Restoring important process information from complex optical spectra with MCR-ALS: Case study of actinide reduction in spent nuclear fuel reprocessing

In industrial processes, information on the transformation kinetics of various components is of ultimate importance for technological control. When metals are involved in a technological process, optical spectroscopy, which yields spectra with characteristic bands for metals is an obvious candidate for system monitoring. However, while being simple in the lab, optical spectroscopy often fails to provide practical results when installed in on-line mode at an industrial site. This failure is due to numerous chemical and physical effects that complicate the observation and analysis of real multicomponent media. Among these effects are the presence of interfering species, gas formation, non-transparency, temperature variations, etc. which make direct selective speciation of metals difficult. All these issues are faced when addressing the problem of on-line monitoring in PUREX (Plutonium Uranium Extraction), a technology intended for reprocessing of spent nuclear fuel, in order to recover uranium and some actinides for further use. In the current work, four sets of solutions closely mimicking real media of the PUREX stage devoted to reduction of actinides were studied with optical spectroscopy, and four different concentrations of nitric acid were investigated. As initial optical spectra were of very poor quality, pre-processing of the raw data was required in order to remove the spectral variation that is uncorrelated to the process information. A penalized least squares procedure was used for baseline correction and Savitsky-Golay smoothing was applied to improve the signal-to-noise ratio. Partially constrained Multivariate Curve Resolution-Alternating Least Square (MCR-ALS) was then applied to extract important process information such as the time required for maximum accumulation of reduced forms of plutonium and neptunium.