Experimental Data-Driven Reduced Order Model for Non-Isothermal Vertical Sloshing

This article outlines an experimental-based methodology for the development of reduced-order models in the context of non-isothermal vertical sloshing. Given operational constraints with managing liquid hydrogen, water was employed as a surrogate fluid. A lumped capacity mathematical model featuring a time-dependent Nusselt number was derived. Leveraging experimental data, the Nusselt number was mapped onto the characteristics of the harmonic seismic excitation, including frequency and amplitude. The same data were used for the training of a reduced-order model based on time-delay neural networks to predict the Nusselt number as output from the tank vertical acceleration. Simplif ied numerical simulations can reproduce the thermodynamic response of the system and are compared with reasonable success with experimental data.