the overgrid interface for computational simulations on overset … · 2016-10-12 · computational...

THE OVERGRID INTERFACE FORCOMPUTATIONAL SIMULATIONS ONOVERSET GRIDS

William M. ChanNASA Ames Research CenterMoffett Field, CA 94035

32nd AIAA Fluid Dynamics Conference24-26 June, 2002 / St. Louis, Missouri

For permission to copy or to republish, contact the copyright owner named on the first page.For AIAA-held copyright, write to AIAA Permissions Department,

1801 Alexander Bell Drive, Suite 500, Reston, VA 20191-4344.

AIAA 2002-3188

AIAA 2002-3188

THE OVERGRID INTERFACE FOR COMPUTATIONAL

SIMULATIONS ON OVERSET GRIDS

William M. Chan�

NASA Ames Research Center

Mo�ett Field, California 94035

Abstract

Computational simulations using overset grids typ-ically involve multiple steps and a variety of softwaremodules. A graphical interface called OVERGRIDhas been specially designed for such purposes. Datarequired and created by the di�erent steps include ge-ometry, grids, domain connectivity information and ow solver input parameters. The interface providesa uni�ed environment for the visualization, process-ing, generation and diagnosis of such data. Generalmodules are available for the manipulation of struc-tured grids and unstructured surface triangulations.Modules more speci�c for the overset approach in-clude surface curve generators, hyperbolic and alge-braic surface grid generators, a hyperbolic volume gridgenerator, Cartesian box grid generators, and domainconnectivity pre-processing tools. An interface pro-vides automatic selection and viewing of ow solverboundary conditions, and various other ow solverinputs. For problems involving multiple componentsin relative motion, a module is available to build thecomponent/grid relationships and to prescribe and an-imate the dynamics of the di�erent components.

1. Introduction

In recent years, overset grid methods1 have beensuccessfully used to compute both steady and un-steady ows for many complex con�gurations. Thesecomputations have contributed to the design andanalysis of a variety of aerospace and marine ve-hicles at government laboratories2�11 as well as inindustry.12�18 One of the critical elements in suchwork is the time required to perform a complete con-�guration analysis. Clearly, trimming the time neededfor the entire process will result in signi�cant costsavings. The complete simulation process typicallyconsists of virtual geometry construction (CAD work),geometry processing, surface and volume grid genera-tion, domain connectivity, ow calculation, and solu-tion post-processing. Up until recently, there has been

�Computer Scientist, Senior Member AIAA

Copyright c 2002 by the American Institute of Aeronautics and Astro-

nautics, Inc. No copyright is asserted in the United States under Title 17, U.S.

Code. The U.S. Government has a royalty-free license to exercise all rights un-

der the copyright claimed herein for Governmental Purposes. All other rights

are reserved by the copyright owner.

little e�ort spent on trying to streamline the varioussteps of the process. Each step requires a number oftools and there is no one place from which all the toolscan be accessed.Two main approaches are being pursued to reduce

the overall process time. The �rst is to develop au-tomated algorithms and software tools that reduce oreliminate the user's input at each step of the process.Examples of recent e�orts in grid generation and do-main connectivity are given in Refs. 19-21. The secondapproach to reduce process time is to incorporate allessential and robust tools into a single graphical userinterface environment (GUI). Clearly, if every step inthe process is completely automated, a GUI is not nec-essary to produce a �nal result. Since the currentstate-of-the-art still requires user's inputs at variousstages,22 it is most convenient to work through thedi�erent steps in a graphical interface. This paperdescribes OVERGRID which is one of the very fewinterfaces available today that has been specially de-signed for applying the overset approach to complexgeometries. A parallel e�ort has also been developedin the OVERTURE23 suite of tools.

2. Overview of Interface

2.1. Software Design

The OVERGRID interface belongs to a larger soft-ware package called Chimera Grid Tools (CGT).24

The CGT package consists of about 40 independentgrid generation and solution analysis modules that runin batch mode, the OVERGRID graphical interface,a suite of scripts that can be used for automatingoverset-grid computations on complex con�gurations,and several libraries of common routines shared by thevarious tools (Fig. 1).OVERGRID serves as a central portal to many of

the modules which are called as batch processes, aswell as a visualization tool for the working data (ge-ometry and grids). By keeping each module separatefrom the graphical interface, a sequence of grid op-erations can be recorded in a script and reproducedeasily without manual intervention. OVERGRID of-fers a script generation capability which automatically

1

American Institute of Aeronautics and Astronautics

DOCUMENTATION EXAMPLES

CHIMERA GRID TOOLS

GRID UTILITIES

LIBRARIES OF COMMON ROUTINES

GRID GENERATION TOOLS

SOLUTIONANALYSIS TOOLS

GRAPHICALINTERFACE OVERGRID

CONFIGURATION SCRIPTS

Fig. 1 Structure of Chimera Grid Tools.

records all of the user's actions, excluding those re-lated to visualization transforms. Comment lines areautomatically inserted to identify the action types andbatch modules used. The script is written in the ToolCommand Language Tcl25 which provides high levelfunctions and modularity. If a simple modi�cation inone of the steps is desired (e.g., alter the number ofpoints in a grid), the script can be edited and re-runin batch mode to reproduce all the steps rapidly.Another advantage of calling batch mode modules

from OVERGRID is the handling of single and doubleprecision data. The input data type is automaticallydetermined by OVERGRID and appropriate single ordouble precision arrays are dynamically allocated fordata storage. Preference settings in OVERGRID areused to specify directories where single and doubleprecision executables of the batch mode modules arelocated. When the user initiates an action requiringa batch mode module, executables of the appropriateprecision are automatically selected to operate on thedata in batch mode. This procedure allows OVER-GRID to operate on both single and double precisiondata without the need for code re-compilation.OVERGRID is primarily written in C and Tcl/Tk.25

The data input and output routines are written inFortran for eÆcient execution speed. Rendering ofobjects is accomplished with OpenGL calls from theC program, while the interface widgets (e.g., buttonsand entries) are built with Tcl/Tk. OpenGL render-ing from a Tcl/Tk window is made possible by theTogl widget. The software has been ported to variousUNIX workstations and personal computers runningLinux, Macintosh OS-X or Windows NT. With itsease of use and conciseness, Tcl/Tk has been found tobe enormously valuable in rapidly creating the desiredprofessional look and feel of the interface. Moreover,the quick learning time allows the developers of gridgeneration codes to easily build the graphical interfacethemselves. The resulting software tends to be morepractical and user-friendly than one built by GUI ex-perts with limited grid generation and computational uid dynamics experience.

