改善卷烟烟气味觉的模型方法研究
VIP免费
摘 要
卷烟由诞生之时的奢侈品转化为大众消费品,消费量激增,成为普通大众生
活中随处可见的物品,同时带来了卷烟产业的发展和行业内竞争的加剧。消费者
对卷烟的喜好主要是通过卷烟的口感,也就是卷烟烟气吸入口腔后的所产生的味
觉感受。因此研究卷烟烟气味觉特征及其调控技术,进而提出改善卷烟烟气味觉
感受的解决方案,成为构筑中式卷烟核心技术、形成卷烟产品特色的重要研究方
向。
目前卷烟感官味觉质量的评定主要依靠人工感官评吸,感官评吸是通过评吸
人员的味觉感受而引起的反应并感知到烟草或卷烟中其它物质的特征或性质的一
种方法。由于受到多种主观、客观因素的影响,人工感官评吸评定结果的准确性
往往难以保证。
本文是国家烟草专卖局项目“卷烟烟气味觉特征及其调控技术研究”(项目编
号:110200901001)的重要组成部分。本文从分析卷烟烟气中各种化学成分含量
的角度来研究卷烟烟气的味觉特征。首先采用无信息变量消除法(UVE)和连续
投影算法(SPA)联合算法对影响卷烟酸、甜、苦、咸四种基本味觉的化学成分变
量分别进行选取,基于偏最小二乘法(PLS)建立了由卷烟烟气化学成分含量对卷
烟烟气酸、甜、苦、咸四种基本味觉值的预测模型。然后利用卷烟烟气的电子舌
味觉浓度数据和人工感官味觉数据建立 BP 神经网络模型,用于对卷烟烟气味觉特
征感官的评定。最后,利用 PLS 模型中变量的系数和变量之间的相关性关系,建
立卷烟烟气中各个成分变量的含量对卷烟味觉特征的调控方法与模型理论;利用
神经网络的平均影响值(MIV),建立卷烟烟气味觉浓度值对卷烟烟气感官味觉的
调控理论。
基于以上理论与模型研究结果,利用 Matlab GUI 工具箱开发了操作简单、易
于使用、具有友好交互界面的卷烟烟气感官味觉评定与调控系统雏形。该系统通
过对大量卷烟样本的电子舌测量的味觉数据和人工感官味觉评吸数据的学习,获
取卷烟烟气的物理味觉值与经过人的味觉神经后给人的主观感受效果之间的关
系,从而用于其它新品牌卷烟烟气感官味觉的评定。利用神经网络的平均影响值
(MIV)对评定样本与目标值之间的差异给出合理的调控建议,从而实现了理论
研究结果的实际应用。
关键词:卷烟感官味觉特征 无信息变量消除法 连续投影算法 BP 神经
网络 Manlab GUI
ABSTRACT
The conversion from luxury goods when it comes into being to consumer goods for
ordinary people nowadays brings cigarettes sharp increase in sales as well as rapid
development and fierce competition of this industry. The preference of consumers on
cigarettes mainly depends on the mouth feel, namely the taste of the smoke when it is
breathed into mouth. Consequently, it is important to investigate the taste characteristics
and adjusting techniques of cigarettes smoke, a step further, to propose a resolution of
improving the taste of the smoke. This has become a significant research direction to
form core technologies and particular style of Chinese cigarettes.
Currently, the judge on the taste of the smoke is mainly depended on manual
sensory assessment, which is a way to obtain the characters of other substance in
tobacco or cigarettes by the reaction of taste from the adjudicator. The results always
satisfactory due to the influence by multiple factors both subjectively and objectively.
This thesis is a part of the project “The Study on Taste Feature and Its Adjusting
Technology of Cigarette Smoke” (Grant No. 110200901001) of State Tobacco
Monopoly Administration. The taste feature of cigarette smoke is investigated from the
perspective of analyzing the contents of chemical compositions in the cigarette smoke.
Firstly, active ingredients which have impacts on cigarettes taste characteristics like
acidity, sweet, bitterness and salty taste in cigarette smoke are selected by uninformed
variable elimination method (UVE) and successive projections algorithm (SPA). Then
the sense character prediction model based on partial least squares method (PLS) is
established to predict the values of the four kind of taste above through the chemical
compositions. After that, BP neural network model is constructed from the data of taste
concentration in cigarette smoke obtained both by electronic tongue and manual taste to
assess the taste characteristics of cigarette smoke. Finally, the adjusting method and
model theory of the composition content to the taste characteristics are established by
the coefficients and correlations of variables. Moreover, theory of the adjusting effect of
taste concentration values on sense taste is formed by the mean impact value (MIV) of
neural network.
Based on the research of theory and model, a miniature system which is simple to
use and friendly designed to assess and adjust the sense taste of cigarette smoke is
developed by Matlab GUI tool box. This system can learn by itself through the data
collected both from electronic tongue and manual taste from large amount of statistics
to obtain the relation between the physical taste value and the subjective sense after
which passing through human neural network, then it can be used to judge the sense
taste of some new brand of cigarettes. The practical uses of theoretical results are
realized the adjusting advices given with the deviation between the sample value and
object value by the MIV of neural network.
