A memory-efficient neutron noise algorithm for reactor physics

Abstract

The neutron noise equation contains a fixed source term where the neutron angular flux is present. Deterministic noise solvers have tended to solve for and store this angular flux term in a preprocessing step. This can be a substantial memory burden as the angular flux is a function of space, energy, and angle. This can limit these solvers to smaller problems and can limit the obtainable angular resolution. This paper proposes solving the neutron noise equation and the standard eigenvalue neutron transport equation simultaneously, reducing the memory footprint of a noise solve while making frequency-domain method of characteristics solvers practical. This approach is demonstrated using the random ray method but is applicable to any sweep-based methods. Using the random ray method also has the advantage of exact geometry representation, no ray effects, and cheaper transport sweeps, all of which have been problematic. These advantages are illustrated on several noise problems.