OpenMC's Simulated Solutions for the Kobayashi Dogleg Void Duct Benchmark

Simulations of radiation transport of neutrons through matter are a critical component of fission and fusion reactor design. Deterministic methods are favored for simple geometries, as they can be modeled and validated analytically. The new, simplified equations describe the average behavior of the neutrons throughout the entire system. Monte Carlo (MC) methods can be utilized for more complex geometries, by simulating the random paths of large numbers of neutrons to achieve highly accurate flux estimates; however, this may require expensive computational resources to achieve [1].

Variance Reduction methods introduce biasing in order to mitigate this cost due to the repeated random sampling. We discuss the weight windows variance reduction method now available in OpenMC, an open-source MC neutron and photon transport simulation code. We demonstrate this method with an evaluation of the Kobayashi Dogleg benchmark problem [2].

We compare the accuracy of OpenMC’s simulated average flux distributions against the benchmark’s deterministic results for regions with large dynamic ranges of expected neutron flux. We observe the expected reduction in statistical uncertainty in deeply shielded regions, and a possible over-estimate of neutron flux in those deeply shielded regions. Verifying the computational accuracy of OpenMC to other MC codes is crucial for improving neutronic models.