Data-driven reduced order models of sloshing in horizontal cylindrical tanks

The use of alternative fuels for future aerial or naval mobility poses several challenges, particularly for liquid fuels such as cryogenic hydrogen. When stored in reservoirs and subjected to external excitation, they experience sloshing, the movement of the free surface. This phenomenon is particularly critical from both an inertial perspective, as forces are exerted on the tank walls, and from a thermodynamic perspective, as cryogenic liquids suffer phase changes, leading to fluctuations in gas pressure inside the tank. Investigating these phenomena and understanding their underlying dynamics is essential for developing diagnostic tools and creating a digital twin, enabling control techniques to be implemented to manage these fluctuations and optimize them for our purposes. This work aims to investigate the feasibility of identifying a reduced-order model (ROM) of sloshing in horizontal cylindrical tanks. The proposed approach is to identify the model from 2D computational fluid dynamics (CFD) simulations in OpenFOAM. The relevant testing conditions were selected from the scaling of the problem according to the relevant non-dimensional numbers. Then, the closure terms of the ROM are identified from the numerical database through a data-driven inverse method. This approach requires a non-linear optimization step to minimize the deviations between the model predictions and high-fidelity CFD, while the ROM’s closure terms are expressed via parametric functions with varying degrees of complexity. As a major outcome of this work, we assess the performance of such functions for the most relevant sloshing condition.