Bayesian optimization of curved grooves for reducing flow-induced loads on a circular cylinder

A novel passive flow control for flow-induced load reduction on a circular cylinder is introduced by incorporating two symmetrically arranged curved grooves. Here low Reynolds numbers (Re, based on the cylinder diameter D) ranging from 100 to 180, where the flow remains two-dimensional, are investigated numerically by solving two-dimensional incompressible Navier–Stokes equations using a finite volume method with 2nd order accuracy both in time and space. The Bayesian Optimization with Gaussian Process Regression method has been successfully applied to this flow configuration for the first time, demonstrating high efficiency in identifying the optimal angular width (θw=90.9°) and angular position (θp=88.8°) of the groove for maximum drag reduction. These optimal values were found within the broad admissible ranges of θw∈[20°,120°] and θp∈[65°,115°], measured clockwise from the cylinder front. We found that i) the optimized groove configuration generates two small corner vortices near groove corners, which effectively weaken vortex shedding and displace the wake farther downstream compared to a smooth cylinder, mainly contributing to a large pressure drag reduction; ii) the upstream corner vortices result in significant viscous drag reduction and decrease in the pressure difference between cylinder top and bottom, thus leading to a significant lift reduction; iii) both drag and lift reductions relative to a smooth cylinder grow monotonically with increased Re, peaking at 16.7% and 42%, respectively. These findings provide fundamental insights in designing effective passive flow control devices with simple geometric modifications.