Aiolos Engineering saved time and costs in the development of a 100-meter long wind tunnel by using CFX computational fluid dynamics (CFD) software instead of physical testing. Building and testing a scale model was ruled out because it was expensive and required more time than was allotted for the design of the entire facility. By creating a CFD model of the wind tunnel, Aiolos was able to evaluate many design parameters in a fraction of the time required to design and build a model wind tunnel and to implement a test program. Engineers used CFD to evaluate different wall divergence angles and roughness conditions of the test section to meet a key design challenge: to achieve a reasonable trade-off between the axial static pressure and velocity gradients over the full range of train shapes and sizes. "Ideally, both the static pressure and velocity should be constant along the entire length of the test vehicle," says Trevor Bender, aerodynamicist at Aiolos Engineering. "With CFD, we could perform many trade-off studies to find the optimal balance between those two factors."

Every climate on the planet

Rail Test and Research, GmbH, of Vienna, Austria, hired Aiolos Engineering and three partner companies, VATech, Elin EBG, and Voest MCE, to design and build a climate research facility for trains. The primary purpose of the facility, which included two wind tunnels, is to test full-size trains under varying climatic conditions. Such huge wind tunnels are still necessary, even in the days of computer simulation, because full-scale train testing is too big a problem to handle computationally. These two wind tunnels are used for evaluations of the train's climate control system, tests of heat transfer through the train shell, tests of product robustness under varying climatic conditions, and windscreen loading of rain, snow, and ice. "Both tunnels had to be capable of simulating most of the climatic conditions found on the planet," says Bender. The specifications for the wind tunnels called for a temperature range of -50 degrees C to +60 degrees C and the ability to generate wind speeds of up to 250 km/hour. In non-freezing temperatures, the wind tunnels produce humidity levels up to 95 percent relative humidity. Both are also equipped with nozzles to produce rain, snow, or ice. The only difference between the two wind tunnels is the length. The longer one, at 100 meters, can accommodate four full-sized train cars. The shorter wind tunnel is 30 meters long and is limited to one train car.

The design of the 100-meter long wind tunnel was particularly challenging because of its length. To make this wind tunnel economically viable, the designers had to limit the height and width to just slightly larger than the train itself. The remaining free space was quite a bit smaller than in conventional wind tunnels. The relative size of the train, called blockage, was a concern to the aerodynamicists on the project, as Bender explains. "With a train in a wind tunnel that size, the geometry is essentially a three-sided annulus," he says. "Flow through the annulus experiences a drop in total pressure and the development of boundary layers due to frictional effects at all surfaces. The existence of boundary layers effectively reduces the flow area in the free-stream direction, resulting in a further decrease in static pressure as velocity increases." The affects of boundary layer growth are a problem for this particular wind tunnel because effective testing of an on-board climate control system requires constant static pressure along the length of the train. A non-uniform static pressure compromises the test by altering the balance between the ventilation intake and exhaust flows. For a more correct simulation of the heat transfer to the train shell, however, a constant axial velocity is also required. Finding the best trade-off between pressure and velocity, then, was a critical goal for the aerodynamicists.

The main design parameter they could manipulate was the wall divergence angle, which is the amount of slope in the side walls and ceiling from the front to the back of the wind tunnel. "Instead of using walls that are straight and parallel, you can angle them slightly from the front to the back of the test section to increase the downstream area by a small amount," says Bender. "That would help minimize the effects of the boundary layers. But the big question was how much of a divergence angle we needed and what were the influences of train size on the selected angle." In other situations, Aiolos would have built a model wind tunnel to answer this question. It was ruled out in this case because of its cost, and the time required. "The time required to design the scaled-down wind tunnel, procure the components, build it, and perform the required tests was beyond that allotted for the design of the entire facility," Bender says.

