The final proceedings of our work at the Summer Program 2024 at Stanford University’s Center for Turbulence Research (CTR) has been published.
This work presents an efficient framework for shape optimization to control flow instabilities and coherent structures in laminar and turbulent flows by combining a Bayesian optimization approach with adjoint-based gradient information (BOA). Linear stability and resolvent analyses, implemented in FELiCS, are used to yield physically meaningful cost functions. In addition, adjoint methods are employed to compute the sensitivity of the cost functions to shape parameter changes. These physics-based function and gradient information are embedded into a global Bayesian optimization framework, which also handles uncertainties that may arise from the provided data or linear model. The BOA framework is tested on the task of designing a hydrofoil to control the wake instability and it is compared against a gradient-free Bayesian optimizer and a purely gradient-based method.
Link to paper: https://web.stanford.edu/group/ctr/ctrsp24/ii05_MULLER.pdf