MarSCoDe Martian Material Analysis Based on a PSO–SVR Approach

作者：Xiong Wan, Peipei Fang, Yian Wang, Yingjian Xin, Mingkang Duan, Hongpeng Wang, Xinru Yan, Chenhong Li, Yanhua Ma, Zhiping He · 发表于：ACS Earth and Space Chemistry · 年份：2024 · DOI：10.1021/acsearthspacechem.4c00100 · 被引用次数：4 · 研究领域：Planetary Science and Exploration、Astro and Planetary Science、Laser-induced spectroscopy and plasma

Laser-induced breakdown spectroscopy (LIBS) has been used for deep space exploration in recent years. The advantages of LIBS include high efficiency, stand-alone detection, and the ability to analyze multiple elements simultaneously. However, due to the fluctuation of laser energy, matrix effect, and instrumental noises, the quantitative prediction of LIBS instruments for planetary exploration is not satisfactory, especially for unknown targets. Therefore, comprehensive methods with higher adaptability and prediction accuracy must be developed to meet the needs of LIBS planetary material analysis. In this paper, we proposed an approach, which is mainly based on a particle swarm optimization (PSO)–support vector regression (SVR) analysis model, for material analysis of MarSCoDe, the LIBS payload of the Chinese Zhurong Mars rover. The model adopts a PSO algorithm to optimize the parameters and hence improve the prediction accuracy of traditional SVR equations. The training of the model was completed with 3600 LIBS spectra, which involved 60 standards and were obtained in the ground simulated Martian chamber before the launch of MarSCoDe. The quantitative performance of the model was evaluated by the coefficient of determination ( R 2 ) and root-mean-square error between real contents and predicted contents. Comparison with convolutional neural network and partial least squares showed that the PSO–SVR model has the highest prediction accuracy and the best robustness. After the l...