
[Sponsors] 
February 13, 2006, 08:43 
I have created a moving mesh s

#1 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Sponsored Links
The timestep is 0.005. Up to time=1.225 nice covergence is obtained, but at time=1.225 the solution begins to diverge, as can be seen from the Courant numbers shown below. Furthermore the number of iterations of solving motionUx and motionUy is increasing every iteration. Therefore I think that the motionSolver causes this problem, but I have no clue why. Any suggestions? Thanks. Greetings, Frank ================================================== Output ================================================== Mean and max Courant Numbers = 0.023978 0.948195 Time = 1.225 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.00074787, Final residual = 9.94298e06, No Iterations 280 ICCG: Solving for motionUy, Initial residual = 0.000497424, Final residual = 9.96133e06, No Iterations 187 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.06953e15, global = 1.90209e18 BICCG: Solving for Ux, Initial residual = 0.000139335, Final residual = 3.63493e06, No Iterations 1 BICCG: Solving for Uy, Initial residual = 0.00080448, Final residual = 4.75049e07, No Iterations 2 ICCG: Solving for p, Initial residual = 0.84611, Final residual = 9.43944e07, No Iterations 290 time step continuity errors : sum local = 4.11744e11, global = 1.12633e14, cumulative = 6.75289e10 ICCG: Solving for p, Initial residual = 0.859509, Final residual = 9.95797e07, No Iterations 290 time step continuity errors : sum local = 4.79839e11, global = 6.3742e15, cumulative = 6.75296e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.282238 0.717762 5.26235e28) Aref = 1.44219 Lref = 0.564692 DragCoefficient = 5.82549 pressureDragCoefficient = 4.83947 viscDragCoefficient = 0.986017 LiftCoefficient = 2.19606 MomentCoefficient = 8.31774 ExecutionTime = 9948.97 s Mean and max Courant Numbers = 0.0239854 1.13739 Time = 1.23 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000752869, Final residual = 9.99251e06, No Iterations 281 ICCG: Solving for motionUy, Initial residual = 0.000500919, Final residual = 9.95567e06, No Iterations 187 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.09456e15, global = 2.59593e18 BICCG: Solving for Ux, Initial residual = 0.000173378, Final residual = 4.67296e06, No Iterations 1 BICCG: Solving for Uy, Initial residual = 0.00112283, Final residual = 7.11956e07, No Iterations 2 ICCG: Solving for p, Initial residual = 0.890628, Final residual = 9.45922e07, No Iterations 295 time step continuity errors : sum local = 5.84773e11, global = 7.53318e14, cumulative = 6.7522e10 ICCG: Solving for p, Initial residual = 0.901723, Final residual = 9.32168e07, No Iterations 296 time step continuity errors : sum local = 6.38908e11, global = 1.57041e13, cumulative = 6.75377e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.284429 0.715571 5.79729e29) Aref = 1.44221 Lref = 0.559905 DragCoefficient = 6.26111 pressureDragCoefficient = 5.26338 viscDragCoefficient = 0.997727 LiftCoefficient = 2.30624 MomentCoefficient = 10.1442 ExecutionTime = 10008.3 s Mean and max Courant Numbers = 0.0240167 1.89597 Time = 1.235 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000757996, Final residual = 9.99529e06, No Iterations 281 ICCG: Solving for motionUy, Initial residual = 0.000504364, Final residual = 9.90425e06, No Iterations 188 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.10338e15, global = 1.22539e18 BICCG: Solving for Ux, Initial residual = 0.000229742, Final residual = 6.51542e06, No Iterations 1 BICCG: Solving for Uy, Initial residual = 0.00160232, Final residual = 1.13808e06, No Iterations 2 ICCG: Solving for p, Initial residual = 0.923514, Final residual = 9.85298e07, No Iterations 299 time step