Home

Edge -- User Guide. Produced from Rev: 1069

Contents

1. Files required Application Files generated Edge ainp bound Edge aboc Edge bmsh Edge ainp preprocessor Edge bedg _p Edge bmsh Edge aboc Edge ainp edge Edge bout Edge bedg Post bout Edge aboc Edge bres Edge bini Edge log Additional files for modal Aeroelastic applications Edge bmos edge Edge_timestep bout Edge amot Edge bres modal data Edge amop Additional files for adjoint flow calculations Edge flow bout edge_ Edge adjoint bout Table 8 Required generated files for Edge applications 3 4 Files Generated Required by Edge To start Edge the only essential input data is a grid file bmsh A number of example grid files can be found in the directory examples In Table 8 the required input files and generated files are listed for Edge and the two main supporting programs As far as possible the file names are as given in the default input file Edge HOME lib default_edge ainp The main files listed table 8 in are explained below The additional files for aeroelastic applications are described in Section 3 8 All these files are FFA format files as described in Section 3 3 The file names in the main Edge input file can be set to any value provided that the same four character extension like bmos is used The main Edge input file is Edge ainp The main Edge input file is most easily set up using the Edge graphical user interface which is started using the command xedge however it can a
2. Input Variables Issue 4 1 0 A 1 General Parameters for Several Programs These file names are used in several programs Name Description Value Default CFIMSH Mesh file name bmsh amsh Edge bmsh CFIEDG Edge file name from the preprocessor bedg aedg Edge bedg CFIBOC Boundary condition file aboc bboc Edge aboc CFIRES Residual file bres ares Edge bres CFIOUT Solution output file also used as bout aout Edge bout initial solution for restart INPRES 2 CFIMEAN Name of time averaged solution file bout aout Edge_mean bout CFIINI Initial solution file INPRES 1 bini aini Edge bini CFIPOST Post processing file name bout Post bout Table 17 File names The parameter NPART applies to the preprocessor the flow solver and the programs that split and merge solution files for the parallel computations The first six of the variables in are used by the flow solver to output lift drag and moment during the iterations The program force uses all of the parameters to compute a complete set of forces and moments Name Description Value Default NPART Number of partitions NPART gt 1 specifies a 1 parallel calculation Used by the preprocessor and flow solver 1 Table 18 Parallel calculation 67 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default IDCLP Direction in which the force is projected
3. 14 5 5 5 5 Edge User Guide dual_grid_data title 12 2_adapt_1 n_dim 2 grid edge_nodes 112 volumes 2 4808954890E 10 n_nodes 88703 coordinates 1 1408840417861 n_edges 220397 edge_surfaces n_edge_colors 8 195638E 08 2 color_indices 1 1262 n_b_colors 1 n_bound 4 boundary b_name slat p n_b_nodes 37 b_nodes 279 280 281 b_surfaces 1 011956E 04 b_inner_nodes 1519 1520 boundary b_name wing p n_b_nodes 59 b_nodes 650 651 652 b_surfaces 1 328219E 04 b_inner_nodes 1890 1891 boundary b_name flap p n_b_nodes 278 273 b_nodes 1 b_surfaces 6 411895E 04 b_inner_nodes 1241 1242 boundary b_name outer n_b_nodes b_nodes b_surfaces b_inner_nodes 21 46090 46091 28 4852609634 88690 88700 Figure 5 A typical edge bedg file Issue 4 1 0 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 N 0x0 8 solution L x 1 brand Edge 4 0 0 www edge foi se LI x 1 title Edge an edge based flow solver from FFA LI x 1 date 070204 23 19 41 LI x 1 author LI i1 program Edge LI x 1 grid_ref Grid collected from ffanet files N 0 0 8 free stream data DC xa pressure 37643 8007812500 DC x 1 temperature 300 000000000000 DC x 3 velocity 261 516998291016 DC x 1 gamma 1 39999997615814 DC x 1 rgas 286 999987781048 DC x 1 mu_reference 1 845999940996990E 005 DC x 1 prantl_nu
4. The program aetribend is used to produce a deformed bmsh mesh file from a base mesh file and a bdis surface_movement file The aetribend command line and arguments are as follows Usage aetribend in bmsh ut bmsh in bdis Options Arguments 1 in bmsh input mesh filename 2 ut bmsh output mesh filename 3 in bdis input surface_movement filename This program is a Bourne shell script which runs three Tritet programs from the Edge mesh adaption system ffatri x triffa x and moving x and two special helper programs aedis2tri see help text and aetribend tag x The mesh deformation program is moving x and aetribend tag x transfers mode attribute data from the bdis surface movement file to the output mesh file These programs can all be used directly but aetribend reduces the task to a single line com mand The most important of the five options is scale which allows the user to set maximum absolute displacement size for the output deformed mesh This is useful when the sur face movement data is a large amplitude deflection which the mesh deformation would be slow to converge or might result negative cell volume errors making the mesh unusable The scale option would typically be used for mass normalised elastic modes where the modeshape is an unphysically large surface deflection Data is added to the output mesh which allows the original scaling to be recovered This is used by the program aeputpert to rescale the
5. Thus a dataset nodes sliding as in figure 10 is not needed and the presence of such a dataset will simply be ignored by meshdeform 3 Unconstrained motion A dataset free_surface specifies that the motion at selected boundary portions should be unconstrained in the minimization of expression 5 For instance if the boundary displacement file contains the sub dataset form N 0x0 Ly free_surface L 1x1 boundary_name outlet 1 L 1x1 boundary name outlet 2 the displacement vector u for each node on boundaries labeled outlet 1 and out let_2 will be treated as an interior node and will thus be let free to attain the values that minimize expression 5 The above list indicates also the hierarchy in assigning boundary condition for meshdeform For instance if the same boundary node appears in nodes_moving and in the list of nodes associated with sliding planes it will be treated as completely constrained nodes moving The remaining boundary nodes that is those nodes that are not specified in nodes moving or contained in sliding planes or free surface will be constrained to zero displacement The option IMDOP specifies the exponent p in expression 5 The value should be p 1 or p 2 p 2 gives better mesh quality for difficult mesh deformations but the convergence rate for the iterative solver is faster when p 1 There are two options for solving the discrete Laplace like equation using the sparse direct solver
6. bmot file 3 8 7 1 The signal table File Edge amot The mottab system supports two signal formats timeseries and harmonics of which timeseries the most efficient and versatile Either signal format can be used to generate on two signal types pulse and periodic a third type step will be added soon The mottab subroutine is a table lookup function which extracts a modal coordinates and velocities from a signal table dataset which contain data for a set of modes each of which has a unique integer identifier The base mode set in Edge is defined by the bmos modal shape data dataset The signal table controls only those modes which match an identifier in the base mode set The coordinates of any other unidentified other modes in the base mode set are set to zero Examples of mottab signal table files are supplied with Edge and are stored in the direc tory examples ae inputfiles Time history plots from these files are stored in a subdirectory gpics The easiest way to learn how to use mottab is to try plotting some of these examples using the program plotmot or preferably it s Matlab equivalent and tinker with some of the variables by manual editing Similar mottab input files can easily be written in binary form using functions in the FFA Matlab Toolbox FOI dnr 03 2870 Edge User Guide Issue 4 1 0 N 0x0 7 signal table L x title sinusoid half wave serie s x signal_format
7. 001 to obtain the lift IDCDP Direction in which the force is projected 100 to obtain the drag IDCMP Direction in which the pitching moment is 010 projected IXMP Point on the axis for all moments 000 SREF Reference area to non dimensionalize 1 forces and moments CREF Reference length for the moment 1 IDCCP Direction in which the side force is 010 projected IDCNP Direction in which the yaw moment is projected 001 IDCRP Direction in which the roll moment is 100 projected BREF Reference width for the moment 1 68 Table 19 Integration of forces and momentums FOI dnr 03 2870 Edge User Guide Issue 4 1 0 A 2 Preprocessor Name Description Value Default CRDIM Coarsening ratio for multigrid in each gt 1 2 dimension 2 is recommended NLEVEL Maximum number of coarse grids produced gt 1 5 by the preprocessor RMSING Removes singletons in the agglomeration 1 0 1 1 NPLAYER Number of prismatic layers Used for 1 1 semicoarsening in the prismatic region and if semicoarsening is desired should be put to the maximum number of prismatic layers in the grid CRATIOB Coarsening ratio in the prismatic layer gt 1 4 when NPLAYER gt 1 CRATIOB 4 is recommended If NPLAYER 1 CRATIO apply PDUAL Output the dual grid in EnSight 5 format 0 1 0 to be plotted COLTYPE Type of edge colouring 0 1 2 1 O no action 1 Cache reordering trying all algorithms 2 Vector colouring 3 Cache reordering using saw alg
8. 1 0 36 Viy ATV Ix 3 5 Hints and Recommendations 3 5 1 Specifying the Free Stream Viscosity and Reynolds Number The free stream viscosity should be specified such that the desired Reynolds number is obtained for a viscous calculation INSEUL 1 This can be done two ways If the free stream temperature pressure and hence density and velocity are specified such that the desired free stream Mach number is obtained the free stream viscosity Hoo RMU is computed from 1 Note that L represents a length in the dimension of the grid on which the Reynolds number is based Uco Lpoo Ho Alternatively the free stream viscosity can be computed from the Sutherland law see 2 commonly used for situations where no Reynolds number is given e g internal flows Re 1 3 2 00 _ S 1 458 10 S 110K 2 To 5 1 2 2 fxs 91 The viscosity can vary with the temperature ICOVIC 0 as in 3 o TSF E a TE T 110 or it can be constant ICOVIC 1 3 5 2 Turbulent Flow Calculations There are three main types of turbulent calculations that can be performed by Edge RANS DES and other hybrid RANS LES methods as well as LES The parameter ITURB in the input parameter file determines which model is used The DES hybrid RANS LES and LES modeling methods are unsteady approaches by na ture When used i e setting ITURB 3 or 4 the option for time accurate computation must be switched on by setting the pa
9. 4x1 BOX ROOT 27 55 38 6 RF 4x1 BOX TIP 36 63 44 4 RF x 1 THICK 2 50000000 I Lx 1 DEGX 1 E 1x1 DEGS 2 I x 1 IEDGE 0 I Lx 1 OEDGE 1 N 0x0 7 BOX_N RF 4x1 BOX_ROOT 36 63 44 44 RE 4x1 BOX TIP 40 48 46 90 RE x 1 THICK 2 50000000 E x 1 DEGX 1 T x 1 DEGS 2 I x IEDGE 0 T Lx 1 OEDGE 1 N 0x0 7 BOX_N RF 4x1 BOX_ROOT 40 48 46 9 RF 4x1 BOX_TIP 45 97 50 42 RF Lox 1 THICK 2 50000000 E x 1 DEGX 1 I x 1 DEGS 3 T x 1 IEDGE 0 I x 1 OEDGE 0 Figure 12 A typical inter grid interpolation aigi file 3 8 4 2 The Inter Grid Interpolation File aigi The aigi inter grid interpolation file is used the define a transformation mapping dis placements from the structural grid onto the wetted surface This is used by the program aedbelast to produce a set of modeshapes on the wetted surface defined by the bset move able_surface file This transformation definition data will in the future be used by other programs and subroutines for more general spatial coupling operations Currently only one interpolation method is supported polyfit a piecewise polynomial fit The aigi input file can be prepared by manual editing using the example file provided in the directory examples ae_inputfiles A listing of this file is shown in Figure 12 The dataset igi method is mandatory and selects the interpolation method Everything following this dataset is specific to the interpolation method in t
10. An optional output file new tetrahedra mesh bmsh contains all the tetrahedra created by the h refinement which is useful to check what the refinement has done The grid adaption is controlled by the following parameters in the input file e CELMAX Maximum cell size dimensional Edges longer than this value will always be divided into two edges with a new node in the middle e CELMIN Minimum cell size dimensional Edges shorter than this value will never be divided e CELFLO Maximum indicator change in one cell Normal value is 0 02 0 04 This means that the edges will be divided if the difference in the flow variables normalized with the free stream values between the two end nodes is larger than 2 4 used only if SENSOR equals 0 or 1 23 FOI dnr 03 2870 24 Edge User Guide Issue 4 1 0 RMAGNI Maximum change number of nodes maximum by this factor The grid adaption will create a grid with more nodes than the previous grid This factor sets the maximum allowed number of nodes compared to the previous grid If the adaption gives more nodes than this value no new grid will be created The user can then either increase this value or increase for example CELFLO in order to reduce the adaption sensitivity If every tetrahedra are refined into 8 new ones RMAGNI must be larger than 8 0 IPROJ If IPROJ equals 1 the nodes created on the boundaries are projected to an approximation of the original geometry Ot
11. Conway 2004 The bound ary may be connected to the nacelle exhaust boundary 3 6 4 12 Nacelle Exhaust This boundary condition is intended for an engine nacelle exhaust boundary It is required to specify pressure and temperature ratios Details about this boundary condition is given in Conway 2004 3 6 4 13 Mass Flow Inlet The total temperature the flow direction and the mass flow through the boundary are spec ified The Mach number is extrapolated upstream and the total pressure adjusted to get the required mass flow The boundary condition is intended as a subsonic inflow condition for internal aerodynamics when the mass flow is known If local varying free stream data is desired at the boundary the following minimum data is required N 0x0 3 boundary_data RF NB x 1 total_temperatur RF NB x 3 flow_direction Other dependent variables like turbulence etc may be given and are used if found 45 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 46 3 6 4 14 Mass Flow Outlet A boundary condition based on the weak static pressure condition and intended for subsonic outflows in internal aerodynamics The mass flow through the boundary is specified and the static pressure adjusted so that the mass flow through the boundary matches the target Local boundary data cannot be supplied 3 6 4 15 Mass Flow Outlet JWS A boundary condition based on the weak static pressure condition and intended for subsonic outflows in
12. File Edge amop with parameters for Edge amop coupled solutions IAEOPT O rigid grids 0 100 101 102 O 100 modal rigid fixed grids 101 modal prescribed motion 102 modal coupled orthogonal modes IFORCED 0 no forcing 0 1 0 1 external modal force MOTTAB file NITIN Structural equations updated in every gt 0 1 NITIN inner iteration 1 IFOFFSET Fluid force offset gt 0 0 O no offset Jig Shape 1 offset at last timestep of res history gt 1 offset at numbered timestep Table 32 Aeroelastic control parameters Name Description Value Default IPREPA Turns on the low speed preconditioning 1 0 1 0 RKBET 2 The parameter 3 gt 0 4 RMO The parameter Mo gt 0 1 78 Table 33 Low speed preconditioning FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default IWRSOL Output solution and residual file every 1 500 IWRSOL iteration IMULOU Outputs a unique solution 1 every IWRSOL 0 1 0 iteration by adding the iteration number to the suffix of the output file specified in CFIOUT EPS Small number to avoid division with zero 1 0e 28 ISTOUT Output convergence data to standard output 1 1 every ISTOUT iteration 1 1 density NRRES Residual for which convergence is monitored in RESRED RESFMG RESTAQ Table 34 Miscellaneous 79 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 80 FOI dnr 03 2870 Edge User Guide Issue 4 1
13. and used at supersonic conditions 3 6 4 6 Pressure Outlet Weak Static Pressure A boundary condition intended for subsonic outflows where the static pressure is specified It may also be used at local supersonic outflow conditions in that case the pressure is not specified all variables extrapolated If local varying free stream data is desired at the pressure needs to be specified N 0 0 1 boundary_data RE NB 1 x x pressure mr FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 6 4 7 Extrapolation Weak Extrapolation All variables are extrapolated with this condition Note that theoretically this boundary condition is only stable for supersonic outflow 3 6 4 8 Free stream Weak Free Stream All variables specified weakly to free stream values Local varying free stream data may be supplied at the boundary any variable may be spefied locally Note that theoretically this boundary condition is only stable for supersonic inflow 3 6 4 9 Propulsion Inlet Boundary condition for coupling to a ram jet module This boundary condition is not avail able in the official Edge distribution 3 6 4 10 Propulsion Outlet Boundary condition for coupling to a ram jet module This boundary condition is not avail able in the official Edge distribution 3 6 4 11 Nacelle Inlet This boundary condition is intended for an engine nacelle infow to which the capture area e is specified Details about this boundary condition is given in
14. be supplied 3 6 4 3 Total States Inlet Weak Total States Total pressure total temperature and a flow direction are specified to constant or varying conditions The Mach number is extrapolated upstream The boundary condition is intended as an subsonic inflow condition for internal aerodynamics If local varying free stream data is desired at the boundary the following minimum data is required N Om 0 3 boundary data RF NB x 1 total pressure RF NB x 1 total temperatur RF NB x 3 flow direction Other dependent variables like turbulence etc may be given and are used if found 3 6 4 4 Total States Inlet Strong total states pressure This boundary condition is similar to the condition above The difference is that the static pressure is extrapolated upstream the other variables are extrapolated An additional dif fernece is the strong condition implied on all variables Varying free stream data may be supplied as well 3 6 4 5 Incompressible Inlet Weak Inlet Density and velocity specified to constant or varying free stream conditions Pressure is extrapolated if the flow is locally subsonic it is specified if it is supersonic The boundary condition is intended for an internal inflow boundary If local varying free stream data is desired at the boundary the following minimum data is required N 0x0 2 koundary_data RF NB x 1 density RF NB x 3 velocity A local pressure field may be given as well
