[1C13-18] Analytical Technique for Fuel Debris

[1C18] Development of Quick and Remote Analysis for Severe Accident Reactor-9

(6) Analysis of Laser-Induced Breakdown Spectra using Machine Learning -2

*Katsuaki Akaoka¹, Takahiro Karino¹, Hironori Ohba¹, Ikuo Wakaida¹ (1. JAEA)

Keywords:fuel debris, machine learning, LIBS, laser, quantitative analysis, uranium, plutonium

We have been evaluating quantitative analysis by machine learning as analytical method of Laser Induced Breakdown Spectrometry (LIBS) for in-situ analysis of fuel debris and other materials generated by the Fukushima Daiichi Nuclear Power Plant accident. In our previous presentation, we evaluated prediction lines for interpolation/extrapolation and noise effects on spectral intensity using spectra of U/Pu mixture samples obtained by LIBS. In this presentation, we report on the evaluation of the prediction line when the wavelength of the spectrum deviates from the training data.