Setup Guide: Steady RANS for External Aerodynamics

Draft. These settings are drawn from SU2 source, tutorials and regression cases. Recommendations marked (to be reviewed) still need an experienced user’s review.

Use this guide to choose settings and understand their consequences. Follow the linked Tutorials for worked examples and consult V&V for numerical and physical comparisons. Example settings are starting points, not evidence that a case is converged or validated.

The documentation guide explains the source tags used below: (code), (test case), (tutorial), (docs), (advice) and (to be reviewed).

• When to use this guide
• Start from
• Settings that matter
• Check convergence and results
• Common problems
• Related tutorials and test cases

When to use this guide

Use this guide for a steady, compressible, turbulent flow around a body in open air: an airfoil, a wing, a fuselage or a full aircraft. “Steady” means that the flow settles to one state that does not change in time.

Use another guide if:

• the flow does not settle, for example with vortex shedding or large separated regions. See Setup Guide: Unsteady Flows and DES.
• large laminar regions matter, for example at low Reynolds numbers. See Setup Guide: Laminar-Turbulent Transition.

Start from

This config comes from the ONERA M6 wing tutorial: compressible_flow/Turbulent_ONERAM6/turb_ONERAM6.cfg in the Tutorials repository. We removed the adjoint, design and output file name options, and the options that have no effect with these settings. All other values are unchanged. The mesh file is in the same tutorial folder.

% ------------- Problem ------------------------------------------------%
SOLVER= RANS
KIND_TURB_MODEL= SA
MATH_PROBLEM= DIRECT
RESTART_SOL= NO
%
% ------------- Free stream (from Mach and Reynolds number) ------------%
MACH_NUMBER= 0.8395
AOA= 3.06
SIDESLIP_ANGLE= 0.0
INIT_OPTION= REYNOLDS
FREESTREAM_OPTION= TEMPERATURE_FS
FREESTREAM_TEMPERATURE= 288.15
REYNOLDS_NUMBER= 11.72E6
REYNOLDS_LENGTH= 0.64607
REF_DIMENSIONALIZATION= FREESTREAM_VEL_EQ_ONE
%
% ------------- Gas and viscosity --------------------------------------%
FLUID_MODEL= STANDARD_AIR
VISCOSITY_MODEL= SUTHERLAND
MU_REF= 1.716E-5
MU_T_REF= 273.15
SUTHERLAND_CONSTANT= 110.4
CONDUCTIVITY_MODEL= CONSTANT_PRANDTL
PRANDTL_LAM= 0.72
PRANDTL_TURB= 0.90
%
% ------------- Reference values for the coefficients ------------------%
REF_ORIGIN_MOMENT_X= 0.25
REF_ORIGIN_MOMENT_Y= 0.00
REF_ORIGIN_MOMENT_Z= 0.00
REF_LENGTH= 0.64607
REF_AREA= 0
%
% ------------- Boundary conditions ------------------------------------%
MARKER_HEATFLUX= ( WING, 0.0 )
MARKER_FAR= ( FARFIELD )
MARKER_SYM= ( SYMMETRY )
MARKER_PLOTTING= ( WING )
MARKER_MONITORING= ( WING )
%
% ------------- Numerics -----------------------------------------------%
NUM_METHOD_GRAD= GREEN_GAUSS
CFL_NUMBER= 100.0
CFL_ADAPT= NO
ITER= 999999
LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= ILU
LINEAR_SOLVER_ERROR= 1E-10
LINEAR_SOLVER_ITER= 10
MGLEVEL= 0
CONV_NUM_METHOD_FLOW= JST
JST_SENSOR_COEFF= ( 0.5, 0.02 )
TIME_DISCRE_FLOW= EULER_IMPLICIT
CONV_NUM_METHOD_TURB= SCALAR_UPWIND
MUSCL_TURB= NO
TIME_DISCRE_TURB= EULER_IMPLICIT
%
% ------------- Convergence: stop when the drag stops changing ---------%
CONV_FIELD= DRAG
CONV_STARTITER= 10
CONV_CAUCHY_ELEMS= 100
CONV_CAUCHY_EPS= 1E-6
%
% ------------- Input and output ---------------------------------------%
MESH_FILENAME= mesh_ONERAM6_turb_hexa_43008.su2
MESH_FORMAT= SU2
SCREEN_OUTPUT= (INNER_ITER, WALL_TIME, RMS_DENSITY, RMS_NU_TILDE, LIFT, DRAG)

The tutorial page Turbulent ONERA M6 explains the free-stream and convergence options of this config.