15. for the spatial discretization of the momentum equations must be used namely IUPWIN 0 For the turbulence transport equation both schemes second order central or upwind schemes can be selected by setting IUPWKZ 0 or IUPWKZ 1 correspondingly but the second order central scheme is recommended for DES hybrid RANS LES and LES computations With the central scheme in DES and LES the two parameters that con trol the artificial dissipation VIS2 and VIS4 for the momentum equations and VIS2KZ and VIS4KZ for turbulence transport equation s may be specified with a value slightly smaller than that in RANS A typical value of between 0 01 0 02 can be taken for VIS4 and VIS4KZ respectively For low speed low Mach number flow calculations it is recommended to have a nega tive value of the variable TURFIX for an upwind discretization of the turbulent equations TURFIX 1 1 is recommended A positive value may add too much dissipation and hence destroy the accuracy 3 5 2 5 Specification of Transition It is possible to specify explicitly the position at which the transition from laminar to turbu lent flow takes place The user should then specify the variable NUAMREG to the number FOI dnr 03 2870 Edge User Guide Issue 4 1 0 of laminar regions in the computational domain A laminar region is a region of a wall boundary that can be specified by a number of limiting planes LAMINAR REGION de fined by a coordinate and a direction into t
16. mottab harmonics file The first few lines of the standard listing for one of the example files is shown in Figure 15 As for the timeseries format the signal_table dataset comprises a block of header datasets and a mode set containing several mode datasets The signal type for the harmonics has a default setting of periodic but it may also be set to puls but if so a t duration input is also required The inputs t start and amplitude factor are optional and have the same use as for timeseries input Each mode dataset has four mandatory attributes identifier frequency hz cos coefficients and degrees_phases and one defaulted attribute mean_position The identifier is as usual a unique positive integer Referring to 4 the dataset names and terms match up as follows mean position ag frequency hz f cos coefficients ap and degrees phases d If the mean_position dataset is not present the term ao is set to zero The Fourier series of each mode are independent and can be of any size 3 9 Edge Modal Coupled Solutions 3 9 1 Structural Model and Modal Parameters Structural coupling in Edge is implemented in a simple formulation based on orthogonal structural modes For details of this formulation refer to the Edge Theoretical Formula tion manual This modal coupled simulation option is selected by setting the main input file parameter IAEOPT 102 which is located in the AEROELASTICS dataset The Edge
17. partition per processor 3 1 4 Running the Flow Solver To run the solver execute the solver script For sequential execution the name of the script is edge run For parallel execution using MPI the name is edge mpi run Both scripts requires the input file as parameter the parallel script also require the number of processors partitions To control the execution of Edge one can store the file update ainp in the running directory The file should contain the UPDATE dataset defined as follows UPDATE N 0 0 1 IBREAK I 1 1 0 3 where the value of the parameter IBREAK determines what Edge will do The following values of IBREAK apply 1 Converged for internal use only 2 Error crashed for internal use only 3 Save and stop 4 Kill In this manner any Edge parameter can be updated by adding the parameter in the UPDATE dataset with the new value When Edge has encountered the file it will be removed FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 1 5 Postprocessing There is no postprocessing program contained in the Edge system There are however three programs to export data to the commercial flow visualization tool EnSight The program ffa2engold outputs the grid solution on the EnSight Gold format and uvol ubody output data on the EnSight 5 format The former format is recommended by the EnSight developers The EnSight Gold converter also has some additional features There is a program called pres to output t
18. widely used models Some what more sophisticated than the Spalart Allmaras model and also more general It should give reasonable results for attached and separated flows e W amp J EARSM std k omega Wallin amp Johansson 2000 A model that is able to capture more of the complex physics associated with rotation curvature three dimensionality non equilibrium flow and separation Reasonable robust and reliable in most flows The default model in previous versions of Edge e W amp J EARSM Hellsten k omega Hellsten 1998 Comparable with the Wallin 2000 model but the omega equation is redefined and recalibrated fully consistently with the EARSM assumption This model is thus better behaved and slightly more accurate The new default model In addition there are two low Reynolds number LRN k w models included in Edge Wilcox LRN k omega and PDH LRN k omega which may be good alternatives for computing internal turbulent flows with moderate separation and or reattachment The other models listed in Appendix A 3 may be tried if the above models fail to converge A typical comparison of several RANS models for in Edge for a computation on an aircraft configuration can be found in Peng amp Eliasson 2004 Performing several computations with different turbulence models is good practice in order to assess the possible sensitivity to the modeling Large differences between models indi cates that the flow is high
19. 0 A 4 Grid Adaption Name Description Value Default CELMAX Maximum cell size gt 0 0 7 CELMIN Minimum cell size gt 0 0 1 CELFLO Maximum indicator change in one cell gt 0 0 05 RMAGNI Change number of cells maximum by this factor gt 0 4 0 IPROJ Project new points on surface 0 1 1 YPLUS Desired y value for the first prismatic cell gt 0 1 0 Table 35 Grid adaption parameters 81 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 82 FOI dnr 03 2870 Edge User Guide A 5 Adjoint Euler Issue 4 1 0 Name Description Value Example CFIEUL Euler flow solution file Edge_flow bout IUPWIN Discretization scheme 2 for adjoint 2 adjoint ADJOPT Type of calculation flow or adjoint 0 flow 1 or 2 adjoint FUNTYP Type of functionnal drag lift side pitch yaw roll drag Table 36 Adjoint Euler parameters 83 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 84 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 B Appendix Datasets in FFA format B 1 Datasets for Input and Output Data Below follows a list of some of the datasets used to define the input and output data to most of the files in both Edge and Euranus Dataset Type Dim Size Nsub Explanation grid N 0 0 Grid dataset contains the mesh boundary data N 0 0 Dataset with boundary data solution N 0 0 Contains the solution time_history N 0 0 Time histor
20. 0 Edge User Guide Issue 4 1 0 22 If vector coloring is selected the edgelist is permuted and divided into a number of colors in order to improve performance on vector machines This is done in the coloring application which is called from the preprocessor script Node order is not changed by this algorithm If the input parameter PDUAL is set to 1 EnSight files for each grid level with the extension dgrd will be created For 2D grids the dual grid will be plotted and for 3D grids the dual boundary grid will be plotted If cache reordering is selected both nodes and edges are reordered to improve performance on a modern CPU with cache memory This is done in the reorder application which is called from the preprocessor script Cache reordering is only done if also NPART is larger than 1 The reason is that there is no mechanism that restores the node order of the solution file for serial computations If the node order was changed for a serial computation the node order of the solution file would not match the node order of the mesh file For parallel com putations merge_partitions restores the node order when merging the partitioned solution files into a single file The output filename is given by the variable CFIEDG For a parallel computation NPART gt 1 NPART files are output with extension _p1 pNPART to the file name for each parti tioning Each of the processors in the flow solver will read a separate file and thus treat one
21. 003349 N 0x0 4 block DF 226515 x 1 density 1 491653856239447E 004 DF 226515 x 3 velocity 8 664490392305818E 007 DF 226515 x 1 pressure 8 148238739742469E 010 DF 226515 x 3 grad 6 890308927563643E 004 Figure 9 A typical solution file from an adjoint calculation bout including nodal gradients 3 4 8 The Sample Solution File Edge adjoint bout The result from an adjoint flow calculation is saved in a usual Edge adjoint bout file where the density velocity and pressure are the adjoint variables instead of the flow primitive vari ables They are in fact the co state variables of the flow equations expressed in conservative variables The adjoint variables may be used for further calculating the shape gradient for example of the functional J associated with the adjoint problem the type of the functional J is given via the input variable FUNTYP in the ainp file dedicated to this calculation If the ADJOPT variable is set to the value 2 in the file ainp a post processing of the adjoint variables delivers an additional field called gradient_set in the result file such as in Figure 9 This is the gradient of the functional J with respect to the nodal coordinates of the mesh It has therefore the size and dimension of the field coordinates as found in the mesh file bmsh or the edge file bedg associated to this problem This result is usually used for calculating gradients of functionals with respect to design paramaters In that
22. 1 0 52 N 0x0 4 modes definition L 1x1 title Elastic MDO wing IF 32 1 grid idents 123 RF 32 x 3 str_coordinates 23 1191006 31 074 N 0x0 37 mode_set N 0x0 5 mode T x 1 identifier 101 L x 1 description ist mode R x 1 frequency_hz 0 839999974 R x 1 generalised_mass 10306 0000 RF 32 x 3 str displacement 0 000 N 0x0 5 mode I x 1 identifier 102 L x 1 description 2nd mode R x 1 frequency_hz 2 16400003 R x 1 generalised mass 4552 79980 RF 32 3 str displacement 0 000 N 0x0 5 mode I x 1 identifier 103 L x 1 description 3rd mode R x 1 frequency hz 3 56399989 R x 1 generalised mass 2457 00000 RF 32 x3 str displacement 0 0000 Figure 11 A typical modes definition amod file where lt range gt is the required range of the field extension in the same units as the bmsh grid Warning This field extension function can be very slow There is also at present no way to check the quality of the mesh for a given perturbation field The recommended procedure for creating modal perturbation field data is to use the programs aetribend and aeputpert 3 8 4 1 The modes definition File amod The modes definition dataset is used to import structural modes into the Edge system It is the Edge native format for this data Conversion routines will be created in due course to transfer data from various structures codes to this amod bmod format The amod f
23. 93520 4 42009 DT 764 x Cm 6 286313173730319E 02 DT 764 x cpu_time 1 63099999999394 DT 764 x work units 0 00000000000E 00 DT 764 x max_heattransfer 3 869822E 09 DT 764 x 1 max_temperature 299 9999999 DT 764 x 1 rho rmsresidual 6 683030961 DT 764 x rho maxresidual 8 2024096726 DT 764 x rhou rmsresidual 8 497851336 DT 764 x rhou maxresidual 10 21657562 DT 764 x rhov_rmsresidual 7 837199928 DT 764 x rhov_maxresidual 9 571859964 DT 764 x rhow_rmsresidual 20 00000000 DT 764 x 1 rhow maxresidual 20 00000000 DT 764 x rhoe rmsresidual 12 16430666 DT 764 x rhoe maxresidual 13 67740256 DT 764 x rhoq 01 rmsresid 6 373938039 DT 764 x rhoqg 01 maxresid 8 087178792 DT 764 x rhoq 02 rmsresid 16 08406770 DT 764 x rhoq 02 maxresid 18 28947518 Figure 8 A typical residual bres file 34 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 N 0x0 8 solution L x brand Edge 4 0 0 www edge foi se LI x title Edge an edge based flow solver LI x date 060713 18 24 17 LI x author LI x program Edge LI x grid ref Grid converted from CGNS N Ox 0 8 free stream data DC x pressure 100000 000000000 DC x 1 temperature 262 938300000000 DC x3 velocity 272 478929818100 DC x gamma 1 39999997615814 DC x rgas 286 999987781048 DC x mu_reference 1 993280000000000E 005 DC x prantl number 0 720000028610229 DC x 1 dynamic pressure 49333 2196
24. FFA file viewer editor FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Directory Contents adaption Mesh adaption source code ae_programs Aeroelastic utility programs source code ae_util Various aeroelasticity related source code files bin Some miscellaneous utility scripts mostly used when building Edge config Makefile snippets for different platforms compilers doc Documentation files examples Example files for running Edge including a generic default input file gui Xedge graphical user interface source code lapack A LAPACK implementation matlab Matlab functions preprocessor Preprocessor source code programs Various utility programs source code solver The flow solver source code src util Various source code files util Various source code files dealing with the FFA file format Table 2 Directories in the root directory 2 4 Building Edge The installation procedure given here is also found in the file INSTALL in the root directory Compiling Edge requires three steps configuring building and installing 2 4 1 Configuring Using the Configure Script The makefiles named Makefile should not be edited Instead the script configure must be run before building Edge It configures Edge for building on your specific platform The configure script is located in the root directory of the source code It is recommended to prepend the configure command with a path e g configure in order to avo
25. FOI HELLSTEN A 1998 On the solid wall boundary condition for w in the k w type turbu lence modes Scientific Report Report No B 50 Series B Helsinki University of Tech nology MENTER F R 1994 Two equation eddy viscosity turbulence models for engineering ap plications AZAA Journal 32 8 1598 1605 PENG S H amp ELIASSON P 2004 A comparison of turbulence models in prediction of flow around the DLR F6 configuration AZAA Paper 2004 4718 SMAGORINSKY J 1963 General circulation experiments with the primitive equations Monthly Weather Review 91 3 99 164 SMITH J 2005 Aeroelastic functionality in Edge Initial Implementation and Validation Scientific Report FOI R 1485 SE Computational Physics Department Aeronautics Di vision FOI SPALART P R amp ALLMARAS S R 1994 A one equation turbulence model for aerody namic flows La recherche Aerospatiale 1 5 21 WALLIN S amp JOHANSSON A V 2000 An explicit algebraic Reynolds stress model for incompressible and compressible turbulent flows J Fluid Mech 403 89 132 WINZELL B 1996 Generation of grid perturbations for aeroelastic CFD applications with special discussion of the use within Euranus Tech Rep TUDLF 0 96 27 Issue 1 Saab Military Aircraft YOSHIZAWA A 1986 Statistical theory for compressible shear flows with the application to subgrid modeling Phys Fluids 29 2152 2164 FOI dnr 03 2870 Edge User Guide A Appendix
26. FOI Eig User Guide E Defence and Security Systems and Technology March 2007 _sEdge FOI dnr 03 2870 Issue 4 1 0 Edge User Guide FOI dnr 03 2870 Defence and Security Systems and Technology March 2007 ISSN 1650 1942 FOI dnr 03 2870 Issue 4 1 0 FOI dnr 03 2870 Edge User Guide Contents 1 Introduction 2 Installation 21 nissen 2 ss puras er ee we dee Ps 2 2 Dieregisiies a 4 ad sc ws den neues 22 Begins s gg nee ps Bea oh ke un 2 2 2 Opel p Se E a a 223 Directory Structure 2g ka rs e ana 24A Building Edge os es an ce ale ee a a ee ee 2 4 1 Configuring Using the Configure Script 242 COMME arena 243 Installing sus ee an 2 4 4 Build Java Based GUI optional 2 4 5 Testing your Installation 2 2 2222 3 Running Edge 3 1 Steps to Take to Run Edge 22 2 paga gama aged 3 1 1 Setting Up an Input File 3 1 2 Specifying Boundary Conditions 3 1 3 Running the Preprocessor 3 1 4 Running the Flow Solver 3 1 3 Postprocessing ss mea dew ee 3 1 6 Running Edge in Parallel 2 222 3 1 7 Running the Grid Adaption 2 22 2 3 18 Running Xedpe creon coca run 3 2 List of all Programs in Edge 3 3 The FFA Dataset Fileformat 3 4 Files Generated Required by Edge 3 4 1 The Input Grid File Edge bmsh 3 4 2 The Boundary Condition File Edge aboc 3 4 3 The E
27. J Gen EARSM std k o WAJ E Kok TNT k o W amp J E Menter BSL k o W amp J CC E std k o W amp J CC E Menter BSL k o W amp J CC E Hellsten k o W amp J DRSM std k o SSG DRSM std k o SSG DRSM Hellsten k o ITURB 3 DES model Spalart Allmaras DES model ITURB 4 LES model Smagorinsky SGS model Spalart Allmaras SGS model Yoshizawa k eq SGS model IUPWKZ 1 for upwind discretization of turbulence 0 1 1 0 for central artificial dissipation discretization VIS2KZ Amount of second order artificial 0 1 dissipation IUPWKZ 0 VIS4KZ Amount of fourth order artificial 0 0 05 dissipation IUPWKZ 0 TURFIX Entropy fix for the turbulent equations lt 1 gt 1 1 1 IUPWKZ 1 Scaled with spectral radius gt 1 or the magnitude of the velocity lt 1 1 25 means a lowest value of 25 of spectral radius of the eigenvalue 1 25 means 25 of velocity IORDKZ Order of accuracy for the upwind 1 2 2 scheme IUPWKZ 1 72 Table 24 Turbulence model parameters Note that E is a short notation for EARSM for some models and that k o stands for k omega FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default INPRES O start from free stream 0 12 10 1 initial solution available specified in CFIINI 2 restart from previous solution specified in CFIOUT IPOST Post processing option output in file name gt 0 0 spe
28. N 0x0 4 koundary_data NB 0x0 37 boundary L x b name slat p L x b_class wall L x 1 b_type weak adiabatic NB 0x0 37 boundary L x b_name wing p L x b_class wall L x 1 b_type weak adiabatic NB O x O Sy boundary L x b_name flap p L x b_class wall L x 1 b_type weak adiabatic NB 0x0 3 boundary L x b name outer L x b class external L x b_type weak characteristic Figure 4 A typical boundary condition aboc file 3 4 3 The Edge File Edge bedg The edge file is produced by the preprocessor The element information from the grid file has been processed and complemented or replaced with edge and node information For each edge the two nodes are given edge nodes as well as the control surface edge surfaces building up the control volumes For each boundary the boundary nodes b_nodes the most orthogonal inner node b_inner_nodes and a control surface is computed The preproces sor also agglomerates the control volumes to coarser grids so the same type of information node_f2c is repeated for each grid level grid together with connectivity information be tween the grids In case when a parallel computation is to be carried out NPART gt 1 then NPART files are produced for each partition processor with names Edge bedg p1 Edge pNPART This file is usually given in binary due to its size being larger than the grid file an example file is listed in Figure 5 Due to the am
