During electrorefining, fission products, such as Sr and Cs, accumulate in a eutectic LiCl-KCl molten salt and degrade the efficiency of the separation process by generating high heat and decreasing uranium capture. Thus, the removal of the fission products from the molten salt bath is essential for reusing the bath, thereby reducing the additional nuclear waste. While many studies focus on techniques for selective separation of fission products, there are few studies on processing monitoring of those techniques. In-situ monitoring can be used to evaluate separation techniques and determine the integrity of the bath. In this study, laser-induced breakdown spectroscopy (LIBS) was selected as the monitoring technique to measure concentrations of Sr and Cs in 550°C LiCl-KCl molten salt. A laser spectroscopic setup for analyzing high-temperature molten salts in an inert atmosphere was established by coupling an optical path with a glove box. An air blower was installed between the sample and lenses to avoid liquid splashes on surrounding optical products caused by laser-liquid interaction. Before LIBS measurements, experimental parameters such as laser pulse energy, delay time, and gate width were optimized for each element to get the highest signal-to-noise ratio of characteristic elemental peaks. LIBS spectra were recorded with the optimized conditions from LiCl-KCl samples, including individual elements in a wide concentration range. Then, the limit of detections (LODs) for Sr and Cs were calculated using calibration curves, which have high linearity with low errors. In addition to the univariate analysis, partial least-squares regression (PLSR) was employed on the data plots to obtain calibration models for better quantitative analysis. The developed models show high performances with the regression coefficient R2 close to one and root-mean-square error close to zero. After the individual element analysis, the same process was performed on samples where Sr and Cs were dissolved in molten salt simultaneously. The results also show low-ppm LODs and an excellent fitted regression model. This study illustrates the feasibility of applying LIBS to process monitoring in pyroprocessing to minimize nuclear waste. Furthermore, this high-sensitive spectroscopic system is expected to be used for coolant monitoring in advanced reactors such as molten salt reactors.

Molten salt reactors and pyroprocessing are widely considered for various nuclear applications. The main challenges for monitoring these systems are high temperature and strong radiation. Two harsh environments make the monitoring system needs to measure nuclides at a long distance with sufficient resolution for discriminating many different elements simultaneously. Among available methodologies, laser-induced breakdown spectroscopy (LIBS) has been the most studied. The LIBS method can provide the required stand-off and desired multi-elemental measurable ability. However, the change of the level for molten salts induces uncertainty in measuring the concentration of the nuclides for LIBS analysis. The spectra could change by focusing points due to the different laser fluence and plasma shape. In this study, to prepare for such uncertainties, we evaluated a LIBS monitoring system with machine learning technology. While the machine learning technology cannot use academic knowledge of the atomic spectrum, this technique finds the new variable as a vector from any data including the noise, target spectrum, standard deviation, etc. Herein, the partial least squares (PLS) and artificial neural network (ANN) were studied because these methods represent linear and nonlinear machine learning methods respectively. The Sr (580–7200 ppm) and Mo (480–4700 ppm) as fission products were investigated for constructing the prediction model. For acquiring the data, the experiments were conducted at 550°C in LiCl-KCl using a glassy carbon crucible. The LIBS technique was used for accumulating spectra data. In these works, we successfully obtained a reasonable prediction model and compared each other. The high linearities of the prediction model were recorded. The R2 values are over 0.98. In addition, the root means square of the calibration and cross-validation were used for evaluating the prediction model quantitatively.