近红外光谱结合偏最小二乘回归法测定白酒酒精度  

Determination of Alcohol Content in Spirits by Near Infrared Spectroscopy Combined with Partial Least Square Regression Method

作　　者：廖昱博[1] 刘小清[1] 陈娟[1] 廖靓 

机构地区：[1]赣南师范学院物理与电子信息学院,江西赣州341000 [2]韶关市质量计量监督检测所,广东韶关512026

出　　处：《赣南师范学院学报》2013年第3期7-9,共3页Journal of Gannan Teachers' College(Social Science(2))

基　　金：江西省自然科学基金项目(2010GQW0006);江西省教育厅科技项目(GJJ11584)

摘　　要：酒精度是白酒的主要理化指标.传统的酒精度检测方法繁琐、费时.采取近红外光谱技术结合偏最小二乘回归法建模,通过相关光谱的方法选择最佳建模光谱区域(955 nm-1 047 nm和1 108 nm-1 347 nm),得出线性回归方程.结果显示,主成分个数为3时,模拟训练集的线性相关系数达0.980 2,白酒校验集的相关系数为0.957 4,总体方均根误差为1.800.该方法方便快捷、高效准确、廉价无损,有望成为检测白酒酒精度的新方法.Alcohol content is the main physical and chemical index of spirits.The traditional methods for detection of alcohol content are tedious and time-consuming.In this paper,near infrared spectroscopy was adopted in combination with the partial least square regression method for modeling,and the correlation coefficient method was used for selecting the optimal spectral regions（i.e.955 1047 nm and 1108-1347 nm）.The results show that the number of principal components is 3,the correlation coefficient of the training set is 0.980 2,the r-value and the root-mean-square-error of the validation set are found to be 0.957 4 and 1.80°,respectively.This technique is rapid,convenient,efficient,accurate,cheap and nondestructive,and can be a potential method for determination of alcohol content in spirits.

关 键 词：近红外光谱 偏最小二乘回归 酒精度 

分 类 号：O433.1[机械工程—光学工程]