2.2. Data Formats and Main Windows

Both geometry and grid data can be manipulatedand created in OVERGRID. For the rest of thispaper, the words `entity' and `grid' will be usedinterchangeably to denote any object stored inOVERGRID as part of a surface geometry descrip-tion (surfaces) or a computational grid (volumes,surfaces, or curves). Surface geometry can be readinto OVERGRID as multiple structured panelswhere each panel is a rectangular array of points inPLOT3D26 format, or as an unstructured surface tri-angulation in CART3D (http://www.nas.nasa.gov/~aftosmis/cart3d/cart3dTriangulations.html) orFAST27 formats. For surface triangulations, a tag isassociated with each triangle which could be used tostore component information. The CART3D formatfurther supports a �le containing both grid and scalarfunction data at the vertices of the triangulation (seedescription on DIAGNOS module in Section 3.1).Computational grids are always read and written inPLOT3D format - unformatted or formatted, singleor multiple zone, with or without iblanks, single ordouble precision. The above attributes for an input�le are automatically determined by OVERGRID.Structured entities in OVERGRID are stored as

discrete points and a J/K/L convention is employedto denote the index directions in a volume grid.Surface geometry de�nitions derived from CAD arefrequently represented by Non-Uniform Rational B-Spline (NURBS) surfaces or solids. With OVERGRIDcurrently limited to reading structured panels or un-structured triangles, another software package such asGRIDGEN28 is needed to convert other geometry for-mats into panel networks or triangles prior to usingOVERGRID. Future plans for OVERGRID to readother data formats are given in Section 9.On execution, OVERGRID will bring up the follow-

ing four main windows shown in Fig. 2.

(1) The Display window contains a graphical displayof the entities currently in memory.

(2) The Controls window contains widgets for settingvarious display options, resetting views, showinginformation on the number of volume, surface andcurve entities currently in memory.

(3) The Main window contains widgets for input andoutput of entities, script creation, access to thevarious modules, and general on-line help. Moredetailed on-line help is also available for the indi-vidual modules.

(4) The Selection window contains widgets for per-forming entity selection, deletion, and blank-ing/unblanking from view. More direct entityselection mechanisms are also available by clickingon or drawing a drag box over the desired entitiesin the Display window.

2


Fig. 2 OVERGRID main windows: upper left - Display, lower left - Controls, upper right - Main, lowerright - Selection.

3. General Grid Tools

General grid tools in OVERGRID fall into twoclasses: diagnostic tools for analyzing entity at-tributes, and manipulation tools for modifying enti-ties. These are described in more details below.

3.1. Diagnostic Tools

The Controls window o�ers widgets for toggling thedisplay of entity attributes such as tangent and normalvectors for surfaces and entity identi�cation numbersas illustrated in Fig. 3.An important attribute of overset grids not typically

found in other gridding methods is an `iblank' valueassociated with each grid point. The iblank value isused to denote whether the point is a �eld point, ahole point, or a fringe point. Field points are wherethe ow equations are solved and the dependent owvariables are computed. Hole points are points thatlie outside the ow domain, e.g., points inside the solidsurface of an object. The ow equations are not solvedat these points. Fringe points are points where thedependent variables are interpolated from stencils inneighboring grids. Such points arise on boundaries ofholes and on the outer boundaries of a grid. Points

on grid outer boundaries are fringe points only if no ow solver boundary conditions are applied. A fringepoint without a valid interpolation stencil is called anorphan point.

It is clear that visualization of the iblank value at thegrid points is immensely helpful in checking and de-bugging the results of the domain connectivity process.OVERGRID can be used to display grid planes, where�eld points are connected by wireframes, hole pointsare not drawn, fringe points are colored by the gridnumber of the grid containing the interpolation sten-cil, and orphan points are highlighted in black againsta white display background (Fig. 4). The x; y; z coor-dinates, grid number, J, K, L indices and iblank valueof a vertex can also be interrogated by picking thevertex via a hot key. For surface triangulations, in-formation on speci�c vertices and faces can also besimilarly obtained.

The DIAGNOS module accessible from the MainMenu window allows the user to check various gridquality functions. Wireframe representations of thegrid surfaces are colored by the value of the grid qual-ity function. Locations and values of the minimumand maximum are also reported to the user.

3


Fig. 3 Display of entity attributes in OVERGRID.(a) Surface tangent vectors (arrow: J direction,line: K direction). (b) Surface normal vectors. (c)Entity identi�cation numbers.

For structured surface and volume entities, gridquality functions include stretching ratios and turn-ing angles in the J and K directions, cell sizes (areasfor surface grids, volumes for volume grids) and gridinduced truncation error estimates (Fig. 5a). Thestretching ratios and truncation error estimates can beused to identify where more grid clustering is needed.The cell size diagnostic is useful in checking grid res-olution compatibility in regions of grid overlap whichis a critical requirement for good inter-grid communi-cation.

For surface triangulations, available grid quality

Fig. 4 Iblank display in OVERGRID. Unblankedpoints are connected by wireframe. Fringe pointsare rendered with symbols colored by the interpo-lation stencil donor grid number. Black symbolsdenote orphan points.

functions include vertex valence (number of trianglesconnected to a vertex), the minimum and maximumangle at a vertex, and a surface curvature estimate.High quality triangulations should have a vertex va-lence of around 6 everywhere, and the minimum andmaximum angles at a vertex should not have large ex-trema. The surface curvature estimate could be usedto identify surface features but a more robust formulathat is independent of grid resolution needs to be de-termined. Further grid function display is provided foran annotated surface triangulation which contains oneor more scalar functions at the vertices, e.g., pressurecoeÆcient, density, and others (Fig. 5b).Other features in the DIAGNOS module include

a report on the number and percentage of blankedpoints, the number of orphan points from domain con-nectivity and the number of negative Jacobians involume grids. A utility is provided to check the topol-ogy (periodic, axis, constant plane) of a surface gridand allow manual resetting of the topology if neces-sary. For example, a periodic grid with non-coincidentstart and end planes in the periodic direction can havethese planes reset to being coincident.

3.2. Manipulation Tools

Four modules are available in OVERGRID for gen-eral grid manipulation.

GRIDED - Structured Grid Editing Tool

The GRIDED module provides the following listof commonly used functions for operations on one ormore structured grids.

