Rapid and Direct Spectrofluorometric and Chemometrics Methods for the Simultaneous Determination of Two Dansyl Derivatives

ABSTRACT

In this study the simultaneous molecular spectrofluorometric determination of ultratrace amounts of two dansyl chloride derivatives, DMNPS (5-(dimethylamino)naphthalene-1-sulfonyl 4-phenylsemicarbazide) and DMNPH (2-(5-(dimethylamino)naphthalen-1-ylsulfonyl)-N-phenylhydrazinecarbothioamide), was accomplished using a genetic algorithm joint partial least squares (GA-PLS) technique that leads to very low detection limits (lower than 10^?6 mol/L) The linear dynamic ranges of the compounds were 1–6 µ mol L^?1 and 1–7 µ mol L^?1 for DMNPS and DMNPH, respectively. Quantification was performed using the emission wavelength range from 360 to 600 nm with an optimum calibration sample number of 25 and prediction sample number of 7. The technique was proved to be beneficial.     

GA-PLS结合PC-ANN算法提高奶粉蛋白质模型精度

提出一种偏最小二乘法(PLS)和人工神经网络(ANN)结合用于近红外光谱(NIRS)的分析方法,以提高奶粉蛋白质模型的预测精度。首先采用基于遗传算法的波长选择法(RS-GA)优化光谱数据,建立GA-PLS模型预测奶粉蛋白线性部分;然后在RS-GA法选择的波段上进行主成分分析(PCA),以主成分的得分矩阵作为ANN模型输入层,以GA-PLS预测值与真实值之差作为输出层,建立PC-ANN模型预测其非线性部分。最终预测结果为两个模型预测值之和,以模型的预测标准偏差(RMSEP)作为评价指标,以便考察新方法的有效性。同时建立线性的全谱模型(Fr-PLS),其Fr-PLS、GA-PLS和GA-PLS+PC-ANN模型的RMSEP分别为0.511,0.440和0.235。结果表明：考虑奶粉蛋白含量近红外模型的非线性部分,可以显著提高模型的预测精度,该方法也可为其它复杂体系模型精度的提高提供借鉴。