Development and integration of a tool for physics-based shape and topology optimization in the MOOSE multiphysics simulation framework

We developed a C++ computational tool for physics-based shape and topology optimization and integrated it into the MOOSE multiphysics simulation framework. The tool implements combinatorial and discrete optimization algorithms, and includes performance enhancements like solution caching, tabu lists, and multi-run restarts. We demonstrate the tool's flexibility with two applications that utilize different MOOSE physics modules. We implemented a Simulated Annealing search engine in our new tool. The first application is novel, adopting a two-dimensional Cartesian geometry representation of a pin-cell aiming for the optimal distribution of fuel and moderator material on a fixed mesh that maximizes neutron multiplication and coolant's hydraulic diameter. Constraints were applied to the search procedure, and we explored their effect on the realized optimal shape, identifying a set that includes preliminary manufacturability constraints and that produces a Cartesian approximation of annular fuel pins, previously proposed by physical intuition. The second is a traditional PWR fuel shuffling application at the full-core scale aiming at minimizing peak power over the core. This capability was not available in MOOSE and is used to illustrate the flexibility of our new optimization capability to address other types of discrete optimization demands. In our test case, we obtained a 1250 pcm improvement in the multiplication factor and a reduced assembly power peaking of more than 30% relative to the initial unoptimized state comprising an IAEA-2D benchmark-based core. The loading patterns generated were consistent with established literature. This work enables multi-scale reactor design improvements, from the individual fuel pin level to the full core level. Future work will leverage MOOSE's multiphysics capabilities to execute coupled-physics optimization exercises.