b'Three-dimensional finite element model generated by injecting random elliptical pores to represent both pores and cracks in addition to tristructural isotropic particles distributed based on a fitted distribution coming from the x-ray computed tomography scans. (a) space exploration and training data generation. (b) experiment-informed modeling from Advanced Gas Reactor compacts. (c) using Raven and Python to build an Abaqus model that matches the tristructural isotropic particles distribution. (d) resulting distribution to be compared to (b). (e) resulting finite element model to be used in training the machine learning models. (f) reference simulation of the finite element model.32'