Cf considerably higher when using the compressible solver

kboychev · July 5, 2024, 09:41

Hi everyone, this is my first post in this forum. I'm relatively new to SU2, but I have a background in CFD using in-house (academic) solvers. My background has been mostly in high-speed flows, but currently, I am interested in simulating low-speed flows.

__________________________________________________

Test case

A NACA4412 finite wing with a chord of 1 m and a span of 6 m (AR=6). The Mach number is 0.09 and the Reynolds number is 1.52E6. The wing has a blunt trailing edge (1% of the chord length).

__________________________________________________
Issue

High skin-friction coefficients are predicted by the compressible solver on the same mesh compared to the incompressible solver. Does anyone observe similar discrepancies or am I missing something? The predicted pressure coefficient agrees well between the solvers and the experiment (see Results).
__________________________________________________
What I've tried

- I've tried adjusting the Venkatakrishnan limiter coefficient value (increased to 0.2). This resulted in a less overprediction of Cf, but the drag was still overpredicted compared to the incompressible solution.
- I've tried the Van-Albada limiter for the flow. For turbulence, MUSCL was set to NO. This resulted in a Cf and drag prediction comparable to the default Venkatakrishnan limiter coefficient (0.05). Drag is still overpredicted.
- Y+ in both solutions (compressible/incompressible) is < 1.
__________________________________________________

Domain and boundary conditions

- FARFIELD
- SYMMETRY
- HEATFLUX

domain.png
__________________________________________________

Setup

Compressible (Roe, MUSCL=YES, Venkatakrishnan=0.05, Entropy Fix=1e-5)

Code:

   
SOLVER= RANS
KIND_TURB_MODEL= SST
SST_OPTIONS= V2003m
MATH_PROBLEM= DIRECT
RESTART_SOL= NO

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

MACH_NUMBER= 0.09
AOA= 0
SIDESLIP_ANGLE= 0.0
INIT_OPTION= REYNOLDS
FREESTREAM_OPTION= TEMPERATURE_FS
FREESTREAM_TEMPERATURE= 297.222
REYNOLDS_NUMBER= 1.52E6
REYNOLDS_LENGTH= 1.0

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

FLUID_MODEL= STANDARD_AIR
GAMMA_VALUE= 1.4
GAS_CONSTANT= 287.058

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

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

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

REF_ORIGIN_MOMENT_X = 0.25
REF_ORIGIN_MOMENT_Y = 0.00
REF_ORIGIN_MOMENT_Z = 0.00
REF_LENGTH= 1.0
REF_AREA= 3.0
REF_DIMENSIONALIZATION= FREESTREAM_VEL_EQ_ONE

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

MARKER_HEATFLUX= ( tri_WING, 0.0, quad_WING, 0.0 )
MARKER_FAR= ( FARFIELD)
MARKER_PLOTTING= ( tri_WING, quad_WING)
MARKER_MONITORING= ( tri_WING, quad_WING)
MARKER_SYM= ( tri_SYMMETRY, quad_SYMMETRY)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

NUM_METHOD_GRAD= GREEN_GAUSS
CFL_NUMBER= 5.0
CFL_ADAPT= NO
CFL_ADAPT_PARAM= ( 0.5, 1.5, 5.0, 25.0 )
ITER=10000

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= ILU
LINEAR_SOLVER_ERROR= 1E-10
LINEAR_SOLVER_ITER= 10

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_NUM_METHOD_FLOW= ROE
ENTROPY_FIX_COEFF=1E-5
MUSCL_FLOW= YES
SLOPE_LIMITER_FLOW= VENKATAKRISHNAN
VENKAT_LIMITER_COEFF= 0.05
TIME_DISCRE_FLOW= EULER_IMPLICIT

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_NUM_METHOD_TURB= SCALAR_UPWIND
MUSCL_TURB= YES
SLOPE_LIMITER_TURB= VENKATAKRISHNAN
TIME_DISCRE_TURB= EULER_IMPLICIT

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_FIELD= LIFT
CONV_STARTITER= 10
CONV_CAUCHY_ELEMS= 100
CONV_CAUCHY_EPS= 1E-6

