Experimental data-driven reduced-order modeling of nonlinear vertical sloshing for aeroelastic analyses

This thesis focuses specifically on the study of nonlinear sloshing effects caused by large tank motions in a direction perpendicular to the free liquid surface with emphasis on aeronautical applications. Sloshing is a phenomenon that typically occurs in aircraft tanks as they are subjected to loads caused by gusts, turbulence and landing impacts. This type of sloshing leads to a noticeable increase in overall structural damping, yet it is generally not modeled in the design phase of modern aircraft. The identification and study of such dissipative effects may enable the development of less conservative aircraft configurations in the future, allowing for increasingly lighter structures and reduced environmental impact. The present thesis proposes a combined experimental and numerical approach aimed at obtaining reduced-order models for vertical sloshing, to be subsequently integrated into aeroelastic modeling and applications for the assessment of their effects on overall performance. An experimental campaign is first carried out to characterise the nonlinear dissipative behaviour of vertical sloshing for different filling levels. Specifically, a controlled electrodynamic shaker is employed to provide vertical displacement by means of sine-sweep excitation. By exploiting vertical harmonic motion, it is shown how the frequency and amplitude of the imposed excitation significantly influence the dissipative capabilities of the sloshing liquid. The same experiment is used to create a database - with an acquisition phase that considers vertical sloshing as an isolated system - to build a neural-network-based reduced-order model. The dynamics to be modeled is considered as a black box process, leading to the identification of a surrogate model driven only by input/output signals, regardless the knowledge of the internal dynamics. In order to assess the capability of the identified reduced order model for sloshing, the same tank used to generate the training data is mounted at the free end of a cantilever beam to create a new experimental setup in which a fluid-structure interaction scenario is expected. Indeed, this experiment provides experimental data for the validation of the identified dynamic model by comparison with numerical data. The comparison is carried out using a dynamic virtual simulation model corresponding to the experiment, in which the numerical model of the beam interacts with the reduced-order model simulating the sloshing dynamics. Finally, the experimentally validated reduced-order model is used in two different aeroelastic applications - wing prototype and flying wing model - to finally predict the dissipative effects induced by vertical sloshing on the aeroelastic response. Aeroelastic response analyses under pre- and post-critical conditions showed how the vertical sloshing dynamics helps to alleviate the dynamic loads due to severe gusts while providing limit cycle oscillation beyond the flutter margin.