4


(a)

(b)

Fig. 5 (a) Grid induced truncation error estimateon the Space Shuttle Launch Vehicle structuredoverset grid system. (b) Pressure coeÆcient on asurface triangulation of the aft-attach hardware.

(1) Swap J and K, K and L, or J and L grid indices.

(2) Reverse one or more of J, K and L indices.

(3) Mirror about X = 0, Y = 0, or Z = 0 plane.

(4) Scale or translate.

(5) Rotate about X , Y , or Z Cartesian axes.

(6) Resequence the identi�cation numbers of a collec-tion of grids.

(7) Extract a subset.

(8) Add extra layers of points at �J , �K, �L bound-aries by extrapolating in the tangential, x, y, or zdirection using a speci�ed stretching ratio.

(9) Concatenate any two volume, surface or curvegrids in the speci�ed direction (J, K, or L).

(10) Split an entity into two along a constant J, K,or L direction at a speci�ed index where the splitentities may overlap in one or more points.

(11) Automatically concatenate any number of sur-face or curve grids in arbitrary relative orien-tations. A tolerance parameter allows adjacentgrids separated by small gaps to be concatenated.

(12) Smooth any subset of a grid in one or more di-rections in J, K, and L.

(13) Generate surface or volume of revolution from acurve or surface, respectively.

(14) Find intersection curve(s) between an intersectorsurface and one or more intersectee surface grids.

TRIGED - Surface Triangulation Editing Tool

The TRIGED module provides the following list ofcommonly used functions for operations on an unstruc-tured surface triangulation.

(1) Swap common edge between two adjacent trian-gles.

(2) Reverse direction of normal on all triangles.

(3) Mirror about X = 0, Y = 0, or Z = 0 plane.

(4) Scale or translate.

(5) Rotate about X , Y , or Z Cartesian axes.

(6) Remove un-used vertices.

(7) Extract all triangles on one side of a Cartesiancutting plane.

(8) Extract all triangles belonging to a given list ofcomponents.

SRAP - Grid Redistribution Tool

The SRAP module is used to redistribute grid pointson a structured curve, surface or volume entity. In thecase of a surface or volume entity, it is treated as acollection of curves in the J, K or L index direction.Points along each curve are �tted to a cubic spline, andthen redistributed based on user input speci�cations.There is an option to project the new points back ontothe original piece-wise linear de�nition of each curve.Redistribution can occur in one or more segments ineach direction where each segment is de�ned by a startand end index. The user has four input speci�cationoptions for redistributing points in a segment.

(1) Specify the new number of points and grid spac-ings at end points (OVERGRID reports the max-imum stretching ratio).

(2) Specify the maximum stretching ratio and gridspacings at end points (OVERGRID reports thenew number of points needed).

(3) Same as (2) but a maximum grid spacing is alsospeci�ed.

(4) Specify a uniform spacing (OVERGRID reportsthe new number of points needed so that the gridspacing in a uniform mesh will not exceed thespeci�ed spacing).

5


(c)

(a)

(b)

Fig. 6 SRAP functions in OVERGRID. (a) Orig-inal surface. (b) Redistribution in one directionwith clustering at end points. (c) Redistributionto uniform spacing in both directions.

The input grid spacings can be in absolute units orrelative to the total arc length of the segment. Thecurrent absolute end grid spacings of the segment aredisplayed as information for the user. Results of redis-tribution options (2) and (4) are shown in Fig. 6.

PROGRD - Grid Projection Tool

The PROGRD module is used to project a set ofactive entities onto a set of reference entities. Thereference entities may consist of structured surfaces orunstructured surface triangulations, but not a mixtureof the two. Active entities may be structured curves orsurfaces. Members of the reference and active entitysets can be graphically selected in OVERGRID. Gridpoints on the active entities are projected to a bilinearrepresentation of the reference entities. The projectioncan be performed in the surface normal direction ofthe reference entities or in the X, Y, or Z directions.After the projection, the maximum distance moved bya point in the active entity set is reported to the user.The user can choose from one of the following threeoptions if an active point falls outside the referencesurfaces.

(1) Do not move point.

(2) Project point to closest cell on reference surfaces.

(3) Project point to tangentially extrapolated refer-ence surfaces.

Fig. 7 illustrates the use of PROGRD to model abump on a blade. A surface grid was originally gen-erated on a clean blade. Subsequent design changesintroduced a small bump on the blade. The user de-cided that it is not critical to model the bump/bladeintersection line exactly and that keeping the �nalcon�guration in a single grid is more important. PRO-GRD is used for this task to project the original bladegrid (active entity) onto the bump grid (reference en-tity). Points on the blade grid outside of the bump areundisturbed by selecting option 1 above.

(a) (b)

original blade grid (active entity)

bump (reference entity)

Fig. 7 PROGRD function in OVERGRID. (a)Original reference entity (bump grid) and activeentity (blade grid). (b) Active entity (blade grid)after projection.

4. Overset Grid Generation Tools

A current strategy for creating overset grids arounda complex con�guration consists of the followingsteps.22

(1) A high �delity de�nition of the surface geometry isobtained in the form of multiple panel networks(structured patches) or an unstructured surfacetriangulation. Conversion from other data typessuch as IGES �les or solid models may be neces-sary.

(2) Surface feature curves and other surface curves areconstructed from the surface geometry, e.g., inter-section curve between components, sharp surfacediscontinuities, high surface curvature contours,and open boundaries.

(3) The geometry surface is decomposed into foursided domains. Some are bounded by one ormore surface feature curves while others are notbounded by any. Concatenation and splitting offeature curves may be necessary.

(4) Surface grids are generated on the decomposeddomains by hyperbolic or algebraic methods.

(5) Body-conforming volume grids in the near �eldare created from the surface grids by hyperbolicmarching.

(6) O�-body Cartesian box grids are generated to en-close the near-�eld volume grids and to extend thecomputational domain to the far �eld.

(7) Hole cutting and interpolation stencil search be-tween the volume grids are performed using do-main connectivity software.21;29�31

To the frustration of overset grid users for manyyears, most common grid generation packages do notcontain tools that are convenient for accomplishing theabove steps. OVERGRID was speci�cally designedto connect steps 2 to 7, taking full advantage of thefreedom allowed by the overset approach to grid gen-eration. Implementation of the various overset relatedtools in OVERGRID are discussed in the subsections

6


Fig. 8 Automatically generated feature curves forthe V-22 fuselage and wing. Surface panels for thewing are not shown.

below. In each case, a code with the same name thatruns in batch mode also exists in the Chimera GridTools package.

