质量、圆周、通风度、烟支吸阻等综合反映卷烟的物理质量和烟支工艺质量的物测数据。烟支吸阻是消费者对于烟支感官的一项重要指标。烟支吸阻的稳定性会对整体烟支的理化指标稳定性产生很大影响;同时,烟支吸阻也受多种因素影响。针对烟支吸阻形成因素多变、稳定性控制多样的问题,从卷烟工艺数据角度出发,确定工艺数据中数据变化规律及数据特征间相互影响的强弱关系;运用斯皮尔曼系数分析数据特征间的线性相关程度,通过随机森林方法分析数据特征之间的非线性相关程度,确认了影响烟支吸阻的关键工艺参数为单支圆周值、质量偏差平均值和内部目标质量。该研究为公司后续建设信息物理系统(CPS)的进一步数据采集规划提供参考,为公司后期构建的智能化控制模型提供参数输入。Cigarette process data is a set of physical measurement data including cigarette weight,circumference,ventilation,and cigarette draw resistance that comprehensively reflect the physical and process quality of the cigarette.Draw resistance is an important sensory indicator ofcigarettes.The stability of cigarette draw resistance will have a great impact on the stability of the overall physical and chemical index of cigarettes.At the same time,draw resistance is also affected by many factors.Aiming at the variable factors and difficult stability control of cigarette draw resistance,the variation law of the cigarette process data and the relationship between data features were determined,the linear correlation between data features was analyzed by Spearman coefficients and the non-linear correlation by Random Forest.It was confirmed that the circumference value,mean weight deviation and internal target weight were key process parameters that affecting draw resistance.The research can be used as a reference for the company s follow-up construction of cyber-physical system(CPS)for further data acquisition and planning,and can provide parameter input for the later construction of the company's intelligent control model.

关 键 词：卷烟工艺数据 斯皮尔曼系数 随机森林 信息物理系统(CPS) 智能化控制 相关性分析 

分 类 号：TH162.1[机械工程—机械制造及自动化]