29. The moveable surface dataset is used by the programs aedbelast and aedbrotate In the future it may be used directly by Edge The aexbset command line and arguments are as follows Usage aexbset in bmsh ut bset moving txt sliding txt Arguments 1 in bmsh input Edge mesh file Edge bmsh 2 ut bset output moveable_surface dataset file 3 moving txt input list of boundary names for movable surface 4 sliding txt input list of boundary names for sliding planes optional The two text files arguments 3 and 4 specify the boundaries comprising the moveable surface and optionally any sliding planes one boundary name per line These files can 49 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 50 N 0x0 9 surface movement L x brand Edge 3 2 www edge foi se L x title AEDBELAST surface displacement L x parent_data mdo_e05 bmsh mdo_e05 bset N 0x0 1 moving_surfaces L x 1 boundary_name wing N 0x0 Ly sliding_planes L x boundary name symmetry IBN 9881 x nodes moving 266 968 969 IBN 2512 x 1 nodes sliding 123 N 0x0 4 mode I x identifier 101 L x description lst elastic mode D x frequency_hz 0 8399999737739563 D x 1 generalised_mass 10306 00000000000 DBN 9881 x 3 displacement 0 0000000000E 00 0 0 Figure 10 A typical moveable surface bdis file most easily be constructed using the program ffalist with the output piped to grep bound ary_name The boundaries must be selected v
30. Umfpack IMDSOL 1 or a Preconditioned Conjugate Gradient PCG algorithm FOI dnr 03 2870 Edge User Guide Issue 4 1 0 IMDSOL 2 Using Umfpack requires that the Edge system is built with Umfpack as an external library Umfpack can be downloaded free of charge whereas the PCG algorithm is build in The sparse direct solver is robust and fast for 2D problems However the memory requirements grow significantly with the size of the problem and become prohibitive for 3D for all but very coarse meshes Moreover the boundary condition options sliding_planes and free_surface is only implemented for the PCG solver The memory requirements for the PCG algorithm are modest but its convergence rate will deteriorate as the problem size grows to alleviate this a multigrid algorithm is under development There are several tests that terminate the PCG algorithm The iterations stop when either of the following conditions have been fulfilled 1 The maximum number of iterations ICGMIT has been reached 2 The quotient of the square root of the current residual 5 to the square root of the zero displacement residual is less then RCGTOL The zero displacement residual is the residual for zero displacements in the unknown displacements 3 The maximum scaled update computed at the current iteration step is less than RUPTOL Maximum is taken over all points in the mesh and each update is scaled with the min imum length of the edges that are ass
31. _energy RFC NVI 0 The NVI vibrational energies vibr temperature RFC 1 2 0 The vibrational temperature 86 Table 39 Field datasets that might occur in the solution file FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Dataset Type Dim Size Nsub Explanation iteration_number IT 1 0 Iteration number Cl RT 1 0 Non dimensional lift Cd RT 1 0 Non dimensional drag Cm RT 1 0 Non dimensional moment cpu_time RT 1 0 CPU time work_units RT 1 0 Work units max_heattransfer RT 1 0 Maximum heat transfer max_temperature RT 1 0 Maximum temperature total_time RT 1 0 Time elapsed for time accurate calculation inner_iterations IT 1 0 of inner iteration when time accurate ITIMAQ 1 mode_identifiers IT Mode shape identifying numbers modal_coordinate RT Generalized modal coordinates modal velocity RT Generalized modal velocities modal force flow RT Generalized forces rho rmsresidual RT Density rms residual rho maxresidual RT Density max residual x momentum rms residual x momentum max residual y momentum rms residual y momentum max residual z momentum rms residual z momentum max residual Energy rms residual Energy max residual Turbulent energy rms residual Turbulent energy max residual Turbulent dissipation rms residual Turbulent dissipation max residual Turbulent time scale rms residual Turbulent time scale max residual Species rms residual Species max residual Electronic vibration rms residua
32. also the preprocessor and many of the helper programs The description below relies on editing the input file with a text editor The editing may also be done with the GUI Xedge if available An example input file with default setup can be found in examples default_edge ainp and the user should copy this file to his working directory and give it a relevant name and keep the suffix ainp The most relevant parameters that usually need to be modified are 1 File names assigned to the variables CFIMSH CFIEDG CFIOUT CFIINI CFIBOC CFIRES 2 Free stream values of the primitive variables PFREE TFEE UFREE VFREE WFREE 3 Parameter for an Euler Navier Stokes calculation INSEUL 4 For a viscous calculation the free stream viscosity RMU and specification for lami nar turbulent calculation ITURB 5 Sequential or parallel computation by specifying NPART 1 19 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 20 Code Meaning Connectivity boundary conditions 1 Periodic boundary Rotation symmetry condition 3 Propeller disk boundary Wall boundary conditions 11 Euler wall 12 Adiabatic wall 13 Isothermal wall Symmetry boundary conditions 21 Symmetry External boundary conditions 31 Farfield 32 Farfield with vortex correction in 2D 33 Total states inlet original type 34 Total states inlet pressure extrapolated 35 Incompressible inlet 36 Pressure outlet 37 Extrapolation 38 Free stream 39 P
33. ape data 47 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 48 is further restricted to perturbation field whole grid displacement field This because on line mesh deformation functions have not yet implemented A further restriction is that aeroelastic solutions are limited to single processor execution 3 8 1 Overview of the Aeroelastics ae Programs Helper programs for the aeroelastics functions in Edge are denoted by the prefix ae and re ferred to as the AE programs or Aeroelastics programs There are just six such programs but the series will be extended in future versions of Edge as further aeroelastics functions are added Several Matlab functions are also provided Each of the AE programs including plotmot have a detailed help text text shown when the program command is entered with h or nothing There is also a program ae help which prints a summary of all the AE programs This should allow the user to run all these programs without opening this manual The present set of AE programs and their associated files are listed in Table 14 Note that the program plotmot is a general purpose utility used with the MOTTAB system described in Section 3 8 7 It is included here because this system is at present used only for Aeroelastics data These AE programs are used to prepare modal Aeroelastic calculation With the perturba tion field method for grid deformation the usual sequence of
34. ated from the command line by entering the program name with any required arguments In many cases the program command has only one argument the name of the main Edge input file To obtain help text for any of these programs simply enter the program name For the more recent programs the help text gives complete instructions for running the program For the aeroelastics programs there is also an ae help command which prints a summary of the ae program set 3 3 The FFA Dataset Fileformat With a few exceptions all data files for Edge are FFA format The acronym FFA stands for Flexible Format Architecture which is a general datastandard defined by FFA Sweden The FFA format defines a structure for hierarchical datasets of a recursive linked list form These datasets can be represented in most programming languages and file systems The Fortran 90 representation of FFA format is used in Edge and the FOI developed IR ray tracing code Sigge A Matlab representation based on cell arrays is used in the FFA Matlab Toolbox Data files in FFA format may be binary IEEE bigendian or represented in a plain ASCII format One of the many advantages of the FFA format is that binary files are generally portable The basic component in the FFA format is the FFA dataset A dataset is divided into an identifier set comprising five descriptors and the data itself A dataset can have one or 25 FOI dnr 03 2870 Edge User Guide Issu
35. ature For turbulent flows the turbulent dependent variables are specified strongly as well Unless otherwise stated it is assumed that a wall resolved grid is used and that yt 1 in the first layer of nodes outside the wall However so called wall functions may also be used in which larger values of yt may be used A wall function approach denoted universal wall boundary conditions is employed in which y in the first layer of nodes very well can be within the boundary log layer i e yt gt 30 and where a finite velocity component is specified strongly In principle any value of y in the first layer is accepted As the grid is refined the wall functions boundary conditions approach the standard no slip conditions To use wall functions the user has to actively choose this in the program bound that specifies the boundary conditions for each wall boundary where wall functions shall be applied 3 6 3 Symmetry Boundary Conditions The symmetry boundary condition in Edge is the same condition as an Euler condition It belongs to a separate class though to separate it from a wall boundary 3 6 4 External Boundary Conditions The available external boundary conditions are given in the list below The external bound ary conditions are in general weak boundary conditions i e the variables on the boundaries FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Function weak characteristic weak charact vortcor weak tota
36. bal scale factor both of which apply to the whole mode set The interpolation used in the timeseries table lookup is an overlapping three point quadratic scheme which ensures that the first derivative of the signal is continuous For signal_type periodic the timeseries data is read cyclically with periodic end conditions For sig nal_type pulse the end conditions are such that the signal is zero outside the pulse time slice 3 8 7 3 The mottab harmonics Signal Format The mottab harmonics format represents each of the mode in the signal table as Fourier series see 4 N q t gt gt ar cos 2rk ft dx 4 k 1 where q t is the modal coordinate This format can be rather cumbersome and is included mainly to reproduce the functionality of the Euranus code The timeseries format is both more versatile and faster to compute than the harmonics format especially for large N Fourier series 57 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 58 N 0x0 6 signal_table L x 1 title appr square wave S x 1 signal_format harmonics S xl signal type periodic R x I t_start 5 00000007E 02 R x 1 amplitude_factor 1 00000000 N 0x0 6 mode_set N 0x0 5 mode I x 1 identifier 101 R x 1 frequency_hz 1 00000000 R x 1 mean_position 0 00000000 R 64 x 1 cos coefficients 1 0 0 33 R 64 x 1 degrees phases 180 0 0 00 N 0x0 5 mode T x 1 identifier 102 Figure 15 A typical
37. ble in the FFA Matlab Toolbox and in Xedge the Edge GUI Finally there is a python program ffaview py which can also view and manipulate FFA files through a GUI requires Python 2 with the pygtk package installed FFA file naming convention dictates that FFA binary files have names ending in b and FFA ASCII files have names ending in a where denotes any single character Examples of this convention can be seen in Table on page 14 This filename convention is strictly enforced FFA format binary files are portable between Unix Linux machines with IEEE bigendian format To achieve this the whole Edge system must be compiled using the appropriate compiler settings defined using the files in the config directory An FFA dataset is written to a binary file using the following Fortran statements 27 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 28 Q Q au O Meaning binary 1 integer 4 real 4 real 8 complex 4 4 double complex 8 8 character 1 character 16 character 72 No data zrunm gt NNUrTU Table 5 FFA format type 1 1 data type Code Meaning Descriptor Field Boundary information Constants Data tables Data list Tam mm Table 6 FFA format type 2 2 data function WRITE UNIT CNAME CTYPE NDIM NSIZE NSUB WRITE UNIT DATA JSIZE JDIM JSIZE 1 NSIZE JDIM 1 NDIM Each dataset is written as two Fortran records a descriptor recor
38. c solution is required However some checks are necessary to ensure that the initial conditions are correct The dynamic simulation should not be started directly INPRES 1 from a steady state solution Instead the modal rigid option IAEOPT 100 should be used to ensure that the modal forces are constant The dynamic solution is then started as an INPRES 2 restart from this solution Using the setting IFOFFSET 1 offsets the modal force with the initial steady state value This does not however define an exact state of static equilibrium For this reason it is recommended to run the dynamic simulation for a short time interval without deliberate structural excitation On restart from the rigid solution the flexible structure will usually show some fluctuations Provided that the system is in a stable non flutter state then this motion should be very small The structure can then be excited via the mop initial conditions or more physically using a mot tabulated force input If the system is in an unstable aeroelastic state then the transition from from the rigid solution may be enough to excite a flutter response The initial fluctuations about the offset state would then increase in amplitude The presence of a flutter can verified by examining the effect of increasing the structural damping This procedure for monitoring and analysis of a design shape simulation is otherwise similar to
39. case a mesh deformation program is required in order to map the displacements of the nodes in the en tire domain of calculation from the displacements of the nodes on the shape only It is also usual to parameterize the nodes on the shape to optimize instead of using the coordinates themselves as design parameters Suppose that an explicit linear mapping is used to deform the mesh for given coordinates at the boundary being optimized We may denote it X Ay where X stands for the vector of nodal coordinates in the entire flow domain y stands for the vector of nodal coordinates of the shape that is optimized and A is the matrix of real coefficients associated with the linear mapping For a given shape y the inviscid flow Euler equations is solved giving the state variables w The solution of the adjoint equations associated with the Euler equations the flow state w the shape y and the functional J FUNTYP is denoted w Having set the ADJOPT variable to 2 in the ainp file for the adjoint calculation tells Edge to compute the gradient of J with respect to X which is done without computational effort once the adjoint problem has been solved w As named above the gradient of J is saved in the result file Edge bout under grad we call it here V Jx The gradient with respect to the shape that is with respect to the vector y of coordinates can then be simply obtained by the chain rule as 35 FOI dnr 03 2870 Edge User Guide Issue 4
40. ce Movement 51 3 8 5 Program aetribend Static Mesh Deformation 54 3 8 6 Program aeputpert Modal Perturbation Fields 55 3 8 7 The mottab Signal Tabulation System 56 3 9 Edge Modal Coupled Solutions 0 4 58 3 9 1 Structural Model and Modal Parameters 58 3 9 2 Coupled Simulation Procedures 59 3 10 Edge Adjoint Calculation for Inviscid Flow 61 3 10 1 Serial Computation nc essa e s ce me Dee RE OE 62 3 102 Parallel Computation ss e cw set ence ke ee ke ed 63 3 11 The Meshdeform Utility s 2 2 242 ee da ara 63 A Appendix Input Variables 67 A l General Parameters for Several Programs 67 Aue PIE ee ee ee Poa we we CS 69 A3 Flow Solver 2a 20 sara eR we ees 70 Ad Grid Adaption 2 ar a ee ee 81 AS Adom Buler wa chk re sets ace aa eGo ae ane nahe 83 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 B Appendix Datasets in FFA format 85 B 1 Datasets for Input and Output Data 2 Comm nen 85 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Abstract This is the users manual for the computational fluid dynamics CFD solver Edge Edge is a parallelized CFD flow solver for 2D 3D viscous inviscid compressible flow on unstructured grids with arbitrary elements Edge can be used for both steady state and time accurate calculations including ma
41. ce vector associated with the current edge in the reference mesh The minimization of expression 5 can be constrained at the boundary as specified in a boundary displacement file The variable CFIDIS in the input file specifies the name of the boundary displacement file Basically the same structure as described in Section 3 8 is used for the boundary displacement file There are three ways to control the boundary points by information given in the boundary displacement file 1 Completely constrained motion The dataset nodes_moving specifies which nodes on the boundary that should be displaced and the dataset displacement specifies corre sponding displacement all x coordinates first then all y coordinates with respect to the boundary coordinates of the reference mesh Each node number specified in nodes_moving should be a node identified as a boundary node in the mesh file 2 Constrained in the local tangent plane A dataset sliding_planes specifies that the motion at selected boundaries should be constrained to the local tangent plane of the reference mesh For instance if the boundary displacement file contains the sub dataset form 1 sliding_planes boundary_name symmetry pe xx ro the displacement vector u for each node on the boundary portion labeled symmetry will be constrained to be orthogonal to corresponding boundary normal vector nj Note that this constraint is specified only through the name of the boundary
42. cified by CFIPOST Several options can be combined by summation O No output 1 Skin friction heat transfer 2 Mach number Cp and energy 4 Distance to the wall and laminar turbulent field 8 Residual to all unknowns 16 Reynolds stresses 32 Gradients of the dependent variables 64 Flow curvature vector PFREE Free stream pressure gt 0 100000 TFREE Free stream temperature gt 0 300 UFREE Free stream velocity in x direction 271 15 VFREE Free stream velocity in y direction 0 WFREE Free stream velocity in z direction 0 RMU Free stream viscosity gt 0 1 7894e 5 TUFREE Free stream turbulence level gt 0 0 001 VRFREE Ratio of eddy viscosity and molecular viscosity gt 0 1 SPMFREE Free stream values for the species 0 1 1 NSPEC ITYGAS 2 NSPEC gt 1 Table 25 Free stream initial solution data post processing options 73 FOI dnr 03 2870 74 Edge User Guide Issue 4 1 0 Name Description Value Default IUPWIN Central 0 0 1 0 or upwind discretization of mean flow VIS2 Amount of second order artificial gt 0 0 5 dissipation for central scheme VIS4 Amount of fourth order artificial gt 0 0 02 dissipation for central scheme VISO Coarse grid 2nd order dissipation when gt 0 0 15 simplification are used MGSIMP 1 IARDIS Type of upwind scheme for the mean flow 1 2 2 1 flux difference splitting 2 symmetric TVD IORDER Order of accuracy for upwind scheme 1 2 2 for the mean flow LIMITE Type of l