4.1. Surface Curve Creation Tools

The SEAMCR module is used to create surfacecurves from a surface de�nition consisting of multi-ple panel networks or a surface triangulation. Forboth types of geometry de�nitions, two methods areavailable for making surface curves: automatic extrac-tion of surface feature curves, and intersection with aCartesian cutting plane. In regions of low surface �-delity, e.g., a curved region represented by just a fewpanels or triangles, many small curves may be au-tomatically extracted. A simple �lter is available insuch cases to reduce the number of curves generated.Persisting extraneous curves can then be deleted in-teractively. Fig. 8 shows feature curves automaticallyextracted by SEAMCR for the panel network geome-try of the V-22 tiltrotor fuselage, wing and tail. Someextraneous sharp-edge curves are produced near thejunction between the horizontal tail and the fuselagedue to poor local surface geometry resolution. Fig. 9shows a surface curve created by intersecting a givenCartesian plane with a surface triangulation. In gen-eral, multiple curves may be generated depending onthe location of the cutting plane.Two additional methods are available for creating

surface curves on surface triangulations: connectingtwo speci�ed vertices along existing triangle edges, andconnecting two speci�ed vertices via a direct surfacepath (Fig. 10). The �rst method is particularly usefulfor constructing curves along high curvature contourssuch as wing leading edges, while the second can beutilized to create a surface curve between two pointsvisible to each other along the surface.

Fig. 9 Surface curve created by intersection withspeci�ed Cartesian cutting plane on a triangula-tion.

(a) (b)

Fig. 10 Surface curve creation between two givenvertices on triangulation. (a) By connecting exist-ing edges. (b) By constructing direct path.

4.2. Surface Grid Generation Tools

After appropriate surface curves have been created,the user must determine a decomposition of the sur-face geometry into domains suitable for surface gridgeneration. The decomposition process frequently re-sults in domains bounded by only one feature curve.Since neighboring grids are allowed to overlap arbi-trarily, the other three boundaries of these domainscan be freely oated. Such exible requirements areideally suited for hyperbolic surface grid generation.In cases where two or more feature curves bound adomain, algebraic methods are more appropriate. TheSURGRD module described in Section 4.2.1 below hasbeen designed for the above gridding strategy.After creating surface grids in domains bounded by

one or more feature curves, there may be regions ofthe surface that have not yet been covered. The usercan de�ne more surface curves and then use SURGRDto create more surface grids to �ll the gaps; or usethe SBLOCK module described in Section 4.2.2 to �llthe gaps with automatically generated overlapping al-gebraic grids. In certain applications, it is desirableto create a wake cut behind an airfoil shape geometry

7


to form a C-grid. The WKCUT module discussed inSection 4.2.3 has been designed for this purpose.

4.2.1. SURGRD - Surface Grid Generator

The SURGRD32 module is used to create surfacegrids from 1, 2, 3, or 4 initial curves, using hyperbolicmarching, algebraic marching, or trans�nite interpo-lation (TFI) methods. Surface grids are created toconform to a bilinear representation of the surface ge-ometry de�ned by a collection of panel networks or asurface triangulation.For domains bounded by one initial curve, hyper-

bolic or algebraic marching is used. In most situations,hyperbolic marching is selected to provide orthogo-nality for the grid lines emanating from the initialcurve. However, algebraic marching is sometimes moresuitable to create skewed grid lines due to geometricconstraints, e.g., marching from a wing/body intersec-tion curve onto a swept wing by following a family ofisoparametric lines on the wing surface de�nition.For both hyperbolic and algebraic marching, OVER-

GRID provides widgets to specify the marching dis-tance, initial and/or end spacings, and either thenumber of points to use or the maximum stretchingratio. Also, for hyperbolic marching, limited control ofgrid lines emanating from the end points of the initialcurve is given via boundary condition speci�cations,e.g., constant plane, periodic, free oating, and oat-ing along a speci�ed curve. The marching distancecan be made to vary for di�erent points along theinitial curve. Smoothing parameters are available foradjustment but are rarely needed except in very diÆ-cult cases. Fig. 11 shows the SURGRD window withwidgets for setting input parameters and the Displaywindow with several V-22 surface grids created by hy-perbolic marching.For domains bounded by two opposite, two adjacent,

three, or four initial curves, trans�nite interpolation isused. Additional straight lines are automatically con-structed by SURGRD for two-curve and three-curvecases to �ll in the missing bounding curves. Fig. 12shows the automatically simpli�ed SURGRD windowfor two opposite initial curves and the Display windowwith a TFI grid created between the two curves. TheSURGRD window simpli�es even further for two adja-cent, three or four curves since no stretching functionneeds to be speci�ed.

4.2.2. SBLOCK - Surface Gap Grid Generator

Given the surface geometry and a set of surface gridscreated around the feature curves, the SBLOCK mod-ule can be used to automatically generate algebraicsurface grids to �ll in regions on the surface not al-ready covered. Details of the SBLOCK algorithm andcode are found in Ref. 19. The OVERGRID interface

is very simple, and provides widgets for the input ofthe uniform global grid spacing to be used for the alge-braic grids. In practice, this module is rarely utilizedsince it tends to generate a large number of small gridsthat results in poor ow solver eÆciency.

4.2.3.WKCUT - Wake Cut Surface Grid Generator

The WKCUT module is used to generate and add awake cut to the surface grid of an airfoil shape such as awing, ap, slat, �n, or pylon to form a C-grid. Defaultparameters are automatically set for the streamwiseextent of the wake, the number of points used andthe grid spacing. For more diÆcult cases such as ahigh/low wing and fuselage, the user can select pa-rameters to modify the de ection angle of the wakecut such that the cut intersects the fuselage. This in-tersection requirement is needed for the constructionof a collar grid33 in the wing/fuselage junction.

4.3. Volume Grid Generation Tools

After creating a set of overlapping surface grids,body-conforming volume grids have to be generated.Again, the overset approach only requires neighboringvolume grids to overlap. This allows the speci�cationof just the surface grid while the other �ve faces ofthe volume domain are free to oat. A hyperbolicmarching scheme is particularly suited for this type ofgrid. Signi�cant user and computer time savings overiterative elliptic methods are possible using a march-ing scheme. Only one instead of six faces needs to bede�ned. Moreover, hyperbolic methods naturally pro-vide the tight clustering needed near the surface forviscous computations, as well as high quality nearlyorthogonal grids everywhere. The HYPGEN34;35 mod-ule described in Section 4.3.1 has been designed toperform hyperbolic volume grid generation.Body-conforming volume grids are usually grown

a constant distance from the body, typically a frac-tion of the body length so that the outer boundariesof the volume grids are well clear of wall-boundedviscous e�ects. The BOXGR module described in Sec-tion 4.3.2 can then be used to automatically createstretched Cartesian box grids around the near �eldvolume grids and extend the computational domainto the far �eld. An alternative to the BOXGR ap-proach is to use multiple layers of adaptive o�-bodyCartesian grids that are automatically generated bythe OVERFLOW-D36;37 module as discussed in Sec-tion 4.3.3.

