Developing Best Practices for Transonic-Truss Braced Wing Aircraft Simulation

Overview

The Boeing Transonic Truss-Braced Wing (TTBW) aircraft concept originated with Subsonic Ultra-Green Aircraft Research (SUGAR), a joint NASA/Boeing effort to develop advanced subsonic commercial transport concepts to meet NASA-defined metrics in terms of reduced noise, emissions, and fuel consumption. The Mach 0.8 variant of the TTBW delivers a nine percent reduction in fuel consumption per seat, and features a high-span, high-aspect-ratio strut-braced wing. Due to the complexity of this configuration, the development of new computational fluid dynamics (CFD) best practices was required to accurately predict aircraft performance. Researchers from NASA’s Langley and Ames Research Centers have been collaborating on testing various approaches and numerical models using the NASA-developed USM3D solver and Launch Ascent and Vehicle Aerodynamics (LAVA) framework.

"Without NASA’s high-performance parallel computing resources, these simulations could take many months to run."

Project Details

Initial efforts for the development of NASA CFD best practices for the TTBW took place with the simulation of the Mach 0.745 variant. This involved a code-to-code verification process simulating the flight configuration, as well as validation comparing against experimental results from the NASA Ames 11-by 11-foot Transonic Wind Tunnel (11-ft TWT). LAVA was used to simulate the TTBW in both free air conditions and within a computational representation of the wind tunnel model to increase simulation fidelity.

Recent efforts have been focused on simulating the Mach 0.8 variant of the TTBW, using the lessons learned from the Mach 0.745 study, which includes minimizing the differences between the experimental data and computational predictions where possible. CFD analysis of the Mach 0.8 concept included angle-of-attack sweeps for flight and wind tunnel configurations, grid refinement studies, and component build-ups to determine the influence each component has on vehicle performance and solution convergence. Recent refactorization of the LAVA flow solver prompted a comparison with the production version of LAVA to determine improvements in speed and accuracy. This required re-simulation of the TTBW in free air and within a computational model of the 11-ft TWT, which incorporated all wind tunnel components and utilized a newly implemented porous wall boundary condition that models the porous tunnel surfaces.

Results and Impact

When simulating the aircraft near the cruise angle of attack, the difference between solvers for the asymptotic value of the grid convergence study was around one percent for lift and pitching moment coefficient and one drag count in drag coefficient. For both of the tested variants of the TTBW, LAVA predicts a shift in the lift curve relative to experimental data. However, this shift is drastically reduced after geometric differences between the computational geometry and experimental geometry are removed. The refactored LAVA solver outperforms the production version of LAVA in terms of speed and accuracy, and eases the setup and post-processing of these simulations. For instance, the shape of the lift curve from the CFD simulations better matches experimental data and there is also a noticeable drag reduction when using the refactored solver, bringing the solutions closer to experimental data. Lastly, in terms of the pressure coefficient, the refactored solver does a much better job predicting the location of the shock, which had been seen in all previous simulations as a large source of load discrepancy from experimental results.

The in-tunnel simulations were run using the refactored version of LAVA with the improved wind tunnel computational model and the porous wall boundary condition, resulting in a higher fidelity simulation. The loads from these runs show an even better comparison with experimental data. For the pressure coefficient profiles, there is no significant improvement over the wing but there are some minor improvements in capturing the strut pressure profiles.

Why HPC Matters

Without NASA’s high-performance parallel computing resources, these simulations could take many months to run. The in-tunnel simulations using the production version of LAVA required eight days and 14 hours across 400 Intel Ivy Bridge cores of the Pleiades supercomputer at the NASA Advanced Supercomputing facility at Ames. Using refactored LAVA, in-tunnel simulations of double the size can run on 1,000 cores for only 12 hours to arrive at a final solution. Not only is this a 12x reduction in resource cost per grid point, there is also significant (17x) time savings, allowing results to be available the next day.