43. d followed by a data record The detailed byte level structure of these records is visible in the code for programs ffa_forti m and ffa forto m in the FFA Matlab Toolbox Attempts to use Edge on Windows platforms may at present result in incompatible Fortran record structures Occasionally one may encounter non standard smallendian FFA format files usually due to an incorrect compilation The endian status of any FFA format file can be checked or switched using the programs ffa_checkendian and ffa_swapendian which are supplied with Edge FFA format ASCII files are used mainly for input to Edge and other programs One im portant example is the default main input file for Edge in examples directory The data structure and ordering of these files are exactly the same as the binary equivalents except that they may contain comment lines A comment line always starts with an asterisk in the first column Comment lines are ignored by the Edge subroutine that reads FFA data The task of creating and editing FFA format ASCII files can be made easier by the use of syntax highlighting colored text Syntax highlighting definitions for the Nedit editor are available For creating and editing the main Edge input file the use of the Edge GUI is recommended FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Code Meaning N Node values F Face values C Cell centre values Table 7 FFA format type 3 3 data location
44. del constant C g s in the context of EDGE When the grid is sufficiently fine the S A SGS model should perform similarly to the Smagorinsky SGS model Another SGS model is the SGS k equation model by Yoshizawa 1986 The grid resolution requirement in LES is very high in all three directions particularly in the wall boundary layer so that it may become infeasible for computations of typical aerodynamic configurations at high Reynolds numbers When a computationally affordable grid can be applied for fundamental flows or for flows at moderate Reynolds number LES may provide with much richer information and usually more accurate modelling than RANS approaches Note that LES should be performed in the context of time dependent and three dimensional calculations It is recommended that LES is only used by experienced users or in consultancy with FOI 3 5 2 4 Specification of Related Parameters For external aerodynamic flow simulations the free stream turbulence intensity TUFREE has a default value of TUFREE 0 001 for RANS computations For DES and LES a value of about 10 of this value should be used The parameter VRFREEis the ratio of the free stream eddy viscosity and the molecular viscosity A value of VRFREE 1 0 is taken as the default value in RANS With the Spalart Allmaras RANS model as well as with DES and LES a value of VRFREE 0 1 is recommended For DES hybrid RANS LES and LES computations the second order central scheme
45. dge File Edge bedg Issue 4 1 0 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 4 4 The Solution File Edge bout Edge bini 31 3 4 5 The Sample Solution Files Edge_timestp bout 31 3 4 6 The Solution File for Post Processing Post bout 34 3 4 7 The Residual File Edge bres 34 3 4 8 The Sample Solution File Edge_adjoint bout 34 3 5 Hints and Recommendations 2 2 be dee ede ee eee 36 3 5 1 Specifying the Free Stream Viscosity and Reynolds Number 36 3 5 2 Turbulent Flow Calculations 36 35 3 Convergence Problems s lt s 4 6 64 0 mass cw ae eb a es 39 3 5 4 Time Accurate Simulation and Inner Iterations 39 3 6 Boundary Conditions inEdge 2 004 40 3 6 1 Connectivity Boundary Conditions 41 3 6 2 Wall Boundary Conditions 4 42 3 6 3 Symmetry Boundary Conditions 2 22222 42 3 6 4 External Boundary Conditions 22 2 2 nme 42 3 7 Using the Vortex Generator Model 2 2 2 Cm nme 46 3 7 1 The VG Geometry File Edge avg 47 3 8 Aeroelastics Options and Programs 2 2 2 Cm nme 47 3 8 1 Overview of the Aeroelastics ae Programs 48 3 8 2 Program aexbset Moving Surface Definition 49 3 8 3 Program aedbrot Rigid Rotation Surface Movement 51 3 8 4 Program aedbelast Modal Elastic Surfa
46. dge bmsh aetribend _ident bmsh _ident bdis _ident bmsh aeputpert Edge bmos Edge bmos Edge bmos edge Edge timestep bout Edge amot Edge bres modal data Edge amop Edge amot plotmot plotmot_ dat Table 14 Required generated files for Edge Aeroelastics programs The program aedbelast is used to transform structural modeshapes onto the movable sur face grid It requires two input files one with the modal data and one defining the interpo lation scheme For some interpolation schemes it is possible to produce a perturbation field and a mos file directly This is described fully in Section 3 8 4 Depending on the required solution the signal_table and modal_parameters datasets must be prepared in the files Edge amot and Edge amop respectively At present these files must be prepared by manual editing The same applies to the mod file This is however quite straightforward Example files are provided in the directory examples ae inputfiles Pro grams for constructing some of these input files may be added in future versions of Edge The rest of this section contains a more detailed description of the AE programs and data files 3 8 2 Program aexbset Moving Surface Definition The program aexbset is used to set up a moveable surface dataset defining set of boundaries which forms the moving wetted surface grid This data is extracted from the mesh file and written to a bset file format described in Section 3 8
47. dimension of the data 4 nsize integer 4 the size of the data 5 nsub integer 4 the number of sub datasets The data in each dataset is effectively a matrix with nsize rows and ndim columns the elements of which may be of numeric or character type The four character type code denotes the type function and location of the data The fourth character is so far unused Any dataset for which nsize or ndim is zero is a null dataset however a special type N is also defined The source code module for the functions which handle FFA format datasets within the Edge code is located in the directory utils Programmers who need to use the FFA format in their own code for example a format conversion routine need only this code A short programmers guide is provided in doc edge_util html Python programmers can use the module ffa py located in the bin directory This module contains functionality to read write and manipulate FFA datasets and files FFA format data files can be inspected using the Edge helper program ffalist which prints a summary of the file data structure including the first three elements of each data matrix The program is the most frequently used in the whole Edge system and examples of output from the ffalist can be found throughout this manual Two other useful programs are the ASCII binary converters ffaa2b and ffab2a Many more advanced functions for comparing and manipulating FFA format data are availa
48. e not specified at a local outflow 3 6 4 1 Farfield Weak Characteristic This boundary conditions may be used for all external boundaries it can handle both sub supersonic in outflow boundaries It is mainly intended for external aerodynamics It is based on a one dimensional analysis of the Euler equations in the direction normal to the wall Depending on the sign of the eigenvalues the characteristics are either speci fied or extrapolated See the section Farfield Boundary using Characteristics in the Edge Theoretical Formulation manual for more information If local varying free stream data is desired at the boundary the following minimum data is required N 0x0 3 boundary_data RF NB x 1 density RF NB x 3 velocity RF NB x 1 pressure where NB corresponds to the number of nodes on the boundary Other dependent variables like turbulence etc may be given and used if found 43 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 44 3 6 4 2 Farfield with Vortex Correction Weak Charact Vort Exactly the same boundary condition as the one above with the difference that the free stream values are no longer constant but depend on the integrated lift using a so called vortex correction method This boundary condition is intended for 2D subsonic external problems where the external boundary is located relatively close to the computed object It gives in general an increased accuracy of the solution Local boundary data cannot
49. e 4 1 0 Name Function aedbelast Transforms structural grid modeshapes to moving surface displacements aedbrot Constructs moving surface displacements for general rigid body rotations aedis2tri Transfers surface displacement data to a Tritet format used by aetribend ae help Print a summary of all the ae series programs aeputpert Constructs a modal shape data file containing perturbation fields aetribend Constructs a deformed mesh matching a given matching surface displacement aexbset Extracts a movable surface set of nodes and boundaries bound Interactive program to set the boundary conditions cgns2ffa Converts grid solution from CGNS format to FFA format edge mpi run Starts flow solver in parallel with mpi edge run Starts flow solver in single precision ffa23 Converts a 2D grid and or solution to 3D by adding a parallel plane in z ffa2cgns Converts grid solution from FFA format to CGNS format ffa2cgns_fluent Converts an unstructured file from FFA format to Fluent CGNS format ffa2engold Exports grid and solution to the EnSight Gold format ffa2tau Converts a 3D grid solution to the tau NetCDF format ffa2tgrid Converts a grid to the fluent tgrid format ffa32 Converts a 3D grid with two planes to a 2D grid ffaa2b Converts any ASCII FFA format file to binary ffab2a Converts any binary FFA format file to ASCII ffa_checkendian ffad2r ffajdiff ffajedit ffalist ffamirror ffanode ffaucut ffauinterpo
50. e Issue 4 1 0 12 implicit dual time stepping scheme which exploits convergence acceleration technique via a steady state form inner iteration procedure Every program in the Edge system can be used via a command line interface which is im plemented as set of simple Unix shell scripts However there is also a platform independent graphical user interface GUI written in Java which allows configuring a complete Edge computation with on line help The GUI is recommended especially for new users of Edge mainly because it greatly simplifies the process of setting up the input for an Edge computation This manual is divided into two main parts a User guide describing how to install and use Edge and an Appendix containing reference information such as file formats and definitions of input variables FOI dnr 03 2870 Edge User Guide Issue 4 1 0 2 Installation 2 1 Introduction Installation from source code involves configuring compiling and installing Care has been taken to make the installation process conform with common practice used in other free software distributions for Unix platforms This section does not apply for users installing the binary distribution of Edge The binary distribution is installed simply by unpacking the compressed tar file 2 2 Prerequisites 2 2 1 Required To compile and install Edge the following tools are required e A Unix platform e g Linux OSF1 Irix AIX Any common workstation sh
51. e timestep at and after which the flow variable recoded with time at a typical location fluctuates around a constant mean value The parameters NPTHIS specifies the number of the locations where the time history of the flow variables is recoded The coordinates of the locations x y z are specified in the sub dataset POINT_COOD and another sub dataset POINT_INDEX is used to specify the type of the recoded points bound point indicates a point on the boundary inter_point for an interior point FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Function Periodic A connectivity through a translational vector Rotation A connectivity through a rotational matrix Propeller Propeller disk model connectivity Table 9 Connectivity boundary conditions 3 6 Boundary Conditions in Edge The boundaries in the computational domain have been determined already when the grid was generated and can be found in the bmsh file Section 3 4 1 on page 29 The program bound creates an additional file with more detailed data for each boundary containing type of boundary condition local free stream boundary values etc This file has the suffix aboc as described in Section 3 4 2 Below all available boundary conditions are listed class wise with a short description to what they do and are intended for In common for all of these boundary conditions is that the program bound should take care of all specifications no editing
52. edbelast 3 8 4 Program aedbelast Modal Elastic Surface Movement The program aedbelast is used to transform a set of mode shapes on a given structural grid into a set of surface_movement datasets defining modeshapes on the wetted surface Usage aedbelast in bmsh in bset in amod in aigi field extension option Arguments 1 in bmsh input bmsh mesh filename 2 in bset input bset moveable_surface filename 3 inamod input amod modes definition filename 4 in aigi input aigi inter grid interpolation filename Input file 1 is a standard Edge mesh file and file 2 is produced by program aexbset Input files 3 and 4 are defined in Section 3 8 4 1 and 3 8 4 2 respectively In normal use aedbelast produces a set of bdis surface movement files with file names of the form modname ident bdis where modname is the name of the amod input file and ident is an integer mode identifier Other mode attributes in the amod file such as frequency hz are automatically copied into the bdis output file So far only one interpolation method is available polyfit a piecewise polynomial fit With this method the displacement can be extended out into the volume mesh to give a perturba tion field This can be used to generate a bmos moda shape data file directly To do this these two inputs are added to the aedbelast command line see Usage above as lt filename gt bmos lt range gt 51 FOI dnr 03 2870 Edge User Guide Issue 4
53. efault damping is optional and if it is absent the default damping is zero Warning init_velocity amp init_coordinate The initial state is set on restart and it there that the variables init_coordinate and init_velocity are applied This means that if you re doing a routine restart you need to save the initial mop file and replace it with one that does not have any of the init_ variables set If you don t then you will restart with the initial modal velocity and coordinates but with the current flow field which is probably not what you want There is another possibly irksome feature in this version Setting init_coordindate or init_velocity for any one mode automatically sets a default initial value for all modes The default ini tial values are zero for both the velocity and coordinate inputs This is sensible for most applications 3 10 Edge Adjoint Calculation for Inviscid Flow An example of adjoint calculation is given in examples rae_euler_adjoint 61 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 62 Solving an adjoint problem is an efficient way to compute gradient of a functional J that depends on the flow solution and the nodal coordinates of the computation domain we note it J w X where X is the set of nodal coordinates and w the set of nodal values of the vector of conservative variables In the current state of developement of the code Edge only the inviscid flow equations Euler have an adjo
54. efault values are set frequency 0 generalized mass 1 It is however recommended that all modal parameters are set using the mop file as described in Section IILS In any coupled solution the time discretized structural equation of motion is updated within the inner iteration loop controlled by ITMXAQ ITMNAQ At each update the grid and dual mesh metrics are recomputed which is a time consuming process The number of updates can be reduced using the input variable ITIN to coarsen the adjustment of the grid during the inner loop pseudo time iterations This is echoed to standard out and to the log file under the heading loop synchro Inner loop convergence should be monitored using the residuals and forces on standard output 3 9 2 2 Simulation types Jig Shape and Design Shape Coupled simulations may be performed in one of two scenarios jig shape and design shape selected by the input parameter IFOFFSET With the default setting IFOFFSET 0 it is assumed that the geometry and structural model is jig shape that is the shape applies when all loads including structural weight are zero For a wing this would be its shape when supported in the jig which is used to build it This is usually very different to its shape in flight With the setting IFOFFSET 1 using an unsteady solution restart INPRES 2 the initial aerodynamic load is subtracted from the modal force vector This force offset a
55. esidual File Edge bres The residual file Edge bres is output from the flow solver it is written together with the flow solution file Edge bout This file is intended to monitor the rate of convergence e g with the program plotres and contents of this file contains residuals integrated forces and moments plus a few additional quantities An example file is given in Figure 8 For time accurate solutions and aeroelastic applications the file also contains time series for modal forces modal coordinates and time information 33 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 N 0x0 8 solution L x 1 brand Edge 3 2 www edge foi LI x 1 title Edge Re 9m alpha 12 2 LI x 1 date 040506 11 14 39 LI x author LI x program Edge LI x grid ref N 0x0 T free stream data RC x 1 pressure 100000 000 RC x 1 temperature 300 000000 RC x 3 velocity 67 8771973 RC x gamma 1 39999998 RC x rgas 287 000000 RC x mu_reference 8 9608638E 06 RC x 1 prantl_number 0 720000029 N 0x0 6 block RF 88703 x 1 density 1 14204168 RF 88703 x 3 velocity 0 00000000E 00 RF 88703 x pressure 99098 6484 RF 88703 x turb kin energy 0 000E 00 RF 88703 x turb omega 3 79523712E 09 RF 88703 x turb viscosity 0 000E 00 Figure 7 A typical intermediate solution bout file N 0x 0 22 time history IT 764 x iteration number 0 10 20 DT 764 x Cl 2 715464574425870E 02 DT 764 x cd 0 58778952277
56. fied in the input file are needed as input to the program e a reference edge file file name in variable CFIEDG bedg format associated with the reference mesh e afile containing the given boundary displacements with respect to the reference mesh file name in variable CFIDIS bdis format and e amesh file file name in variable CFIMSH bmsh format containing an initial guess for the deformed mesh The last file typically represents the same mesh as the one associated with the reference edge file but this is not necessary the initial guess mesh file provides initial values for the iterative solver in meshdeform so this file can be used to continue a previously aborted calculation The meshdeform program outputs a mesh file file name in variable CFIOUM bmsh for mat that contains the deformed mesh This file will be a copy of the input file given in CFIDIS except that the coordinates will be changed according to the mesh deformation The coordinate vector for each mesh vertex coordinate given in the output mesh is given by zi oF us where gre is the coordinate vector for corresponding vertex in the reference mesh con tained in CFIEDG The displacements u are computed by minimizing the quadratic ntf 2 3 reste tal 5 ij a 63 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 64 ref where the sum is taken over all edges in the mesh Moreover us or ant and n 7 is the dual mesh surfa
