相似度计算法在汽油溢油识别中的应用

摘要

将相似度计算法引入到汽油的溢油识别中,将油品的谱图信息转化为二维向量,利用夹角余弦法计算二维向量间的相似度,并以此来表征谱图之间的相似度.对2个组成相似的催化裂化汽油进行了模拟溢油识别.结果表明,催化裂化汽油萃取油E-1与样品油C-1、C-2的相似度分别为0.996、0.974,前者大于后者,萃取油E-2与样品油C-1、C-2的相似度分别为0.983、0.992,前者小于后者,2个萃取油与对应原样品油的相似度结果均高于其与另外一个样品的相似度.在重复性限法和t检验法均无法完全识别的情况下,利用相似度计算法可以准确找到萃取油对应的原样品油,解决了相似油品识别、汽油识别的难题.%Similarity algorithm was used for the gasoline oil spill identification.The chromatogram information was first converted to a two-dimensional vector,and the similarity between two-dimensional vectors was then calculated by the angle cosine method,by which the similarity of oil spill fingerprint chromatograms can be measured.The extracted oils referring as E-1,E-2 from two gasoline+water samples(C-1,C-2) from different FCC units were used as simulated oil spills.The results showed that the similarity between E-1 and C-1,E-1 and C-2 is 0.996,0.974 respectively,The similarity between E-2 and C-1,E-2 and C-2 is 0.983,0.992 respectively.It is clear that the similarities difference between extracted oils and its corresponding original gasoline(E-1 and C-1,E-2 and C-2)can clearly distinguish two different gasoline samples.In the case of the repeatability limit method and t test method invalid,the original oil corresponding to extracted oil can be recognized accurately by similarity calculation method.