卤汁中氯化钠添加量的电子舌预测方法研究  

Research on Predicting the Amount of Sodium Chloride Added in Brine Solution Using Electronic Tongue

作　　者：韩方凯 张东京 段腾飞 张晓瑞 张兴桃 HAN Fangkai;ZHANG Dongjing;DUAN Tengfei;ZHANG Xiaorui;ZHANG Xingtao(School of Biological and Food Engineering,Suzhou University,Suzhou 234000,China;School of Food and Biological Engineering,Jiangsu University,Zhenjiang 212013,China)

机构地区：[1]宿州学院生物与食品工程学院,安徽宿州234000 [2]江苏大学生物与食品工程学院,江苏镇江212013

出　　处：《宿州学院学报》2024年第6期27-32,共6页Journal of Suzhou University

基　　金：安徽省高校优秀青年人才支持计划项目(gxyq2022105);安徽省高等学校教育教学改革研究重点项目(2022jyxm1592);宿州学院科研平台(2021XJPT);宿州学院第四批学术技术带头人及后备人选、优秀学术技术骨干项目(2020XJHB04);宿州学院校企合作实践教育基地建设项目(szxy2022xqhz02)。

摘　　要：氯化钠是食盐的主要成分,开发一种客观、准确且快速的卤汁中氯化钠添加量预测方法对卤汁咸度的智能调控具有重要的意义。以符离集烧鸡生产用卤汁为对象,将氯化钠按梯度浓度添加至卤汁基质溶液中,制备电子舌供试液,采集脉冲伏安型味觉传感器信号。以竞争自适应重加权采样法(Competitive adaptive reweighted sampling,CARS)筛选关键特征变量,形成自变量数据集,对比采用偏最小二乘(Partial least squares,PLS)、支持向量机(Support vector machine,SVM)和极限学习机(Extreme learning machine,ELM)构建检测模型。结果显示,CARS法成功筛选出93个关键特征变量,电极重要性排序为W电极>Pd电极>Au电极>Ti电极>Pt电极;所建PLS、SVM和ELM模型测试集相关系数分别为0.969、0.994和0.9996,预测均方根误差分别为0.0356 g/L、0.0898 g/L和1.67×10-5 g/L,其中ELM模型表现最优。得出,基于裸惰性金属电极的脉冲伏安型电子舌结合ELM算法可用于卤汁中氯化钠添加量的快速预测。Sodium chloride serves as the primary ingredient of salt.It is of great significance to develop an objective,accurate and rapid prediction method of sodium chloride addition in brine for intelligent control of brine salinity.Taking the brine used in the production of Fuliji roast chicken as the object,sodium chloride was added to the brine matrix solution according to the gradient concentration to prepare the electronic tongue test solution.Signals from the pulse voltammetric taste sensor are then collected for analysis.Competitive Adaptive Reweighted Sampling(CARS)method was used to screen key feature variables to form independent variable datasets.This data set is then used to construct and compare detection models employing techniques such as Partial Least Squares(PLS),Support Vector Machine(SVM)and Extreme Learning Machine(ELM).The results demonstrate that the CRAS method successfully identified 93 key feature variables.The electrodes are ranked in order of importance as follows:Tungsten(W)electrode>Palladium(Pd)electrode>Gold(Au)electrode>Titanium(Ti)electrode>Platinum(Pt)electrode.The constructed PLS,SVM and ELM models yielded test set correlation coefficients of 0.969,0.994,and 0.9996 respectively.The root mean square errors in prediction were 0.0356 g/L,0.0898 g/L,and 1.67×10-5 g/L respectively,with the ELM model demonstrating the best performance.It can be concluded that the combination of a pulse voltammetric electronic tongue,based on bare inert metal electrodes,and the ELM regression algorithm can be effectively utilized for rapid prediction of sodium chloride addition in brine.

关 键 词：卤汁 氯化钠 咸度调节 电子舌 化学计量学 

分 类 号：TS251.1[轻工技术与工程—农产品加工及贮藏工程]