Tips for Analyzing Large Aircraft in Flow Simulation
In honor of America’s upcoming elections, let’s take a look at one of my favorite bits of engineering associated with the presidency—Air Force One, or more specifically, a Boeing VC-25.
The VC-25 is a heavily customized version of the commercial Boeing 747-200B. Modifications include electromagnetic pulse shielding and anti-aircraft countermeasures. The base version is capable of flying one-third of the distance around the world before refueling. This is distance largely gained due to engine power, efficiency, and drag reduction.
Let’s use the VC-25 and learn from it. I’ll take it into SOLIDWORKS Flow Simulation and show three good practices for analyzing large aerospace models in this computational fluid dynamics (CFD) program.
Scaled Models and Input Parameters
The 747 tested in a wind tunnel is not full scale. For high-speed tests like the plane at cruising speed, the model is 3% of full size and tested in a high-density fluid. Is it necessary to run CFD on a scaled model, too?
There is one major positive and one major negative to running a scaled version of the aircraft.
- On the positive side, you will not have to calculate the Reynolds numbers and decide on input parameters to correlate with your scaled model.
- On the negative side, a model with a 60 m wingspan will eat up more RAM than a model with a 1.8 m wingspan.
In general, mesh size is relative to model size and can be manually refined and unrefined, so technically mesh on a full-scale versus a 3% model wouldn’t affect RAM requirements. As an example, a 3% model versus a 1% model of the 747, using the automatic meshing capabilities, produced an equal number of total mesh cells ±0.5%. Realistically, however, a user would likely end up creating more mesh cells on the larger model. If we do use a scaled model, we have two things to consider.
The Reynolds number for the scaled model should be equal to the Reynolds number of the full-sized aircraft. However, the Reynolds number is not automatically calculated, largely because characteristic length is a value dependent on the model. See the Reynolds number definition below:
However, all other variables can be obtained using the engineering database, initial conditions, and goals. Velocity and density are inlet conditions. Set these up in either “Boundary Conditions” or “Initial and Ambient Conditions”. If we use “Boundary Conditions,” density is calculated based on Pressure, Temperature, and the Gas Law. If we use “Initial and Ambient Conditions,” density can be directly input.
To obtain dynamic viscosity go to “Tools, Flow Simulation, Tools, Engineering Database.” Within the materials list, find a specific fluid, and then select “Tables and Curves.” Dynamic viscosity will be one of the curves listed and is dependent on fluid temperature.
Finally, to avoid having to manually calculate the Reynolds number, set up “Goals” for velocity and density, then use an “Equation Goal” to calculate the Reynolds number. Remember, goals have the added benefit of forcing the solver to converge on these results, so the results are more precise and easily accessed.
Internal or External Fluid Flow
We can model a wind tunnel around the aircraft. This will require an internal flow project. If we do this, inlet conditions are specified via “Boundary Conditions.”
More likely, we will adjust the computational domain around the aircraft. Computational domain size follows the same rules of thumb as the walls of a wind tunnel. There must be enough room around the model to fully capture pressure effects.
Inlet conditions are applied using “Initial and Ambient Conditions.”
Analyzing Portions of the Aircraft
To save RAM and calculation time, two studies should be run. The first encapsulates the entire aircraft and the second captures just a portion. A lower-quality mesh or simplified geometry can be used in the first general run, and a very fine mesh intended to capture turbulence, accurate pressure readouts. Other important results will be used on the second detailed run. In this example, I use the entire aircraft as my general run and a single winglet as my detailed run.
Use “Transferred Boundary Conditions” to take the results of the entire aircraft and apply them as inlet conditions to the winglet. The inlet conditions are thus much less assumptive and allow me to dedicate more RAM to this important area of the wing.
There are many different meshing tools and options within Flow Simulation. There are some that lend themselves better to aircraft testing. I will discuss the ones that I used on the VC-25 and will organize them based on mesh accuracy for the skin of the aircraft and mesh accuracy for turbulence and vortices.
I recommend running leak testing on the aircraft. We do not want any openings in the outer shell for air to flow through. Besides being unrealistic, this wastes calculation time because mesh cells will appear within the aircraft and require calculation.
Within “Local Initial Mesh” I recommend “Level of Refining Fluid/Solid Boundary, Curvature Refinement, and Equidistant Refinement.” Fluid/Solid boundaries, or partial cells, are mesh cubes that encompass both the model and the fluid around it. The skin of the aircraft will always be made of partial cells. Refining along the curvature works similarly. This slider bar will automatically divide mesh cells in regions of high curvature. Both these options are necessary to produce an accurate representation of the skin.
Equidistant refinement is imperative in producing a refined mesh in the region just outside the aircraft skin. For my second test run on the winglet, I specified any mesh cell within 0.008 m of the winglet should be 5 levels smaller than the global mesh. I chose this offset distance because it equated with the airfoil thickness of my scaled model. This is an extremely useful tool because it is easy to use, saving the user a great deal of setup time.
Vortices and Turbulence
Winglets reduce drag by shrinking the size of wingtip vortices. To ensure we correctly determine their drag effects, the mesh must be sufficiently refined where we expect vortices and turbulence.
- First, forcibly refine the mesh in a particular fluid region by creating a “dummy solid,” a generic, solid shape that you insert into the assembly, like a cylinder, and place where you want to refine the mesh. Use “Local Initial Mesh” to refine the mesh cells within the dummy solid via the slider “Level of Refining Fluid Cells.” Fluid will still flow through this dummy solid as if it were intangible.
- Second, we cannot always predict exactly where vortices and turbulence will occur, and we cannot predict if our mesh cells are sufficiently small. Automatic refinement covers these bases. Find this within “Calculation Control Options” under the “Refinement” tab. Essentially, we allow the solver to pinpoint where results have a high rate of change and then decrease the mesh cell size in those regions. This automatic refinement can be done multiple times throughout the calculation. Luckily, we also set the maximum amount of cells we will allow in our computational domain so that we do not exceed our RAM limitations.
Regardless if using a scaled or full-size aircraft, we have a method to calculate the Reynolds number. Because of RAM limitations, use methods to narrow down the analysis to portions of the aircraft, and use the handful of meshing techniques I mentioned to correctly mesh curvy aircraft models.
The B747-200 was a very interesting aircraft to run through Flow Simulation, and there was a lot to be learned from it. However, the VC-25 is aging and has become less cost-effective to operate. In January 2015, Boeing’s proposal of a modified 747-800 was accepted as the replacement, chosen for its long fuselage, quieter and more economical engines, and drag-reducing raked wingtips.
Design modifications are currently underway, some of which are probably being run through CFD software today. Perhaps, in a few years, we will run a CFD simulation on a modified 747-800. In the meantime, though, we can use Flow Simulation on our own aircraft designs.
For more information on Flow Simulation for aircraft design, see the video link: