基于特征波长快速筛选的润滑油金属元素的激光诱导击穿光谱定量分析研究

Quantitative Analysis Method of Metal Element in Lubricating Oil Based on Laser-Induced Breakdown Spectroscopy and Characteristic Wavelength Fast Selection

润滑油中的金属元素能够直接反映机械结构磨损状态和位置,对其定量分析是实现故障预警和诊断的有效手段。基于激光诱导击穿光谱(LIBS)技术,采用相关系数初遴选(CCPS)快速缩小特征波长范围,再结合迭代预测权重偏最小二乘法(IPW-PLS)、无信息变量消除法(UVE)、竞争性自适应重加权法(CARS)等算法精确提取特征波长,实现特征波长的快速有效筛选,最后基于偏最小二乘法(PLS)建立润滑油金属元素的定量分析模型,实现润滑油金属元素定量分析。针对7种润滑油标样开展元素分析实验,结果表明,提出的CCPS方法能有效提高特征波长选择效率,运行时间减少50%以上,CCPS-IPW-PLS的相关系数RP2、预测集均方根误差RMSEP值分别为0.9945、25.1678μg/g,CCPS-UVE的RP2、RMSEP值分别为0.9790、52.7363μg/g,CCPS-CARS的RP2、RMSEP值分别为0.9939、25.0996μg/g,证明了所提方法具有较好的准确度和特征波长选择效率,为实现润滑油的快速、便携、准确检测提供了新途径。

Metal elements in lubricating oil can directly reflect the wear status and position of the mechanical structure,and analyzing them quantitatively is an effective means of realizing fault warning and diagnosis.Based on laserinduced breakdown spectroscopy(LIBS)technology,the correlation coefficient method and threshold setting are used to narrow the range of feature wavelengths rapidly.The feature wavelengths are extracted accurately by the iterative predictive weighted partial least squares(IPWPLS),uninformative variable elimination(UVE),and competitive adaptive reweighted sampling(CARS)methods.Finally,based on partial least squares(PLS)method,a quantitative analysis model of metal elements in lubricating oil is established to analyze metal elements in lubricating oil quantitatively.The experimental results show that the proposed CCPS method can improve the efficiency of characteristic wavelength selection and reduce the running time by more than 50%.The correlation coefficient R2 P and rootmeansquare error prediction(RMSEP)values of CCPSIPWPLS are 0.9945 and 25.1678μg/g,respectively.The R2 P and RMSEP values of CCPSUVE are 0.9790 and 52.7363μg/g,respectively,and the R2 P and RMSEP values of CCPSCARS are 0.9939 and 25.0996μg/g,respectively.These results prove the accuracy and efficiency of the proposed method.The approach provides a new way to perform the rapid,portable,and accurate detection of lubricating oil.