The Use of Computational Adjoint Design Processes in Optimization of Olympic Luge Sleds

The combination of CFD with adjoint design processes offers the prospect of automated optimization of aerodynamic shapes. In this work we investigate the application of an adjoint design process to the minimization of aerodynamic drag on Olympic luge sleds. We compare two approaches: 1. Using the adjoint process to automate changes in the aerodynamic shape; and 2. Using the adjoint process to inform human choice in the changes in the aerodynamic shapes. The computational results are compared to wind tunnel drag measurements. The results indicate that the automated adjoint process resulted in shapes that increased, rather than decreased, the drag in all cases. This appeared to be due to the inability of the automated process to add or remove material from the shape. Specifically, the total amount of the physical material was conserved resulting in warped versions of the original shape. The trends in the automated shape were qualitatively similar to the human designed shapes, which were shown to decrease aerodynamic drag. This indicates that automated computational adjoint design processes are currently capable of informing by not automating the design process.