论文部分内容阅读
利用偏最小二乘法将婴儿乳粉的近红外光谱技术(NIRS)数据与多不饱和脂肪酸含量建立校正模型,并通过交互验证和外部检验2种方式来考察模型的可靠性。选择不同的波长范围和采用平滑(Smoothing)、矢量归一化(SNV)、一阶求导(First derivative)对光谱进行处理,所得到的较优模型的结果,决定系数(R2)为0.8719,校正标准偏差(SEC)、预测标准偏差(SEP)及相对分析误差(RPD)分别为5.95、6.35和2.40。应用模型得到的预测值与化学测定值之间经配对t检验分析,P=0.457>0.05,表明2种方法得到的检验结果无显著差异。说明所建立的婴儿乳粉多不饱和脂肪酸测定的NIRS模型具有很高的预测准确性,可用于乳粉品质分析的快速检测。
The partial least squares method was used to establish a calibration model of infant milk powder NIRS data and polyunsaturated fatty acid content. The reliability of the model was examined by two methods: interactive verification and external inspection. After selecting different wavelength ranges and using the first derivatives of smoothing, vector normalization (SNV) and first derivative processing, the result of the optimal model obtained is that the determination coefficient (R2) is 0.8719, The corrected standard deviation (SEC), prediction standard deviation (SEP) and relative analytical error (RPD) were 5.95, 6.35 and 2.40, respectively. The predicted values obtained from the application model were analyzed by paired t-test between the predicted values and the chemical values, P = 0.457> 0.05, indicating no significant difference between the two methods. It shows that the established NIRS model of the determination of PUFA in baby milk powder has high predictive accuracy and can be used for the rapid detection of milk powder quality analysis.