Isotope Ratio Features For Classification Of Dissolution Events Using Effluents Measurements

Gamma-ray measurements of effluents at the off-gas stack of a radiochemical processing facility are used to train a set of classifiers to identify events associated with radioactive material dissolution and processing. Datasets from two Pu-238 subsequent campaigns involving the dissolution of irradiated Np-237 targets of possibly different source levels are utilized. Gamma-ray count rates estimated through spectral analysis for isotopes of Iodine, Krypton, and Xenon, have been utilized as classifier features in the past. The cumulative distribution functions (cdf) of select isotopic count ratios during the dissolution of material and other periods show separability, indicating their potential use as classifier features. The isotopic ratios, when properly selected, have some inherent invariance and could be more stable indicators of dissolution source activity than (solely) count rate estimates of individual isotopes. Hence classifiers using them as features are expected to be less sensitive to count rates which may vary across dissolution campaigns especially given the difference in irradiation time and cooling time. Multiple classifiers are trained to detect the dissolution events using seventeen isotopic count rates. Additionally, classifiers are trained using six isotope ratios as features and the results of various classifiers are compared. Under 5-fold cross validation, the classification performance with isotope ratios closely matches the performance with count rates, for example, within 0.53% of the classification error for the best classifier and 0.03% for the classifier-fuser.