continuity errors : sum local = 8.725e11, global = 8.65098e13, cumulative = 6.74512e10 ICCG: Solving for p, Initial residual = 0.932281, Final residual = 9.66381e07, No Iterations 301 time step continuity errors : sum local = 9.52416e11, global = 8.27088e13, cumulative = 6.75339e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.286627 0.713373 7.71628e28) Aref = 1.44222 Lref = 0.55508 DragCoefficient = 7.31698 pressureDragCoefficient = 6.28853 viscDragCoefficient = 1.02845 LiftCoefficient = 2.55426 MomentCoefficient = 14.3745 ExecutionTime = 10067.6 s Mean and max Courant Numbers = 0.0240893 3.38602 Time = 1.24 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000763107, Final residual = 9.99048e06, No Iterations 280 ICCG: Solving for motionUy, Initial residual = 0.000507756, Final residual = 9.93261e06, No Iterations 188 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.08207e15, global = 1.90597e18 BICCG: Solving for Ux, Initial residual = 0.000328665, Final residual = 1.3123e06, No Iterations 2 BICCG: Solving for Uy, Initial residual = 0.00234532, Final residual = 2.49132e06, No Iterations 2 ICCG: Solving for p, Initial residual = 0.945383, Final residual = 9.50596e07, No Iterations 306 time step continuity errors : sum local = 1.20953e10, global = 1.79658e13, cumulative = 6.7516e10 ICCG: Solving for p, Initial residual = 0.953924, Final residual = 9.49527e07, No Iterations 308 time step continuity errors : sum local = 1.33918e10, global = 5.25042e14, cumulative = 6.75212e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.288832 0.711168 5.26306e28) Aref = 1.44221 Lref = 0.550215 DragCoefficient = 9.85785 pressureDragCoefficient = 8.74477 viscDragCoefficient = 1.11308 LiftCoefficient = 3.1429 MomentCoefficient = 24.2969 ExecutionTime = 10126.9 s Mean and max Courant Numbers = 0.0242311 6.68236 Time = 1.245 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000768247, Final residual = 9.8493e06, No Iterations 282 ICCG: Solving for motionUy, Initial residual = 0.000511277, Final residual = 9.82936e06, No Iterations 193 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.09053e15, global = 8.99978e19 BICCG: Solving for Ux, Initial residual = 0.000517687, Final residual = 2.34858e06, No Iterations 2 BICCG: Solving for Uy, Initial residual = 0.00348416, Final residual = 5.18035e06, No Iterations 3 ICCG: Solving for p, Initial residual = 0.955581, Final residual = 9.54336e07, No Iterations 312 time step continuity errors : sum local = 1.69908e10, global = 1.25309e12, cumulative = 6.76465e10 ICCG: Solving for p, Initial residual = 0.96498, Final residual = 9.80104e07, No Iterations 314 time step continuity errors : sum local = 1.91259e10, global = 1.58831e12, cumulative = 6.74877e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.291044 0.708956 7.71984e28) Aref = 1.44217 Lref = 0.545312 DragCoefficient = 15.0292 pressureDragCoefficient = 13.489 viscDragCoefficient = 1.54019 LiftCoefficient = 4.10684 MomentCoefficient = 48.2943 ExecutionTime = 10187 s Mean and max Courant Numbers = 0.0244966 14.8136 Time = 1.25 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000773288, Final residual = 9.89781e06, No Iterations 282 ICCG: Solving for motionUy, Initial residual = 0.000514758, Final residual = 9.9967e06, No Iterations 188 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.04741e15, global = 8.71871e19 BICCG: Solving for Ux, Initial residual = 0.000889756, Final residual = 2.69979e06, No Iterations 3 BICCG: Solving for Uy, Initial residual = 0.00519126, Final residual = 1.8881e06, No Iterations 5 ICCG: Solving for p, Initial residual = 0.950479, Final residual = 9.4596e07, No Iterations 318 time step continuity errors : sum local = 2.22139e10, global = 1.45724e12, cumulative = 6.76334e10 ICCG: Solving for