Publications

Aerodynamic shape optimization has become well established, with designers routinely performing wing and full aircraft optimizations with hundreds of geometric design variables. However, increased geometric design freedom increases optimization difficulty. These optimizations converge slowly, often taking hundreds of design iterations. In addition, designers have to manually scale design variables through trial and error to achieve a well-behaved optimization problem, which is tedious and time-consuming. In this work, we propose a sensitivity-based geometric parametrization approach that maps the design space onto one better suited for gradient-based optimization while keeping the same optimization problem. At the same time, the process can automatically determine design variable scaling so that the new optimization problem can be solved more effectively. We demonstrate the approach on two aerodynamic shape optimizations and show improved terminal convergence trends compared to the traditional approach, without requiring manual adjustments to the design variable scaling.