PNNL

Abstract

The fission time projection chamber (fissionTPC) was developed to provide precise neutron-induced fission measurements for several major actinides. As fission fragments lose energy in one of the gas volumes of the fissionTPC, energy loss information is captured and may be used to determine fission product yields as the stopping power of an ion is dependent on atomic number. The work presented demonstrates the ability to apply machine learning techniques for Bragg curve classification. A set of one million energy loss curves for 24 different fission fragment elements was generated using common stopping power software. A Resnet architecture optimized for 1-D data was used to train, test, and validate a model for light and heavy fission fragments using the simulated data. The resultant classification accuracy for the light and heavy fragments indicates that this could be a viable method for elemental classification of data from the fissionTPC. Further development of this technique is recommended.