论文部分内容阅读
将二维相关谱与化学计量学结合实现掺杂牛奶与纯牛奶的判别分析。配制40个掺杂尿素牛奶、40个掺杂三聚氰胺牛奶和80个纯牛奶样品,在室温下采集各样品的一维红外光谱。以掺杂物浓度为外扰,分别构建了900~1 700 cm-1和900~1 200 cm-1vs.1 200~1 700 cm-1同步红外二维相关谱。在此基础上,分别建立了基于两个区间相关谱掺杂牛奶的多维偏最小二乘判别模型,并与传统一维的偏最小二乘判别模型的分析结果进行对比分析。结果表明:基于900~1 200 cm-1vs.1 200~1 700 cm-1二维相关谱模型优于900~1 700 cm-1二维相关谱模型,相关谱的多维偏最小二乘判别模型优于传统一维谱的偏最小二乘判别模型。
Discriminant Analysis of Doped Milk and Pure Milk by Combining Two - Dimensional Correlation Spectroscopy and Chemometrics. 40 doping urea milk, 40 doping melamine milk and 80 pure milk samples were prepared. One-dimensional infrared spectra of each sample were collected at room temperature. Using the dopant concentration as external disturbances, two-dimensional synchronous infrared spectra of 900 ~ 1 700 cm-1 and 900 ~ 1 200 cm-1vs.1 200 ~ 1 700 cm-1 were constructed respectively. On this basis, a multidimensional partial least squares discriminant model based on two interval correlation spectrum doping milk was established and compared with the traditional one-dimensional partial least squares discriminant analysis results. The results show that based on the two-dimensional correlation spectrum model of 900 ~ 1 200 cm-1vs.1 200 ~ 1 700 cm-1, the multi-dimensional partial least squares discriminant model of correlation spectrum is superior to the two-dimensional correlation spectrum model of 900 ~ 1 700 cm- Which is better than the traditional one-dimensional partial least squares discriminant model.