为了解烟草行业质量数据分析利用现状以及应用需求,选取卷烟、烟叶、烟用材料等12类产品对行业重点工商企业质量检测机构开展调研.结果表明:①行业质量数据分布层级清晰,数据量大,覆盖率高,能够满足大数据分析需求;②主要采用传统描述性统计方式完成数据分析,数据分析需求集中在卷烟产品的质量状况、质量趋势和预测、实验室能力水平和资源配置情况等方面;③数据分布分散且孤立,信息化系统多且系统间协同性差是存在的主要问题.在构建行业质量大数据时要重点围绕质量数据库建设、数据标准体系制定、大数据分析技术应用、数据共享与安全以及技术人才培养等方面开展工作.该研究可为充分利用质量数据资源提升卷烟产品质量提供支持.

为快速、准确地识别叶丝、梗丝、膨胀叶丝、再造烟叶丝等烟丝类型,利用各类烟丝图像特征差异,以残差神经网络为基础构建了识别模型,并对模型的预训练权值、优化算法、学习率等超参数进行了研究,结果表明:①基于残差神经网络的识别方法可以有效识别4种类型烟丝,相比基于卷积神经网络的识别方法,模型具有更高的识别率、泛化能力与鲁棒性.②较优超参数对模型的训练速度及表现影响显著,通过训练得到的模型在测试集上的准确率及召回率均高于96％,且与训练集表现差异较小.该方法可为提高烟丝类型识别效率和准确性提供支持.

针对实验室检测能力建设,遵循普遍适用的设计理念,浅述实验室检测能力设计的必要性、原则和相关步骤,阐述了设计是保证检测结果质量的前提,并从方法选择、测量原理的理解、试验、资源和供给、检测流程设计、控制策略、实验等方面,介绍了如何通过规划、方案设计、详细设计、改进设计等阶段进行检测能力的设计.

The collaborative data including the puff number, TPM, water, nicotine, tar and CO contents of four brands of cigarette which was labeled 3, 8, 12, 15mg tar per cigarette on packs, respectively and smoked by SM400, RM20, ASM500 and RM200 smoking machines in 2004 were statistically analyzed. The results showed that there was no significant difference between the data except that the puff numbers detected by RM200 was significantly lower than that by ASM500.

The object, integral structure and main content of information network construction of quality supervising system in tobacco industry were introduced, the constitute of quality supervising information network (WAN) , the developing tendency of informationization construction of test laboratory at all levels (LAN) , and the method of data exchange between WAN and LAN were discussed.