Simulation of laser-induced plasma temperature based on machine learning

Laser-induced breakdown spectroscopy (LIBS), a rapid and nondestructive elemental analysis method, is widely applied across various fields. This technique is used for elemental analysis of materials by analyzing the light emitted from laser-induced plasma, with plasma temperature being a crucial parameter. This study focused on obtaining time-resolved spectra of Cu samples at different distances from the lens to the sample using LIBS in conjunction with the following machine-learning algorithms to simulate plasma temperature: decision trees, support vector machine, Gaussian process regression, and neural network. The study compared the accuracy of these four algorithms before and after normalization based on the Cu (I) spectral line at 510.55 nm. All four algorithms achieved a correlation coefficient greater than 0.95 after normalization. The Gaussian process regression model provided excellent predictions before and after normalization with a correlation coefficient of 0.99, demonstrating the advantages of this model in plasma temperature simulation. This study confirmed that machine learning can effectively, conveniently, and accurately predict laser-induced plasma temperature, providing significant utility in LIBS research. (c) 2024 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license