Settings that matter

Turbulence model: SA or SST

Set SOLVER= RANS and pick the model with KIND_TURB_MODEL. With SOLVER= RANS, SU2 stops with an error if KIND_TURB_MODEL is NONE (code).

Spalart-Allmaras (SA). SA solves one transport equation, for a modified eddy viscosity.

• With an empty SA_OPTIONS, SU2 runs SA without the ft2 term (“SA-noft2”) (docs, code).
• SA_OPTIONS accepts NEGATIVE, EDWARDS, WITHFT2, QCR2000, COMPRESSIBILITY, ROTATION, BCM and EXPERIMENTAL (code).
• SA-neg needs the ft2 term: write SA_OPTIONS= NEGATIVE, WITHFT2. Without WITHFT2, SU2 stops with an error unless you add EXPERIMENTAL (code). The test cases use both forms (test case).
• The free-stream value of the SA variable is FREESTREAM_NU_FACTOR times the free-stream kinematic viscosity. The default is 3.0 (code, docs).
• Most SU2 external aerodynamics tests use SA: ONERA M6, the SA versions of RAE2822, and 30P30N (test case).

Menter SST (SST). SST solves two transport equations: one for the turbulent kinetic energy k, one for the specific dissipation rate omega.

• With an empty SST_OPTIONS, SU2 runs the 2003 version, “SST-2003m”, as set in ParseSSTOptions in option_structure.hpp (code). Some older pages give V1994m as the default; this is out of date.
• The other SST_OPTIONS values are V1994m, V2003m, SUSTAINING, VORTICITY, KATO-LAUNDER, UQ, COMPRESSIBILITY-WILCOX, COMPRESSIBILITY-SARKAR and DIMENSIONLESS_LIMIT. Write KATO-LAUNDER with a dash (code).
• SST needs free-stream values for k and omega. They come from FREESTREAM_TURBULENCEINTENSITY (a fraction: 0.001 means 0.1 %) and FREESTREAM_TURB2LAMVISCRATIO (eddy viscosity over laminar viscosity). The defaults are 0.05 and 10.0 (code). These are parser defaults, not calibrated inlet conditions. Choose values for the actual flow and check their decay to the body; see the TMR SST boundary conditions (reference, advice).
• External aerodynamics tests use, for example, FREESTREAM_TURBULENCEINTENSITY= 0.001 with FREESTREAM_TURB2LAMVISCRATIO= 100.0 (RAE2822 SST), or 0.001 with 0.1 (NACA0012 SST with SUSTAINING) (test case).
• In SST, k decays between the far field and the body. Two options control this decay. SST_OPTIONS= SUSTAINING supplies a floor to production using the ambient values; SU2 uses a maximum rather than the additive sustaining terms of the TMR definition (code). See the implementation difference. TURB_FIXED_VALUES= YES holds k and omega at the far-field values in an upstream region set by TURB_FIXED_VALUES_DOMAIN (code). The test rans/naca0012/turb_NACA0012_sst_fixedvalues.cfg uses the second one (test case).

Which one to pick.

• SA is robust and widely used for attached flows over wings and airfoils (advice).
• SST is often preferred when adverse pressure gradients and separation matter (advice, to be reviewed).
• Differences between SA and SST are evidence of model sensitivity, not a quantified uncertainty bound. First check iterative and mesh convergence, then examine the regions where the models differ (advice).

See Turbulence Models for the full option list.

Convective scheme: JST or an upwind scheme

The convective scheme computes the flux of mass, momentum and energy through each cell face. Set it with CONV_NUM_METHOD_FLOW. It has no default, so always set it (code).

JST (Jameson-Schmidt-Turkel) is a central scheme with added artificial dissipation.