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

MESH_FILENAME= naca4412_wing.cgns
MESH_FORMAT= CGNS
SOLUTION_FILENAME= restart_flow.dat
TABULAR_FORMAT= CSV
CONV_FILENAME= history
RESTART_FILENAME= restart_flow.dat
VOLUME_FILENAME= flow
SURFACE_FILENAME= surface_flow
OUTPUT_WRT_FREQ= 250
SCREEN_OUTPUT= (INNER_ITER, WALL_TIME, RMS_DENSITY, RMS_NU_TILDE, LIFT, DRAG)
HISTORY_OUTPUT= (ITER, RMS_RES, AERO_COEFF)
OUTPUT_FILES = (RESTART, PARAVIEW_ASCII, SURFACE_PARAVIEW_ASCII)

Incompressible (FDS, MUSCL=YES, Venkatakrishnan=0.05)

Code:

   
SOLVER= INC_RANS
KIND_TURB_MODEL= SST
SST_OPTIONS= V2003m
MATH_PROBLEM= DIRECT
RESTART_SOL= NO

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

INC_DENSITY_INIT=0.8956
INC_VELOCITY_INIT=(31.1051, 0.0, 0.0)
INC_NONDIM=INITIAL_VALUES
INC_DENSITY_REF=1.0
INC_VELOCITY_REF=1.0
INC_TEMPERATURE_REF=1.0

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

VISCOSITY_MODEL= CONSTANT_VISCOSITY
MU_CONSTANT=1.8327E-5

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

REF_ORIGIN_MOMENT_X = 0.25
REF_ORIGIN_MOMENT_Y = 0.00
REF_ORIGIN_MOMENT_Z = 0.00
REF_LENGTH= 1.0
REF_AREA= 3.0

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

MARKER_HEATFLUX= ( tri_WING, 0.0, quad_WING, 0.0 )
MARKER_FAR= ( FARFIELD)
MARKER_PLOTTING= ( tri_WING, quad_WING)
MARKER_MONITORING= ( tri_WING, quad_WING)
MARKER_SYM= ( tri_SYMMETRY, quad_SYMMETRY)

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

NUM_METHOD_GRAD= GREEN_GAUSS
CFL_NUMBER= 5.0
CFL_ADAPT= NO
CFL_ADAPT_PARAM= ( 0.5, 1.5, 5.0, 25.0 )
ITER=10000

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

LINEAR_SOLVER= FGMRES
LINEAR_SOLVER_PREC= ILU
LINEAR_SOLVER_ERROR= 1E-10
LINEAR_SOLVER_ITER= 10

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_NUM_METHOD_FLOW= FDS
MUSCL_FLOW= YES
SLOPE_LIMITER_FLOW= VENKATAKRISHNAN
VENKAT_LIMITER_COEFF= 0.05
TIME_DISCRE_FLOW= EULER_IMPLICIT

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_NUM_METHOD_TURB= SCALAR_UPWIND
MUSCL_TURB= YES
SLOPE_LIMITER_TURB= VENKATAKRISHNAN
TIME_DISCRE_TURB= EULER_IMPLICIT

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CONV_FIELD= LIFT
CONV_STARTITER= 10
CONV_CAUCHY_ELEMS= 100
CONV_CAUCHY_EPS= 1E-6

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

MESH_FILENAME= naca4412_wing.cgns
MESH_FORMAT= CGNS
SOLUTION_FILENAME= restart_flow.dat
TABULAR_FORMAT= CSV
CONV_FILENAME= history
RESTART_FILENAME= restart_flow.dat
VOLUME_FILENAME= flow
SURFACE_FILENAME= surface_flow
OUTPUT_WRT_FREQ= 250
SCREEN_OUTPUT= (INNER_ITER, WALL_TIME, RMS_DENSITY, RMS_NU_TILDE, LIFT, DRAG)
HISTORY_OUTPUT= (ITER, RMS_RES, AERO_COEFF)
OUTPUT_FILES = (RESTART, PARAVIEW_ASCII, SURFACE_PARAVIEW_ASCII)

__________________________________________________

Results

Cp_comparison.png

CLCD_history.png

comparison_Cf.jpg

bigfootedrockmidget · July 7, 2024, 05:01

What is the difference in CL and CD for both cases? It looks very similar. Do you have some values, and the corresponding value from literature?

Is this the case from the nasa larc website?
https://turbmodels.larc.nasa.gov/naca4412sep_val.html
The nasa validation was done for high angle of attack, and you simulate it for AoA=0, right? It would help here if the expected CL and CD are given. It would then be possible to say if all nondimensionalization was done correctly.

