The Study On Identification Of Single Oil And Blend Oil Using A Multi-dimensional Chemical Fingerprinting Model

With the rapid development of offshore oil exploration and shipping-industries,the cases of oil spill accidents has become more frequent.The currently reported oil fingerprinting identification methods are mainly used to identify the type of single oil,and the identification of blend oil is rarely done.However,when an oil spill accident occurs,it is often not a single oil leakage,which may be accompanied with a mixture of multiple oil,so the identification of blend oil is also very important.In order to solve related problems,In this study.the n-alkanes,fluorescence characteristics and Ni/V in the crude oil or marine fuel were used to establish a multi-dimensional chemical fingerprinting for quantitative identification of oil samples.The purpose is to quickly and accurately identify crude oil,marine fuel and both blend oil.A total of 27 kinds of oil samples were used as the study objects,including 6 kinds of crude oil,6 kinds of marine fuel,and 15 kinds of blend oil of crude oil and fuel oil.These 27 kinds of oil samples were weathered,and their characteristics of multi-dimensional chemical fingerprinting before and after weathering were analyzed in detail.At the same time,the data were processed with chemometrics.First,the principal component analysis was used to process the data of the n-alkanes n-C₁₆～n-C₃₅,and 5 principal components were extracted with eigenvalues greater than 1,the cumulative explanatory contribution rate of these 5 principal components to the original data is up to 95.63%;Secondly,the fluorescence characteristics of27 oil samples were detected by the constant wavelength synchronous fluorescence spectrometry,and the db7 wavelet basis function was selected to perform 6-layer discrete wavelet transform on the fluorescence spectra and the d3 wavelet coefficient was obtained as the study object,and the wavelet coefficients at characteristic wavelengthsλ=255±2 nm,λ=280±2 nm,λ=302±2 nm,λ=332±2 nm andλ=354±2 nm are extracted,respectively;Then,combined with the ratio of nickel to vanadium,the most representative parameters of n-alkane and fluorescence were selected as modeling variables by using variable cycle screening method.All combinations of variables are exhausted,and the most suitable combination of variables is determined by comparing the identification accuracy of each combination.Finally,the third principal component(n-C₃₃～n-C₃₅),the d3 wavelet coefficient at 255±2 nm,and the ratio of nickel to vanadium were selected as modeling variables from 13 variables,and the Fisher discriminant model was established.The comprehensive discrimination accuracy rate of the established model on the training set is 92.6%,among which the discrimination accuracy rate of 12 kinds of single oil（6 kinds of marine fuel and 6 kinds of crude oil）is as high as 100%,and the discrimination accuracy rate of 15 kinds of blend oil is 86.7%.The established model was used to identify 37 non-modeled and short-term weathered oil samples,the results showed that the model’s identification accuracy to blend oil samples was 84.0%,and the identification accuracy of crude oil and marine fuel are 100%and 83.3%,respectively.It can be seen that the model established in this study can effectively identify fuel oil,crude oil and their blend oil samples.