Pointwise Reliable People. Reliable Tools. Reliable CFD Meshing. C F g L M P t Y +



One Of Larger CFD Models Ever Helps Optimize Advanced Fighter Aircraft

Mike Malone, Engineer/Specialist, Northrop Grumman Corporation, Pico Rivera, California

Silicon Graphics World, July 1998, page 14

One of the larger computational fluid dynamics (CFD) models ever developed helped to optimize the performance of an advanced fighter aircraft. The 5,000,000 grid point model was used by Northrop Grumman Corporation to investigate straight vertical landing for the Joint Strike Fighter, the U. S. Air Force's next generation combat plane. The model, which covered the entire exterior of the proposed aircraft, helped engineers investigate the effects of entrainment, which produces negative lift and must by counteracted by additional engine thrust. A special preprocessor automated most of the model creation process, including strategically distributing grid points for high accuracy while minimizing computation time.

Northrop Grumman Corporation, headquartered in Los Angeles, Calif., is a leading designer, systems integrator and manufacturer of military surveillance and combat aircraft, defense electronics and systems, airspace management systems, information systems, marine systems, precision weapons, space systems and commercial and military aerostructures. The company was formed in 1994 when Northrop Corporation acquired Grumman Corporation. Since then, it has acquired Vought Aircraft, the defense and electronics systems business of Westinghouse Electric Corporation, and Logicon Inc. Northrop Grumman employs about 52,000 people and reported sales last year of $8.1 billion.

Northrop Grumman is a principle member of the Lockheed Martin team in the competition to develop the Joint Strike Fighter (JSF). The JSF employs a direct lift system for short takeoffs and vertical landings, with uncompromised up-and-away performance. The JSF is an affordable, multi-service aircraft that will enter service in the next century with the U.S. Air Force, Maine Corps, Navy and the United Kingdom Royal Navy. America's armed forces will need as many as 3,000 JSFs to replace several different aircraft in service today.

The Advanced Short Takeoff/Vertical Landing (ASTOVL) design was a previous concept tested at near full-scale in the wind tunnel. One of the critical issues that arose during the development of the ASTOVL, and the JSF, was a concern over negative lift caused by close ground effects during vertical landing. When the plane is hovering close the ground, the jets at the front and rear of the craft hit the ground, move towards each other along the surface, then form a fountain when they meet that rises to hit the bottom of the aircraft. The result is a recirculation flow that creates a low pressure zone around the bottom of the aircraft, often producing negative lift that would cause the aircraft to drop to the ground if it were not offset by sufficient engine thrust.

Solution showing streamlines from the jets (one fore and two aft) with the ground plane colored by temperature.
Streamlines in Ground Effect

Hot gas ingestion is another related problem that can occur during vertical landing. It occurs when the jet engine inlet draws in hot gas from the fountain described above rather than clean air. As temperature of the gas ingested by the engine rises, the performance of the engine drops. this effect can interact with the negative lift phenomenon described earlier to cause serious problems during vertical landing. Another related concern is that the hot recirculating gas striking the bottom of the plane could raise its temperature high enough to damage its skin.

While a prototype of the ASTOVL had been built and tested in a wind tunnel, engineers felt that computer simulation would greatly streamline the design process. Wind testing provides flow and temperature information only at the limited number of points where sensors can be placed. Evaluating a different geometry requires an expensive and time-consuming modification of the prototype. A CFD analysis, on the other hand, provides fluid velocity, pressure, temperature and species concentration values throughout the solution domain. Engineers can usually modify the model in a matter of an hour or two in order to investigate a different geometry or boundary conditions. Simulation allows engineers to evaluate many more alternative designs in a short period of time and provides more information about each design they evaluate. The result is a better design in less time.

Simulating an object as complex as the ASTOVL, however, is a difficult challenge. The problem is the analytical model must include the entire aircraft exterior and at the same time capture many small details in order to achieve an accurate simulation. Conventional CFD preprocessors are not suited to the task. Meshing the entire aircraft is not difficult but maintaining the level of detail required to define such complex areas as the engine inlets would require a model with an enormous number of grid points. Such a model couldn't be solved in a reasonable period of time, even on the Cray C90 computers at NASA Ames Research Center that Northrop Grumman engineers have available.

But, Northrop Grumman engineers were aware of a preprocessor, Gridgen from Pointwise (http://www.pointwise.com), Bedford, Texas, that had been designed in cooperation with NASA for modeling tasks of this magnitude. A key advantage of this software program is its ability to divide a structure into contiguous subdomains called blocks that make it possible to use a fine mesh where needed to capture details and a coarse mesh elsewhere in order to minimize computational time. Gridgen also provides a number of powerful tools that automate many of the more difficult aspects of the meshing such as the asymmetrical allocation of grid points and smoothing the mesh to eliminate negative volume cells. Northrop Grumman engineers run Gridgen on an R8000 Indigo II workstation from Silicon Graphics, Inc., Mountain View, California.

To produce the mesh, engineers start with an IGES file defining the plane's geometry with NASA IGES surfaces, surfaces that conform to the IGES subset defined by NASA. After reading these surfaces onto Gridgen, engineers then decomposed the geometry into about 30 blocks. They defined block boundaries around areas that require close mesh spacing because of high pressure or flow gradients. Areas requiring close mesh spacing include areas of geometric complexity, such as the engine inlet and auxiliary inlet guide vanes, as well as the boundary layer around the surface of the aircraft. They drew connectors on the geometry to define the edges of the block domains, grouped the connectors to form surfaces (domains) and finally grouped the surfaces to form volumes defining blocks. In cases where the grids did not adhere precisely to the domains, the engineers used a Gridgen feature that automatically projects a grid onto the geometry.

The next step was distributing grid points along the connectors. Uneven spacing is desirable along most of these connectors. For example, fine spacing is usually needed at sharp and trailing edges of aerospace surfaces while the area in between can usually be quite coarse. Ideally, the mesh should start fine at the leading edge, then become continually finer. This type of mesh distribution is very tedious to produce by hand because of the need to define each grid point one by one on order to provide a smooth transition from a fine to coarse mesh. Instead, on this project engineers used a Gridgen feature that automatically distributes grid points along a connector based on any of a wide range of functions that can be specified by the user. In most cases, engineers used a hyperbolic tangent function to provide the spacing described above.

Top view of grid.
Top View of Grid

The irregularity of the ASTOVL geometry meant that the initial grid had areas of negative and zero volume that would have made it impossible to analyze. With a conventional grid generator, Northrop Grumman engineers would have been forced to modify the grid element by element to improve its quality, a process that would have taken months or even years. Fortunately, Gridgen provides an elliptic smoother that allowed the engineers to improve the quality of the mesh automatically applying elliptic partial differential equation methods. Engineers applied smoothness, clustering and orthogonality controls to improve the mesh. With each iteration of mesh improvement, the program provided a graphical display of negative and skewed volume cells. In only a few hours, they had produced an excellent quality mesh ready for analysis with Northrop Grumman's proprietary CFD solver.

The analysis took about 60 hours on a Cray C90 supercomputer. The results correlated very well with wind tunnel testing. Oil droplets placed on the floor of the wind tunnel were used to determine the flow streamlines at the ground plane. They matched up very well to the ground plane streamlines at the ground plane. They matched up very well to the ground plane streamlines produced as one of the results of the analysis. The most critical area, the stagnation zone where the two jets meet, was precisely predicted by the analysis. Confident in the accuracy of the model, engineers used the results to determine the amount of thrust required to achieve a safe vertical landing.