• The two dissipation coefficients are JST_SENSOR_COEFF= (2nd order, 4th order). The default is ( 0.5, 0.02 ) (code, docs).
• JST is robust and allows high CFL numbers. It over-predicts viscous drag on coarse meshes or at low Reynolds numbers (docs).
• JST_MAT uses matrix dissipation and gives better viscous drag on such meshes (docs).
• The ONERA M6 tutorial and the RAE2822 tests use JST (tutorial, test case).

Upwind schemes use the direction of the waves in the flow. ROE is the most used (test case). The others are L2ROE and LMROE (low Mach number variants), AUSMPLUSUP, AUSMPLUSUP2, SLAU, SLAU2, HLLC and MSW (docs, code).

• An upwind scheme is first order unless you turn on MUSCL reconstruction, see the next section (docs).
• ENTROPY_FIX_COEFF sets a lower bound on the Roe dissipation. The default is 0.001 (code). The 30P30N validation case uses 1e-5 (test case).
• The tests rans/oneram6/turb_ONERAM6.cfg and rans/naca0012/turb_NACA0012_sst.cfg use ROE (test case).

Turbulence equations. Use CONV_NUM_METHOD_TURB= SCALAR_UPWIND, as most external aerodynamics tests do (test case). The code also accepts BOUNDED_SCALAR; the tests turb_SA_RAE2822.cfg and turb_SST_RAE2822.cfg use it (code, test case). MUSCL_TURB defaults to NO, so the turbulence convection is first order unless you change it (code). Most tests keep MUSCL_TURB= NO (test case).

See Convective Schemes.

MUSCL reconstruction and slope limiters

MUSCL reconstruction makes an upwind scheme second order. It uses the solution gradient to compute the values on each side of a face. A slope limiter reduces this gradient near shocks and steep changes, so that the solution does not oscillate.

• MUSCL_FLOW defaults to YES and SLOPE_LIMITER_FLOW defaults to VENKATAKRISHNAN (code). MUSCL has no effect on central schemes such as JST (docs).
• For transonic flows with shocks, keep a limiter. The ONERA M6 test uses VENKATAKRISHNAN with VENKAT_LIMITER_COEFF= 1.0 (test case).
• For subsonic flows without shocks, you can turn the limiter off. The NACA0012 SST test at Mach 0.15 uses SLOPE_LIMITER_FLOW= NONE and MUSCL_KAPPA_FLOW= 0.5 (test case).
• VENKAT_LIMITER_COEFF defaults to 0.05. Larger values limit less (code). For VENKATAKRISHNAN, its effect depends on the cell size in mesh units. So the right value changes with the mesh and with the non-dimensional scaling (docs).
• VAN_ALBADA_EDGE is another option. The 30P30N validation case uses it (test case).
• LIMITER_ITER freezes the limiter after this many iterations. The default is 999999 (code). Freezing can help when the residuals stall because the limiter keeps switching (advice, to be reviewed).
• NUM_METHOD_GRAD sets the gradient method (default WEIGHTED_LEAST_SQUARES). NUM_METHOD_GRAD_RECON can set a different method for MUSCL only. By default MUSCL uses NUM_METHOD_GRAD (code).

See Slope Limiters and Shock Resolution.

CFL number and adaptive CFL

SU2 marches a steady case in pseudo-time. The CFL number sets the local pseudo-time step. A higher CFL can reduce the iteration count, but can also make the linear solve harder or destabilize the nonlinear iteration.

• The default CFL_NUMBER is 1.25 (code). Implicit RANS runs use much higher values: 100 in the ONERA M6 tutorial, 1000 in the RAE2822 SA and NACA0012 SST tests (tutorial, test case).
• CFL_ADAPT= YES lets SU2 change the CFL during the run. CFL_NUMBER is then the starting value (code).
• CFL_ADAPT_PARAM has up to six values: factor down, factor up, minimum CFL, maximum CFL, acceptable linear solver residual, start iteration. Missing values take the defaults ( 0.1, 1.2, 10.0, 100.0, 0.001, 0 ) (code).
• The comments in config_template.cfg explain the rules. The CFL goes up by the “factor up” while the solution update is not limited and the linear solver converges well enough. It goes down when the update is limited, when the linear solver does not converge enough, or when the residuals stall or oscillate. It returns to the minimum if the linear solver diverges or the residuals grow too much (code).
• Examples: the E387 transition tutorial uses CFL_NUMBER= 1 with CFL_ADAPT_PARAM= ( 0.1, 2.0, 40.0, 1e10, 0.001 ). The 30P30N case uses CFL_ADAPT_PARAM= ( 1, 1.05, 10, 400 ) (tutorial, test case).
• CFL_REDUCTION_TURB scales the CFL of the turbulence equations. The default is 1.0 (code). The RAE2822 SST test uses 0.5 (test case).
• Add AVG_CFL to SCREEN_OUTPUT to see the CFL during the run (test case).

