NASA Trap Wing Research Articles

Aerodynamic optimization of high-lift devices using a 2D-to-3D optimization method based on deep reinforcement learning and transfer learning

The computational fluid dynamics is the main method to evaluate the aerodynamic performance for the optimization of high-lift devices. Currently, the direct three-dimensional (3D) optimization requires significant computational resources. Additionally, the commonly used heuristic algorithms can not extract the experience of two-dimensional (2D) optimization to accelerate the 3D optimization process. In order to resolve these issues, a novel 2D-to-3D optimization method based on the coupling of Deep Reinforcement Learning (DRL) and Transfer Learning (TL) is proposed to conduct the aerodynamic optimization of the 3D high-lift devices and tested on the NASA Trap Wing model. The 2D optimization is first carried out, and then its neural networks in DRL and the optimal configuration are extracted by TL to turn into the 3D optimization. Compared with the direct 3D optimization, the proposed 2D-to-3D optimization method can result in improved aerodynamic performance for the same computational cost, or can save 51%-81% of the computational cost to obtain a similar performance.

Coupling of turbulence wall models and immersed boundaries on Cartesian grids

An improved coupling of immersed boundary method and turbulence wall models on Cartesian grids is proposed, for producing smooth wall surface pressure and skin friction at high Reynolds numbers. Spurious oscillations are frequently observed on these quantities with most immersed boundary wall modeling methods, especially for the skin friction which is found to be very sensitive to the solid surface's position and orientation against the Cartesian grids. The problem originates from the irregularity of the wall distance on the stair-step grid boundaries where the immersed boundary conditions are applied. To reduce this directional error, several modifications are presented to enhance the near wall solution. First, the commonly used interpolation for the flow velocity is replaced by one for the friction velocity, which has much less variation near wall. The concept of using a fictitious point to retrieve flow fields in the wall normal direction is abandoned and the interpolation is performed in the wall parallel plane with existing fluid points. Secondly, the velocity gradients at the approximated boundary are computed with advanced schemes and the normal gradient of the tangential velocity is reconstructed from the wall laws. To further protect the near wall solution, the normal velocity gradient and the working viscosity from the Spalart-Allmaras turbulence model are enforced by their theoretical solutions in the interior fluid close to the wall. Additionally, various post-processing algorithms for reconstructing wall surface quantities and force integrations are investigated. Other related factors are also discussed for their effects on the results. The validity of present method has been demonstrated through numerical benchmark tests on a flat plate at zero pressure gradient, both aligned and inclined with respect to the grid, as well as aerodynamic cases of NACA 23012 airfoil and NASA trap wing.

Mechanism/structure/aerodynamic multidisciplinary optimization of flexible high-lift devices for transport aircraft

Mission adaptive variable camber wing in both chord-wise and span-wise directions that can improve the aerodynamic performance during takeoff, landing and cruising flight, will be the state-of-the-art high-lift system for next generation airliners. Based on NASA TrapWing model released on the 1st AIAA CFD High-lift Prediction Workshop, a smart high-lift system with “Flexible Droop Nose & Single Slotted Hinge Flap combined with spoiler deflection & Flexible Trailing Edge Flap” is proposed in this paper. The Flexible Droop Nose is actuated by kinematic chains mechanism, the Single Slotted Hinge Flap is actuated by simple hinge mechanism and the Flexible Trailing Edge Flap is actuated by link/track mechanism.A mechanism/structure/aerodynamic multidisciplinary optimization platform based on iSIGHT software is constructed for this smart high-lift system. This platform consists of stress analysis, high-lift configuration generation, high-lift configuration structure grid generation, computational fluid dynamics and optimization algorithm modules. The optimal takeoff and landing configurations with comprehensive performance of mechanism, aerodynamic and structure is then obtained after multidisciplinary optimization. Finally, the CFD results show that the aerodynamics performance of this smart high-lift system is more effective than the original NASA TrapWing model.

The Numerical Simulation of Civil Transportation High-lift Configuration

The complex multiform flow phenomenon including boundary layer transition, lamination separation, turbulence attach line and re-lamination, boundary shock interaction and so on will appear when the air flow over the high lift configuration, which makes it very difficult to predict by numerical simulation. In order to validate the CFD code prediction capability of high lift configuration of swept wing with high aspect ratio, AIAA held two workshops of CFD high lift prediction in 2010 and 2012. The NASA Trapwing model and DLR-F11 model were chosen to research the mesh resolution, turbulence model, transition prediction and Reynolds effect by numerical simulation. Based on the published work of the1st and 2nd high lift prediction workshop, some numerical simulation research to validate the rationality of mesh generation are done in this paper. The two configurations of NASA Trapwing and DLR F-11 are simulated by solving the Reynolds-averaged Navior-Stokes equations with the S-A turbulence model. The numerical data in linear region agree very well with the experiments but the discrepancy between numerical data and experiment data near the stall region is perceptible. It can be conclude generally from the numerical results of two configurations that in the engineer design phase the Reynolds-averaged Navior-Stokes equations can give correct results of the aerodynamic characteristic with carefully generated mesh and appropriate turbulence model.

Results from the Second AIAA CFD High-Lift Prediction Workshop Using Edge

The results presented at the Second AIAA High-Lift Prediction Workshop, using the flow-solver Edge, are summarized for the DLR, German Aerospace Center F11 model. A comparative study of the results, using three turbulence models, is carried out, including the Spalart–Allmaras model, an explicit algebraic Reynolds-stress model, and a curvature correction to the explicit algebraic Reynolds-stress model. The comparisons include a grid-convergence study on a simplified model without slat- and flap-track fairings, and polar calculations including the fairings. The grid-convergence study shows relatively small differences due to different grid resolution and turbulence models, but the differences are larger than those obtained in the first workshop for the NASA trap wing. The prediction has fairly large discrepancies from experimental measurements at low Reynolds numbers for which the computations were carried out assuming fully turbulent flow. The explicit algebraic Reynolds-stress model (with or without curvature correction) tends to exaggerate flow separation and to underpredict the lift, as compared to the experiment. The Spalart–Allmaras model, on the other hand, overpredicts the lift at higher incidences. At the higher Reynolds number, the predictions agree well with the experimental measurements for different turbulence models. The curvature correction to the explicit algebraic Reynolds-stress model has an insignificant effect on the computed results in comparison with the explicit algebraic Reynolds-stress-model predictions.