Instead of physical testing, Aiolos opted to simulate the performance of the wind tunnel using CFD analysis. CFD involves the solution of the governing equations for fluid flow, heat transfer, and chemistry at many thousand discrete points on a computational grid representing the flow domain. The use of CFD enables engineers to obtain solutions for problems with complex geometries and boundary conditions. A CFD analysis yields values for fluid velocity, fluid temperature, and fluid pressure throughout the solution domain. Based on the analysis, a designer or an engineer can optimize fluid flow patterns by adjusting either the geometry of the system or the boundary conditions such as inlet velocity or temperature. Aiolos chose CFX-5 from AEA Technology in part, because its strong meshing capabilities would simplify the creation of this problem's complex analysis mesh. "CFX can automatically create an unstructured mesh, simplifying a tedious step in the CFD process and yet, through well designed mesh controls, it still gives the user a lot of control over the final mesh." Bender explains. "But where most unstructured meshing utilities produce only tetrahedral elements, which adapt to complex geometries more easily, CFX also has a capability called 'inflation' that creates high aspect ratio prism elements in the near wall region. These prism elements let you capture the high velocity gradient and shear effects along the walls, without having to use a prohibitively fine mesh of tetrahedral elements."

Creating the analysis model

Bender began the CFD analysis by importing a CAD model of a train into CFX-Build, the software's preprocessor. The model, supplied by Siemens GmbH, was not suited to CFD analysis in its native state because not all surfaces were completely closed and there were some highly detailed features such as a windscreen and spoiler that would have slowed the solution time. The extraneous details were removed and a single smooth body surface was created. Using CFX-Build's geometry manipulation tools, Bender created the geometry of the wind tunnel test section. Most of this geometry was fixed, with the exception of the diverging wall angles. Bender created four different wall angles: 0.1 degree, 0.15 degree, 0.2 degree, and 0.27 degrees.

CFX's meshing utility created the analysis mesh. It was an unstructured mesh consisting of tetrahedral elements everywhere except the near wall region, which used prism elements, as mentioned above. There were approximately 1,150,000 nodes in the entire flow domain with 625,000 of them being tetrahedral elements and 525,000 being prism elements. The flow domain was defined by boundary conditions set at the exterior and interior boundaries. The majority of the boundaries were stationary solid walls. "In the CFX program, the user simply identifies a surface as being a wall, either smooth or rough, and the software takes care of the rest," says Bender. "It applies the correct boundary conditions for all transport equations, including the proper use of the wall functions." Other boundaries included the inlet, the outlet, and the plane of symmetry. Bender used the k-epsilon turbulence model.

The CFD analyses were run on an IBM computer with dual 800 MHz Pentium III processors. Preliminary 2D cases took about six hours each to solve, while the subsequent 3D cases took 30 hours each. The initial 3D analyses of the four different wall angles indicated that a 0.15 degree angle would give the most effective trade-off between velocity and pressure. But Bender and his colleagues knew that the roughness of the wind tunnel walls would affect the aerodynamics as well. "The walls themselves are smooth painted construction but there are areas such as the concrete floor, the interface of the glass covering the solar simulation lights, and the expansion joints, that could affect boundary layer development," says Bender. They repeated the analyses to see the effect of surface roughness on pressure and velocity. The results showed that a small amount of roughness has a fairly significant effect on pressure drop, similar to decreasing the wall angle by 0.1 degree. "Initially a wall angle of 0.15 degree seemed to offer the best trade-off between velocity and pressure gradients," says Bender. "However, the roughness studies indicated that we needed to increase the angle to account for the surface roughness in the actual wind tunnel. Therefore it was decided to make the wall angle 0.2 degree."

The ability to simulate different design parameters allowed Aiolos to accommodate the geometric challenges of this one-of-a-kind wind tunnel. "CFX saved time and money we would have spent on model testing," says Bender. "Not only did it allow us to complete the design on schedule, it also gave us an extensive look at the entire flow field, something that is more difficult to do experimentally."

A drawing of the Rail Test & Research 2 wind tunnel facility.

Numerical mesh of tetrahedral and prism elements. The seven layers of thin prism elements show-up as a dark line at the edge of the surfaces.

CFX was able to model both centerline velocity vectors and train surface pressures even though the geometry of the wind tunnel and train came from separate sources.