Using Feature-Based Mesh Adaptation to Improve the Adjoint Optimisation of Transonic Compressor Blades

Abstract

The benefit of mesh adaptation to improve the optimisation process of turbomachinery components is here demonstrated for the first time. Mesh movement is used to automatically cluster and align the cells with significant flow features such as shocks, shock-induced separation and wakes for every geometry tested during a transonic compressor blade optimisation. Using mesh movement means that the same size grid is used while significantly improving the accuracy of the simulation and resulting adjoint gradients. A method is demonstrated to automatically carry out feature based mesh movement during every step of an adjoint optimisation process. Optimisations are carried out using the adaptation method and also using the starting mesh as a comparison. It is shown that (when tested on a very fine grid) the adaptation-optimisation process results in a better design, due to more accurate flow and gradient prediction throughout the optimisation process. A cost breakdown of the process is given to show that using adaptation during the optimisation process only increases the overall optimisation cost by a small amount, but results in greater efficiency of the final blade design.