4.3.1. HYPGEN - Hyperbolic Field Grid Generator

The HYPGEN34;35 module is utilized to create athree-dimensional volume grid by marching from a sur-face grid. For two-dimensional cases, a �eld grid on a

8


Fig. 11 Hyperbolic surface grids created by SURGRD for part of the V-22. Points on initial curves areindicated by symbols.

Fig. 12 A surface grid created by trans�nite interpolation between two opposite curves in SURGRD.

plane is generated by marching from a curve. Themarching distance, initial and/or end grid spacingsand the number of points to use in the marching di-rection are speci�ed by the user. Boundary conditionsare automatically selected by OVERGRID based onthe topology of the given surface grid, e.g., periodic-ity, singular axis point, constant plane, and others. Afree oating boundary is selected if no special topol-ogy is detected. The user also has the choice to enforcedi�erent boundary conditions and to adjust smoothingparameters if needed for diÆcult cases such as highlyacute concave corners. Since all near-body volumegrids employ the same stretching function in the nor-mal direction, OVERGRID allows the user to generatethe volume grids for a group of surface grids using thesame parameters with a single click of a button. Aseach volume grid is created, a table is displayed whichshows if any negative Jacobians are found in each grid.

Fig. 13 shows the HYPGEN window with widgetsfor setting input parameters, and the Display win-

dow with several volume grids created by hyperbolicmarching. For all volume grids shown here, a geo-metric stretching is used in the normal direction withthe same marching distance, initial spacing, numberof points and smoothing parameters. Default bound-ary conditions are employed for all grids (free oatingfor the cases shown). The eight volume grids showncontain a total of about 460000 points. It takes lessthan 15 seconds of wall clock time (user's labor timeplus CPU time) to generate all eight grids on a SiliconGraphics R12000 workstation.

4.3.2. BOXGR - Stretched Cartesian Grid Generator

The BOXGR module is used to create a Cartesianbox grid consisting of an interior core with uniformspacing, and with optional stretched outer layers in theplus and minus X, Y, and Z directions. In BOXGR'sautomatic mode, the user selects one or more near-�eld volume grids, and BOXGR automatically creates

9


Fig. 13 Samples of hyperbolic volume grids for the X-38 Model-G created under the HYPGEN interface.

a Cartesian box grid with uniform spacing that com-pletely encloses all the selected volume grids. Theuniform spacing is automatically chosen to match theaverage grid spacing at the outer boundaries of thevolume grids, thus providing good inter-grid commu-nication. In BOXGR's manual mode, coordinates ofthe corners of the interior core can be explicitly pre-scribed. In both automatic and manual mode, extrastretched layers in all directions can easily be added byspecifying a distance and a stretching ratio (Fig. 14a).The computational domain can be extended to the

far �eld via the stretched layers or the ellipsoidal shelloption in BOXGR. In the latter, BOXGR automati-cally generates an ellipsoidal surface grid that �ts oneor more cells inside the outer boundaries of a givenstretched Cartesian box grid, thus providing properoverlap; and the grid spacings in the tangential direc-tion are made to match those of the Cartesian grid. Ahyperbolic volume grid can then be grown from thissurface to extend the computational domain to the far�eld. The ellipsoid topology allows a uniform expan-sion of the grid, resulting in grid point savings over thestretched Cartesian box option which keeps the tightuniform core spacing out to the far �eld (Fig. 14b).However, a drawback of the ellipsoid topology is thepresence of the singular polar axes which could reduce ow solver stability for time accurate computations.

4.3.3. OVERFLOW-D Cartesian Grid Generator

The OVERFLOW-D36;37 program provides an op-tion to automatically create o�-body Cartesian grids.Starting from a collection of near-body curvilinear vol-ume grids, o�-body Cartesian grids of successive levelsof re�nements are created. The �nest level (level 1) isclosest to the body and encloses all near-body gridswith a grid spacing that matches the outer boundary

(a)

(b)

ellipsoidal surface

box grid

Fig. 14 Automatically generated grids by theBOXGR module. (a) Stretched Cartesian box gridwith uniform core. (b) Far �eld ellipsoidal surfaceand volume grid.

10


Fig. 15 O�-body Cartesian grids for the X-38 cre-ated by the OVERFLOW-D module.

spacings of the near-body grids. Progressively coarserlevels of grids are generated based on proximity tothe body. OVERGRID provides an interface for thespeci�cation of the level 1 grid spacing and special con-trol planes at the outer boundaries of the Cartesiandomain, e.g., a symmetry plane or a ground plane.OVERFLOW-D can be called from OVERGRID inthe o�-body Cartesian grid generation mode and theresults displayed in the graphics window (Fig. 15).

5. Boundary Conditions Selection

After creating the volume grids, it is advantageousto determine the boundary conditions to be appliedto each grid prior to performing domain connectivityand ow solution computation. This is because bound-ary condition information constitutes a signi�cant partof the input required for both of these procedures.OVERGRID o�ers a module to automatically selectboundary conditions for each grid based on grid topol-ogy (periodic, singular axis, constant plane, symmetryplane, wake cut). Furthermore, an intelligent estimateon the locations of viscous and inviscid walls is madebased on heuristic rules. A menu is also available inthe interface for manual override of any automaticallyselected inputs (Fig. 16). Additionally, a fast visualcheck on the automatically selected values is availablein the Display window where grid surfaces can be col-ored by boundary condition type. The resulting setof boundary conditions for each grid can be writtento input �les for domain connectivity tools and owsolvers such as OVERFLOW.38 Moreover, existing in-put �les for such tools where the boundary conditionswere selected manually can be read into OVERGRIDand their validity checked on the graphical display.Overall, signi�cant time savings can be achieved with

this interface in the laborious and error-prone processof boundary conditions speci�cations for grid systemsinvolving a large number of grids.

