We introduce FREDS, a multi-objective optimization Python framework for energy grid discretization in reactor neutronics calculations. FREDS uses sensitivity profiles derived from the Serpent2 Monte Carlo code to guide the allocation of energy groups, aiming to exploit the trade-off between accuracy and computational efficiency. Multi-objective optimization is performed using genetic algorithms, allowing FREDS to optimize the energy grids for the keff sensitivity to multiple reactions simultaneously while minimizing the number of energy groups in the mesh. We demonstrate its capabilities on both fast and thermal benchmarks, namely Godiva, Jezebel, and UAM, optimizing the keff sensitivity to fission, capture, and elastic scattering reactions. We demonstrate the application of FREDS through the optimizations performed in this work, showing its effectiveness in generating accurate and computationally efficient discretizations across a variety of reactor systems and reactions of interest, in agreement with the underlying physics.
Casas-Molina et al. (Thu,) studied this question.