2020 journal article

Energy group search engine based on surrogate models constructed with the RAVEN/NEWT/PHISICS sequence

NUCLEAR ENGINEERING AND DESIGN, 356.

co-author countries: United States of America πŸ‡ΊπŸ‡Έ
author keywords: PHISICS; NEWT; High Temperature Gas-cooled Reactors; Group structure; Adaptive sampling; RAVEN
Source: Web Of Science
Added: January 13, 2020

Transient calculations with nodal neutronics codes entail few-group energy structures. In systems other than LWRs, significant efforts are devoted to obtain satisfying group structures. The energy cut-offs available in the literature do not always match the energy boundaries available in lattice codes. This paper demonstrates an automated sequence that searches for suitable coarse-group configurations. The sequence couples the lattice code T-XSEC/NEWT for cross section generation and collapsing, the nodal code PHISICS for core calculations and the software RAVEN for variable sampling and analytical purposes. T-XSEC/NEWT receives an energy group configuration from RAVEN to generate microscopic self-shielded cross sections in a coarse format. PHISICS provides the core solution using the microscopic libraries. The performances of the group structures in the core model are stored to train a Reduced Order Model (ROM) built on-the-fly. The ROM spares the necessity to survey the large input space of all possible energy group structures, or expert judgements. The solution provided by RAVEN is a Limit Surface of group structures fitting success criteria. The approach is tested on a simplified two-dimensional HTTR core model. The Limit Surface obtained by RAVEN derives a few six-group structures fitting the HTTR model.