An adaptive COMET method

Abstract

This thesis presents a formulation for an adaptive COMET method for solving whole reactor eigenvalue and flux distribution problems using a varying flux expansion at mesh interfaces. While COMET solutions have enjoyed accuracy on par with Monte Carlo techniques with a computational efficiency several orders of magnitude greater than stochastic methods, it was desired to extend the efficiency of the method further. Improved efficiency is obtained by allowing the flux expansion at mesh interfaces, which was previously held constant throughout a whole problem, to adapt to different expansion orders depending upon mesh composition and spatial effects due to neighboring meshes. To test the method, two benchmark problems were solved using the standard and adaptive COMET solution methods: the C5G7 benchmark problem and a pressurized water reactor benchmark with mixed-oxide (MOX) fuel assemblies. In both benchmark cases, three different configurations for different insertion of control rods were considered. For all cases, the agreement between the standard and adaptive COMET solutions was excellent, with eigenvalue agreement being 3 pcm or less and average pin fission errors being much less than 0.5% in all cases. Increases in computational efficiency by factors of 2.1 to 3.6 were observed. The strong performance of the adaptive method implies that it can be used to obtain accurate solutions to reactor problems with more efficiency than the standard COMET method.