Implementation of machine learning algorithms for detecting missing radioactive material

The detection of missing radioactive material is an important capability for safeguards measurements. Gamma ray signatures provide sample information, but interpretation is complicated by measurement environments. To determine if machine learning is a viable analysis option, three algorithms are applied to gamma ray detection data to assess their success at identifying missing sources. Preliminary results demonstrate that these algorithms can predict the number and location of missing sources on simple models of spent fuel assemblies. In addition to simulated experiments, a study to investigate if the algorithms can be trained with simulated data and tested on measured data is presented.