Discrimination Of Oil Samples Using Laser Induced Fluorescence Spectroscopy

Oil spill, which is one of the most serious problems among the global marine pollution, has caused enormous loss of national property and environment. Detecting the source of spilled oil quickly and accurately is a significant foundation for hazard decreasing and responsibility division in oil spill accidents. Laser induced fluorescence (LIF) technology is used in discrimination of spilled oil due to the fluorescence activity of polycyclic aromatic hydrocarbon (PAH) in crude oil and petroleum products.In this article, a discrimination method is developed based on LIF and pattern recognition algorithm to identify the type of spilled oil. The background and significance of spilled oil identification is introduced, and the development situation of liDAR and LIF method in oil spill detection is investigated. Then the main work in this article is also introduced, as well as LIF spectrum technique and principle of the algorithm applied. In this thesis, the work content includes three aspects.Firstly, three different algorithms are applied on the discrimination of crude oil and refined petroleum products using emission spectrum. In this part, fluorescence emission spectrum of diesel, gasoline, heavy fuel oil and five kinds of crude oil is detected and identified by PLS-DA, PCA integrated with BP-ANN and SVM algorithm. Considering classification accuracy and algorithm model training difficulty, SVM model is selected for follow-up work.Additionally, SVM method is used in feature extraction of laser induced time-resolved fluorescence (TRF) spectrum. Time-resolved spectra data was descended into two dimensions with selecting appropriate range in time and wavelength domains respectively to form a SVM data base, and an appropriate range is selected in both domains to maximize the classification accurate rate. As a result, the feature ranged is extracted after the data processing procedure mentioned above.Finally, a discrimination method of crude oil and refined petroleum products based on time-resolved fluorescence spectrum is presented. In this part, PCA and statistical parameters are adopted for feature extraction and SVM model is adopted as discrimination algorithm. Moreover, experimental device, samples and spectra data acquisition are also introduced in this article.