Speaker
Description
IFMIF-DONES facility is a deuterium-lithium (d+Li) neutron source driven by a deuteron accelerator (40 MeV and 125 mA) striking at the liquid Li target, and producing neutrons through stripping reactions to provide irradiation data for the construction of a DEMO fusion power plant. The shielding analyses of the accelerator and target demand efficient variance reduction (VR) technique, in particular the global radiation distributions. Due to the complicated particle-material interactions in the accelerator facilities with streaming paths, there are often issues raised during the generation and utilization of the weight-window mesh (WWM). The common methodology such as the deterministic-based FW-CADIS method is widely used, yet still not straightforward to handle the complex source modelling and particle streaming.
An “on-the-fly” global variance reduction technique will be presented. It is based on iterations of flux map and WWM, generating and improving the WWM on-the-fly along with the particle transport simulation. Since it is a Monte Carlo (MC) based method, no extra step is needed for geometry and source modelling and WWM generation. One common issue caused by the streaming path is that, a particle history could be extremely long due to the over-splitting of particle tracks, since the high flux gradient around the streaming path will cause the high gradient of WW. A novel dynamic WW upper bound method has been proposed to limit the splitting by increasing the WW upper bound once sufficient particles have been scored for specific mesh cells. This method has been implemented in the MC transport code MCNP and OpenMC. This method has been tested with computational benchmarks, with promising speed-up achieved compared with the analogue run and FW-CADIS WWM simulation. Similar speed-up have been observed in the application to the IFMIF-DONES shielding calculation, with neutron flux maps being produced over the 3~6 m thick concrete shielding.