Abstract:The performance of aluminum alloys can be affected by the element of magnesium.Therefore, It is necessary to realize fast and accurate quantitative analysis of magnesium in aluminum alloys. Herein, microchip-lasers based high repetition rate spark discharge enhanced laser-induced breakdown spectroscopy (HRR SD-LIBS) was applied to quantitatively analyze magnesium elements in aluminum alloys. Baseline correction was performed using the moving median method to obtain the net spectral data. Multivariate calibration models such as back-propagation neural network (BP), support vector machine (SVM), and random forest (RF) were established, respectively. Compared with the conventional univariate calibration model, the correlation coefficients (RP2) of BP, SVM, and RF models increased from 0.810 to 0.985, 0.997, and 0.994, while the root mean square error (RMSEP) decreased from 0.283 to 0.0598, 0.0410, and 0.0542, the average relative error decreased from 26.62% to 3.13%, 2.82%, and 3.61%. It indicated that machine learning combined microchip laser-based HRR SD-LIBS can significantly improve the quantitative analysis ability of LIBS, and it is promising to achieve portable, fast, and accurate quantitative elemental analysis of alloy samples and other materials.