57. file where it finds the name of the file containing the definition of the VGs CFVGNAME VG_identify inputfile ainp The format of the file CFVGNAME is described below By looking at the geometry defi nition of the VGs in CFVGNAME VG identify marks all the edges where the source term should be added and update the edge file If CFVGNAME contains data for several VG configurations all of them are added to the edge file VG_identify automatically projects the points defining the VGs on the closest wall and if VG data already exists in the edge file it is saved in a separate file FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 7 1 The VG Geometry File Edge avg The vortex generator VG geometry file Edge avg is an FFA format file that contains in formation on the geometry of the VG configurations used by the VG model Each VG configuration can be made of several VGs each of them being defined by a set of points on the wall and corresponding normal and height Each VG is approximated by a zero thickness surface defined by the points corners of the surfaces at the wall and normals sides of the surface rising from the corner points If the VGs are straight and short only one surface is necessary but for longer VGs that follow the curvature of the wall several surfaces might be necessary to better represent the VGs Below the content of the file ffalist for a configuration with one 4mm high VG is shown the rectangular VG being define b
58. fine several refinement boxes BOXCOOR If RADIUS equals zero BOXCOOR contains the coordinates of two points defining a diagonal of the refinement box If RADIUS is positive BOXCOOR contains the coordinates of the two end points of the cylinder axis The format is X min Xmax Ymin Ymax Zmin Zmax If Xmin Xmax and Yin Ymax and Zmin Zmax NO box is created ADAPTCASE H refinement and or y adaption for SENSOR 0 1 Only h refinement 2 Only y adaption of prismatic layer 3 Both h refinement and y adaption this is the same as alternative 1 for a grid without a prismatic layer YPLUS Desired y value for the first prismatic cell FOI dnr 03 2870 Edge User Guide Issue 4 1 0 In order to run the flow solver with the adapted grid the preprocessor must first be executed on the new grid Update the grid file name CIFMSH in the inputfile and then run the preprocessor 3 1 8 Running Xedge To start Xedge write xedge When working with Xedge the same steps are followed as when running Edge The work flow goes from left to right in the GUI menus and the relevant options will get accessi ble as the necessary files are generated A detailed description of Xedge is found in the HTML help pages accessible in Xedge in the Edge Graphical User Interface manual or in Chevalier 2005 3 2 List of all Programs in Edge Below is a list with all the programs available in the current Edge distribution All these programs are oper
59. gure 10 The dataset comprises a copy of the moveable_surface definition to which is added a mode dataset containing a set of mode attributes and a displacement dataset containing the dis placement vectors for all nodes contained in the moving surfaces The mode attributes in FOI dnr 03 2870 Edge User Guide Issue 4 1 0 this example are for an elastic mode but any attributes can be used such as rigid body rotation parameters 3 8 3 Program aedbrot Rigid Rotation Surface Movement The program aedbrot is used to create a surface_movement dataset for a rigid rotation This is written to a bdis file as described in Section 3 8 The command line and arguments are as follows Usage aedbrot in bmsh in bset ut bdis angle Nng nyn xo Yo 20 Arguments 1 in bmsh input mesh filename 2 in bset input movaeble_surface filename 3 ut bdis output surface_movement filename 4 angle rotation angle in degrees positive for right handed screw 5 Ng NyNz rotation axis direction vector size doesn t matter 6 zo Yo zo reference point on the rotation axis The first three arguments are file names and the remaining three define the rotation The two vectors are entered as three numbers enclosed in single quotes This applies even for 2D meshes The rotation parameters are written to the output file as attributes of a mode dataset These mode attributes are automatically carried forward in the dis output file as described for program a
60. he laminar region for each plane The laminar region is then defined by the convex region built up by the planes It is possible to limit the laminar region close to the wall by the parameter MAXLAMDIST so that turbulent flow is obtained at nodes at larger distances Specification of laminar regions is error prone and it is therefore recommended that the user uses the post processing option IPOST 4 in the flow solver to output the array specifying laminar turbulent flow 0 1 respectively as well as the distances to the walls 3 5 3 Convergence Problems The default setting for the numerical parameters usually provides good convergence rates and stable solutions It may happen though that the convergence is poor or there might even be cases that diverge There may be many reasons behind convergence problems One of the first tests to find the reason behind it is to lower the CFL numbers CFL CFLVIS CFLREF to see if this stabilizes and improves the convergence Another parameter may be to reduce the number of multigrid levels NGRID to see if the coarser grids has an overall influence It may be that the numerical dissipation is too small so the parameters VIS4 VIS2 VISO VIS4KZ VIS2KZ should be increased For upwind discretization the parameters LIMITE ENTRFX TURFIX are concerned In viscous calculations the coarsening ratio CRATIOB in the boundary layer can occa sionally lead to instabilities This quantity may then be reduced t
61. he residual reduction at each physical time 39 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 40 step which should carry on in an approximately cyclic manner when the solution moves from one time step to the next Furthermore with increasing sub iteration number at each time step the variation in the output parameters e g the integrated lift and drag should be negligibly small DELTAT also influences the number of inner iterations ITMXAQ In general reducing the value of DELTAT a smaller time step may result in a reduction of the inner sub iteration number ITMXAQ In DES and LES the time resolution is important for which an appropriate specification of the DELTAT value may affect both the numerical solution procedure and the resolution of the resolved turbulent structure Therefore caution must be taken in DES and LES with the value of DELTAT which in some occasions may have to taken a smaller value than in RANS calculation for the same problem in order to get accurate physical resolution of the flow evolution It is also possible to make time accurate calculations with explicit Runge Kutta time march ing by imposing ITIMAQ 2 The Runge Kutta coefficients IRKCO have to be adapted so that the desired temporal accuracy is obtained the default coefficients are optimal fr steady state calculations and are provide only first order accuracy Explicit time marching requires fairly equal sizes of the grid elements since the maximum al
62. he solution on a boundary in ASCII format on a data file The program plotres can be used to monitor the convergence and plots the residual and integrated forces and moment It requires the open source software Xmgr 3 1 6 Running Edge in Parallel A parallel computation can be carried out by splitting the input grid into a desired number of partitions NPART gt 1 The intention is that a parallel calculation shall produce the same output and be handled in the same way as a sequential calculation i e a user should not notice any difference except for the specification of NPART gt 1 choice of flow solver script edge_mpi_run in stead of edge_run and the computational speed up The preprocessor outputs a file for each partition processor to be read by each of the flow solver processes During the iterations each of the processes outputs separate solution files that at the end of the calculation are merged into a single solution file The resulting flow solution is therefore the same as the flow solution from a sequential computation 3 1 7 Running the Grid Adaption When a solution is obtained the grid can be adapted to the solution with the command hadaption inputfile adapted_mesh bmsh The original grid is then h refined to a new output grid adapted mesh bmsh The old cells are divided into smaller cells In each adaption run the edges can only be divided once A tetrahedra is divided into either two four or eight new tetrahedron
63. he user to change free stream values on the boundaries It is only necessary to supply these values if they deviate from the free stream values given in the input file If no values are supplied the values from the input file are used The output from this program is written to a file with the name given by the variable CFIBOC 3 1 3 Running the Preprocessor The preprocessor is started with the command preprocessor inputfile In the first step of the preprocessor the elements are discretized to control volumes sur rounding each vertex The volume elements accepted so far are triangle quadrilateral tetra hedron hexahedron and prismatic element The enumeration of these elements is depicted in Figure 2 The boundary elements are bar triangle and quadrilateral The orientation of boundary elements is automatically adjusted by the preprocessor to point outward from the computational domain In the second step the preprocessor fuses control volumes to coarser grid cells to be used for multigrid The input parameter NLEVEL determines how many grid levels that are created If the input parameter NPART is greater than 1 then a partitioning of the grid is performed enabling parallel computations on NPART processors The input parameter COLTYPE controls the reordering algorithm for nodes and edges Reordering is done in order to improve performance of the solver by utilizing cache memory or vector computers better 21 FOI dnr 03 287
64. herwise the nodes lie on the surface elements of the original grid SENSOR Indicates which method to use to select the cells to be refined 0 Same as 1 but with additional y adaption for grids with prismatic layers First layer is adapted to have y equal to 1 0 1 Default method the edges are divided in two equally pieces if the difference in the flow variables p u v w p between the two end nodes is larger than CELFLO 2 The edges are divided if their two end nodes have a total pressure ratio lower than THRESHOLD 3 The edges are divided if their two end nodes have an entropy production larger than THRESHOLD 4 An eigenvalue analysis of the velocity gradient tensor is performed at each node The edges are divided if their two end nodes have a reduced velocity lower than THRESHOLD This method is only relevant for flows with vortices and requires a postprocessing file CFIPOST that contains the gradients of the velocity See the Edge Theoretical Formulation manual for more details THRESHOLD Threshold value to start the refinement Only used if SENSOR is larger than 1 Typical values are 0 950 0 999 for the total pressure ratio 1 5 for the entropy production and 0 1 0 8 for the reduced velocity RADIUS Defines the shape of a box that limits where the refinement is performed If RADIUS is positive it is the radius of a cylinder If RADIUS equals zero the box is an hexahedra Several RADIUS parameters can be given to de
65. his case the polyfit method Winzell 1996 In the polyfit method a number of boxes are defined such that the whole structure and the moving boundary is contained in the union of the boxes In each box there will be generated a least squares fit of those displacements from Edge amod in points belonging to the box using a set of up to 64 polynomials in streamwise and spanwise relative the box direction and this fitted polynomial expansion is used to set the displacements on all moving boundary points within the particular box Since the boxes can and often should 53 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Option Type Description scale real Maximum size of surface displacement in the bmsh grid units maxit integer Iteration limit for mesh deformation delta real Maximum residual for mesh deformation D Delete TRITET files at start and finish T Use the Tritet moving x prompt directly Table 15 aetribend options overlap it is always the last box that contains a particular mesh point that determines the displacements in this point This polyfit method is quite general Its drawback is that there might occur minor discon tinuities over box boundaries It is also very important to choose the number of polynomials few enough not to introduce the overshoots that are typical of least squares fit with to high degree of polynomials 3 8 5 Program aetribend Static Mesh Deformation
66. icit dual time stepping Default 2nd BETNMI order backward difference GAMNPI The two components for the residual in the 1 0 GAMN implicit dual time stepping Default 2nd order backward difference ITMEAN Time averaging switcher for unsteady 0 1 0 computations particularly DES and LES 1 time averaging desired 0 no time averaging will be made ITERM Time step at which time averaging starts function only if ITMEAN 1 NOTE If INPRES 2 and ITMEAN 1 this gt 1 1000 indicates that time averaging restarts from a previous run In this case ITERM lt ITER NPTHIS Number of points where time history is gt 1 1 recorded for unsteady computations POINT_COOD Coordinates of the point x y z POINT INDEX The indices of the points point and Table 30 Unsteady time marching 77 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default IMANEUV 0 No manoeuvre 0 1 0 1 Arbitrary manoeuvres IMVDTF 0 MOVDAT format 0 12 1 Time v1 v2 v3 wl w2 w3 1 MOVDAT format Time v1 v2 v3 w1 w2 w3 W1 W2 W3 2 MOVDAT format Time vl v2 v3 wl w2 w3 O11 O12 O13 O21 033 MOVDAT Give description of rigid movement of grid See format above Table 31 Manoeuvre parameters Name Description Value Default CFIMOS File Edge bmos with mode shape data Edge bmos CFIMOT File Edge amot with prescribed motion Edge amot parameters CFIMOP
67. id mix up with any other configure script that may exist in your PATH The configure script requires bash to run If bash is not in the default location bin bash the script will fail In this case you should run bash configure instead of configure You may choose to compile Edge in a directory tree separated from the source code tree This can be convenient when building Edge for multiple platforms from the same source code directory or from a source code directory to which you have no write access To accomplish this change directory to the build directory and run the configure script from there E g mkdir edge build cd edge build edge configure The following subsections describe the command line options that can be given to configure In most cases you don t need to specify any option to configure the script then attempts to determine all settings automatically 15 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 16 2 4 1 1 help The configure script takes several optional parameters type configure help for a complete listing of available options 2 4 1 2 config The config option determines which file in the config directory contains the makefile environment for your build E g to use the settings in config config i686 type configure config i686 If you give no config option the default value is taken from the environment variable MACHINE If MACHINE is not set the output of
68. ile can also be created manually using the example file supplied in examples ae inputfiles The listing of this three mode example file is given in Figure 11 The dataset str coordinates is the matrix of x y z coordinates for the selected subset of the structural grid in this case 32 points No element information is stored The structural grid points are optionally la belled with a set of unique integer identifiers grid idents default is 1 2 The modal data is gathered in a mode set and comprising mode datasets Each mode has a number of attributes stored as sub datasets The identifier is mandatory and must be a unique positive integer The str displacement is also mandatory and defines the modeshape The remaining mode attributes are optional and none of them are used by the program aedbelast The parameters generalised mass and frequency hz may be set or overridden by input from the amop modal parameter file If the generalized mass is absent a default value of 1 is assumed All mode attributes are automatically carried forward to the bmos modal shape data file FOI dnr 03 2870 Edge User Guide Issue 4 1 0 N 0x0 3 igi_interp_data L x 1 title intergrid data L L xl igi_method polyfit N 0x0 4 Boxes N 0x0 7 BOX_N RFE 4x1 BOX ROOT 20 4300003 RE 4x1 BOX TIP 27 552 38 6 RF x 1 THICK 2 50000000 I x 1 DEGX 1 I x 1 DEGS 3 AE x 1 IEDGE 2 I Lx 1 OEDGE 1 N 0x0 7 BOX_N RE
69. imiter for the mean flow 2 1 2 3 2 1 two values for the linear and non linear fields respectively 1 minmod 2 van Leer 3 superbee IROEAV Roe 1 or arithmetic 0 1 1 0 averages for the mean flow ENTRFX Entropy fix for the mean flow 2 1 2 1 1 2 1 05 two values for the linear and non linear gt 1 lt l fields respectively Scaled with spectral radius gt 1 or the magnitude of the velocity lt 1 1 25 means a lowest value of 25 of spectral radius of the eigenvalue 1 25 means 25 of velocity Table 26 Spatial discretization FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default NGRID Number of grid levels gt 1 4 MGRSTR Multigrid strategy 1 7 1 1 V cycles 2 W 3 F 4 V saw tooth cycles 5 W saw tooth 6 F saw tooth IFULMG Full multigrid 1 or not 0 Cannot be 0 1 0 used with restart INPRES 2 NFMGCY Maximum number of full multigrid cycles on 1 500 coarser levels NSWPFC Number of sweeps on the different grid levels 10 0 1 10 1 in coarse to fine sweep One number for each gt 0 level starting with the finest NSWPFC Number of sweeps on the different grid levels 10 0 1 10 1 in fine to coarse sweep One number for each gt 0 level starting with the finest MGSIMP Multigrid simplification on coarser grids 1 0 1 1 Uses a first order upwind scheme on coarser grids or a simplified numerical dissipation algorithm IPROLO Type of prolongati
70. int equivalent The adjoint equations depend on J via a source term At the time the source terms for the following functionals are available lift IDCDP drag IDCLP side IDCCP pitch IDCMP yaw IDCNP and roll IDCRP moments where the variables in parenthesis describe in the flow and adjoint input files the reference directions used to compute the forces and moments The source term is indicated by setting the name under the variable FUNTYP in Adjoint ainp A brief description of the result file from an adjoint calculation is given in Section 3 4 8 In order to carry an adjoint calculation it is recommended to proceed as follows Copy the input file Euler ainp used to carry the Euler flow calculation to Adjoint ainp for example The file Adjoint ainp will be used to carry the adjoint Euler calculation with Edge Since the adjoint equations are linear equations whith coefficients depending on the flow solution w it is of course required to provide the file with the Euler flow solution which name for example Euler bout must be given in the variable CFIEUL in Adjoint ainp The adjoint equations are solved if the variable ADJOPT is 1 or 2 In this version there are no more differences when running parallel or serial adjoint computations that is explained below If the variable ADJOPT is 0 Edge solves the flow equations The adjoint equations that are implemented in the present version of Edge correspond to the scheme IUPWIN 2 a cent
