论文部分内容阅读
采用傅里叶近红外光谱结合偏最小二乘法(PLS)法建立了测定婴儿配方奶粉中的总脂肪酸、饱和脂肪酸和不饱和脂肪酸含量的近红外数学模型,并通过交互验证和外部检验两种方式考察了近红外数学模型的可靠性。通过选择不同的波长范围,采用平滑、矢量归一化、一阶求导、二阶求导和散射校正对近红外光谱进行处理,总脂肪酸、饱和脂肪酸和不饱和脂肪酸的校正模型相关系数(R2)分别为0.9337、0.9374、0.9020,RPD分别为3.63、3.65、2.90。结果表明近红外数学模型具有良好的预测性能。采用建立的模型对验证集中的20个婴儿配方奶粉样品进行预测,总脂肪酸含量、饱和脂肪酸和不饱和脂肪酸的预测值与化学测定值之间经配对t检验分析,与常规化学方法得到的检验结果无显著差异。
A near-infrared mathematical model for the determination of total fatty acids, saturated fatty acids and unsaturated fatty acids in infant formula milk powder was established by Fourier Near Infrared Spectroscopy (PLS) and Partial Least Squares (PLS) method. Through interactive verification and external inspection, The reliability of near-infrared mathematical model was investigated. By choosing different wavelength ranges, the near-infrared spectra were processed by smoothing, vector normalization, first-order derivation, second-order derivation and scatter correction. Correlation coefficients of total fatty acids, saturated fatty acids and unsaturated fatty acids ) Were 0.9337,0.9374,0.9020, RPD were 3.63,3.65,2.90. The results show that the near-infrared mathematical model has good predictive performance. The established model was used to predict the samples of 20 infant formula in the validation set. The total fatty acid content, the predicted value of the saturated fatty acid and the unsaturated fatty acid were analyzed by paired t-test and the results of the conventional chemical method No significant difference.