石油作为重要的能源和工业原料, 在造福人类社会的同时, 其引起的环境污染问题日益严重。 因此针对混合油液的快速、 准确检测成为鉴别溢油来源和保护生态环境的重要内容。 石油类物质一般由具有较强荧光特性的芳香烃成分及其衍生物组成, 荧光光谱分析技术以其灵敏度高、 分析速度快和受风化影响程度小等优点成为了混合油液检测的重要手段之一, 并与二阶校正和三阶校正的各类算法相结合取得了较好的成分鉴别和浓度预测效果。 但二阶校正算法普遍存在对噪声的容忍能力弱和对组分数敏感、 收敛速度慢等不足, 限制了在实际混合油液检测中的应用。 针对上述存在的问题, 将三维荧光光谱技术和交替加权残差约束四线性分解(AWRCQLD)算法相结合, 提出一种用于混合油液检测的新方法。 首先以乙醇作为溶剂, 将航空煤油和润滑油按不同浓度比配制7个校正样本、 4个预测样本和3个空白样本; 然后利用FLS920荧光光谱仪采集拟进行成分检测的混合油液在不同实验温度条件下共42个样本的荧光光谱数据, 并通过空白扣除的方法消除散射的干扰; 再利用核一致诊断法和残差分析法估计出最佳的组分数; 最后分别利用AWRCQLD算法、 4阶平行因子(4-PARAFAC)算法和二阶校正算法解析样本的荧光光谱数据, 做出混合油液样本的定性鉴别和定量预测。 研究结果表明, 经AWRCQLD算法解析后得到的航空煤油预测样本的回收率为96.7%~102.7%、 预测均方根误差为0.015 mg·mL-1; 润滑油预测样本的回收率为96.9%~101.7%、 预测均方根误差为0.009 mg·mL-1; 在不同实验温度条件构建的四维响应数阵能够更为准确地测定出航空煤油和润滑油的组分浓度, 其回收率更高和预测均方根误差更小, 满足准确定量分析的要求; AWRCQLD算法在航空煤油和润滑油样本的荧光光谱严重重叠的情况下, 较之二阶校正算法和4-PARAFAC算法, AWRCQLD算法更能够体现出三阶校正算法所具有的优势, 综合预测能力更强, 达到了对混合油液进行快速检测的目的。 该研究提供了一种不依赖于“物理和化学分离”的快速、 准确的对混合油液进行检测的“数学分离”方法, 为石油类混合油液检测提供了必要的技术支持。

As an important energy and industrial raw material, petroleum brings benefit to human beings and the environment pollution is increasingly serious. Therefore, rapid and accurate detection of mixed oil becomes an important content of identification of its source and protect ecological environment. Petroleum substances are generally composed of aromatic hydrocarbon and its derivatives with strong fluorescence characteristics, and fluorescence spectroscopy is an important means of detecting mixed oil with the advantages of high sensitivity, fast analysis and small weathering effects. And it has obtained good results for components identificationand concentration prediction of oil spill by various algorithms of second-order calibration algorithm and third-order calibration algorithm. Second-order calibration has the shortcomings of weak tolerance to noise, sensitivity to number of components, and limited real application of mixed oil detection. Aiming at these problems, a novel method is proposed to detect mixed oil in this paper based on the combination of three-dimensional fluorescence spectroscopy and alternating weighted residue constraint quadrilinear decomposition (AWRCQLD) algorithm. Firstly, using ethanol as a solvent, 7 calibration samples, 4 prediction samples and 3 blank samples of jet fuel and lube with different volume ratios were prepared. Secondly, the fluorescence spectra of 42 samples of the mixed oil at different experimental temperatures were obtained by FLS920 fluorescence spectrometer, and the effect of scattering was removed by using blank subtraction. Then, the optimum number of components was estimated by core consistency diagnosis and residual analysis. Finally, using AWRCQLD algorithm, 4-PARAFAC algorithm and second-order calibration algorithm to analyze the fluorescence spectra, and the qualitative identification and quantitative prediction of mixed oil samples were made. The research results show that the interval of the obtained recovery rate of jet fuel prediction samples is 96.7%~102.7%, and the root mean square error of prediction is 0.015 mg·mL-1; the interval of the obtained recovery rate of lube prediction samples is 96.9%~101.7%, and the root mean square error of prediction is 0.009 mg·mL-1. The four-dimensional response matrix constructed can more accurately determine the concentration of jet fuel and lube at different experimental temperatures, and the recovery rate is higher, the root mean square error is smaller, and can meet the requirements of accurate quantitative analysis. Compared with the second-order calibration algorithm and 4-PARAFAC algorithm, AWRCQLD algorithm can better reflect the superiority of the third-order calibration algorithm and the comprehensive prediction ability is stronger under seriously overlapped fluorescence spectra of jet fuel and lube. The purpose of rapid detection of mixed oil can be achieved by AWRCQLD algorithm. The study provides a rapid and accurate “mathematical separation” method to detect mixed oil not based on “physical and chemical separation”, and also provides a necessary technological support for detection of petroleum mixed oil.