Multigrid

Multigrid also solves the equations on coarser versions of the mesh. This removes large-scale errors faster.

• MGLEVEL sets the number of coarse levels. The default is 0 (off) (code).
• MGCYCLE is V_CYCLE (default), W_CYCLE or FULLMG_CYCLE (code).
• The RAE2822 tests use MGLEVEL= 3 with JST, MGCYCLE= V_CYCLE, MG_DAMP_RESTRICTION= 0.75 and MG_DAMP_PROLONGATION= 0.75 (test case). The code default for both damping factors is 0.5 (code).
• With a central scheme, the coarse levels always use the Lax-Friedrich scheme. MUSCL has no effect on coarse levels (docs).
• If the run is unstable with multigrid, reduce MGLEVEL or turn multigrid off (advice).

Linear solver and preconditioner

Each implicit iteration solves a linear system. You do not need to solve it exactly: only the final nonlinear residual matters for accuracy (docs).

The defaults are LINEAR_SOLVER= FGMRES, LINEAR_SOLVER_PREC= ILU, LINEAR_SOLVER_ERROR= 1E-6 and LINEAR_SOLVER_ITER= 10 (code).

The Linear Solvers and Preconditioners page gives this advice (docs):

• Use the loosest tolerance that keeps the run stable, usually between 0.05 and 0.001.
• About 10 linear iterations are enough for upwind schemes. Central schemes may need up to 50.
• Use ILU at high CFL. LU_SGS can be faster at low CFL because it has no setup cost.
• If memory is a problem, use RESTARTED_FGMRES or BCGSTAB.

From the test cases (test case):

• Add LINSOL_ITER and LINSOL_RESIDUAL to SCREEN_OUTPUT to see how well the linear solver converges.
• A Newton-Krylov method is also available with NEWTON_KRYLOV= YES. The 30P30N validation case shows its parameters.

Convergence criteria

CONV_FIELD names the output fields that decide when the run stops. For the compressible solver the default is RMS_DENSITY (code). If you list several fields, all of them must meet their criterion (code).

Residual fields (such as RMS_DENSITY): the run stops when the log10 of the residual falls below CONV_RESIDUAL_MINVAL. The default is -14 (code).

Coefficient fields (such as LIFT or DRAG) use a Cauchy criterion. This criterion looks at how much the coefficient changes from one iteration to the next (code):

• At each iteration, SU2 computes the change of the coefficient divided by the coefficient. If the coefficient is smaller than 0.1 in size, SU2 divides by 0.1 instead.
• SU2 averages this value over the last CONV_CAUCHY_ELEMS iterations (default 100).
• The run stops when the average is below CONV_CAUCHY_EPS (default 1E-10).
• CONV_STARTITER (default 5) sets the first iteration where SU2 checks the criteria.

The ONERA M6 tutorial stops on CONV_FIELD= DRAG with CONV_CAUCHY_EPS= 1E-6 (tutorial). A coefficient can look converged while the residuals are still high, so check both (advice).

ITER sets the maximum number of iterations. The default is 1000 (code).

See Solver Setup.

