Efficient Multi-Resolution Approaches for Exploration of External Aerodynamic Shape and Topology

This paper uses previous developments in multi-fidelity shape parameterisation and aerodynamic topology optimisation to develop a framework which brings significant performance enhancements to topology optimisation cases. The combined framework comprises: a multilevel subdivision shape parameterisation used in combination with an adjoint solver and gradient-based optimiser for robust high-fidelity local shape optimisation; and a restricted snake volume of solid parameterisation and differential evolution search algorithm for flexible coverage of a global topological design space. Several key challenges arise when integrating global and local optimisation methods and solutions are developed and presented here. The combined framework is benchmarked on a challenging constrained supersonic drag minimisation problem exhibiting multi-body solutions, discontinuities and multi-modality. Configuration of the integrated method is selected and validated using a test-based methodology and the resulting algorithm demonstrates significant objective improvements when compared to the individual local and global methods applied separately. Moreover the combined framework overcomes the slow convergence of the global optimisation and thereby greatly expands the capability for exploring topological design spaces.