Shape optimisation to enhance flow-induced vibration of a cylinder using Bayesian optimisation

Bayesian optimisation with Gaussian process regression was performed to optimise the shape of an elastically mounted cylinder undergoing transverse flow-induced vibration. The vibration amplitude and mean power coefficient were obtained from two-dimensional numerical simulations, with Reynolds number $Re = 100$ . First, shape optimisation was performed to maximise the amplitude of undamped vibrations. The optimised shape was found to be a thin crescent cylinder aligned perpendicular to the oncoming flow. The optimised shapes exhibited simultaneous vortex-induced vibration and galloping, a response which was not observed for other cylinder geometries at the same Reynolds number. Shape optimisation was also performed to maximise the power coefficient, where the power generation device was modelled as a linear damper. The power-optimised cylinders were also thin crescents, but with greater curvature compared with the amplitude-optimised cylinders. Compared with a circular cylinder, improvements in the power coefficient and efficiency of up to $523\,%$ and $152\,%$ , respectively, were obtained.