Data fusion of Laser-Induced Breakdown Spectroscopy and Diffuse Reflectance Spectroscopy for improved quantitative analysis of EAST-like plasma-facing materials

Abstract

Tungsten impurities deposited on the plasma-facing components (PFCs) of the inner walls of tokamak devices pose a significant risk to steady-state operation and safety. Nevertheless, a quantitative chemical analysis of the co-deposited layer on PFCs is a critical task. Laser-induced breakdown spectroscopy (LIBS) is a promising technique for assessing the deposition of impurities on PFCs, enabling in-situ monitoring, real-time analysis, and simultaneous detection of all elements. However, the accuracy of the quantitative analysis is affected by the effects of the material matrix. Diffuse reflectance spectroscopy (DRS) provides complementary information about the material matrix, such as optical properties. Herein, we present a novel data fusion model of LIBS and DRS techniques, providing more accurate results compared to the models based on each of these single methods. 66 standard samples were prepared to simulate the impurities deposited on PFCs in EAST. 15 samples were randomly selected as prediction set, while the remaining 51 samples were used as calibration set. The average relative errors of the data fusion model for W, Li, Fe, and O were 2.23, 1.95, 2.08, and 0.39%, respectively. Compared with the single LIBS data model, the root mean square errors of prediction(RMSEP) of four elements W, Fe, Li, and O in the data fusion model are reduced by 19.4, 18.5, 21.4, and 20.9%, respectively. The results show that the fusion of LIBS and DRS data enables an improved quantitative analysis of co-deposited impurities on PFCs.