Analytical spectral dependent partial least squares regression: a study of nuclear waste glass from thorium based fuel using LIBS

Abstract

This work demonstrates the use of laser induced breakdown spectroscopy (LIBS) for elemental characterization of power reactor thoria reprocessing facility waste glass. Seven waste constituents (Th, Fe, Cr, Ni, Al, F and Sr) in the glass were simultaneously determined by this method. A multivariative PLSR calibration approach was preferred due to the complex emission spectrum of the waste glass. Investigations were carried out to optimize the experimental conditions such as the number of laser shots accumulating per LIBS spectra, laser energy and acquisition time delay. The optimization of factors for minimizing errors in the PLSR algorithm was investigated for Th, Fe, Cr, Ni, Al, F and Sr. Spectral truncation to only utilize relevant pixel intensities based on the PLSR coefficient guidelines was found to increase the analytical capability of the PLSR analysis. The optimum PLSR coefficient threshold of individual elements was calculated based on the SEP variation. The average %RSD (2σ) was found to be 5.3%, 5.1%, 7.1%, 8.1%, 10.4%, 7.7%, and 6.4% for Th, Fe, Al, F, Sr, Ni, and Cr, respectively.