Multifidelity Airfoil Optimization with Adjoints and Mesh Adaptation
Speaker: Derek Dalle, Raytheon Missile Systems
Presentation
Abstract
This talk will focus on gradient-based aerodynamic shape optimization is investigated using adaptive mesh refinement. Instead of using a mesh with a fixed cell spacing and total mesh size, the mesh is adapted to each design iteration. In the early stages of optimization, the objective function is evaluated only on the coarse mesh, and progressively more mesh adaptations are performed as the design approaches optimality.
This approach has several features that can reduce optimization time. First, the multifidelity aspect allows the early portion of the optimization to be computed on meshes that require much reduced computational resources. Second, output-based adaptive meshing allows the flow calculations to target areas of the flow that most affect the objective function, which reduces total mesh size for a given level of accuracy. Finally, output error estimates reduce the chance of over-optimizing on a coarse mesh, or under-optimizing on a fine one. The approach is tested on two airfoil optimization test cases, and it is compared in computational expense to other approaches, which include a traditional fixed-mesh optimization, optimization using a constant level of fidelity, and multifidelity optimization using uniform refinement.
Biography
Derek Dalle was a Postdoctoral Research Fellow in the Department of Aerospace Engineering at the University of Michigan with co-advisors Krzysztof Fidkowski and James Driscoll. Prior to that, Derek earned his Ph.D. in the same department studying under Professor James F. Driscoll. Currently, Derek works in the Flight Sciences department at Raytheon Missile Systems.