基于红外光谱技术与变量选择算法的柴油纯度检测  

Diesel Purity Detection Based on Infrared Spectroscopy Technology and Variable Selection Algorithm

作　　者：周围[1] 李安吉 吕文敏 俞铁铖 赵丽娟[1] 赵美琪 尹冉 ZHOU Wei;LI An-ji;LV Wen-min;YU Tie-cheng;ZHAO Li-juan;ZHAO Mei-qi;YIN Ran(School of Physics and Electronic Engineering,Northeast Petroleum University)

机构地区：[1]东北石油大学物理与电子工程学院

出　　处：《化工自动化及仪表》2024年第4期592-598,613,共8页Control and Instruments in Chemical Industry

基　　金：东北石油大学引导性创新基金(批准号:2018YDL-07)资助的课题。

摘　　要：为实现柴油纯度的快速检测,采用RS法、KS法和SPXY法对利用红外光谱技术所采集的柴油样本数据进行样本集划分,之后利用蒙特卡洛交叉验证法(MCCV)剔除异常样本,并采用无信息变量消除法(UVE)、竞争性自适应重加权采样法(CARS)、遗传算法(GA)和自举柔性收缩算法(BOSS)对全光谱变量进行优选,根据所选变量建立偏最小二乘(PLS)模型。结果表明:SPXY法划分样本后所建立的模型效果最好,其预测均方根误差RMSEP为0.0412,预测相关系数R_(P)为0.9921;经MCCV法剔除异常样本后,模型的RMSEP降低了4.85%,R_(P)提高了0.50%;经UVE法选择的变量所建立的PLS模型性能最好,其RMSEP可达0.0239,R_(P)可达0.9977,与全谱模型相比,RMSEP降低了39.03%,R_(P)提高了0.06%,变量减少率可达67.27%。For purpose of quickly detecting the purity of diesel oil,having the RS method,KS method and the SPXY method adopted to classify the sample set of diesel oil sample data collected by infrared spectroscopy technology was implemented,including having Monte Carlo cross validation(MCCV)used to eliminate abnormal samples,the partial least squares(PLS)models established based on the optimal selection of the full spectrum variables by using the uninformative variable elimination(UVE),the competitive adaptive reweighted sampling(CARS),genetic algorithm(GA)and the bootstrapping soft shrinkage(BOSS).The results show that,the model established by SPXY method has the best performance,and its root mean square error for prediction(RMSEP)is 0.0412 and related coefficient for prediction(Rp)is 0.9921.After removing abnormal samples by the MCCV method,the RMSEP of the model can be decreased by 4.85%and theR_(P) increased by 0.50%.The PLS model established by the UVE method has the best performance with an RMSEP of 0.0239 and an R_(P) of 0.9977.Compared with the full spectrum model indicates that,the RMSEP decreases by 39.03%and the R_(P) increases by 0.06%and the variables reduction rate can reach 67.27%.

关 键 词：柴油纯度检测 红外光谱技术 变量选择 偏最小二乘模型 

分 类 号：TH83[机械工程—仪器科学与技术]