The Woodruff School

SUMMARY

Monte Carlo (MC) codes are widely used for the accurate modeling of neutron transport in nuclear reactors. However, the efficient inclusion of the thermal-hydraulic (TH) feedback within the MC calculation sequence is still an open problem. Among the techniques proposed to solve this problem is the utilization of stabilized stochastic Picard iteration in conjunction with a low-order acceleration step. In this thesis, a novel approach to perform this reduced-order prediction step and therefore accelerate the convergence of the Picard scheme. The prediction relies on the use of generalized transfer functions to predict the change in macroscopic cross-sections due to spatial variations in TH conditions. This method may also be used to improve the fidelity of traditional 2-dimensional cross-section generation procedure that use single valued T/H values, by accounting for the spatial dependence of T/H conditions and their effect on the cross-section values.

SUBJECT:	Ph.D. Proposal Presentation

BY:	Stefano Terlizzi

TIME:	Monday, December 16, 2019, 10:00 a.m.

PLACE:	G. H. Boggs Building, 3-47

TITLE:	On-the-fly Generation of Spatial Transfer Functions for Efficient Monte Carlo-T\H Coupled Calculations

COMMITTEE:	Dr. Dan Kotlyar, Chair (NRE) Dr. Bojan Petrovic (NRE) Dr. Weston M. Stacey (NRE) Dr. Pavel V. Tsvetkov (Texas A&M, Department of Nuclear Engineering) Dr. Joel A. Kulesza (Los Alamos National Laboratory)

SUMMARY Monte Carlo (MC) codes are widely used for the accurate modeling of neutron transport in nuclear reactors. However, the efficient inclusion of the thermal-hydraulic (TH) feedback within the MC calculation sequence is still an open problem. Among the techniques proposed to solve this problem is the utilization of stabilized stochastic Picard iteration in conjunction with a low-order acceleration step. In this thesis, a novel approach to perform this reduced-order prediction step and therefore accelerate the convergence of the Picard scheme. The prediction relies on the use of generalized transfer functions to predict the change in macroscopic cross-sections due to spatial variations in TH conditions. This method may also be used to improve the fidelity of traditional 2-dimensional cross-section generation procedure that use single valued T/H values, by accounting for the spatial dependence of T/H conditions and their effect on the cross-section values.