MarSCoDe Martian Material Analysis Based on a PSO-SVR Approach

Xiong Wan,Peipei Fang,Yian Wang,Yingjian Xin,Mingkang Duan,Hongpeng Wang,Xinru Yan,Chenhong Li,Yanhua Ma,Zhiping He

DOI: https://doi.org/10.1021/acsearthspacechem.4c00100

2024-01-01

ACS Earth and Space Chemistry

Abstract: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 launch, we used the LIBS spectra of the LC-005 calibration standard on a Zhurong rover to further evaluate the prediction accuracy of the model. The main element contents of LC-005 predicted by the model are basically consistent with its real contents. Since then, the model has been used in the onboard quantitative element analysis of MarSCoDe. Finally, quantitative analysis results of eight different unknown Martian targets on different Mars days are selected and shown, which reflects the main geological composition of the landing area of the Martian Utopia plain.