p, Initial residual = 0.957963, Final residual = 8.6076e07, No Iterations 320 time step continuity errors : sum local = 2.21044e10, global = 1.29554e12, cumulative = 6.75039e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.293263 0.706737 1.80998e28) Aref = 1.44214 Lref = 0.54037 DragCoefficient = 26.3516 pressureDragCoefficient = 23.0254 viscDragCoefficient = 3.32619 LiftCoefficient = 6.17307 MomentCoefficient = 104.607 ExecutionTime = 10246.6 s Mean and max Courant Numbers = 0.0249365 40.0673 Time = 1.255 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000778549, Final residual = 9.97842e06, No Iterations 282 ICCG: Solving for motionUy, Initial residual = 0.000518411, Final residual = 9.89038e06, No Iterations 193 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.08214e15, global = 1.5051e18 BICCG: Solving for Ux, Initial residual = 0.00178422, Final residual = 3.98896e06, No Iterations 7 BICCG: Solving for Uy, Initial residual = 0.00891935, Final residual = 7.66749e06, No Iterations 8 ICCG: Solving for p, Initial residual = 0.924408, Final residual = 9.65802e07, No Iterations 322 time step continuity errors : sum local = 2.86659e10, global = 4.15199e13, cumulative = 6.75454e10 ICCG: Solving for p, Initial residual = 0.940882, Final residual = 9.8166e07, No Iterations 322 time step continuity errors : sum local = 3.33021e10, global = 5.05583e13, cumulative = 6.74948e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.295489 0.704511 2.74279e28) Aref = 1.44219 Lref = 0.53539 DragCoefficient = 61.6947 pressureDragCoefficient = 52.343 viscDragCoefficient = 9.35167 LiftCoefficient = 18.4061 MomentCoefficient = 201.47 ExecutionTime = 10307 s Mean and max Courant Numbers = 0.0259063 127.598 Time = 1.26 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000783877, Final residual = 9.81292e06, No Iterations 284 ICCG: Solving for motionUy, Initial residual = 0.000521931, Final residual = 9.95579e06, No Iterations 194 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.0859e15, global = 2.14512e18 BICCG: Solving for Ux, Initial residual = 0.00572169, Final residual = 7.68117e06, No Iterations 12 BICCG: Solving for Uy, Initial residual = 0.0344701, Final residual = 7.918e06, No Iterations 14 ICCG: Solving for p, Initial residual = 0.863708, Final residual = 9.84167e07, No Iterations 324 time step continuity errors : sum local = 3.85863e10, global = 8.52682e13, cumulative = 6.74096e10 ICCG: Solving for p, Initial residual = 0.89722, Final residual = 9.61545e07, No Iterations 323 time step continuity errors : sum local = 4.96108e10, global = 1.1644e12, cumulative = 6.72931e10 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.297722 0.702278 4.16906e28) Aref = 1.44221 Lref = 0.53037 DragCoefficient = 194.264 pressureDragCoefficient = 178.762 viscDragCoefficient = 15.5021 LiftCoefficient = 68.4224 MomentCoefficient = 491.493 ExecutionTime = 10368 s Mean and max Courant Numbers = 0.0287663 420.85 Time = 1.265 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000788983, Final residual = 9.7713e06, No Iterations 285 ICCG: Solving for motionUy, Initial residual = 0.00052554, Final residual = 9.99632e06, No Iterations 194 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.07999e15, global = 4.6044e19 BICCG: Solving for Ux, Initial residual = 0.0340294, Final residual = 3.50395e06, No Iterations 20 BICCG: Solving for Uy, Initial residual = 0.290113, Final residual = 1.78799e06, No Iterations 22 ICCG: Solving for p, Initial residual = 0.691357, Final residual = 9.72633e07, No Iterations 380 time step continuity errors : sum local = 1.24537e09, global = 6.79261e12, cumulative = 6.79724e10 ICCG: Solving for p, Initial residual = 0.997215, Final residual = 8.99401e07, No Iterations 393 time step continuity errors : sum