电子舌预测不同体积分数牛奶的表观黏度

该文为建立牛奶的电子舌响应信号与其表观黏度的关系，在单因素方差分析和主成分分析的基础上，提出了比较多元线性回归、逐步多元线性回归和偏最小二乘回归3种模型对牛奶表观黏度的预测效果的方法。结果显示，单因素方差分析表明体积分数对牛奶的表观黏度和各个传感器响应信号都具有极显著性的影响；主成分分析（PCA）可以用来区分牛奶的5种不同体积分数；偏最小二乘回归模型预测效果最好，模型预测值与实际值的相关系数R达到0.9659，平均相对误差（MRE）和预测均方根误差（RMSEP）分别为4.5499%和8.4645×10-5，建模最佳主成分数为3。研究结果表明，偏最小二乘回归模型是电子舌预测牛奶表观黏度的有效方法，该方法为牛奶表观黏度的科学研究提供参考。

Prediction of apparent viscosity of milk with different

In order to establish the relationship between response signals of electronic tongue and apparent viscosity of milk, the method of comparing with the effect of MLR, SMLR and PLS model was presented based on one-way analysis of variance (One-Way ANOVA) and principal component analysis (PCA). The results of one-way analysis of variance (One-Way ANOVA) showed that volume fraction had significant effect on apparent viscosities and sensors signal of pure milk; it was also found that principal component analysis (PCA) could be used to distinguish five different volume fractions of pure milk; the results of MLR, SMLR and PLS model showed that the best forecasting effect was the PLS model. Its R was 0.9659, its mean relative error (MRE) was 4.5499%, its RMSEP was 8.4645×10-5, and its principal element was 3. It had been indicated that PLS was an efficient method to predict apparent viscosity of milk with electronic tongue. This method will provide reference for scientific research of apparent viscosity of milk.