Plutonium(IV) quantification in acidic process solutions using partial least-squares regression applied to UV-Vis spectrophotometry

In this study, a Partial Least-Squares (PLS) regression approach mainly based on a lab-produced calibration set enhanced with a few process samples is proposed. Comparison with spectrophotometric reference offline measurements and predictions by PLS is presented on a validation set composed of real process solutions. The mean difference between the presented model predictions and offline measurements is less than 5%. The results show that this approach can be effective for process control and can successfully replace usual sample preparations and time-consuming calibration procedures if the model is constructed properly.

Automated summary

In this study, a Partial Least-Squares (PLS) regression approach is proposed, which is enhanced with a few process samples. The model predictions of PLS are compared with spectrophotometric reference offline measurements on a validation set composed of real process solutions. The results show that this approach can be effective for process control, with less than 5% difference between model predictions and offline measurements, if the model is constructed properly.