6. Domain Connectivity Tools

The task of domain connectivity involves surfaceprojection, hole cutting and fringe points interpola-tion. Interpolation stencils are adjusted in surfaceprojection to remove the e�ect of slight mismatchesbetween overlapping surface grids due to discretiza-tion. Hole cutting is the identi�cation and taggingof �eld grid points from one grid that lie inside thesolid boundary of a body. Fringe points interpolationis needed to provide communication between neigh-boring grids. Such points arise at the boundaries ofa grid that overlap with another grid (outer bound-ary points), and at the boundaries of holes left by thehole cutting process (hole boundary points). Interpo-lation stencils from neighboring grids are sought forthese points which require inter-grid communication.OVERGRID provides connection to two do-

main connectivity programs: PEGASUS521 andOVERFLOW-D.29;37 Both programs utilize the com-plete list of boundary conditions for each grid to deter-mine the locations of fringe points. Such informationis supplied by OVERGRID's boundary conditions se-lection module (Section 5).PEGASUS5's approach to hole cutting relies on an

automatic capability plus additional manual hole cutswhich are sometimes needed for diÆcult cases. The au-tomatic hole cutting surfaces can be derived directlyfrom the list of solid wall boundary conditions that areset for each grid. After writing an OVERFLOW38 in-put �le from OVERGRID, one of the PEGASUS5 toolswill take this �le and generate a PEGASUS5 input �leand also create the directory structure necessary tostart the run.OVERFLOW-D's approach to hole cutting is based

on object X-rays.29 Hole cutters have to be user-de�ned, where a hole cutter is a collection of surfacesthat forms a closed volume. Object X-rays are thenbuilt from the hole cutters. Each object X-ray consistsof a set of pierce points on the hole cutter's surface.The list of grids that are cut by each hole cutter alsohas to be speci�ed. OVERGRID provides a graphicalutility for specifying the hole cutters and generatingthe object X-rays under the GEN X utility. Resultingobject X-rays are visualized by sweeping through con-stant X or Y planes with the surface pierce points dis-played by symbols (Fig. 17). A menu is also availableto create the input �le and execute the domain connec-tivity module in OVERFLOW-D. After execution, theresulting grid �le with connectivity iblanks is automat-ically read into OVERGRID. Fringe points that failedto �nd interpolation stencils (orphan points) are auto-matically displayed. The number of orphan points for

11


Fig. 16 Left panel: boundary conditions display with outline of volume grids. Surfaces with prescribedboundary conditions are colored by type. Boundary conditions for only one grid are shown here. Rightpanel: boundary conditions speci�cation widgets.

Fig. 17 Object X-rays for the X-38 Model-G cre-ated by the GEN X module. Pierce points ongeometry using constant-X cutting plane are shownby black symbols.

each grid and the overall total are reported. With thisinteractive tool, the user can then make appropriatemodi�cations to the volume grids and domain connec-tivity inputs and re-run OVERFLOW-D to eliminatethe orphan points.

7. Flow Solver Input Preparation

OVERFLOW38 and OVERFLOW-D37 are com-pressible Navier-Stokes ow solvers for overset gridsthat are widely used for both research and productionpurposes. Many options have been implemented to al-low for a variety of ow physics, numerical methods,geometrical complexities, and computer architectures.The general nature of the code allows a potentiallycomplicated input �le. For example, about 30 param-eters can be speci�ed globally and about 70 parameterscan be speci�ed locally for each grid in OVERFLOW.A namelist input format is utilized where defaults are

automatically set for all parameters and only the pa-rameters that apply to a particular problem need tobe prescribed. However, it is still a formidable taskfor a novice user to understand exactly which param-eters, out of the long speci�able list, are required fora particular problem.OVERGRID o�ers a simpli�ed interface for the se-

lection and entry of the most commonly used optionsand inputs such as ow conditions, numerical meth-ods, time steps, and turbulence models (Fig. 18). Oneimportant feature is the automatic selection of bound-ary conditions with optional manual override underthe boundary conditions module described in Sec-tion 5. With the OVERGRID interface, the namelistinput preparation time for the OVERFLOW andOVERFLOW-D ow solvers can be reduced by a fac-tor of at least 2 to 5 for most complex problems.

8. Multiple-Component Dynamics

Steady ow computations on complex geometriesare now routinely performed in both research and pro-duction environments. Although more expensive, un-steady ow computations are also becoming more com-mon as a result of advances in computer technology.However, computations involving multiple componentsin relative dynamic motion remain scarce. This is dueto three main reasons. First, this class of problemscan only be computed in time-accurate mode, and thusare inherently costly. Second, domain connectivity be-tween grids needs to be re-computed at every timestep, thus incurring more cost. Third, the dynamicsinvolved can be extremely complex, and there is nosimple method for the user to specify the motion. Aninterface is designed in OVERGRID to help alleviate

12


Fig. 18 OVERFLOW ow solver input optionsinterface.

the third problem.

Components can be de�ned in OVERGRID whereeach component may be stationary or move relativeto an inertial frame. The interface allows the user tode�ne the grids that belong to each component. Also,each component is allowed to move in one of two ways:(1) six-degrees-of-freedom motion under the in uenceof aerodynamic forces and moments, gravity, and anyexternally applied loads, or (2) prescribed motion.

In case (1), the ow solver computes the resultantaerodynamic forces and moments acting on each com-ponent and the dynamics is determined by solving theequations of motion. The input parameters requiredare component properties such as mass, weight, mo-ments of inertia, and externally applied loads. Widgetsare provided in OVERGRID for the entry of these pa-rameters. This information is then communicated tothe ow solver via a namelist input �le.

In case (2), the motion of each component needs tobe prescribed by the user and the information trans-mitted to the ow solver. OVERGRID provides aninterface for specifying a simple class of motions con-sisting of a sequence of constant translational velocitiesand angular speeds for each component. The pro-grammed motion can then be viewed inside OVER-GRID via an animation of the de�ned components(Fig. 19). Work is in progress to represent this in-formation in a data �le that can be read by di�er-ent ow solvers (also see Section 9). Currently, theOVERFLOW-D solver requires a user to supply a sub-routine containing the prescribed motion information.While this allows the most general types of motion pos-sible, it is diÆcult for a novice user to get started evenfor very simple cases. The ability to prescribe a wideclass of motions via a data �le will signi�cantly reducethe user work required for problem setup and allow

Fig. 19 Component dynamics animation modulein OVERGRID showing a �ctitious Space ShuttleLaunch Vehicle separation scenario.

simulations involving multiple components in relativemotion to be more common place.

