1. 燕山大学信息科学与工程学院，河北 秦皇岛 066004
2. 燕山大学电气工程学院，河北 秦皇岛 066004
3. Department of Telecommunications and Information Processing, Ghent University, B-9000 Ghent, Belgium
4. 北京化工大学信息科学与技术学院，北京 100029
An Oil Identification Method Based on Reconstructed 3D Fluorescence Spectra Combined With Partial Least Squares Discriminant Analysis
CUI Yao-yao1, KONG De-ming2, 3*, KONG Ling-fu1, WANG Shu-tao2, SHI Hui-chao4
1. School of Information Science and Engineering, Yanshan University, Qinhuangdao 066004, China
2. School of Electrical Engineering, Yanshan University, Qinhuangdao 066004, China
3. Department of Telecommunications and Information Processing, Ghent University, B-9000 Ghent, Belgium
4. School of Information Science and Technology, Beijing University of Chemical Technology, Beijing 100029, China
Abstract：Oil pollution is becoming more and more frequent, which brings a serious threat to human health and the ecological environment. Therefore, it is of great significance to study effective oil identification methods to protect the ecological environment. Three-dimensional (3D) fluorescence spectra are one of the most effective analytical methods for oils identification. 3D fluorescence spectra data are analyzed by using second-order calibration method. And then the concentration score matrix in the analysis results of the second-order correction method is classified by using pattern recognition, which can realize the qualitative identification of unknown samples. However, in the process of classifying and identifying unknown samples, the above methods only apply the concentration score matrix, which is essential to classify the unknown samples by using the relative content difference of the chemical components contained in the samples. The qualitative load matrix is not used, that is, the qualitative analysis of the sample is not achieved from the chemical components contained in the sample. Thus, a new identification method for oil samples was proposed by combining the reconstructed 3D fluorescence spectra with partial least squares discriminant analysis (PLS-DA). First, 80 oil samples were prepared by using four oils (gasoline, diesel, jet fuel and lubricating oil) in different backgrounds (sodium lauryl sulfate solvent prepared from purified water, tap water, river water and sea water); The 3D fluorescence spectra data of the sample was collected by using FS920 fluorescence spectrometer, and the data were preprocessed by de-scattering and standardized; Then, the abnormal spectra data was identified and deleted by using the Leverage value, and the remaining spectral data was reconstructed by using parallel factor analysis algorithm (PARAFAC); Finally, a classification model of reconstructed 3D fluorescence spectra was established by PLS-DA. The classification model established by reconstructing 3D fluorescence spectra was compared with the classification model established by unreconstructed 3D fluorescence spectra. The results show that, after the reconstruction of the 3D fluorescence spectrum, the correct classification rates of the four oils can be increased from 100%, 50%, 60% and 20% to 100%, 100%, 100% and 100%, respectively. It indicates that the reconstructed 3D fluorescence spectra have obvious intra-class characteristics. The sensitivity (SENS), specificity (SPEC) and F-scores of the classification model established by reconstructing the 3D fluorescence spectrum were 100%, 100%, and 100%, respectively. It indicates that the model established has robust and reliable analysis results. In this paper, 3D fluorescence spectra were reconstructed by using concentration score matrix and load matrix in the PARAFAC analysis results. Therefore, the PLS-DA classification model established by reconstructing 3D fluorescence spectra qualitatively identified samples not only from the difference in the relative content of chemical components, but also from the chemical components itself. Its results were convincing. This study provides a reliable method for oil identification.
Key words：Reconstructed 3D fluorescence spectrum; PARAFAC; PLS-DA; Oil identification
崔耀耀，孔德明，孔令富，王书涛，史慧超. 基于重构三维荧光光谱结合偏最小二乘判别分析的油类识别方法研究[J]. 光谱学与光谱分析, 2020, 40(12): 3789-3794.
CUI Yao-yao, KONG De-ming, KONG Ling-fu, WANG Shu-tao, SHI Hui-chao. An Oil Identification Method Based on Reconstructed 3D Fluorescence Spectra Combined With Partial Least Squares Discriminant Analysis. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2020, 40(12): 3789-3794.
[1] Danovaro R，Carugati L，Berzano M. Front. Mar. Sci.，2016, 3：213.
[2] Laffon B, Eduardo Pásaro, Valdiglesias V. Journal of Toxicology & Environmental Health Part B, 2016, 19(3-4): 105.
[3] Langangen O, Olsen E J, Stige L C, et al. Marine Pollution Bulletin, 2017, 119(1): 102.
[4] KONG De-ming, ZHANG Chun-xiang, CUI Yao-yao, et al(孔德明, 张春祥, 崔耀耀, 等). Acta Optica Sinica(光学学报), 2018, 38(11): 1130005.
[5] ZHOU Yan-lei, ZHOU Fei-fei, JIANG Cong-cong, et al(周艳蕾, 周飞飞, 姜聪聪, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2018，38(2): 475.
[6] Silvana M Azcarate, Adriano de Araújo Gomes, Mirta R Adriano, et al. Food Chemistry, 2015, 184: 214.
[7] Gu Huiwen, Wu Hailong, Yin Xiaoli, et al. Analytica Chimica Acta, 2014, 848: 10.
[8] Camilo L M Morais, Kássio M G Lima, Francis L Martin. Analytica Chimica Acta, 2019, 1063: 40.
[9] Callejóna R M, Amigo J M, Pairo E, et al. Talanta, 2012, 88: 456.
[10] Lenhardt L, Bro R, Zekovic I, et al. Food Chem., 2015, 175: 284.
[11] Camilo L M Morais, Kássio M G Lima, et al. Chemometrics and Intelligent Laboratory Systems, 2019, 188: 46.
[12] Morais C L M, Lima K M G. Chemometr. Intell. Lab. Syst., 2017, 170: 1.
