PLS-ATR-FTIR用于牛奶主要成分的快速测定.doc

  PLS-ATR-FTIR 用于牛奶主要成分的快速 测定 殷秀秀1,3，穆海波2，艾连中2，顾小红1** 5 10 15 20 25 30 （1. 江南大学食品科学与技术国家重点实验室，江苏 无锡 214122； 2. 光明乳业股份有限公司乳业生物技术国家重点实验室，上海 200436； 3. 江南大学食品学院，江苏 无锡 214122） 摘要：利用衰减全反射傅里叶红外光谱技术(ATR-FTIR)结合偏最小二乘法(PLS)实现牛奶成 分的快速测定。考察了光谱预处理、主成分数对模型的影响，分别建立了牛奶中蛋白质、脂 肪和乳糖的定量模型，蛋白质、脂肪和乳糖模型的校正集均方根误差(RMSEC) 分别为 0.030、 0.136、0.114，验证集均方根误差(RMSEP)分别为 0.043、0.178、0.163。利用模型对预测集 样品进行定量分析，预测值和实测值的相关系数(R2)分别为 0.932、0.941、0.989。说明利用 PLS-ATR-FTIR 对牛奶成分进行快速测定具有可行性。 关键词：中红外；牛奶；定量模型；偏最小二乘法 中图分类号：O 657.33 Rapid Analysis of Main Contents in Milk Based on PLS-ATR-FTIR YIN Xiuxiu1,3, MU Haibo2, AI Lianzhong2, GU Xiaohong1 (1. State Key Laboratory of Food Science Technology, Jiangnan University, JiangSu WuXi 214122; 2. State Key Laboratory of Dairy Biotechnology, Bright Dairy Food Co.,Ltd, ShangHai 200436; 3. School of Food Science and Technology,Jiangnan University, JiangSu WuXi 214122) Abstract: Attenuated total reflectance fourier transform infrared spectroscopy (ATR-FTIR) combined with partial least squares (PLS) was employed to the rapid analysis of main contents in milk. The protein, fat and lactose quantitative models were established separately after preprocessing spectrum and selecting principal component scores (PCs).The root-mean-square-error of calibration (RMSEC) of the protein, fat and lactose quantitative models were 0.030、0.136、0.114, respectively, and the root-mean-square-error of prediction (RMSEP) were 0.043、0.178 、0.163, respectively. The quantitative models were validated for independent test samples, the correlation coefficients (R2) between the predicted values and actual values were 0.932、0.941、0.989, respectively. This indicated that PLS-ATR-FTIR would be a valid method for rapid analysis of main contents in milk. Keywords: mid-infrared spectroscopy (MIR); milk; quantitative model; PLS 35 0 引言 牛奶是一类营养全面的理想食品，含有较丰富的蛋白质、脂肪和乳糖。这三个成分的含 量是牛奶品质检验的重要指标，同时也是监测奶牛健康状况的参考值[1]，因此对其进行准确 的测定具有重要意义。化学分析法虽然是测定牛奶成分的主要方法，但存在耗时费力、操作 40 繁琐及污染环境等不足，寻找快速准确的测定方法是当前面临的一个难题[