牛乳清粉掺伪羊乳粉的近红外光谱法快速无损检测
发布时间：2018-10-23 08:41
【摘要】：为快速无损检测掺伪羊乳粉中牛乳清粉的含量,采用近红外光谱法(NIR)结合ν-支持向量回归(ν-SVR)检测197个掺伪牛乳清粉的羊乳粉,并与偏最小二乘法(PLS)对比,光谱经平滑、标准变量变换及导数预处理。随机选取132个样本作为校正集建立模型,其余65个样本作为测试集,评估模型性能。结果显示:校正集最优模型为标准变量变换、Savitzky-Golay平滑和二阶导预处理ν-SVR模型,其交叉验证误差均方根(RMSECV)为0.586,交叉验证相关系数(RCV)达到0.9947。预测集验证结果ν-SVR法比PLS法更优,ν-SVR模型预测值与真实值R达到0.9958,RMSEP为0.526。试验结果表明NIR结合ν-SVR模型可用于掺伪羊乳粉中牛乳清粉的快速无损检测,而且操作简便,可为实际应用提供参考。
[Abstract]:In order to detect the content of bovine whey powder in adulterated sheep milk powder quickly, 197 sheep whey powder adulterated with false bovine whey powder were detected by near infrared spectroscopy (NIR) combined with 谓 -support vector regression (SVR), and compared with partial least square method (PLS), the spectrum was smooth. Standard variable transformation and derivative preprocessing. 132 samples were selected randomly as calibration set to establish the model, and the other 65 samples were selected as test sets to evaluate the performance of the model. The results show that the optimal model of calibration set is standard variable transformation, Savitzky-Golay smoothing and second-order derivative preprocessing v-SVR model. The root-mean-square (RMSECV) of cross-validation error is 0.586, and the correlation coefficient (RCV) of cross-validation is 0.9947. The prediction set verifies that the SVR method is better than the PLS method, and the predicted value and the real value R of the 谓 SVR model are 0.9958 and 0.526 respectively. The experimental results show that NIR combined with v-SVR model can be used for fast nondestructive detection of bovine whey powder in adulterated sheep milk powder, and it is easy to operate and can be used as a reference for practical application.
【作者单位】： 北京食品营养与人类健康高精尖创新中心/食品添加剂与配料北京高校工程研究中心北京工商大学;吉林省食品检验所;
【分类号】：O657.33;TS252.7
，
本文编号：2288727
[Abstract]:In order to detect the content of bovine whey powder in adulterated sheep milk powder quickly, 197 sheep whey powder adulterated with false bovine whey powder were detected by near infrared spectroscopy (NIR) combined with 谓 -support vector regression (SVR), and compared with partial least square method (PLS), the spectrum was smooth. Standard variable transformation and derivative preprocessing. 132 samples were selected randomly as calibration set to establish the model, and the other 65 samples were selected as test sets to evaluate the performance of the model. The results show that the optimal model of calibration set is standard variable transformation, Savitzky-Golay smoothing and second-order derivative preprocessing v-SVR model. The root-mean-square (RMSECV) of cross-validation error is 0.586, and the correlation coefficient (RCV) of cross-validation is 0.9947. The prediction set verifies that the SVR method is better than the PLS method, and the predicted value and the real value R of the 谓 SVR model are 0.9958 and 0.526 respectively. The experimental results show that NIR combined with v-SVR model can be used for fast nondestructive detection of bovine whey powder in adulterated sheep milk powder, and it is easy to operate and can be used as a reference for practical application.
【作者单位】： 北京食品营养与人类健康高精尖创新中心/食品添加剂与配料北京高校工程研究中心北京工商大学;吉林省食品检验所;
【分类号】：O657.33;TS252.7
，
本文编号：2288727
本文链接：https://www.wllwen.com/kejilunwen/huaxue/2288727.html
教材专著
