基于近红外光谱法的多组分复杂溢油源定量建模分析

研究利用近红外光谱分析方法进行模拟复杂混合溢油源的定量分析问题。选取汽油、柴油、煤油三种轻质石油类产品，按照不同浓度比例配置成40个模拟混合溢油样本，利用傅里叶变换近红外光谱仪采集其在4 000～12 000 cm-1谱区范围内的近红外光谱;采用不同预处理方法，利用偏最小二乘算法建立混合溢油样本三组分各自的浓度定量模型。汽油、柴油和煤油的最优预处理方法均为二阶导数方法，分别在8 501.3～7 999.8 cm-1，6 102.1～4 597.8 cm-1，6 549.5～4 597.8 cm-1，7 999.8～7 498.4 cm-1和6 102.1～4 597.8 cm-1谱区范围内，预测模型的决定系数R2分别为0.998 2，0.990 2和0.993 6;RMSEP值分别为0.474 7，0.936 1和1.013 1;RPD值分别为25.126 9，10.517 3和13.072 0。实验结果表明：利用近红外光谱分析技术结合化学计量学方法能够定量确定模拟混合溢油样本中各组分的浓度，为海洋复杂溢油源的定量检测与分析提供有效手段。

Quantitative Analysis Model of Multi-Component Complex Oil Spill Source Based on Near Infrared Spectroscopy

Near infrared spectroscopy technology was used for quantitative analysis of the simulation of complex oil spill source. Three light petroleum products, i.e. gasoline, diesel fuel and kerosene oil, were selected and configured as simulated mixture of oil spill samples in accordance with different concentrations proportion, and their near infrared spectroscopy in the range of 8 000～12 000 cm-1 was collected by Fourier transform near infrared spectrometer. After processing the NIR spectra with different pretreatment methods, partial least squares method was used to establish quantitative analysis model for the mixture of oil spill samples. For gasoline, diesel fuel and kerosene oil, the second derivative method is the optimal pretreatment method, and for these three oil components in the ranges of 8 501.3~7 999.8 and 6 102.1~4 597.8 cm-1;6 549.5~4 597.8;7 999.8~7 498.4 and 102.1~4 597.8 cm-1, the correlation coefficients R2 of the prediction model are 0.998 2, 0.990 2 and 0.993 6 respectively, while the forecast RMSEP indicators are 0.474 7, 0.936 1 and 1.013 1 respectively; The experimental results show that using near infrared spectroscopy can quantitatively determine the content of each component in the simulated mixed oil spill samples, thus this method can provide effective means for the quantitative detection and analysis of complex marine oil spill source.