Key Word :Sensory taste feature of cigarette, uninformed variable
elimination, successive projections algorithm, BP neural network,
Matlab GUI
目 录
中文摘要
ABSTRACT
第一章 绪 论 ..................................................... 1
§1.1 课题的来源 ............................................... 1
§1.2 研究背景 ................................................. 1
§1.3 国内外相关研究现状 ....................................... 2
§1.3.1 关于卷烟感官质量的研究 ............................... 2
§1.3.2 人工智能味觉系统的应用研究 ........................... 3
§1.3.3 卷烟烟气化学成分的研究 ............................... 4
§1.4 本文的研究思路及创新点 ................................... 4
§1.4.1 论文的研究思路 ....................................... 4
§1.4.2 论文的创新点 ......................................... 5
§1.4.3 论文的研究意义 ....................................... 5
§1.4.4 论文的研究框架 ....................................... 5
§1.5 本章小结 ................................................. 6
第二章 影响卷烟烟气味觉特征的因素分析 ............................ 7
§2.1 卷烟的感官评吸概述 ....................................... 7
§2.1.1 感官评吸的要求和影响因素 ............................. 7
§2.1.2 评吸工作中感官的作用 ................................. 7
§2.1.3 感官评吸工作中存在的问题 ............................. 8
§2.1.4 数字化卷烟质量评价系统的可行性 ....................... 8
§2.2 统计理论与方法 ........................................... 9
§2.2.1 偏最小二乘回归算法 ................................... 9
§2.2.2 无信息变量消除法(UVE) ............................. 10
§2.2.3 连续投影算法(SPA) ................................... 11
§2.3 基于 UVE-SPA 得卷烟烟气成分对味觉特征的作用分析 .......... 12
§2.3.1 相关性选取 .......................................... 12
§2.3.2 利用无信息变量消除法(UVE)进行变量选取 ............. 14
§2.3.3 利用连续投影算法(SPA)进行变量选取 ................... 17
§2.3.4 三种变量选取方法的效果比较 .......................... 20
§2.4 本章小结 ................................................ 21
第三章 基于神经网络的卷烟烟气感官味觉评定 ....................... 22
§3.1 人工神经网络的 .......................................... 22
§3.1.1 人工神经网络的概述 .................................. 22
§3.1.2 人工神经元模型 ...................................... 23
§3.1.3 神经元网络的学习过程 ................................ 24
§3.1.4 BP 神经网络 ......................................... 25
§3.2 基于 BP 神经网络的卷烟烟气感官味觉评定 ................... 27
§3.2.1 评定模型网络结构的设计 .............................. 27
§3.2.2 评定模型激活函数的选取 .............................. 29
§3.2.3 网络模型的 Matlab 实现 ............................... 30
§3.2.4 网络仿真结果分析 .................................... 33
§3.3 本章小结 ................................................ 34
第四章 卷烟烟气味觉特征调控的研究 ............................... 35
§4.1 基于变量相关性和 PLS 系数的模型调控 ...................... 35
§4.1.1 变量相关性理论概述 .................................. 35
§4.1.2 对卷烟烟气味觉值的调控 .............................. 35
§4.2 基于 MIV 的神经网络评定模型调控 .......................... 37
§4.2.1 平均影响值 MIV 算法概述 ............................. 37
§4.2.2 卷烟烟气感官值的调控 ................................ 37
§4.3 本章小结 ................................................ 40
第五章 评定及调控系统雏形的实现 ................................. 41
§5.1 基于 Matlab GUI 系统开发环境 ............................. 41
§5.1.1 Matlab 简介 ......................................... 41
§5.1.2 Matlab GUI 概述 ..................................... 41
§5.1.3 Matlab GUI 对象层次结构 ............................... 42
§5.2 系统分析 ................................................ 43
§5.2.1 系统需求分析 ........................................ 43
§5.2.2 系统功能分析 ........................................ 43
§5.3 系统设计 ................................................ 44
§5.3.1 PLS 模型模块设计 .................................... 45
§5.3.2 BP 神经网络模块设计 ................................. 45
§5.4 系统实现及界面说明 ...................................... 46
§5.4.1 模型选择界面 ........................................ 46
§5.4.2 PLS 模型界面 ........................................ 48
§5.4.3 BP 网络模型界面 ..................................... 52
§5.5 本章小结 ................................................ 54
第六章 结论与展望 ............................................... 55
§6.1 结论 .................................................... 55
§6.2 展望 .................................................... 56
附 录 .........................................................57
参考文献 ........................................................ 59
在读期间公开发表的论文及参加的科研项目 .......................... 62
致 谢 .........................................................63
摘要:
展开>>
收起<<
摘要卷烟由诞生之时的奢侈品转化为大众消费品,消费量激增,成为普通大众生活中随处可见的物品,同时带来了卷烟产业的发展和行业内竞争的加剧。消费者对卷烟的喜好主要是通过卷烟的口感,也就是卷烟烟气吸入口腔后的所产生的味觉感受。因此研究卷烟烟气味觉特征及其调控技术,进而提出改善卷烟烟气味觉感受的解决方案,成为构筑中式卷烟核心技术、形成卷烟产品特色的重要研究方向。目前卷烟感官味觉质量的评定主要依靠人工感官评吸,感官评吸是通过评吸人员的味觉感受而引起的反应并感知到烟草或卷烟中其它物质的特征或性质的一种方法。由于受到多种主观、客观因素的影响,人工感官评吸评定结果的准确性往往难以保证。本文是国家烟草专卖局项目“...
相关推荐
-
VIP免费2025-01-09 5
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 7
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 6
-
VIP免费2025-01-09 7
-
VIP免费2025-01-09 6
作者:牛悦
分类:高等教育资料
价格:15积分
属性:67 页
大小:1.86MB
格式:PDF
时间:2024-11-19