SuperFINIX: A flexible-fidelity core level fuel behavior solver for multi-physics applications

VTT Technical Research Centre of Finland Ltd (VTT) is currently re-building its computational reactor analysis framework. The development of this new framework, Kraken has recently started and the framework builds on in-house developed solvers such as the neutronics solvers Serpent (Monte Carlo continuous energy) and Ants (multi-group nodal) as well as the fuel behaviour solver FINIX (traditional 1.5-dimensional rod representation). The separate solvers communicate with a central multi-physics driver that transfers field data and controls the general solution flow. The individual solvers are coupled in a modular fashion, meaning that the solver for a specific task can be switched to one with a higher fidelity or a lower running time if needed. While the solvers for the neutronics and for thermal hydraulics consider the whole reactor core as their solution domain, the fuel behaviour solver FINIX only models a single fuel rod. As current light water reactor cores contain tens of thousands of individual fuel rods, a separate tool that distributes the task of solving the core level fuel behaviour to tens of thousands of FINIX instances is needed. An additional consideration in the context of the Kraken framework is the different fidelities supported by the neutronics solvers. While Serpent can tally the pin-power distribution using a detailed axial-radial meshing inside each individual fuel rod if needed, the nodal code Ants is limited to a much coarser resolution. SuperFINIX is a recently created wrapper code for multiple FINIX-instances that accepts the core level power distribution from the multi-physics driver, distributes it to FINIX-instances describing the fuel rods in the core, executes the FINIX solvers, collects the fuel behaviour result and passes it back to the multi-physics driver. While this simple task is important in itself, SuperFINIX is tailor made for variable-fidelity multi-physics coupling so that the same SuperFINIX instance can accept power distributions and provide fuel temperature distributions at multiple different levels of fidelity. In the common case, each individual fuel rod is initialized individually and any coarser level power distributions that are provided (e.g. assembly or sub-assembly level power distributions) are passed to the individual fuel rod instances. The fuel behaviour solution is solved for each rod separately and can be provided as axial-radial distributions for each individual rod or homogenized to some lower fidelity, e.g. pin-wise temperatures without radial dependence or assembly-wise (or sub-assembly-wise) temperatures). This paper will describe SuperFINIX including its currently available temperature homogenization procedures and the parallel execution of the individual rod-level FINIX solvers. The use of SuperFINIX will be demonstrated in a minicore case, coupled to the Serpent Monte Carlo code (and potentially to the Kharon channel level thermal hydraulics solver). Comparisons will be made between coupled solutions obtained using different fidelities for the fuel behaviour feedback (different fidelities of fuel temperature) and using different approaches for fuel temperature homogenization.