|
The "virtual flight" and preliminary design of conceptual vehicles require running hundreds or even thousands of aerodynamic simulations. Practical computation of such large test matrices places a premium on automation of the entire simulation process and on automated control of data flow and job execution. In recent years, Cartesian mesh methods have made rapid strides toward the goal of fully automated aerodynamic analysis. Based on this technology, Ames researchers, with partners from the Courant Institute (New York University), have developed systems for automatically running the extremely large matrices of simulations required for preliminary design and virtual flight. When design permutations or control surface deflections modify the shape of a vehicle, the system automatically generates the new geometry, remeshes the design, and runs simulations covering a wide range of flight conditions to populate the test matrix. This capability allows users to evaluate a proposed design over its entire flight envelope, rather than at only a single flight condition. It also provides detailed information about vehicle stability and control and has been used in support of virtual flight by populating an aerodynamic database for Ames' Vertical Motion Simulator.
Traditional structured and unstructured approaches to computational fluid dynamics (CFD) rely upon body-fitted meshes to discretize the space surrounding a vehicle. Since the mesh must be fitted to the geometry, meshing is generally an interactive process and can consume several months of the CFD cycle. In contrast, Ames' Cartesian approach is based on the use of nonbody-fitted Cartesian cells. These cells are free to arbitrarily intersect the vehicle, and such intersections are accurately accounted for by using techniques from computational geometry. Fundamentally, the approach trades the case-specific problem of generating a body-fitted mesh for the more general problem of computing and characterizing intersections of the cells touching the body. This generality means that analysts need not be concerned with details of the surface geometry and leads directly to automation.
With mesh generation, a fully automatic process, analysts are free to control the simulation from a more macroscopic level. A typical simulation matrix may study the effects of a deflection of a control surface (flap, aileron, rudder, etc.) over a range of Mach numbers and angles of attack. For example, testing a configuration with five different flap deflections, each at five Mach numbers, five angles of attack, and five sideslip angles produces a test matrix with 625 cases (5 x 5 x 5 x 5 = 625).
The upper left frame in figure 1 shows a recent example of a space shuttle configuration with 20 separate components and 13 moveable control surfaces. The lower left frame gives an overview of a typical multilevel Cartesian mesh containing approximately 2 million cells. The three frames at the right of the figure show the mesh system's automatic response to deflection of the main body-flap to 0, 15, and 30 degrees. The system will automatically deflect the control surface, regenerate the mesh, and run each new case over a specified sweep of Mach number, angle of attack, and sideslip.
The automated run system drives Ames' Cart3D inviscid simulation software package. It includes tools for surface preprocessing, mesh generation, flow solution, and postprocessing. Recent in-house examples with configurations such as these have populated test matrices of over 400 cases in under 24 hours without dedicated computing time. External users in the Department of Defense (DoD) have reported running over 1,200 cases/day on dedicated computers. Cart3D has been distributed to over 80 sites, including four NASA Centers, several DoD and Department of Energy laboratories, as well as various sites within U.S. industry and academia. The package runs on computers ranging from simple desktops to high-performance parallel platforms and is being commercialized through the Ames Commercial Technology Office.
Point of contact: Michael Aftosmis
(650) 604-4499
aftosmis@nas.nasa.gov
http://www.nas.nasa.gov/~aftosmis/cart3d/
Back To Top
Previous Paper
Return to Advanced Space Transportation
Next Paper |
|
Fig. 1.(a) A recent example of a space shuttle configuration with 20 separate components and 13 moveable surfaces; (b) an overview of a typical multilevel Cartesian mesh containing approximately 2 million cells; and (c) the three frames show the mesh system¥s automatic response to deflection of the main body-flap to 0, 15, and 30 degrees.
|
|
