Computational Challenges of Fusion Neutronics for ITER Ports

Abstract

This paper elaborates computational challenges tackled for providing neutronics service supplied for developing the design of the Diagnostics Equatorial and Upper Port Plugs (EPP and UPP). The aim was to guide and assist the EPP and UPP design developers with optimal shielding solutions which are characterised of maintain the diagnostics purposes of the systems together with adequate radiation shielding performance. The target parameter for the shielding optimization was the minimum of Shut-Down Dose Rate (SDDR) inside the interspace between the port back-side and ITER bioshield. This aim was reached by parametric neutronic analyses of the shielding geometry and material composition, mitigating direct streaming of neutrons from the plasma by arranging the labyrinths and horizontal rails. Variation of many geometrical parameters of the labyrinths was possible only by applying the high performance parallel computations with MCNP5 using pure MPI and hybrid OpenMP/MPI parallelization techniques on several available supercomputers. MCNP5 parallel performance assessments were carried out to find an efficient way to run the code in a parallel regime. It was found a strong scaling (up to 4096 cores) performance of the MCNP5 jobs running with analogue Monte Carlo sampling and weak scaling for the tasks with biased sampling as a variance reduction technique, such as the MCNP5 intrinsic weight window generator. Deep penetrating radiation in the complex ITER tokamak geometry combined blocks of strong attenuation of the radiation together with the void gaps along which the particles are streamed freely contributes to computation challenges of radiation transport.