9. Conclusions and Future Plans

A uni�ed and practical graphical user interface foreÆcient and streamlined handling of the overset gridcomputational simulation process is presented. Thegoal is to reduce the manual e�ort needed in thesteps required prior to running the ow solver. Pro-cedures covered include geometry processing, surfaceand volume grid generation, boundary conditions spec-i�cation, domain connectivity, diagnostics, ow solverinput preparation, and treatment of multiple compo-nent dynamics.The OVERGRID project started a few years ago

when there was no satisfactory interface that containedall the tools necessary for eÆcient overset grid gener-ation. Today, it's capabilities have expanded beyondgrid generation to the above list, and remains one ofthe few options available to overset grid users withcomplex geometry applications. Typically, a three tofour-fold reduction in manual process time is achievedusing the software. Further signi�cant time savingsare obtained using the scripting capability. The soft-ware is being utilized by numerous U.S. organizationsin various government laboratories, industry groups,and universities under a non-disclosure agreement.The future development plan for OVERGRID con-

sists of a long list. One of the major items is toimplement the capability to create surface grids basedon CAD geometry de�nitions such as trimmed NURBSsurfaces and solid models. Another area that requireswork is the automatic coverage of the surface domainfrom multiple hyperbolically generated surface grids.

13


As domain connectivity tools evolve, OVERGRID willkeep hooks to the most current and robust modules.Input generation for ow solvers other than OVER-FLOW can be easily added. Work is in progress todesign a general framework for multiple component hi-erarchy and relative motion dynamics, expanding thesimpli�ed approach described in this paper. Commu-nications to the ow solver will most likely be made viaan XML �le. A exible format will be designed to al-low interfacing with ow solvers from both structuredand unstructured grid approaches.

Acknowledgements

The development of OVERGRID has vastly bene-�tted from the expertise of the author's colleagues atNASA. Mouse transforms and direct screen picking ofentities were borrowed from routines implemented bySteven Nash. Insightful ideas on graphics and GUIswere shared by Dr. Shishir Pandya. Also, the author isgrateful to the many brave testers throughout OVER-GRID's evolution, especially Drs. Robert Meakin, EarlDuque, and Pieter Buning for their patience, encour-agement and constructive comments. Finally, the soft-ware would not be as practical today without the feed-back from numerous users. The X-38 geometry andgrids shown in this paper have been kindly providedby James Greathouse and Reynaldo Gomez of NASAJohnson Space Center. This work was funded by anelement of the DOD High Performance ComputingModernization Program known as CHSSI (CommonHigh Performance Scalable Software Initiative) from5/1996 to 9/2000, the NASA High Performance Com-puting and Communications Program from 9/2000 to9/2001, and the NASA Computer, Information andCommunication Technology Program from 10/2001 tothe present.

References

1 Steger, J. L., Dougherty, F. C. and Benek, J. A.,\A Chimera Grid Scheme," Advances in Grid Genera-tion, K. N. Ghia and U. Ghia, eds., ASME FED-Vol. 5,June, 1983.

2 Smith, M., Chawla K., and Van Dalsem, W. R.,\Numerical Simulation of a Complete STOVL Aircraftin Ground E�ect," AIAA Paper 91-3293, 9th AIAAApplied Aerodynamics Conference, September, 1991.

3 Atwood, C. A. and Van Dalsem, W. R., \Flow�eldSimulation about the SOFIA Airborne Observatory,"AIAA Paper 92-0656, January, 1992.

4 Meakin, R., \Moving Body Overset Grid Meth-ods for Complete Aircraft Tiltrotor Simulations,"AIAA Paper 93-3350, Proceedings of the 11th AIAAComputational Fluid Dynamics Conference, Orlando,Florida, July, 1993.

5 Gomez, R. J. and Ma, E. C., \Validation of a LargeScale Chimera Grid System for the Space ShuttleLaunch Vehicle," AIAA Paper 94-1859, Proceedings

of the 12th AIAA Applied Aerodynamics Conference,Colorado Springs, Colorado, June, 1994.

6 Slotnick, J. P., Kandula, M. and Buning,P. G., \Navier-Stokes Simulation of the Space Shut-tle Launch Vehicle Flight Transonic Flow�eld Using aLarge Scale Chimera Grid System," AIAA Paper 94-1860, Proceedings of the 12th AIAA Applied Aerody-namics Conference, Colorado Springs, Colorado, June,1994.

7 Atwood, C. A., \Computation of a ControlledStore Separation from a Cavity," J. of Aircraft, Vol. 32,No. 4, 1995, pp. 846{852.

8 Duque, E. P. N, Berry J. D., Budge A. M. andDimanlig A. C. B., \A Comparison of Computed andExperimental Flow�elds of the RAH-66 Helicopter,"Proceedings of the 1995 American Helicopter SocietyAeromechanics Specialist Meeting, Fair�eld County,Connecticut, 1995.

9 Gee, K., Murman, S. M. and Schi�, L. B., \Com-putation of F/A-18 Tail Bu�et," J. of Aircraft, Vol. 33,No. 6, 1996, pp. 1181{1189.

10 Srinivasan, G. R. and Klotz, S. P., \Featuresof Cavity Flow and Acoustics of the StratosphericObservatory For Infrared Astronomy," Proceedings ofthe ASME Fluids Engineering Conference, Vancouver,British Columbia, Canada, June, 1997.

11 Meakin, R. L., \Unsteady Aerodynamic Simu-lation of Static and Moving Bodies Using ScalableComputers," AIAA Paper 99-3302, Proceedings of the14th AIAA Computational Fluid Dynamics Confer-ence, Norfolk, Virginia, June, 1999.

12 Gea, L. M., Halsey, N. D., Intemann, G. A.and Buning, P. G., \Applications of the 3-D Navier-Stokes Code OVERFLOW for Analyzing Propulsion-Airframe Integration Related Issues on SubsonicTransports," ICAS Paper 94-3.7.4, 19th Congress ofthe International Council of the Aeronautical Sciences,September, 1994.

13 Wai, J., Herling, W. W. and Muilenburg, D. A.,\Analysis of a Joined-Wing Con�guration," AIAA Pa-per 94-0657, January, 1994.

14 Rogers, S. E., Cao, H. V. and Su, T. Y., \GridGeneration for Complex High-Lift Con�gurations,"AIAA Paper 98-3011, 29th AIAA Fluid DynamicsConference, Albuquerque, New Mexico, June, 1998.

