A local fission matrix correction method for heterogeneous whole core transport with RAPID

Abstract

The recently developed RAPID code uses a unique pre-calculated fission matrix method in order to solve the whole-core eigenvalue problem very quickly and accurately. However, in strongly heterogeneous cores, errors are introduced at the boundary between different assembly types. In this paper, two methods are discussed that uses a set of small, 2D, 2×2 assembly fixed source fission matrix calculations in order to correct the RAPID 3-D whole-core fission matrix. The methods are applied to the BEAVRS benchmark hot zero power case with 1.6%, 2.4%, and 3.1% enriched assemblies with varying amounts of burnable absorbers. The standard and corrected RAPID methods are compared to a Serpent reference case on this highly heterogeneous core. Compared to the standard, the locally-corrected RAPID drops the 2D RMS pin-wise fission source error from 6.3% to 0.54% (compared to a Serpent RMS uncertainty of 0.09%). The 3D, pin-wise, 100 axial level RMS error drops from 6.6% to 2.2% (Serpent RMS uncertainty 1.9%). The k-eigenvalue difference drops from 157 pcm to 26 pcm (Serpent uncertainty 0.5 pcm). In order to obtain these levels of uncertainty, the Serpent reference required a calculation time of 80 h on 20 cores, compared to a RAPID time of 2.4 min on the same system. Though the RAPID database requires roughly 16 h on 20 cores, it can be used for any other RAPID calculations without performing any new Monte Carlo calculations.