There are also some pressure peaks at the trailing edge, but this might be due to the mesh quality there, the mesh needs some special attention in that region.

kboychev · July 13, 2024, 10:19

Thank you for the prompt reply! The purpose of these simulations was for me to benchmark the compressible solver for low-Mach problems. More specifically - if the compressible solver can reproduce the solution from the incompressible solver. I decided to use the freestream conditions from the NACA 4412 TMR case. This case, as you noted, is 2D and is at an AOA of 10 deg. Using the freestream conditions from the NACA4412 TMR case was an arbitrary decision (I am interested in comparing the solvers). Before performing the 3D simulations, I have done the 2D simulations for the NACA4412 and NACA0012 2D TMR cases.

The 3D simulations of the NACA4412 wing result in the following CL and CD:

CL: 0.3275 (compressible)
CD: 0.0341 (compressible)
CL: 0.3145 (incompressible)
CD: 0.0206 (incompressible)

The differences in the CL are reasonable, but the difference in CD is substantial. The peaks at the trailing edge are due to the "bluntness" of the trailing edge. Most SU2 cases use sharp trailing edges, but for complex 3D geometries sharp trailing edges usually result in poor mesh quality. Therefore the simulations above employ a blunt trailing edge. I have 3 cells across (coarse mesh). The location of the blunt trailing edge is at 99% of the chord.

PS: I have noted that when the Reynolds number is lowered the results for CD from the incompressible and compressible solver begin to grow apart. I've done 2D simulations of the NACA0012 TMR case at 6e6, both solvers agree on the CD although the compressible solver predicts a slightly higher CD. I've lowered the Re to 6e5 and repeated the simulations. The differences in CD between the solvers increased to 35%.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
foam-extend-4.1 release	hjasak	OpenFOAM Announcements from Other Sources	19	July 16, 2021 06:02
Fail to converge when solving with a fabricated solution	zizhou	FLUENT	0	March 22, 2021 07:33
Blow of compressible solver while using K-epsilon model in openfoam	Amit Mathur	OpenFOAM	16	October 6, 2013 12:09
Creating New Solver: For particle-laden compressible jets	sankarv	OpenFOAM	0	April 4, 2010 19:06
Liquid instead of gas in a compressible solver !?	iggor	OpenFOAM	15	October 14, 2009 12:58

July 7, 2024, 05:01		#2
bigfootedrockmidget Senior Member bigfoot Join Date: Dec 2011 Location: Netherlands Posts: 676 Rep Power: 21	What is the difference in CL and CD for both cases? It looks very similar. Do you have some values, and the corresponding value from literature? Is this the case from the nasa larc website? https://turbmodels.larc.nasa.gov/naca4412sep_val.html The nasa validation was done for high angle of attack, and you simulate it for AoA=0, right? It would help here if the expected CL and CD are given. It would then be possible to say if all nondimensionalization was done correctly. There are also some pressure peaks at the trailing edge, but this might be due to the mesh quality there, the mesh needs some special attention in that region.

July 13, 2024, 10:19		#3
kboychev New Member Join Date: Jul 2024 Posts: 2 Rep Power: 0	Thank you for the prompt reply! The purpose of these simulations was for me to benchmark the compressible solver for low-Mach problems. More specifically - if the compressible solver can reproduce the solution from the incompressible solver. I decided to use the freestream conditions from the NACA 4412 TMR case. This case, as you noted, is 2D and is at an AOA of 10 deg. Using the freestream conditions from the NACA4412 TMR case was an arbitrary decision (I am interested in comparing the solvers). Before performing the 3D simulations, I have done the 2D simulations for the NACA4412 and NACA0012 2D TMR cases. The 3D simulations of the NACA4412 wing result in the following CL and CD: CL: 0.3275 (compressible) CD: 0.0341 (compressible) CL: 0.3145 (incompressible) CD: 0.0206 (incompressible) The differences in the CL are reasonable, but the difference in CD is substantial. The peaks at the trailing edge are due to the "bluntness" of the trailing edge. Most SU2 cases use sharp trailing edges, but for complex 3D geometries sharp trailing edges usually result in poor mesh quality. Therefore the simulations above employ a blunt trailing edge. I have 3 cells across (coarse mesh). The location of the blunt trailing edge is at 99% of the chord. PS: I have noted that when the Reynolds number is lowered the results for CD from the incompressible and compressible solver begin to grow apart. I've done 2D simulations of the NACA0012 TMR case at 6e6, both solvers agree on the CD although the compressible solver predicts a slightly higher CD. I've lowered the Re to 6e5 and repeated the simulations. The differences in CD between the solvers increased to 35%.