71. internal aerodynamics The mass flow through the boundary is specified and the static pressure relaxed in time to desired value so that the mass flow through the bound ary matches the target The value of Relaxation_constant must be supplied in boundary condition file Local boundary data cannot be supplied 3 6 4 16 Mach Outlet A boundary condition based on the weak static pressure and intended for subsonic outflows in internal aerodynamics An average Mach number through the boundary is specified and the static pressure adjusted so that the Mach number at the boundary matches the target Local boundary data can not be supplied 3 7 Using the Vortex Generator Model The vortex generator VG model simulates the presence of VGs in the solution by adding a source term to the governing equations Thus the VGs do not need to be physically present in the mesh The use of the vortex generator model is governed by the parameter VGCONFIG in the input file VGCONFIG is the number of the VG configuration used for the computation By default it is zero and the VG model is not used By VG configuration is meant a set of VGs Only one set of VGs can be used per computation but the edge file bedg file that contains the information on the VG configurations can have data for several configurations To import the VG geometry data in the edge file bedg file the program VG_identify has to be run after the preprocessor This program reads the input
72. internal structural model is a set of 1D harmonic oscillators one for each mode which are coupled to the flow via a generalized force term the modal force vector Each modal force component is an integral of the fluid forces over the whole movable surface weighted by the mode shape a surface displacement distribution defined in the mos input file see Section 3 8 6 1 The input data defining the modal structural model and the initial conditions for the simu lation is given in the modal shape data mos file Section 3 8 6 1 and the modal_parameters FOI dnr 03 2870 Edge User Guide Issue 4 1 0 dataset the mop file which is described in Section 3 9 2 5 The main modal parameters are natural frequency damping ratio and generalized mass These may be set in the mos data but it is recommended to use the mop file input Any value set in the mop file override those in the mop shape data file All modal parameters and initial conditions data are echoed to the Edge log file 3 9 2 Coupled Simulation Procedures 3 9 2 1 General Start up and Execution For all coupled simulations the initial set up procedure is the same the program aedbelast is used to construct a modal_shape_data dataset mos file as described in Section 3 8 4 This contains the mode shapes as full grid perturbation fields plus the basic modal param eters such as frequency and generalized mass If these parameters are not present in the mos data the d
73. iscosity 1 0 1 0 variable viscosity 0 RMU Free stream viscosity gt 0 1 7894e 5 PRREF Prandtl number 0 72 IFULNS Full Navier Stokes 1 0 1 0 thin layer approximation 0 Table 22 Viscosity related data Parameters except INSEUL apply to viscous calculations only In addition to the different turbulence models available according to the table below there are specific parameters for each turbulence model that are not given here These parameters may be found in the default input file Note that laminar regions are specified as part of the preprocessing FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default IROT 1 for a computation in a relative frame 0 1 0 of reference OMGROT Angular velocity and direction for the rotation 0 0 0 OMGCOO Coordinates for a point on the axis 0 0 0 Table 23 Computation in a relative frame of reference 71 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default ITURB O for a laminar calculation 0 2 3 4 0 2 for a differential turbulence model 3 for DES 4 for LES TURB_ Name of the turbulence model W amp J EARSM Hellsten k o MOD_ ITURB 2 RANS model Spalart Allmaras one eq model NAME Menter SST k o W amp J DRSM Hellsten k o std Wilcox k o Kok TNT k o Menter BSL k o Wilcox LRN k o Rung LLR k o PDH LRN k o W amp J EARSM std k o GW
74. isually for example using EnSight or the Edge Matlab program mshplot3d The definition of the moving and sliding nodes follows the Tritet program moving x which is run by the script aetribend For the movable surface the mesh nodes move as prescribed by the modal structural coordinates For a sliding plane for example symmetry plane in a half aircraft mesh the nodes are constrained to move in plane The motion of all other mesh nodes is computed by the mesh deformation program so far usually aetribend and moving x 3 8 2 1 The movable_surface file bset The moveable surface definition file usually created using program aexbset comprise the following datasets moving surfaces and optionally sliding planes containing lists of boundary names nodes moving and optionally nodes sliding containing the correspond ing lists of unique node numbers The FFA listing of a typical bset file is given below N 0x0 7 moveable_surface L t g 1 brand Edge 4 0 www edge foi se L 1x1 title surface data from aexbset L 1x1 parent data mdo e05 bmsh N 0x0 1 moving_surfaces L 1x1 boundary name wing N 0x0 1 sliding_planes L 1x1 boundary name symmetry IBN 9881 x 1 nodes moving 266 968 969 IBN 2512 x 1 nodes_sliding 123 3 8 2 2 The surface movement File bdis The bdis surface movement file format is used to store surface displacements based on the moveable surface dataset A typical file is shown in Fi
75. ithm of the current residual NRRES Default the density residual 76 Table 28 Steady state time stepping Name Description RSMPAR Implicit residual smoothing parameter gt 1 to be active IRESMO Type of residual smoothing coefficients 1 account for grid stretching NSMCYC Number of Jacobi smoothing cycles SMCOR Smoothing of corrections in multigrid Defined as RSMPAR and IRESMO 0 is applied for this smoothing O standard smoothing with constant Value Default gt 1 1 3 0 1 1 gt 1 2 gt 1 1 5 Table 29 Residual smoothing bound_point point on a boundary surface or inter point interior point FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default ITIMAQ Time stepping option 0 1 2 0 O steady state marching 1 time accurate dual time stepping 2 explicit time accurate ITMAX Number of time steps gt 1 1000 ISGLTS O use global time step DELTAT when 0 1 0 ITIMAQ 2 1 compute it when ITIMAQ 2 DELTAT Time step when ITIMAQ 1 gt 0 0 005 RESTAQ Convergence criterion for the inner 1 iteration in dual time stepping ITIMAQ 1 The logarithm of the maximum norm of the chosen residual NRRES is used default is the density residual ITMXAQ Maximum number of inner iterations gt 1 100 ITMNAQ Minimum number of inner iterations 1 ITMXAQ 3 BETNPI The three components for the unknown in the 1 5 2 0 5 BETN impl
76. ition To do this a series of static solutions must be obtained for example as a function of incidence angle Once the static solution is available for the required flight condition the dynamic simulation can be run using a restart INPRES 2 in the main input file To restart from the static solution however the timestep in the initial res and out files must first be set to the value of DELTAT for the dynamic simulation The easiest way to do this is to use the Edge Matlab program timeset One day soon a simpler procedure will be made available using and INPRES 1 start up On restarting from the static solution the initial velocity for all modes must be set to zero using the input init_velocity as described in Section 3 9 2 5 The structural excitation input can be set either via the initial conditions in the mop file or by using a modal force input supplied as a mot timeseries input The force input is selected via the variable FORCED in the main input file The output from a modal coupled dynamic simulation is the time history data in the residuals res file modal coordinates velocity and forces and the solution sample fields in the out files The solution sampling frequency can be set using the variable IMULOU in the main input file 3 9 2 4 Procedure for Design Shape Simulation The procedure for a design shape simulation is simpler than for jig shape in the no initial stati
77. ive major components e Preprocessor constructs the dual mesh and edge based datasets e Edge flow solver performs the main flow computation e Helper programs a general toolbox for manipulating input and output data e Application programs programs for shape optimization mesh deformation and aeroe lasticity e Xedge a graphical user interface GUT generator for Edge The flow solver employs an edge based formulation which uses a node centered finite volume technique to solve the governing equations The control volumes are non overlapping and are formed by a dual grid which is computed from the control surfaces for each edge of the primary input mesh The relationship between the dual and primary input mesh is illustrated in Figure 1 In this example a set of hexagonal control volumes are constructed from a simple triangular input mesh In any Edge mesh all the mesh elements are connected through matching faces Edge meshes therefore may not contain hanging nodes vi noso V2 V3 Figure 1 The input grid solid also denoted triangular grid and its dual grid dashed forming the control volumes In the flow solver the governing equations are integrated explicitly towards steady state with Runge Kutta time integration Convergence is accelerated using agglomeration multigrid and implicit residual smoothing Time accurate computations can be performed using a semi 11 FOI dnr 03 2870 Edge User Guid
78. l Electronic vibration max residual Electronic energy rms residual Electronic energy max residual rhou_rmsresidual RT rhou_maxresidual RT rhov_rmsresidual RT rhov_maxresidual RT rhow_rmsresidual RT rhow_maxresidual RT rhoe_rmsresidual RT rhoe_maxresidual RT rhok_rmsresidual RT rhok_maxresidual RT rhoep_rmsresidual RT rhoep_maxresidual RT rhoet rmsresidual RT rhoet maxresidual RT rhos rmsresidual RT NSP rhos maxresidual RT NSP evs rmsresidual RT NVI evs maxresidual RT NVI eel rmsresidual RT 1 eel maxresidual RT 1 e e eee e i i ne i no oooocoocoocoocoococoocoocococooqcqcoooqoco eo Table 40 Datasets that may occur in the residual file 87 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 88 FOI is an assignment based authority under the Ministry of Defence The core activities are research method and technology development as well as studies for the use of defence and security The organization employs around 1350 people of whom around 950 are researchers This makes FOI the largest research institute in Sweden FOI provides its customers with leading expertise in a large number of fields such as security policy studies and analyses in defence and security assessment of different types of threats systems for control and management of crises protection against and management of hazardous substances IT security an the potential of new sensors FOI Swedish Defence Research Agency Tel 46 8 555 03000 www foi
79. l ffauline ffauplane ffaview ffaxyz fluent_cgns2ffa force merge_boundaries merge_partitions meshdeform plotmot plotres preprocessor pres scale split_partitions stout struct2unstruct tau2ffa tecgridbound tecgridvol tecsolbound Check binary file format whether it is on big or little endian Converts real data in double precision to single precision Java tool to display difference between two FFA format files Java tool to edit an FFA format file Lists the contents of any file in FFA format Mirrors a grid file in its plane s of symmetry thus removed Interactive program to print out the coordinates for a given node Cuts a boundary with a plane in a 3D problem Interpolates the solution based on the output from ffaucut ffauline Interactive program to extract data along a line Main program for cutting the edge mesh file boundaries and volume View and manipulate FFA files using a GUI Requires Python 2 pygtk and numarray packages Interpolates an unstructured grid to user specified XYZ points Converts an unstructured file from Fluent CGNS to FFA format Computes forces and moments Merges and reduces number of boundaries in a mesh file Merge solution files from a parallel run This is called from edge_mpi_run Utility for mesh deformation Plots modal timeseries tabulation data can also be used from Matlab XMGR program to plot convergence history Produces dual grid agglomeration and partitioning for pa
80. l states strong total states pressure weak inlet weak static pressure weak extrapolation weak free stream propulsion inlet propulsion outlet nacelle inlet nacelle exhaust mass flow inlet mass flow outlet mass flow outlet JWS Mach outlet Characteristic boundary conditions Characteristic boundary conditions with vortex correction in 2D Subsonic inflow boundary condition Subsonic inflow boundary condition Subsonic inflow boundary condition Subsonic outflow boundary condition Extrapolation of all variables All variables specified to free stream Propulsion inlet Propulsion outlet Nacelle inlet Nacelle exhaust Subsonic inflow boundary condition Subsonic outflow boundary condition Subsonic outflow boundary condition Subsonic outflow boundary condition Table 12 External boundary conditions are unknowns and the specified conditions are implied trough the flux Usually constant free stream values are used at all boundaries Some of the boundary conditions however may be given local varying data by supplying the bedg file with a dataset N 0x0 boundary_data as subset to the boundary in question and with local data to the boundary supplied as sub sets to this dataset boundary_data Required data is given for each boundary condition below Additional dependent variables like turbulence species etc are always specified weakly from global or local boundary data at a local inflow or extrapolated i
81. llows a simulation to be performed using a geometry approximating to the design shape the shape which the flexible wing assumes in flight 3 9 2 3 Procedure for Jig Shape Simulation The first step in a Jig Shape simulation is to obtain a static aeroelastic solution the flow solution with the fluid loads and elastic forces in static equilibrium Static aeroelastic solutions are obtained by setting the modal damping in the mop input file to an arbitrary positive value lt 1 This is most easily achieved by setting the global default damping value and leaving out the separate modal damping inputs With heavy structural damping like this the structural equation of motion can be integrated with a very coarse timestep DELTAT in 59 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 60 the main input file The static solution can thus be obtained with a much shorter run time than is needed for a dynamic solution To monitor the convergence of a static coupled simulation it is usually sufficient to use the program plotres to display the forces and residuals The modal forces and variables can be extracted from the res file and plotted using the Edge Matlab program plotmores The deformed surface shape and the surface fields can be visualized in EnSight using the converter program ubody For a Jig Shape simulation it is usually necessary to obtain the static solution for a given Cz or other flight cond
82. lowable time step is determined by the smallest control volume Explicit time marching can not be combined with multigrid and implicit residual smoothing The physical time scales of the problem usually give a hint on how large DELTAT should be For instance there should be at least some 30 60 time steps per cycle of a harmonically oscillating component to be resolved Best practice is to plot all available global quantities like aerodynamic forces or gener alized forces to verify that they do converge within the inner iterations In this respect moments are sharper probes than lift components Incorporated into the time accurate simulation with unsteady RANS DES or LES a pa rameter ITMEAN is used to switch on 1 the time averaging For DES and LES this parameter should be set as ITMEAN 1 since the default output of the solution is for the instantaneous resolved large scale field Setting ITMEAN 1 enables to make time averaging of the instantaneous field and gives output of the time averaged mean field Another parameter ITERM specifies the time step at which the time averaging process starts This should be justified by looking into the time history of local solution variables in the solution procedure Up to the specified time step at ITERM the flow should have been fully developed This can be observed by recording the time history of the flow variables at some typical locations A proper value for ITERM is justified by th
83. lso be prepared by direct manual editing The file is an FFA format ASCII file suffix ainp and should initially be a copy of the default input file examples default_edge ainp in which all input variables are set to default values The input file is constructed by manually resetting the appropriate parameters Typically an existing case specific input file is used together with the default file as a template and the editing process is assisted with a graphical difference tool such as tkdiff for example A complete list and explanations of all main input variables is given in Appendix A Many of these variables are also described in the Edge Theoretical Formulation manual 29 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 30 N 0x0 9 unstr_grid_data L x title 12 2_adapt_1 RF 88703 x 2 coordinates 1 14088404 1 140531 N 0x0 4 region L x region_name volume N 0x0 Zh element_group L x element_type tria3 IF 87235 x 3 element_nodes 36736 37929 N 0x0 2 element_group L x element type quad4 IF 44457 x 4 element nodes 123 N 0x0 3 boundary L x boundary name slat p N 0x0 2 belem_group L x bound_elem_type bar2 IF 371 x 2 bound_elem_nodes 279 280 N 0x0 37 boundary L x boundary_name wing p N 0x0 2 belem_group L x bound_elem_type bar2 IF 591 x 2 bound_elem_nodes 650 651 N 0x0 3 boundary L x boundary name flap p N 0x0 2 belem_group L x bound_elem_ty
84. ly critical with respect to the turbulence modeling and that the results are unreliable In this case a comparison with experimental data may be necessary in order to determine which model is most accurate 3 5 2 2 DES and Hybrid RANS LES The DES Detached Eddy Simulation or hybrid RANS LES model is chosen by setting ITURB 3 In this version besides the Spalart Allmaras DES model a hybrid RANS LES model Peng hybrid HYBO model is also available The DES and the hybrid RANS LES model has been developed for unsteady aerodynamic flows characterized usually by massive boundary layer separation and or after body vortex shedding In applications of CFD analysis of flow control aeroacoustics and fluid structure interaction dealing with unsteady time dependent flow phenomena the DES and hybrid RANS LES methods should be taken into consideration in the selection of modelling approaches These 37 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 38 methods are able to give rich resolved flow structures and time dependent turbulent fluctu ations 3 5 2 3 LES LES Large Eddy Simulation is chosen by setting ITURB 4 The model employed in LES is often termed SubGrid Scale SGS model Three SGS models have been imple mented in Edge with the Smagorinsky SGS model Smagorinsky 1963 as the default The other LES model is the SGS variant of the Spalart Allmaras model which has been implemented for testing the mo
