Coupled High-Fidelity Aero-Structural Optimization of the Wing Twist and the Wing Structure of the Common Research Model

Gradient based methods in multidisciplinary optimization are promising to exploit remaining potential in aircraft design. To further improve the aerodynamic performance of the NASA CRM transport aircraft, an aerostructural optimization process has been implemented at the DLR. The analysis accounts for the physical interaction between aerodynamic forces and elastic deformation of the structure. Results close to reality are expected, as the structural model of the complete aircraft configuration leans on industrial standard and the flow is solved at a cruise Mach number of 0.85 using RANS-CFD computations. The necessary gradients are computed using the coupled adjoint approach. However, the gradient computation is only implemented for aerodynamic cost functions and aerodynamic design variables. The structural model is generated within each design iteration. In a three-step approach the model is sized with initial loads based on a rigid structure, then loads based on the flexible structure are computed which are then applied to minimize the structural mass in a sub optimization. To compute gradients for the structural cost functions, a finite difference scheme has been implemented, which treats the structural design process as a black box. Additionally, the structural design process is improved by looping the sizing and optimization executions with the loads computation. For demonstration purposes, the wing's twist is parameterized and the parameters are utilized as global design variables. However, it was noticed that the structural gradients are computed robustly only if the structural optimization is not included. For initial optimization runs, the product of the maximum root bending moment of all load cases and the drag coefficient at cruise flight is used as cost function. The maximum root bending moment could be reduced by 12.22 % which caused a reduction of the wing's mass by 2960 kg. However, the aerodynamic performance decreased drastically from 18.86 to 17.30. Although the cost function did not yield a balanced multidisciplinary optimization, the structural gradient computation could be successfully examined and is now ready to be used for comparable optimization tasks