本文已被：浏览 667次 下载 423次
投稿时间：2022-03-14
投稿时间：2022-03-14
中文摘要: 为探究白酒感官品评与风味成分间的密切关系，辅助从业人员高效完成酒体设计、品质鉴别等工作，提出基于模型融合的白酒感官与风味成分质量关系模型。该模型以成品酒液的感官品评数据与气相色谱-质谱仪测定的风味成分含量数据作为基础，结合AdaBoost、GBDT、Bagging、Stacking 模型融合方法建立质量关系模型，实现通过感官品评数据对酒液中酸、酯、醇、羟基与呋喃化合物含量值的准确预测。结果表明，酸、酯、醇、羟基与呋喃化合物对应质量关系模型的拟合优度分别为0.974 3、0.961 1、0.868 0、0.908 7，相较于传统的机器学习方法建立的质量关系模型具有明显的优势，且稳定性更好。
Abstract:To explore the close relationship between sensory evaluation and flavor components of Baijiu and assist industry professionals in efficiently completing the tasks，such as liquor body design and quality identification，a quality relationship model for Baijiu sensory evaluation and flavor components based on model fusion is proposed.Based on sensory evaluation data of the finished wine and flavor component content measured by gas chromatography-mass spectrometry，this model used AdaBoost，GBDT，Bagging，and Stacking model fusion methods to establish a quality relationship model. The model accurately predicted the values of acid，ester，alcohol，hydroxyl，and furan compounds in the wine through the sensory evaluation data. After testing，the goodness of fit of the mass relationship models corresponding to acids，esters，alcohols，hydroxyls，and furan compounds were 0.974 3，0.961 1，0.868 0，and 0.908 7，respectively. Compared with the mass relationship models established by traditional machine learning methods，the present model had clear advantages and better stability.
keywords: sensory evaluation of Baijiu flavor components model fusion machine-learning mass relationship model
文章编号：202313009 中图分类号： 文献标志码：
基金项目:四川省科技计划项目（2021YFS0339）；四川轻化工大学产学研合项目（CXY2020ZR006）
| 作者 | 单位 |
| 刘鑫1，2，韩强1，2*，李陈杰1，2，张良3，李锦松3，张怀山3，庹先国1，2* | 1.人工智能四川省重点实验室，四川宜宾 644000；2.四川轻化工大学自动化与信息工程学院，四川宜宾 644000；3.泸州老窖集团有限责任公司，四川泸州 646000 |
引用文本：
津ICP备20000603号-1