KOUMOUTSAKOS, Petros; MUELLER, Sibylle; WALTHER, Jens: Biologically inspired optimization algorithms for fluid dynamics

For centuries, engineers have sought inspiration from nature in devising their creations. We are interested not only in developing systems based on natural creations (biomimetics) but also in developing algorithms for the systematic optimisation of engineering problems. We are developing optimisation strategies based on evolutionary behaviour of biological systems ranging in scales from bacterial chemotaxis to social human organisations. Our goal is to develop evolutionary algorithms depleted to the extend possible of ad-hoc heuristics, by adopting biological concepts of processes such as learning, mutation, and selection. These methods are being applied on technically relevant optimisation problems such as drag reduction in bluff body flows, mixing processes and turbine blade design. The robustness and portability of these approaches enable the development of optimisation tools not only in a simulated environment but also in real world applications in an experimental and industrial set-up. In this talk we will discuss results from the application of these strategies to the optimization of bluff body flows and the implementation of Evolution Strategies in a Pareto optimisation of a realistic turbine design. The talk will discuss some of our present research areas in (i) the development of meta-optimisation techniques based on concepts such as re-inforcement learning and (ii) the application of evolutionary concepts to the study and design of systems emerging in fluid (life in moving fluids)