|
[Sponsors] |
January 7, 2023, 22:15 |
Struggling with convergence
|
#1 |
Member
Anders Aamodt Resell
Join Date: Dec 2021
Location: Oslo, Norway
Posts: 64
Rep Power: 4 |
I have tried becoming familiar with SU2 lately, but I am struggling a lot to get convergent results in general.
I tried setting up a very simple 2D cylinder flow case, and I would be happy to get some input as to why it's not converging. If I'm unable to run simple case like this successfully don't see how I will be ever able to run more complicated cases My impression is that these kinds of opensource codes are much harder to converge than for instance ansys, and I wonder why that is the case. The triangular mesh is generated in salome, and converted to .su2 format with a python script. With the current settings, density rms reached a minimum value of about -1.5. Navier stokes equations are being solved and farfield boundaries are used. I added a picture of the mesh and the mesh quality statistics. Config options are pasted below: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Autogenerated config file %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Physical governing equations (EULER, NAVIER_STOKES,WAVE_EQUATION, HEAT_EQUATION, FEM_ELASTICITY,POISSON_EQUATION) SOLVER= NAVIER_STOKES KIND_TURB_MODEL= NONE MATH_PROBLEM= DIRECT RESTART_SOL= NO % -------------------- COMPRESSIBLE FREE-STREAM DEFINITION -------------------- MACH_NUMBER= 2 AOA= 0.0 SIDESLIP_ANGLE= 0.0 INIT_OPTION= TD_CONDITIONS FREESTREAM_OPTION= TEMPERATURE_FS FREESTREAM_TEMPERATURE= 288.15 FREESTREAM_PRESSURE= 101325.0 % ---- IDEAL GAS, POLYTROPIC, VAN DER WAALS AND PENG ROBINSON CONSTANTS ------- FLUID_MODEL= STANDARD_AIR % --------------------------- VISCOSITY MODEL --------------------------------- VISCOSITY_MODEL= SUTHERLAND % --------------------------- THERMAL CONDUCTIVITY MODEL ---------------------- CONDUCTIVITY_MODEL= CONSTANT_PRANDTL % ---------------------- REFERENCE VALUE DEFINITION --------------------------- REF_ORIGIN_MOMENT_X= 0.25 REF_ORIGIN_MOMENT_Y= 0.0 REF_ORIGIN_MOMENT_Z= 0.0 REF_LENGTH= 1.0 REF_AREA= 1.0 REF_DIMENSIONALIZATION= DIMENSIONAL % -------------------- BOUNDARY CONDITION DEFINITION -------------------------- MARKER_HEATFLUX= (projectile,0.0) MARKER_EULER= NONE MARKER_FAR= (farfield,outlet) INLET_TYPE= MASS_FLOW MARKER_INLET= NONE MARKER_OUTLET= NONE MARKER_SUPERSONIC_INLET= NONE MARKER_SUPERSONIC_OUTLET= NONE MARKER_SYM= NONE MARKER_PLOTTING= projectile MARKER_MONITORING= projectile % ------------- COMMON PARAMETERS DEFINING THE NUMERICAL METHOD --------------- NUM_METHOD_GRAD= GREEN_GAUSS CFL_NUMBER= 20 CFL_ADAPT= NO INNER_ITER= 1000 % ------------------------ LINEAR SOLVER DEFINITION --------------------------- LINEAR_SOLVER= FGMRES LINEAR_SOLVER_PREC= ILU LINEAR_SOLVER_ERROR= 1e-10 LINEAR_SOLVER_ITER= 10 % -------------------------- MULTIGRID PARAMETERS ----------------------------- MGLEVEL= 0 % -------------------- FLOW NUMERICAL METHOD DEFINITION ----------------------- CONV_NUM_METHOD_FLOW= ROE ENTROPY_FIX_COEFF= 0.01 JST_SENSOR_COEFF= ( 0.5, 0.02 ) MUSCL_FLOW= YES SLOPE_LIMITER_FLOW= VENKATAKRISHNAN VENKAT_LIMITER_COEFF= 0.05 TIME_DISCRE_FLOW= EULER_IMPLICIT USE_ACCURATE_FLUX_JACOBIANS= YES % -------------------- TURBULENT NUMERICAL METHOD DEFINITION ------------------ CONV_NUM_METHOD_TURB= SCALAR_UPWIND MUSCL_TURB= NO SLOPE_LIMITER_TURB= VENKATAKRISHNAN TIME_DISCRE_TURB= EULER_IMPLICIT % --------------------------- CONVERGENCE PARAMETERS -------------------------- CONV_FIELD= RMS_DENSITY CONV_RESIDUAL_MINVAL= -5 CONV_STARTITER= 10 % ------------------------- INPUT/OUTPUT INFORMATION -------------------------- MESH_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/2Dcyl.su2 MESH_FORMAT= SU2 MESH_OUT_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/configuration_0/flow_output/mesh_out.su2 SOLUTION_FILENAME= solution_flow.dat TABULAR_FORMAT= CSV CONV_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/configuration_0/flow_output/history RESTART_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/configuration_0/flow_output/restart_flow.dat VOLUME_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/configuration_0/flow_output/flow SURFACE_FILENAME= /home/anders/dev/CFD-Traject/Tests/2Dcylinder/configuration_0/flow_output/surface_flow SCREEN_OUTPUT= ( INNER_ITER, WALL_TIME, REL_RMS_DENSITY, RMS_DENSITY, LIFT, DRAG ) HISTORY_OUTPUT= ( INNER_ITER, REL_RMS_DENSITY, RMS_DENSITY, LIFT, DRAG ) |
|
January 7, 2023, 22:18 |
attachments
|
#2 |
Member
Anders Aamodt Resell
Join Date: Dec 2021
Location: Oslo, Norway
Posts: 64
Rep Power: 4 |
attachments
|
|
January 8, 2023, 07:45 |
|
#3 |
Senior Member
Pedro Gomes
Join Date: Dec 2017
Posts: 465
Rep Power: 13 |
What is the diameter? Is this supposed to be laminar? And if so, is it supposed to be steady state?
|
|
January 8, 2023, 08:08 |
|
#4 |
Member
Anders Aamodt Resell
Join Date: Dec 2021
Location: Oslo, Norway
Posts: 64
Rep Power: 4 |
The diameter is 1m, but the point was not really to try and model viscous effects. I was looking for a steady solution. I changed the solver to EULER to simplify things and eliminate the potential for vortex shredding, but still, the solution doesn't converge. It stops converging at density rms = -0.35.
|
|
January 9, 2023, 09:11 |
|
#5 |
Senior Member
Pedro Gomes
Join Date: Dec 2017
Posts: 465
Rep Power: 13 |
Inviscid flow is not potential flow, it will still separate and become unsteady, just in an unphysical way.
If you want any steady solution around a cylinder you need conditions and models that give you steady state. |
|
January 9, 2023, 19:01 |
|
#6 |
Member
Anders Aamodt Resell
Join Date: Dec 2021
Location: Oslo, Norway
Posts: 64
Rep Power: 4 |
Ahh, that was clarifying to hear, thanks! I tried the pseudo time stepping of 1st order and it gave me good convergence for the navier stokes equations, however the Euler equations still flatline at density rms = -0.75. Not really sure what I can do to make the latter case work though. Anyhow I will probably use the time domain stepping as my default setup from now since it seems more robust, but please correct me if that's a dumb idea.
