Source Term Estimation Via Combined Sparse Convex Optimization And Maximum Likelihood Estimation For Nuclear Material Accounting.

In this work we investigate the use of sparse convex optimization and maximum likelihood estimation for the specific problem of source term estimation for neutron sources. We simulate an experimental set up at Los Alamos National Laboratory (LANL), consisting of an array 3He neutron detectors located and glove boxes containing nuclear material of variable intensity and number. We demonstrate that under the correct conditions, the correct location, strength and number of sources can be recovered via this method, without prior knowledge of the number of sources present in the experiment. We investigate the effect of background strength, detector configuration, and optimization constraints on the robustness of the solutions obtained via our algorithm. Based on these factors, we identify “feasibility” regions for our algorithm, and explore how an adversary may exploit knowledge of the detector array configuration to increase errors in source term estimation.