基于SPA—PLSR的回收油鉴别研究  

Identification research in recovery-oil based on SPA—PLSR model

作　　者：赵明富[1] 张政委[1] 冯小平[1] 夏曦[1] 王念[1] 

机构地区：[1]重庆理工大学电子信息与自动化学院,重庆400054

出　　处：《激光杂志》2013年第3期45-47,共3页Laser Journal

摘　　要：近年来回收油(地沟油)事件频繁出现,为寻找快速、无损、高效的鉴别方法,防止回收油再次流回餐桌。本文采集到10种回收油以及10种正常食用油样品,建立光谱技术结合SPA—PLSR组合模型,选取因变量苯环(270nm)、多环芳烃类(225nm),提取特征吸收波长个数为8,分别为:195,197,209,245,255,270,310和830nm。利用最低PRESS法确定潜变量个数:LV225=4,LV270=3,建立PLSR回归方程Y225,Y270,预测相关度分别为R225=0.994119,R270=0.9886,均方根残差分别为RMSEC225=0.033654,RMSEC270=0.0336。最后利用Fisher有监督模式分类器实现对8种正常食用油和8种回收油的投影,以及对2种回收油和2种正常食用油的有效辨别,准确率为100%。可作为鉴别回收油与正常食用油的一种办法。Events of the recycled - oil （waste oil） often happen in recent years. In order to prevent the recycled - oil return to the table again, it is necessary to find a fast, efficient, non - destructive methods of identifying. This paper collected ten kinds of recycled - oil and ten kinds of cooking - oil samples. Set up the model combined spectroscopy and SPA- PLSR. There are benzene ring {270nm） and polycyclic aromatic hydrocarbons （225nm）, which are dependent variables selected. The extracting characteristic absorption wavelength number is 8. There are 195, 197j 209, 245, 255, 270, 310 and 830nm, respectively. The minimum PRESS method was used to determine the latent variable number, there are LV225 = 4, LV270 = 3. And established PLSR regression equations： Y225 and Y270. The correlation of R225 = 0.994 119 and R270 = 0. 9886 are expected, and the root mean square residuals are RMSEC225 = 0. 033654 and RMSEC270 = 0.0336, respectively. Finally, Fisher supervision pattern classifier is used to achieve projection of eight kinds of the cooking- oil and recycled oil samples, And achieve effective discrimination of two kinds of recycled - oil and two kinds of cooking - oil samples at the same time. The accuracy is 100%. So it can be used as a way to identif~ the recycled - oil with cooking- oil.

关 键 词：回收油 SPA--PLSR组合模型 最佳潜变量 FISHER分类器 

分 类 号：TQ646[化学工程—精细化工] TN248[电子电信—物理电子学]