15 Slotnick, J. P., An, M. Y., Mysko, S. J., Yeh,D. T., Rogers, S. E., Roth, K., Nash, S. M. and Baker,M. D., \Navier-Stokes Analysis of a High Wing Trans-port High-Lift Con�guration with Externally BlownFlaps," AIAA Paper 2000-4219, 18th AIAA AppliedAerodynamics Conference, Denver, Colorado, August,2000.

16 Rogers, S. E., Roth, K., Cao, H. V., Slotnick,J. P., Whitlock, M., Nash, S. M. and Baker, M. D.,\Computation of Viscous Flow for a Boeing 777 Air-craft in Landing Con�guration," AIAA Paper 2000-4221, 18th AIAA Applied Aerodynamics Conference,Denver, Colorado, August, 2000.

17 Brand, A., Peryea, M., Wood, T., and Meakin,R., \Flow�eld and Download Measurements and Com-putation of a Tiltrotor in Hover," American HelicopterSociety - International 57th Annual Forum, Washing-ton, D.C., May, 2001.

14


18 Naik, D. A., and Om, D., \Assessment of theOVERFLOW Navier-Stokes Code for Various Air-plane Components," SAE Paper Number 2001-01-2976, World Aviation Conference, Seattle, Washing-ton, September, 2001.

19 Chan, W. M. and Meakin, R. L., \AdvancesTowards Automatic Surface Domain Decompositionand Grid Generation for Overset Grids," AIAA Pa-per 97-1979, Proceedings of the AIAA 13th Compu-tational Fluid Dynamics Conference, Snowmass, Col-orado, June, 1997.

20 Chan, W. M. and Gomez, R. J., \Advances inAutomatic Overset Grid Generation Around SurfaceDiscontinuities," AIAA Paper 99-3303, Proceedings ofthe AIAA 14th Computational Fluid Dynamics Con-ference, Norfolk, Virginia, June, 1999.

21 Suhs, N. E., Rogers, S. E., and Dietz, W. E., \PE-GASUS 5: An Automated Pre-processor for Overset-Grid CFD," AIAA Paper 2002-3186, 32nd AIAA FluidDynamics Conference, St. Louis, Missouri, June, 2002.

22 Chan, W. M., Gomez, R. J., Rogers, S. E., andBuning, P. G., \Best Practices in Overset Grid Gen-eration," AIAA Paper 2002-3191, 32nd AIAA FluidDynamics Conference, St. Louis, Missouri, June, 2002.

23 Brown, D. L., Henshaw, W. D. and Quinlan,D. J.,\Overture: Object-Oriented Tools for Over-set Grid Applications," AIAA Paper 1999-3130, 17thAIAA Applied Aerodynamics Conference, Norfolk,Virginia, 1999.

24 Chan, W. M., Rogers, S. E., Nash, S. M., Buning,P. G., and Meakin, R. L., \User's Manual for ChimeraGrid Tools, Version 1.6," NASA Ames ResearchCenter, September, 2001 (http://www.nas.nasa.gov/~rogers/cgt/doc/man.html).

25 Welch, B., Practical Programming in Tcl and Tk,3rd ed., Prentice Hall, 1999.

26 Walatka, P. P., Buning, P. G., Pierce, L. andElson, P. A., \PLOT3D User's Manual," NASA TM101067, 1990.

27 Walatka, P. P., Clucas, J., McCabe, R. K., Ples-sel, T. and Potter, R, \FAST User Guide," RND-93-010, NASA Ames Research Center, 1994.

28 Chawner, J. R. and Steinbrenner, J. P., \Auto-matic Structured Grid Generation Using GRIDGEN(Some Restrictions Apply)," Proceedings of NASAWorkshop on Surface Modeling, Grid Generation,and Related Issues in Computational Fluid Dynamics(CFD) Solutions, NASA CP 3291, May, 1995.

29 Meakin, R. L., \Object X-rays for Cutting Holesin Composite Overset Structured Grids," AIAA Paper2001-2537, 15th AIAA Computational Fluid DynamicsConference, Anaheim, California, June, 2001.

30 Maple, R. C. and Belk, D. M., \Automated SetUp of Blocked, Patched and Embedded Grids in theBeggar Flow Solver," Numerical Grid Generation inComputational Fluid Dynamics and Related Fields,Ed. N. P. Weatherill et al., Pine Ridge Press, pp. 305{314, 1994.

31 Henshaw, W. D., Chesshire, G. and Henderson,M. E., \On Constructing Three Dimensional Over-lapping Grids with CMPGRD," Software Systems forSurface Modeling and Grid Generation, Ed. RobertE. Smith, Hampton, VA, Langley Research Center,NASA CP 3143, 1992, pp. 415{434.

32 Chan, W. M. and Buning, P. G., \Surface GridGeneration Methods for Overset Grids," Computersand Fluids, Vol. 24, No. 5, 1995, pp. 509{522.

33 Parks, S. J., Buning, P. G., Steger, J. L. andChan, W. M., \Collar Grids for Intersecting Geo-metric Components Within the Chimera OverlappedGrid Scheme," AIAA Paper 91-1587, Proceedings ofthe 10th AIAA Computational Fluid Dynamics Con-ference, Honolulu, Hawaii, 1991.

34 Chan, W. M. and Steger, J. L., \Enhancementsof a Three-Dimensional Hyperbolic Grid GenerationScheme," Appl. Math. & Comput. Vol. 51, 1992,pp. 181{205.

35 Chan, W. M., Chiu, I. T. and Buning, P. G.,\User's Manual for the HYPGEN Hyperbolic GridGenerator and the HGUI Graphical User Interface,"NASA TM 108791, October, 1993.

36 Meakin, R., \Adaptive Spatial Partitioning andRe�nement for Overset Structured Grids," Comput.Methods Appl. Mech. Engrg., Vol. 189, 2000,pp. 1077{1117.

37 Chan, W. M., Meakin, R. L. and Potsdam,M. A., \CHSSI Software for Geometrically ComplexUnsteady Aerodynamic Applications," AIAA Paper2001-0593, Reno, Nevada, January, 2001.

38 Buning, P. G., Jespersen, D. C., Pulliam, T. H.,Klopfer, G. H., Chan, W. M., Slotnick, J. P., Krist,S. E. and Renze, K. J., \OVERFLOW User's Man-ual, Version 1.8s," NASA Langley Research Center,November, 2000.

15


the overgrid interface for computational simulations on overset … · 2016-10-12 · computational...

Documents