激光诱导击穿光谱结合偏最小二乘测定渣油中镍钒  

Determinatin of nickel and vanadium in residue oil by laser-induced breakdown spectroscopy combined with partial least squares

作　　者：刘良 闫春华 李茂刚 张天龙[2] 汤宏胜[2] 李华[1,2] LIU Liang;YAN Chunhua;LI Maogang;ZHANG Tianlong;TANG Hongsheng;LI Hua(College of Chemistry and Chemical Engineering,Xi'an Shiyou University,Xi'an 710065,China;Key Laboratory of Synthetic and Natural Functional Molecule of the Ministry of Education,College of Chemistry&Materials Science,Northwest University,Xi'an 710127,China)

机构地区：[1]西安石油大学化学化工学院,陕西西安710065 [2]合成与天然功能分子教育部重点实验室,西北大学化学与材料科学学院,陕西西安710127

出　　处：《冶金分析》2024年第10期87-94,共8页Metallurgical Analysis

基　　金：国家自然科学基金资助项目(22173071);西安石油大学研究生创新与实践能力培养计划资助项目(YCS22221015)。

摘　　要：渣油中金属元素含量的快速分析对于保证渣油的产品质量和炼油过程的效率具有重要意义。从某炼油厂获得不同渣油样品,煅烧后研磨成粉末,按不同质量比混合获得渣油样本。本研究将激光诱导击穿光谱(LIBS)与化学计量学相结合建立了渣油中金属元素Ni和V的快速定量分析方法。首先,基于20个不同样本的LIBS光谱构建渣油中Ni和V定量分析的偏最小二乘(PLS)模型。其次,探究了不同预处理和变量选择方法对PLS校正模型预测性能的影响。结果表明,对于渣油中Ni和V元素,基于一阶导数-多元散射校正结合和协同区间偏最小二乘的PLS模型以及基于二阶导数-多元散射校正结合和协同区间偏最小二乘的PLS模型获得了最优预测效果,最优的决定系数(R_(P)^(2))分别为0.986 4和0.981 2,均方根误差(RMSE_(P))分别为1.440 2 mg/kg和0.588 8 mg/kg,平均相对误差(MRE_(P))分别为3.89%和1.85%。因此,LIBS技术与PLS算法相结合可为渣油中Ni和V元素的定量分析提供一种可行方法。The rapid analysis of metal elements in residue oil is of great significance for the product quality of residue oil and the efficiency of refining process.Different residue oil samples were obtained from a refinery,calcined and ground into powder,and mixed according to different mass ratios to obtain residue oil samples.In this study,the laser-induced breakdown spectroscopy(LIBS)was combined with chemometrics to establish a rapid quantitative method for the analysis of nickel(Ni)and vanadium(V)in residue oil.Firstly,the partial least squares(PLS)model for the quantitative analysis of Ni and V in residue oil was constructed based on the LIBS spectra of 20 different samples.Secondly,the effects of different preprocessing and variable selection methods on the prediction performance of PLS calibration model were investigated.The results showed that for Ni and V in residue oil,the PLS model based on first derivative-multiplicative scatter correction and synergetic interval PLS(D1st-MSC-siPLS)and PLS model based on second derivative-multiplicative scatter correction and synergetic interval PLS(D2nd-MSC-siPLS)had the optimal prediction results.The optimal determination coefficients(R_(p)^(2))were 0.9864 and 0.9812,and root mean square errors(RMSE_(p))were 1.4402 and 0.5888 mg/kg,and the mean relative errors(MRE_(p))were 3.89%and 1.85%,respectively.Therefore,the combination of LIBS technique with PLS algorithm could provide a feasible method for the quantitative analysis of Ni and V elements in reside oil.

关 键 词：激光诱导击穿光谱(LIBS) 偏最小二乘(PLS) 变量选择 渣油 镍 钒 

分 类 号：O657.38[理学—分析化学] TE624.93[理学—化学]