Tank ships transporting LNG (liquified natural gas) or rockets using cryogenic fuels are exposed to excitations e.g. stemming from the sea waves or the atmosphere. If these excitations are close to the resonance frequency of the liquids in the tanks, strong wave amplifications can occur and may become critical to these systems. For the structure design it is therefore essential to predict these resonances. However, most of them are strongly nonlinear and their prediction extremely challenging.

Our group performs fundamental experiments in a rectangular tank and elucidates the nonlinearity of the flows in detail. We have discovered, for example, that the phase-lag between the driving and the sloshing waves is crucial for predicting the dynamics. In cooperation with the Haller group (ETH) Zurich our observations stimulate the development of predictive models, the spectral submanifold theory.

This method combines dynamical systems theory with machine learning and allows for the first time the correct prediction of non-linearizable flow resonances (e.g., including hysteretic effects).