There is ongoing interest in using Artificial Intelligence (AI) tools to accelerate these types of calculations.
Coupled thermalhydraulic/neutronic calculations provide access to many variables of particular interest for developing a metamodel and its optimization, as they present constraints from an economic, safety, and licensing perspective. Of particular interest are maximum power, duration of the cycle, and boron concentration.
The various computational programs used for this study were PARCS (v3.4.2) and PATHS (v1.08). PATHS is a thermalhydraulic code that can be coupled with PARCS. The latter calculates the neutron part of the problem. PATHS uses a drift flux model for two-phase flow, rather than the typical six-equation two-phase flow model used in system codes. Matlab and Python are also used for tasks such as the automatic generation of input files for PATHS and PARCS, with a random distribution of the loading pattern.
The random generation of a certain number of different fuel configurations must meet several criteria (correct number of each fuel element type, impossibility of two identical fuel elements being side by side, etc.). The outputs of the automatic post-processing programs of the huge amount of results obtained (Boron Concentration, duration of cycle, , PCT, Maximum Power, etc.), make up the database that serves as input for the training and verification of a neural network (NN) (Keras), whose objective is to obtain a metamodel with which to accelerate the calculation of optimization in future works of the optimal fuel configuration based on certain criteria.
