Accidental oil spills occur frequently around the world and they can have serious influence on the human health and the ecosystem due to the compounds in oil.Thus,there is an urgent need for an accurate method for determining the source of spills.In order to meet the rapidly demand identifying methods of oil spills,this paper utilized parallel factor analysis to develop an identification method for the crude oils and fuel oils based on their fluorescence excitation-emission matrixes.Firstly,the fluo-rescence excitation-emission matrixes of six kinds of crude oils(Roncador crude oil,Basra crude oil,Russian crude oil,Saudi crude oil(heavy crude oil),Upper Zakum crude oil,sea two station crude oil)and three kinds of fuel oils(380 CST fuel oil,Fu-el oils No.5—No.7,LanShan fuel oil)were normalized after removing scatter by Blannany triangulation with linear interpola-tion.Afterwards,the parallel factor analysis was used for the fluorescence excitation-emission matrixes.Seven factors can be re-liably extracted from the data set,and then seven fluorescence components were obtained,which made up the characteristic spec-trum.By clustering analysis and Bayesian Discrimination of the characteristic spectrum from the samples that had experienced weathered for 3,15 and 45 days and not experienced weathered,the capacity of the oil fluorescence spectrum analysis was deter-mined and the fluorescence feature spectra library consisted of 12 standard fluorescence spectra of crude oils and 6 standard fluo-rescence spectra of fuel oils was established.In the end,multiple linear regression was used to recognize the samples that had ex-perienced weathered for 0,7 and 30 days by the oil standard fluorescence spectra.The results showed that,except for Russian crude oil,the method could clearly classify 5 kinds of crude oils and 3 kinds of fuel oils which had experienced weathered and had not experienced.Furthermore,the accuracy of the identification for crude oils and fuel oils were 100.0%,and the overall accura-cy of the identification for crude oils and fuel oils were 87.5% and 100%,respectively.%在世界范围内溢油事件频繁发生,溢油的组成成分会影响人类身体健康和生态系统.因此,迫切地需要一种可以快速识别溢油种类的方法.针对溢油污染物现场快速鉴别的需求,利用平行因子分析技术建立了基于三维荧光光谱的原油、燃料油识别方法.首先,利用Delannay三角形内插值法对实验选的6种原油(Roncador原油、巴士拉原油、俄罗斯原油、沙特原油(重质)、上扎库姆原油、海二站原油)和三种燃料油(380CST燃料油、5-7号燃料油、岚山燃料油)的三维荧光光谱去散射,去散射后的三维光谱数据进行归一化处理;之后,对三维荧光光谱进行平行因子解析,确定七个荧光组分为最佳荧光组分,进而得到由7个荧光成分组成的样品荧光特征谱,将风化第3,15和45天的样品及未风化样品的第一平行样的荧光特征谱进行贝叶斯方法(Bayes)判别分析和聚类分析,确定油品荧光特征谱的分析能力和18条荧光标准谱库(12条原油标准谱和6条燃料油标准谱);最后,利用非负最小二乘多元线性回归建立溢油荧光识别方法,对第0,7和30天风化的样品和未风化样品的另一平行样进行识别.实验结果表明,除对风化及未风化的俄罗斯原油识别外,该方法对其余风化和未风化的五种原油和三种燃料油识别正确率均为100.0%,整体识别原油正确率为87.5%,燃料油正确率为100.0%.
展开▼