Here is my cfg file for the Euler problem as reference: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Autogenerated config file %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % Physical governing equations (EULER, NAVIER_STOKES,WAVE_EQUATION, HEAT_EQUATION, FEM_ELASTICITY,POISSON_EQUATION) SOLVER= EULER KIND_TURB_MODEL= SA MATH_PROBLEM= DIRECT RESTART_SOL= NO % ------------------------- UNSTEADY SIMULATION ------------------------------- TIME_DOMAIN= YES TIME_MARCHING= DUAL_TIME_STEPPING-1ST_ORDER UNST_CFL_NUMBER= 3000 TIME_STEP= 0 MAX_TIME= 1 TIME_ITER= 1000 % -------------------- COMPRESSIBLE FREE-STREAM DEFINITION -------------------- MACH_NUMBER= 2 AOA= 0.0 SIDESLIP_ANGLE= 0.0 INIT_OPTION= TD_CONDITIONS FREESTREAM_OPTION= TEMPERATURE_FS FREESTREAM_TEMPERATURE= 288.15 FREESTREAM_PRESSURE= 101325.0 % ---- IDEAL GAS, POLYTROPIC, VAN DER WAALS AND PENG ROBINSON CONSTANTS ------- FLUID_MODEL= STANDARD_AIR % --------------------------- VISCOSITY MODEL --------------------------------- VISCOSITY_MODEL= SUTHERLAND % --------------------------- THERMAL CONDUCTIVITY MODEL ---------------------- CONDUCTIVITY_MODEL= CONSTANT_PRANDTL % ---------------------- REFERENCE VALUE DEFINITION --------------------------- REF_ORIGIN_MOMENT_X= 0.25 REF_ORIGIN_MOMENT_Y= 0.0 REF_ORIGIN_MOMENT_Z= 0.0 REF_LENGTH= 1.0 REF_AREA= 1.0 REF_DIMENSIONALIZATION= DIMENSIONAL % -------------------- BOUNDARY CONDITION DEFINITION -------------------------- MARKER_HEATFLUX= ( NONE, 0.0 ) MARKER_EULER= projectile MARKER_FAR= (farfield,outlet) MARKER_PLOTTING= projectile MARKER_MONITORING= projectile % ------------- COMMON PARAMETERS DEFINING THE NUMERICAL METHOD --------------- NUM_METHOD_GRAD= GREEN_GAUSS CFL_NUMBER= 40 CFL_ADAPT= YES CFL_ADAPT_PARAM= ( 0.1, 2.0, 1.0, 10000000000.0 ) INNER_ITER= 7 % ------------------------ LINEAR SOLVER DEFINITION --------------------------- LINEAR_SOLVER= FGMRES LINEAR_SOLVER_PREC= LU_SGS LINEAR_SOLVER_ERROR= 0.0001 LINEAR_SOLVER_ITER= 7 % -------------------------- MULTIGRID PARAMETERS ----------------------------- MGLEVEL= 0 % -------------------- FLOW NUMERICAL METHOD DEFINITION ----------------------- CONV_NUM_METHOD_FLOW= ROE ENTROPY_FIX_COEFF= 0.01 JST_SENSOR_COEFF= ( 0.5, 0.02 ) MUSCL_FLOW= YES SLOPE_LIMITER_FLOW= VENKATAKRISHNAN VENKAT_LIMITER_COEFF= 0.05 TIME_DISCRE_FLOW= EULER_IMPLICIT USE_ACCURATE_FLUX_JACOBIANS= YES % --------------------------- CONVERGENCE PARAMETERS -------------------------- CONV_FIELD= REL_RMS_DENSITY CONV_RESIDUAL_MINVAL= -3 CONV_STARTITER= 10 |
|
|
|
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Low Mach number wing/body junction convergence | Zen | SU2 | 6 | May 3, 2019 04:51 |
Convergence of CFX field in FSI analysis | nasdak | CFX | 2 | June 29, 2009 01:17 |
Problems with convergence with an easy system | franzdrs | Main CFD Forum | 0 | June 15, 2009 18:17 |
increasing mesh quality is leading to poor convergence | tippo | CFX | 2 | May 5, 2009 10:55 |
convergence problem with SIMPLER | NURAY KAYAKOL | Main CFD Forum | 1 | February 24, 1999 13:43 |