Reconstructing flexible body vortex-induced vibrations using machine-vision and predicting the motions using semi-empirical models informed with transfer learned hydrodynamic coefficients

This work assesses the validity of transfer learning the hydrodynamic coefficient database, consisting of the added mass and lift coefficients, applicable to flexible bodies undergoing vortex-induced vibrations. Specifically, the hydrodynamic coefficient database learned on data collected by Braaten and Lie (2005) are used to predict the motions observed during in house bare riser model experiments at the MIT Towing Tank. A fully immersed vertical flexible riser model with a length-to-diameter ratio of 145 is towed at different flow speeds and top tensions. Motion is tracked using underwater cameras and the motions are reconstructed using a machine-vision framework eliminating the need for expensive sensing hardware. The vibration amplitude, frequency, and mode shape are determined and the results are compared with those in the literature. Finally, blind predictions of the in-house observed experiments are made using the software VIVA informed with transfer learned hydrodynamic coefficients learned on the experiments by Braaten and Lie (2005).