85. mat files CGNSlib Used for conversion utility to from CGNS format files e LAPACK and BLAS Actually not required a stripped implementation is included with the Edge software distribution but optionally an external implementation may be used instead Used by a mesh adaption utility e UMFPACK and AMDLIB Used by a mesh deformation utility Xmgr Used by plotres to plot convergence curves 2 3 Directory Structure The Edge source code is usually supplied as a compressed tar file This file should be un compressed and unpacked in a chosen installation directory for instance using the following Unix command gunzip edge tgz tar xf After unpacking the Edge tarfile the files and directories listed in Table 1 and 2 respectively should appear The source code directories listed here are the first level of a hierarchical structure and compilation system based on Makefiles Configuration settings for the compilation process are determined automatically by running the configure script Some additional configuration parameters are stored in a set of files in the config directory The Xedge GUI is compiled separately from the main Edge system The Matlab programs supplied with Edge are based on the FFA Matlab Toolbox a special set of Matlab functions for handling the FFA data format see Section 3 3 This Toolbox is not specific to Edge and so is supplied separately Python 2 4 GTK and pyGTK Required to use the ffaview py utility
86. mber 0 720000028610229 DC x 1 dynamic_pressure 14980 7743211015 N 0x0 6 block DF 22088 x 1 density 0 434306166672034 DF 22088 x 3 velocity 255 959142902464 DF 22088 x 1 pressure 37609 3537068229 DF 22088 x 1 turb_kin_energy 17 8736499360706 DF 22088 x 1 turb_omega 314 440368967509 DF 22088 x 1 turb_viscosity 4 194888755189772E 002 Figure 6 A typical solution bout file 3 4 5 The Sample Solution Files Edge timestp bout The user has an option to write a sample output file similar to the Edge bout file at spec ified time steps The input variable IMULOU controls this output so that Edge writes an Edge timestep bout every IMULOU th step The file contains a time stamp Figure 7 shows a typical intermediate solution file 3 4 6 The Solution File for Post Processing Post bout The file Post bout is generated whenever additional data on a separate file is desired POST gt 0 The format is exactly the same as for the solution file Edge bout the contents is deter mined from the variable IPOST Several options for additional output of field data to the Post bout exist e g surface skin friction vector surface heat transfer wall distance field laminar turbulent field gradients residuals etc Note that surface quantities like skin friction and heat transfer have values on wall boundaries only and are zero elsewhere These quantities are scaled with dynamic pressure and made non dimensional 3 4 7 The R
87. noeuvres and aeroelastic simulations It can be used to compute both viscous and inviscid flow problems It also contains mesh adaption functionality and an inviscid adjoint flow solver This users manual describes Edge from a user s point of view It contains information on how to install and run the Edge solver and related utility programs such as preprocessor postprocessor various file conversion utilities etc All required input parameters are de scribed as well as the contents of the solution files and other output files This document does not contain a complete theoretical formulation of the numerical meth ods in Edge Nor does it contain details on implementation or design for software develop ers FOI dnr 03 2870 Edge User Guide Issue 4 1 0 10 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 1 Introduction This document is part of Edge a parallel CFD code developed by FOI see Eliasson 2001 and Edge homepage Edge is a parallelized CFD flow solver system for solving 2D 3D viscous inviscid compressible flow problems on unstructured grids with arbitrary elements Edge can be used for both steady state and time accurate calculations including manoeuvres and aeroelastic simulation There is also an inviscid adjoint flow solver that can be used for gradient based shape optimization There is also functionality for mesh deformation postprocessing file format conversion and other utility tasks The Edge system comprises f
88. nput base mesh filename 2 in bmsh input deformed mesh filename 3 iu bmos input output modal shape data filename The program aeputpert automatically either updates and existing bmos file or creates a new file from scratch The perturbation field dataset is computed as the difference X deformed Xbase Where X denotes the coordinates matrix of the corresponding mesh file Any mode attributes stored in the deformed mesh file are automatically transferred to the bmos output The perturbation field is automatically rescaled using the dis_scale_factor dataset if this is present The manual options indent and scale override any information read from the de formed mesh file 3 8 6 1 The modal shape data File Edge bmos The modal shape data bmos file contains the entire set of modeshape data required for an Edge modal Aeroelastic computation The listing of a bmos file with three modes is shown in Figure 13 The modal shape data comprises a header block followed by a series of mode datasets and each of these has a set of attributes in the form of sub datasets In this example the attributes are as described for elastic modes in Section 3 8 4 1 with the addition of the perturbation field dataset The modeshape format dataset in the header block defines type of modeshape data In future versions of Edge with on line mesh deformation the modeshapes will be supplied as displacements on the moveable surface Currently the only su
89. o CRATIOB 2 As an additional consequence of this the coarser grids become finer and the overall computing time increases The residual smoothing has occasionally caused problems and it may be worthwhile to alter the value of RSMPAR to see the effect This is rare though Another reason behind poor convergence is that the flow is actually unsteady in nature and hence there is no steady state solution In this cases the unsteady option TTIMAQ 1 may be considered 3 5 4 Time Accurate Simulation and Inner Iterations Setting ITIMAQ 1 enables time accurate simulation with implicit time stepping There will be ITMAX time steps with global time step DELTAT The driver of the implicit time stepping is a sequence of inner iterations by means of the dual time stepping method The number of such inner steps is controlled by three parameters in the input file RESTAQ see above which is a convergence criterion on the residuals ITMXAQ which is the maximum number of inner iterations to be used and ITMNAQ which is the minimum number of such inner steps Usually ITMXAQ is the important parameter This has to be large enough to ensure suf ficient convergence of the flow variables at each physical time step before going on to the next There is no general perception of what this number should be although 30 100 in ner iterations could be required for a full viscous computation To approximately estimate the number of sub iteration one can observe t
90. ociated with the current node Setting RUPTOL to zero makes this condition inactive Note that the tolerances RCGTOL and RUPTOL and the maximum number of iterations ICGMIT may need to be adjusted if the PCG algorithm is used RUPTOL 0 and a conservatively small value of RCGTOL is used in the example A good enough mesh is obtained with much smaller number of iterations An optional flag in the input file specifies if a log file should be created When IMDLOG 1 meshdeform produces a file in FFA format with some data from the computation No log file is produced when IMDLOG 0 or if IMDLOG is missing in the input file The file name for the log file should be specified in variable CFILOG if IMDLOG 1 The log file is mainly useful when the iterative solver is used in order to record the residual as a function of iteration number 65 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 66 References CHEVALIER M 2005 Xedge v1 2 User Manual Scientific Report FOI D 0210 SE Computational Physics Department Aeronautics Division FOI Conway S 2004 Implementation and Validation of an Engine Nacelle Boundary Con dition in Edge 3 2 Scientific Report FOI R 1320 SE Computational Physics Depart ment Aeronautics Division FOI EDGE HOMEPAGE http www edge foi se ELIASSON P 2001 EDGE a Navier Stokes Solver for Unstructured Grids Scientific Re port FOI R 0298 SE Computational Aerodynamics Department Aeronautics Division
91. of the boundary conditions file should normally be necessary 3 6 1 Connectivity Boundary Conditions There are three different connectivity boundary conditions given in the table below A boundary with a periodic boundary must be connected to another boundary assigned the same boundary condition and there must be a node to node correspondence for the nodes on these boundaries The test of the node to node correspondence is carried out in the preprocessor Inside Edge the fluxes on the connectivity boundaries are computed from averages of the values on two connected boundaries For the rotation boundary conditions the velocity vectors are rotated before averaged 3 6 1 1 Periodic Boundary Conditions Periodicity assumes that the boundary and the connected boundary are connected through a translational vector and that the difference between each node on the boundaries corre sponds to this vector The vector is given as input to the program bound where also some testing is done to establish the correct sign of the vector 3 6 1 2 Rotation Boundary Conditions Rotation boundary condition assumes that the boundary and the connected boundary are connected by a rotation matrix The rotation matrix is established by supplying an axis of rotation a point on the axis and an angle This data is supplied to the program bound where testing is done to establish the correct sign of the angle 3 6 1 3 Propeller Boundary Conditions The propeller bounda
92. on O injection 0 1 2 0 1 account for gradients on coarser grids 2 first order with limiters IRESTR Type of restriction currently only linear 0 0 RESFMG Convergence criterion for the full multigrid lt 0 4 5 cycling Iterations stopped when max and rms residual of the logarithm of the current residual NRRES Default the density residual Table 27 Multigrid options 75 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Description Value Default ITMAX Maximum number of iterations gt 0 1000 NSTAGE Number of Runge Kutta stages gt 1 3 IRKCO Runge Kutta coefficients 0 66667 0 66667 1 ISWV One number for each stage when to compute 10 0 1 1 0 0 viscous fluxes 1 or not 0 ISWS One number for each stage when to compute 10 0 1 1 0 0 source terms 1 or not 0 IBOTH 1 ISWV applies to both numerical dissipation 0 1 1 and the physical viscous dissipation 0 ISWV applies only to physical dissipation Numerical dissipation recalculated in all stages CFL CFL number gt 0 1 25 CFLVIS Viscous CFL number Minimum convective viscous 1 or gt 0 1 time step chosen For negative values only CFL used and harmonic mean of the time steps is used CFLREF CFL reduction factor on coarser grids CFL gt 0 0 8 and CLFVIS are reduced with where IGRID 1 is the grid level 1 being the finest level RESRED Convergence criterion Iterations stopped lt 0 5 5 when max and rms residual of the logar
93. operations is as follows 1 The program aexbset is used to set up a moveable_surface dataset defining set of boundaries which forms the moving wetted surface grid This is stored in a bset file 2 The program aedbelast or aedbrot for rigid rotations is used to generate a series of surface_movement datasets one for each mode These are stored in files labeled with a mode identifier _ident bdis 3 The script aetribend is used to produce for each modeshape a deformed mesh a complete bmsh mesh file conformed to the modal bdis surface displacement 4 The program aeputpert is used to calculate for each mode a perturbation field which is then added to a modal_shape_data dataset stored in a Edge bmos file The Edge bmos file contains the modal perturbation fields and optionally the modal param eters natural frequency damping and mode description This is the main input to Edge for any modal aeroelastic calculation IAEOPT Option description 0 Rigid grids no aeroelastics 100 Modal Fixed Rigid 101 Modal Prescribed Motion 102 Modal Coupled orthogonal modes central difference Table 13 External boundary conditions FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Files required Application Files generated ae help Edge bmsh aexbset bset Edge bmsh aedbrot bdis bset Edge bmsh aedbelast _ident bdis or bmos bset amod w aigi E
94. orithm 4 Cache reordering using saw2 algorithm 5 Cache reordering using bfs1 algorithm 6 Cache reordering using bfs2 algorithm 7 Cache reordering using bfs3 algorithm NLAMREG Number of laminar regions 0 0 LAMINAR SETTING Dataset containing data for all regions No data LAMINAR REGION Dataset for each laminar region No data PLANE_COO Planes with coordinates for the laminar region PLANE_NORMAL Normals of the planes with for the laminar region MAXLAMDIST Maximum distance to the nearest wall for all gt 0 0 003 laminar regions Table 20 Preprocessor variables 69 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 70 A 3 Flow Solver Name Description Value Default ITYGAS Type of gas 1 2 1 1 calorically perfect gas 2 thermally perfect gas GAMMA y gt 1 1 4 CP Cy 1004 5 NSPEC Number of species ITYGAS 2 1 1 SCHMID Schmidt number for NSPEC gt 1 gt 0 0 5 SCHMTU Turbulent Schmidt number for NSPEC gt 1 gt 0 0 9 NGTAB Input table size for thermally perfect gt 1 2 gas ITYGAS 2 IGTTAB Table for variable temperature ITYGAS 2 gt 0 0 2000 IGTAB Table for variable gamma ITYGAS 2 gt 1 141 4 One dimension per specie required IRTAB Table for variable gas constant RGAS ITYGAS 2 287 287 One dimension per specie required Table 21 Gas related options Name Description Value Default INSEUL Viscous calculation 1 0 1 0 Euler calculation 0 ICOVIC Constant v
95. ould work fine MIMD type parallel computers e g Linux clusters are supported e A Fortran 95 compiler e g Intel ifc 7 Intel ifort 8 or 9 lahey f95 GNU gfortran 4 1 0 will not work due to some compiler bugs and incompatibility of unformatted VO e Bash Bourne Again Shell v2 05b or later e GNU make 3 79 1 or later freely available at http www gnu org software make make html e A C compiler 2 2 2 Optional The following tools are optional If an optional component is not available Edge will be built without the functionality that requires the unavailable package s e Java 2 Platform Standard Edition v1 4 Runtime Environment Required to run Xedge 13 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 14 Filename Contents configure A script used to automatically configure Edge COPYRIGHT Copyright statements INSTALL Installation instructions INSTALL BINARY Installation instructions for binary distribution Makefile Top level makefile README Short description of the directory contents rules mk Makefile snippet containing generic build rules Table 1 Files in the root directory e Java 2 Platform Standard Edition v1 4 Development Kit Required to recompile Xedge e An MPI implementation e g LAM MPICH Required to take advantage of a parallel computer e METIS Required to partition the computation problems on parallel computers NetCDF Used for conversion utility to from NetCDF for
96. ount of information the output is truncated and only the fine grid information is given 3 4 4 The Solution File Edge bout Edge bini The file Edge bout is generated from the flow solver and contains the nodal values of all primitive variables in the flow solution The nodal ordering is the same as in the bmsh grid file This file may be used to restart an Edge computation from a the saved solution INPRES 2 This file may also be used as an initialization solution file suffix bini INPRES 1 The contents of the file depend on the physical model used Euler Navier Stokes laminar turbulent steady unsteady etc A listing of an example output from a turbulent steady state solution is given in Figure 6 For time accurate solutions there are two groups of flow field datasets corresponding to the current and previous time steps and these are located under time datasets For a restart both of these time levels are required together with the bres time history file If used for an initialization as a bini file only the first time group is used and it is treated as steady state data 31 FOI dnr 03 2870 NB IF DF DF DF IF NB LB IF DF IF NB LB IF DF IF NB LB IF DF IF NB LB IF DF IF 32 0 220397 88703 88703 220397 37 37 37 59 59 59 278 278 278 21 21 21 xxx KM KM KK KM MK KKM KKK KM KM KKM MK KK KM KM KK KM KK KM EX 0 2 H 3
97. pe bar2 IF 278 x 2 bound_elem_nodes 1 2 3 N 0x0 3 boundary L x boundary_name outer N 0x0 2 belem_group L x bound_elem_type bar2 IF 21 x 2 bound_elem_nodes 46104 Figure 3 A typical grid bmsh file 3 4 1 The Input Grid File Edge bmsh A grid file is required to start any application of Edge and must be prepared using a mesh generator such as the FFA developed system Tritet or commercially available tools such as ICEM The grid file is written in binary FFA format suffix bmsh A typical file listing is given in Figure 3 Illustration 3 is printed using the program ffalist and contains the type of the dataset to the left the size and dimension in the middle and the and either the number of subsets or the first data to the right The grid file contains a list of coordinates a list for each element type triangles tria3 and quadrilaterals quad4 with node numbers referring to the list of coordinates Each of the boundaries four in the example should be given a name boundary_name and a type bound_elem_type and list of nodes bound_elem_nodes for each type 3 4 2 The Boundary Condition File Edge aboc This file is generated by the program bound and contains for each boundary a complete description of the boundary condition to be used This is a small file and is therefore usually an ASCII file An example of it is given in Figure 4 FOI dnr 03 2870 Edge User Guide Issue 4 1 0
98. perturbation field in the bmos modal shape data file so that there is no need to adjust the generalized mass values When processing a series of bdis files such as a set of modeshapes with a common base mesh it is recommended not to use the D option The program moving x can then re use some of the data required for computing the mesh displacement field file MOV TRITET FAC This makes the mesh deformation process significantly faster 54 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Options Type Defaults ident lt integer gt Integer mode identifier 1 2 3 as added scale lt real gt Scale factor for perturbation field 1 Table 16 aeputpert options The option delta sets a convergence criterion for the mesh deformation which is the max imum fractional change in the nodal displacement between successive iterations Setting a coarser value can speed up convergence but may compromise the smoothness of the final deformed mesh 3 8 6 Program aeputpert Modal Perturbation Fields Program aeputpert is used to construct the bmos modal_shape_data file when using the perturbation field method as described in Edge Theoretical Formulation manual For this method the modal shape data is a perturbation dataset The format of the bmos output file is described in Section 3 8 6 1 The aeputpert command line and arguments are as follows Usage aeputpert in bmsh in bmsh iu bmos Options Arguments 1 in bmsh i
