Quantitative study of Fe in lubricating oil by metal substrate and vacuum negative pressure assisted LIBS with SAO-LSSVM

Abstract

The content of metal elements in lubricating oil can reflect the wear state of machinery, and its quantitative analysis can provide a basis for the early diagnosis of equipment. A total of 35 Fe-containing lubricating oil samples were prepared. First, for the 15 high-concentration samples, confined area metal zinc substrate assisted laser-induced breakdown spectroscopy (MZS-LIBS) was employed. The remaining low-concentration samples, where Fe spectral lines were difficult to detect using MZS-LIBS, were analyzed using vacuum negative pressure assisted LIBS (VNP-LIBS). Second, the spectral data of the two methods were analyzed, and the univariable calibration curves of MZS-LIBS and VNP-LIBS were established with Fe II 259.94 nm as the analytical spectral line. The detection limits were calculated to be 18.96 μg g⁻¹ and 0.05 μg g⁻¹, respectively, which verified the limitations of MZS-LIBS in low concentration detection. Finally, a least squares support vector machine (LSSVM) quantitative model was constructed based on the data sets of the two methods, and the snow ablation optimizer (SAO) was introduced to optimize its hyperparameters. After optimization, the determination coefficient (R_P²) of MZS-LIBS increased from 0.64 to 0.92, an increase of 43.75%, and VNP-LIBS increased 130.23% from 0.43 to 0.99. These results validate the potential of LIBS technology in the detection of wear metals.