Fixed lift (fixed CL)

In fixed-CL mode, SU2 changes the angle of attack until the lift coefficient reaches a target. These rules come from the code (code):

• FIXED_CL_MODE= YES turns it on. TARGET_CL is the target (default 0.0). AOA is the starting angle.
• DCL_DALPHA is your estimate of the lift slope, per degree. The default is 0.2, also in config_template.cfg. SU2 changes the angle by (target CL − current CL) / DCL_DALPHA.
• SU2 updates the angle when the convergence criteria are met but CL is still away from the target. It also updates it at the latest every UPDATE_AOA_ITER_LIMIT iterations (default 200).
• The lift counts as reached when it is within CONV_CAUCHY_EPS/2 of the target. So CONV_CAUCHY_EPS sets both the convergence tolerance and the lift tolerance.
• At the end, SU2 runs ITER_DCL_DALPHA more iterations (default 500) to compute dCL/dalpha by finite differences. These are added to ITER. Set ITER_DCL_DALPHA= 0 to skip this step.
• Fixed-CL mode is for compressible solvers only.
• Add the AOA history field to see the angle during the run.

The RAE2822 design tutorial runs SA in fixed-CL mode with TARGET_CL= 0.724, DCL_DALPHA= 0.2, UPDATE_AOA_ITER_LIMIT= 100, ITER_DCL_DALPHA= 500 and CONV_CAUCHY_EPS= 1E-6 (tutorial). The fixed-CL test case fixed_cl/naca0012/inv_NACA0012.cfg is inviscid (test case).

Free stream from Mach and Reynolds number

With INIT_OPTION= REYNOLDS (the default), SU2 builds the free stream from the Mach number and the Reynolds number (code, tutorial):

1. It computes the speed of sound from FREESTREAM_TEMPERATURE and the gas constants.
2. It computes the velocity from MACH_NUMBER, AOA and SIDESLIP_ANGLE.
3. It computes the viscosity with the viscosity model.
4. It computes the density from REYNOLDS_NUMBER and REYNOLDS_LENGTH.
5. It computes the pressure from the gas law. FREESTREAM_PRESSURE is ignored.

Things to check:

• With SYSTEM_MEASUREMENTS= SI (the default), give mesh lengths in metres. REYNOLDS_LENGTH is the length (in mesh units) that the Reynolds number refers to, for example the chord (tutorial).
• With INIT_OPTION= TD_CONDITIONS, SU2 uses FREESTREAM_PRESSURE and FREESTREAM_TEMPERATURE instead, and ignores the Reynolds number (tutorial).
• SU2 prints a table of free-stream values before the first iteration. Check it (tutorial).
• REF_DIMENSIONALIZATION sets the scaling of the equations. The default is DIMENSIONAL (code). The tests also use FREESTREAM_VEL_EQ_ONE, FREESTREAM_VEL_EQ_MACH and FREESTREAM_PRESS_EQ_ONE (tutorial, test case).

Reference values for the coefficients (code). REF_LENGTH is used for moments. REF_AREA is used for all force coefficients. With REF_AREA= 0, SU2 computes the area as the projection of the body on the y plane (2D) or the z plane (3D). REF_ORIGIN_MOMENT_X, _Y and _Z set the moment reference point.

See Physical Definition.

Far-field and wall boundary conditions

• MARKER_FAR sets a far-field boundary. SU2 imposes the free-stream state there with a characteristic-based condition (docs, tutorial).
• Place the far-field boundary far from the body, often 50 to 100 chords for airfoils, so that it does not change the lift (advice, to be reviewed).
• MARKER_HEATFLUX= ( wall, 0.0 ) is an adiabatic no-slip wall. All external aerodynamics tests use it (test case). MARKER_ISOTHERMAL sets a wall temperature instead (docs).
• MARKER_SYM is a symmetry plane, for example at the root of a half wing (tutorial).
• MARKER_MONITORING lists the surfaces whose forces give LIFT, DRAG and the other coefficients (code). MARKER_PLOTTING lists the surfaces written to the surface output files (tutorial).

See Markers and Boundary Conditions.

Near-wall mesh, y+ and wall functions

y+ is the distance of the first point from the wall, in wall units. It tells you if the mesh resolves the thin layer next to the wall.