99. pported format is perturbation field 55 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 56 N 0x0 7 modal shape data L x brand Edge 3 2 www edge foi se L x title Modeshapes for mesh file L X modeshape_format perturbation field I 3 n modes 3 N 0x0 5 mode I x identifier 101 L x description lst elastic mode D x frequency_hz 0 8399999737739563 D al generalised mass 0306 00000000000 DF 76801 x 3 perturbation 0 0000000000 N 0x0 5 mode I x identifier 102 L x description 2nd elastic mode D x frequency hz 2 164000034332275 D ae generalised mass 4552 799804687500 DF 76801 x 3 perturbation 0 000000000000 N 0x0 9 mode P x identifier 103 L x description 3rd elastic mode D x frequency_hz 3 563999891281128 D x 1 generalised_mass 2457 000000000000 DF 76801 x 3 perturbation 0 00000000000 Figure 13 A typical modal shape data bmos file 3 8 7 The mottab Signal Tabulation System The mottab system is used to input tabulated modal vector timeseries to Edge There are several potential applications n the present code mottab is only used for defining modal pre scribed motion solutions In the near future mottab will be used for defining structural force inputs for coupled aeroelastic simulations and for specifying manoeuvres and similar time dependent inputs A mottab modal timeseries tabulation is always stored as signal_table dataset in a amot or
100. ral scheme with a slightly different implementation than that with IUPWIN 0 To be consistent with the flow calculation associated to the adjoint equations the Euler equations should be solved using this scheme too IUPWIN 2 The boundary conditions that are available for the adjoint equations are the counterparts of the symmetry euler wall and farfield boundary conditions without vortex correction Attempts to solve the adjoint equations with other boundary conditions see the file aboc which name is given by the variable CFIBOC in both the flow and adjoint input files will cause Edge to stop and exit with an error message ERROR IN EDGE IN ROUTINE ADJINI BOUNDARY TYPE weak charact vortcor NOT YET IMPLEMENTED if using a farfield boundary condition with vortex correction for example The adjoint calculation can be run using exactly the same options as for solving the flow multigrid time integration CFL 3 10 1 Serial Computation The required edge file and boundary file are the same as for the Euler flow calculation if ADJOPT is 1 The output file described in Section 3 4 8 will contain the adjoint variables Additional calculations can be performed in Edge if ADJOPT is 2 giving the gradient of J with respect to the nodal coordinates X in the output file that is described in Section 3 4 8 The additional calculations ADJOPT 2 require additional data to be present in the edge file CFIEDG is for e
101. rallel computations Extracts data on a boundary column by column Scales a given grid Split solution file into its partitions for parallel computations Called from edge_mpi_run Strips Edge standard output data for plotting Converts a Euranus mesh file with boundary conditions to an Edge mesh file Converts grid solution from tau NetCDF to FFA format Exports the boundaries of a 3D grid to Tecplot format Exports the interior and boundaries of a 3D grid to Tecplot format Exports the boundaries of a 3D solution to Tecplot format Table 4 List of all available programs scripts in Edge FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Name Function tecsolvol Exports the interioir and boundaries of a 3D solution to Tecplot format transform 2 5d Main program to convert a 2D bedg file to compute in 2_5D ubody Exports the boundaries of a 3D grid and solution to EnSight 5 format uvol Exports the interior and boundaries of a grid and solution to EnSight 5 format VG identify Main program for specifying vortex generators VG s in Edge Table 4 Continued List of all available programs scripts in Edge several sub datasets each of which is also an FFA dataset In this way a hierarchical tree like structure is assembled The five descriptors in each identifier set are 1 name character 16 the name of the dataset 2 type character 4 a four character code as in Table 5 3 ndim integer 4 the
102. rameter ITIMAQ gt 0 ITIMAQ 1 is recom mended and consequently other related parameters should be specified to enable a time dependent calculation All RANS models are integrated down to the wall low Re approach and the wall normal grid resolution for the closest near wall nodes must be of the order of yt 1 With DES and other hybrid RANS LES models the wall normal grid resolution is similar to the low Re RANS approaches Note that DES is a hybrid approach combining near wall RANS modeling and off wall LES modeling With the Spalart Allmaras DES model it is preferable to have the near wall RANS switched to LES at a location around the outer edge of the boundary layer FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 5 2 1 RANS RANS Reynolds Averaged Navier Stokes equations models can be chosen by setting ITURB 2 There are several RANS models to choose from Each of these models has a number of parameters It is recommended that these parameters are not changed by the user These parameters are therefore not listed in the Appendix The following four models are recommended for RANS computations e Spalart Allmaras one eq model Spalart amp Allmaras 1994 One of the most widely used models for aeronautical flows With only one turbulence transport equation it is relatively simple and robust and gives reasonable results in attached and slightly separated flows e Menter SST k omega Menter 1994 One of the most
103. ropulsion inlet 40 Propulsion outlet 41 Nacelle inlet 42 Nacelle exhaust 43 Mass flow inlet 44 Mass flow outlet 45 Mass flow outlet time relaxation type 46 Mach number outlet Table 3 Boundary conditions codes There is a large number of parameters gt 100 for different options in the preprocessor numerical schemes in the flow solver turbulence models etc More parameters than the above mentioned may have to be changed A more complete list of available options for all programs is given in Appendix A The values of the different parameters are given default values that should for most cases work well and need therefore not to be adjusted Itis recommended to use different input files and different file names for computations with different conditions 3 1 2 Specifying Boundary Conditions The boundary conditions must be specified for each boundary that is given in the grid To set the boundary conditions execute bound and the program will ask for the input file and a specification of a boundary type for each boundary The available boundary conditions are listed in Table 3 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 P ies N z gt uo ji D w w a 3 wa oo K n w Figure 2 Enumeration of corner points for the volume elements triangle quadrilateral tetrahedron hexahedron prismatic element and pyramid Once all boundary conditions are specified bound allows t
104. ry condition is a disk model that simulates the presence of a propeller by adding a force to the momentum and energy equations Additional data has to be sup 41 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 42 Name Function Weak euler Zero normal velocity Weak adiabatic No slip Assumes zero normal derivative of the temperature Weak isothermal No slip Assumes constant wall temperature Table 10 Wall boundary conditions Name Function Weak euler Zero normal velocity Table 11 Symmetry boundary conditions plied for each propeller boundary e g number of propeller blades size of propeller radial lif force distribution etc An example file of what data to supply is given in the file exam ples propeller_example dat 3 6 2 Wall Boundary Conditions There are three different wall boundary conditions Euler conditions or for viscous flows adiabatic and isothermal conditions The most common type for viscous flow is the adia batic wall boundary condition The Euler conditions can only be applied to inviscid flow The Euler wall condition is a weak condition i e all variables on the wall are unknowns and the specified conditions are implied trough the flux The two viscous boundary conditions use strong conditions for the velocity i e the velocities are specified usually no slip and are hence no longer unknowns At an isothermal wall a weak condition is implied on the constant wall temper
105. se SE 164 90 STOCKHOLM Fax 46 8 555 03100
106. that for a jig shape simulation as described above FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 9 2 5 The Modal_Parameters File Edge amop A typical modal parameters mop file is provided with Edge in the directory examples ae_inpufiles The FFA listing for this file is given below N 0x0 2 modal parameters R 1x1 default damping 1 00000005E 03 N 0x0 3 mode set N 0x0 4 mode I 1x identifier 103 R 1x frequency hz 1 10000002 R 1x generalised mass 1111 00000 R 1x init velocity 0 109999999 N 0x0 4 mode L 1x identifier 101 R 1x frequency hz 2 20000005 R 1x generalised mass 2222 00000 R 1x1 damping ratio 1 99999996E 02 N 0x0 3 mode I 1x identifier 102 R 1x frequency_hz 3 29999995 R 1x generalised_mass 3333 00000 The main dataset modal_parameters contains a single variable dataset default damping and a mode set The modal parameters are entered separately for each mode in the mode set In this example there are just three modes For each mode dataset the following input variables are available Dataset Description notes Default identifier Unique integer identifier None mandatory input frequency _hz Natural frequency From mos file or 0 generalised mass Mass a normalization factor From mos file or 1 damping_ratio e g 1 for critical damping default_damping init velocity Initial modal velocity 0 init_coordinate Initial modal coordinate 0 The dataset d
107. timeseries S x signal_type pulse R x t_step 4 99999989E 03 R x t_start 2 99999993E 02 R x amplitude factor 1 00000000 N 0x0 2 mode_set N Ox 0 2 mode I x identifier 101 RT 1001 x timeseries 0 0000 00 6 28310E 03 1 2566040 N Ox 0 2 mode I x identifier 102 RT 1001 x timeseries 0 0000000 0 000000 0 0000000 Figure 14 A typical mottab timeseries file 3 8 7 2 The mottab timeseries Signal Format The listing of a typical timeseries signal_table is shown in Figure 14 The signal_table dataset comprises a block of header datasets and a mode_set comprising a set of mode datasets Each mode dataset has two mandatory attributes the identifier a unique positive integer and the timeseries a vector of values for the modal variable at uniformly spaced intervals The number of points in the timeseries is the same for all modes The dataset t_step is mandatory and defines the table time step and hence the duration of the tabulated series The dataset signal_type is mandatory and takes the values pulse or periodic The signal_type is periodic then the tabulated series is repeated indefinitely to produce a periodic wavelike signal The effect of these two signal_type inputs is shown by these example files time_1001_halfwave_pulse amot and time_1001_halfwave_periodic amot The inputs t start and amplitude factor are optional and set respectively a delayed onset time and a glo
108. tion Your PATH environment variable must be set to the installation directory in order to run Edge For convenience there is a special target path that prints any necessary PATH envi ronment variable changes gmake path A simple first test to verify that Edge has been built ok is to run ffalist on any binary ffa file to view it s content e g ffalist examples rae_euler mesh bmsh The output should be a listing with columns describing type size and name of the datasets contained in this file If you wish you can run Edge with the test cases found in the examples directory See instructions for running Edge in Section 3 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 Running Edge 3 1 Steps to Take to Run Edge To start a computation with Edge all that is necessary is a grid file The format for that grid file is described below The steps to take to make a computation are the following 1 Set up an input file with relevant parameters for your computation 2 Set proper boundary conditions 3 Run the preprocessor 4 Run the flow solver 5 Post processing Most files input and output are written in the FFA format described in Section 3 3 The different files are explained in Section 3 4 3 1 1 Setting Up an Input File An input file with all global parameters is used to control the execution of all programs available in Edge The input file contains necessary parameters and controls not only the flow solver but
109. uffix debug type configure config i686 _intel _s program suffix _s The command to run this installation of edge will then be edge run s instead of edge run FOI dnr 03 2870 Edge User Guide Issue 4 1 0 2 4 2 Compiling Once Edge is configured compiling is easy Simply type gmake and wait for the build process to finish You must use gmake GNU make Other make implementations are unlikely to work If you only want to build a single directory you can run gmake in that directory cd solver basic gmake To only build a single Fortran file filename f90 write gmake filename o To remove all built object files and executable files do gmake clean To rebuild dependency information write gmake depend This requires Python 2 2 4 3 Installing If the compilation finished without errors Edge can be installed gmake install By default the binaries will be installed into the directory specified with the prefix option to the configure script but this value can be overridden by specifying a new prefix value directly to make gmake install prefix opt local To only install the contents of a subdirectory e g solver you may run gmake install from within that directory 2 4 4 Build Java Based GUI optional If you have the necessary prerequisites you may build Xedge the Java GUI to Edge gmake gui all gmake qui install 17 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 18 2 4 5 Testing your Installa
110. uname m is used Note that some configuration files build Edge in single precision others in double precision some compile Edge with debug flags others compile with heavy optimization Use the config file that best suits your needs If there is no suitable file in config for your platform you may need to create a new config file Use one of the existing files in config as template and save the file e g as con fig config custom Then configure with config custom See the comments inside the config files for more hints 2 4 1 3 prefix The configure script also determines where Edge will be installed The prefix option determines this E g to put the Edge binaries in HOME bin type configure prefix HOME config i686 If you give no prefix option the default value is the build directory itself The prefix value may also be overridden later by specifying a prefix directory as an argument to make install see 2 4 3 2 4 1 4 with and without The configure script searches certain standard directories to find optional external libraries For more fine grained control of where to find such libraries use the with and without options Type configure help for more help on this subject 2 4 1 5 program suffix This option adds a suffix to all executable files This may be convenient when installing several versions of Edge to the same location E g to install a single precision compile with program s
111. xample Adjoint bedg This is obtained by running in the usual way the preprocessor on the input file Adjoint ainp where ADJOPT 2 The mesh file must be the same as the one used for the flow calculation CFIMSH FOI dnr 03 2870 Edge User Guide Issue 4 1 0 3 10 2 Parallel Computation The partitionned edge files CFIEDG and flow solution files CFIEUL are required as well as the boundary file CFIBOC The number of partitions NPART in Adjoint ainp must be equal to the number of partitionned files CFIEDG and CFIEUL Option 2 of ADJOPT is now supported for parallel computations NPART gt 1 This means that the output file CFIOUT in Adjoint ainp contains the adjoint solution and the gradient with respect to the nodal coordinates Run the MPI script of Edge as usual for example edge_mpi_run Adjoint ainp 10 on 10 processors This will split the flow solution CFIEUL if it is not already done and merge the adjoint solution files at the end of the calculation 3 11 The Meshdeform Utility The utility program meshdeform deforms the interior mesh given deformations with respect to a reference mesh of nodes on the boundaries The directory examples rae_meshdeform provides a complete example invoked with the command meshdeform meshdeform ainp The command line argument to meshdeform is the name of an input file with extension ainp that provides the file names and options needed for the meshdeform program Three files speci
112. y the dataset that contains the residuals time N 0 0 Grid solution etc at a specific time grid_level N 0 0 Different grid levels multi grid block N 0 0 2 A block structured domain or unstructured free stream data N 0 0 Free stream data Table 37 Directory datasets Dataset Type Dim Size Nsub Explanation boundary data N 0 0 Dataset with boundary data connectivity IB 1 8 Connectivity between two boundaries rot_symmetry IB 1 8 Symmetry from rotation around an axis trans_symmetry IB 1 8 Symmetry from translation mirror IB 1 6 Mirror boundary no flow through pol_sing_linel IB 1 6 Polar singular line a collapsed boundary wall IB 1 6 Wall boundary external IB 1 6 External boundary boundary type SI 1 1 0 Type of boundary given as data given as Dub dataset to the boundary conditions above Table 38 Boundary condition datasets The real field datasets in could also be given in double precision 8 bytes real with type Dxx 85 FOI dnr 03 2870 Edge User Guide Issue 4 1 0 Dataset Type Dim Size Nsub Explanation density RFC 1 0 Density velocity RFC 3 0 Velocity pressure RFC 1 0 Static pressure temperature RFC 1 0 Static temperature energy RFC 1 0 Total energy turb kin energy RFC 1 2 0 Turbulent kinetic energy k turb_dissipation RFC 1 0 Turbulent rate of dissipation e turb time scale RFC 1 2 0 Turbulent time scale t density_species RFC NSP 0 Density for the NSP species vibr
113. y only two points N 0x0 1 VGENDATA N 0x0 1 VGENCONFIG N 0x0 3 VGEN DI 2x3 coord 1 619999785 DI 2x 3 norm 0 603200000 DI 28 1 height 4 0000E 03 In ASCII format the file looks like VGENDATA N 0 0 1 VGENCONFIG N 0 0 1 VGEN N 0 0 3 coord DI 3 2 0 1 61999978085664 norm DI 3 2 0 0 6032 0 5569 0 1478 height DI 1 2 0 4 000000000000000E 03 Note that in FFA format the coordinates are entered as xy 2 n Y Y2 Yn 21 22 en 3 8 Aeroelastics Options and Programs In this section we present the programs used to prepare data for modal aeroelastic compu tations including the associated file formats and describe the basic procedures for coupled and prescribed motion simulations More details about the implementation and validation of the aeroelastic module can be found in Smith 2005 Aeroelastic simulations are controlled from the main Edge input file by a single dataset AEROELASTICS The variables in this dataset are listed in Table 32 The solution option is selected by the integer variable IAEOPTas described in Table 13 It is intended that the set of solution options will be expanded to include non modal struc tural models including coupling to external structural solvers The options series 1xx is reserved for modal solutions In the present version aeroelastic solutions are always carried out as time accurate un steady simulations ITIMAQ 1 in the main input file The form of modal sh

Download Pdf Manuals

Related Search

Related Contents

Mode d`emploi - Roland Central Europe    User Guide for FEBFSL206MRN_H428v1 Evaluation Board  Old Company Name in Catalogs and Other Documents  取扱説明書 - 株式会社ナスタ  Manuale di Installazione 5MKS90E 4MXS80E 5MXS90E  MSAM Transport DMC SW Upgrade Instructions    Tripp Lite Full Motion Universal Tablet and Phone Desk Clamp  

Copyright © All rights reserved. Failed to retrieve file