Statistical Analysis of Neutron Relative Biological Effectiveness

Corvid Technologies is set to advance DTRA’s predictive capabilities regarding the impact of nuclear weapons by developing a comprehensive database that catalogs historical experiments on neutron Relative Biological Effectiveness (RBE). This initiative will enhance military and civilian readiness against nuclear threats by providing a Gaussian Process (GP) model that estimates neutron RBE values and their experimental uncertainties. The project’s Phase I involves creating a relational database to characterize the RBE distributions from past RBE experiments, formulating a manifold Gaussian Process regression for statistical analysis, and designing a prototype deep learning model for future efforts utilizing synthetic neutronics data. This approach will enable the integration of experimental uncertainties into future machine learning applications, thereby refining the accuracy of radiation effect predictions.