基于导数光谱融合建模的红外光谱定量分析方法

摘要

A derivative spectral estimator ( DSE) based on singular perturbation technique was designed and a quantitative analysis method based on derivative spectra information space, termed derivative spectra fusion interval partial least squares ( DSF-iPLS) modeling was proposed. DSF-iPLS mainly focused on obtaining final fusion model by making full use of derivative spectra information. The glucose spectra dataset with concentrate ranging from 0 . 04% to 5% and the beer spectra dataset with the original extract concentration ranging from 4. 23 to18. 76°P (Plato) were used to evaluate the effectiveness of the proposed quantitative analysis method. The experiment results indicated that DSF-iPLS model for two infrared spectra datasets provided the minimum root mean square error of prediction ( RMSEP ) and the values were 0. 121 and 0. 087, respectively. Compared with other single model, DSF-iPLS model based derivative spectra could provide more excellent predictive performance.%设计了基于奇摄动技术的导数光谱估计器并提出基于不同阶次导数光谱空间的融合建模定量分析方法。方法充分利用导数光谱信息空间、区间最小二乘法和融合建模的优点,挖掘光谱深层次信息进行融合建模。分别利用麦汁浓度范围4.23~18.76° P (柏拉图度)的啤酒红外光谱公共数据集和配制的浓度为0.04%~5%范围的葡萄糖溶液实测光谱数据集进行定量分析方法的对比实验。实验结果表明,融合建模定量分析方法能获得最小的预测均方根误差(RMSEP),其值分别为0.121和0.087,能够准确地进行定量分析。与其它建模方法相比较,基于导数光谱的融合建模方法所建立的预测模型具有明显优越的性能。