• Without wall functions, the first point must be in the viscous sublayer: y+ below 5 (docs). y+ about 1 is the usual target (advice).
• With wall functions, SU2 models the near-wall layer, so the first point can be further from the wall. Turn them on per marker: MARKER_WALL_FUNCTIONS= ( wall, STANDARD_WALL_FUNCTION ). They work with SA and SST, compressible and incompressible (docs).
• For RANS, only NO_WALL_FUNCTION and STANDARD_WALL_FUNCTION are tested. ADAPTIVE_WALL_FUNCTION, SCALABLE_WALL_FUNCTION and NONEQUILIBRIUM_WALL_MODEL stop SU2 with an error (code, test case).
• The wall function is active only where y+ is above WALLMODEL_MINYPLUS (default 5.0) (code, docs). The activation threshold is not a mesh-accuracy target. See Wall Treatment for log-layer placement and the current restrictions on mixed wall treatments.
• Wall functions and WALL_ROUGHNESS cannot be used together (code).
• Wall functions assume an attached boundary layer in equilibrium. They are less accurate near separation and strong pressure gradients (advice).
• Check the Y_PLUS output field on the wall. It is in the default PRIMITIVE output group (code).

See Markers and Boundary Conditions and the test cases in TestCases/wallfunctions/flatplate/.

Check convergence and results

• Residuals. Plot RMS_DENSITY and the turbulence residual (RMS_NU_TILDE for SA, RMS_TKE and RMS_DISSIPATION for SST). A drop of several orders of magnitude is a good sign (advice).
• Coefficients. LIFT, DRAG and the moments must be flat at the end of the run. Add the CAUCHY group to HISTORY_OUTPUT to see the Cauchy values (code).
• Linear solver and CFL. LINSOL_RESIDUAL is log10 of the final linear residual: compare it with log10(LINEAR_SOLVER_ERROR), for example −6 for a tolerance of 1e-6 (code). With CFL_ADAPT= YES, AVG_CFL should grow and stay high (advice).
• Free stream. Read the free-stream table that SU2 prints at the start: Reynolds number, density, pressure, velocity (tutorial).
• y+. Look at Y_PLUS on the walls (code).
• Surface data. Put SURFACE_PARAVIEW or SURFACE_CSV in OUTPUT_FILES, and plot PRESSURE_COEFF and SKIN_FRICTION-X along the wall (code, tutorial).
• Force breakdown. WRT_FORCES_BREAKDOWN= YES writes a file that splits the forces into pressure and viscous parts, per surface (code, tutorial).
• Mesh. Run at least two, better three, meshes of different size. Check that the coefficients change less and less (advice).

See History and Solution Output and the Output Fields Reference.

Common problems

Related tutorials and test cases

Tutorials:

• Turbulent ONERA M6: 3D transonic wing, SA, JST, Cauchy criterion on drag.
• Turbulent Flat Plate: zero pressure gradient flat plate with SA.
• Constrained shape design of a transonic turbulent airfoil: RAE2822 with SA in fixed-CL mode, used for shape design.
• UQ NACA0012: SST with the UQ option.

Test cases (develop branch):

• rans/oneram6/turb_ONERAM6.cfg: ONERA M6 with ROE, MUSCL and VENKATAKRISHNAN.
• rans/rae2822/turb_SA_RAE2822.cfg and turb_SST_RAE2822.cfg: transonic airfoil with JST and multigrid.
• rans/naca0012/turb_NACA0012_sst.cfg: subsonic airfoil, SST, ROE without limiter.
• rans/naca0012/turb_NACA0012_sst_fixedvalues.cfg: TURB_FIXED_VALUES.
• turbulence_models/sa/rae2822/: SA variants (NEGATIVE, EDWARDS, QCR2000, COMPRESSIBILITY).
• vandv/rans/30p30n/config.cfg: high-lift airfoil, ROE with VAN_ALBADA_EDGE, adaptive CFL, Newton-Krylov.
• vandv/rans/swbli/config_sst.cfg: Mach 5 shock-boundary layer interaction with SST.
• fixed_cl/naca0012/inv_NACA0012.cfg: fixed-CL mode (inviscid).
• wallfunctions/flatplate/compressible_SA/turb_SA_flatplate.cfg: wall functions with SA.

User guide pages:

• Physical Definition, Convective Schemes, Slope Limiters and Shock Resolution, Linear Solvers and Preconditioners, Markers and Boundary Conditions, Solver Setup, Governing Equations, History and Solution Output.
• All options: Option Reference. All output fields: Output Fields Reference.

Source review: 22 October 2026, against SU2 docs_site at 41ea6a219a. See the review scope and outstanding validation. Setup recommendations remain draft where indicated.