local = 1.00642e07, global = 4.05372e10, cumulative = 1.0851e09 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.299962 0.700038 2.34445e28) Aref = 1.44222 Lref = 0.525311 DragCoefficient = 3812.82 pressureDragCoefficient = 916.412 viscDragCoefficient = 2896.41 LiftCoefficient = 9426.75 MomentCoefficient = 102740 ExecutionTime = 10430 s Mean and max Courant Numbers = 0.571589 111497 Time = 1.27 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000794283, Final residual = 9.78199e06, No Iterations 285 ICCG: Solving for motionUy, Initial residual = 0.000529197, Final residual = 9.90328e06, No Iterations 196 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.07254e15, global = 1.28338e19 BICCG: Solving for Ux, Initial residual = 0.999564, Final residual = 3.97533e06, No Iterations 164 BICCG: Solving for Uy, Initial residual = 0.993602, Final residual = 7.81896e06, No Iterations 160 ICCG: Solving for p, Initial residual = 0.45766, Final residual = 6.70541e07, No Iterations 664 time step continuity errors : sum local = 1.8692e08, global = 3.35061e11, cumulative = 1.05159e09 ICCG: Solving for p, Initial residual = 0.998543, Final residual = 9.70955e07, No Iterations 750 time step continuity errors : sum local = 1.49225e05, global = 3.6547e08, cumulative = 3.54954e08 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.302209 0.697791 3.63624e28) Aref = 1.4422 Lref = 0.520213 DragCoefficient = 2.70433e+07 pressureDragCoefficient = 2.69181e+07 viscDragCoefficient = 125286 LiftCoefficient = 6.50234e+06 MomentCoefficient = 1.03616e+08 ExecutionTime = 10500.2 s Mean and max Courant Numbers = 426.603 3.37259e+07 Time = 1.275 Correct mesh motion diffusion field. ICCG: Solving for motionUx, Initial residual = 0.000799647, Final residual = 9.80335e06, No Iterations 285 ICCG: Solving for motionUy, Initial residual = 0.00053282, Final residual = 9.96991e06, No Iterations 196 Correcting 2D mesh motion ...done volume continuity errors : sum local = 4.08114e15, global = 1.70846e18 BICCG: Solving for Ux, Initial residual = 0.999992, Final residual = 3.38725, No Iterations 1001 BICCG: Solving for Uy, Initial residual = 0.999993, Final residual = 1.81754, No Iterations 1001 ICCG: Solving for p, Initial residual = 1, Final residual = 0.423881, No Iterations 5001 time step continuity errors : sum local = 5.22248e+06, global = 2.21042, cumulative = 2.21042 ICCG: Solving for p, Initial residual = 0.999492, Final residual = 10.9301, No Iterations 5001 time step continuity errors : sum local = 4.02462e+11, global = 272.815, cumulative = 270.604 Wall patch = 3 Wall patch name = cyl_wall Uav = (0.304463 0.695537 7.76322e28) Aref = 1.44217 Lref = 0.515075 DragCoefficient = 4.22157e+16 pressureDragCoefficient = 4.06296e+16 viscDragCoefficient = 1.58615e+15 LiftCoefficient = 1.55157e+16 MomentCoefficient = 5.22663e+17
__________________
Frank Bos 

Sponsored Links 
February 13, 2006, 13:36 
In my experience, interFoam is

#2 
Senior Member
Xiaofeng Liu
Join Date: Mar 2009
Location: State College, PA, USA
Posts: 118
Rep Power: 10 
In my experience, interFoam is not ready for open channel flow, yet.
__________________
Xiaofeng Liu, Ph.D., P.E., Assistant Professor Department of Civil and Environmental Engineering Penn State University 223B Sackett Building University Park, PA 16802 Web: http://water.engr.psu.edu/liu/ 

February 13, 2006, 13:44 
I use the icoFoamAutoMotion eq

#3 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
I use the icoFoamAutoMotion equations and related algorithms, piso loop and stuff. The moving mesh part is copied from interFoam, since that solver works with the movingFvMesh class.
Regards, Frank
__________________
Frank Bos 

February 14, 2006, 03:14 
Hoi Frank,
Je pisoloop con

#4 
Member
Guido Adriaensen
Join Date: Mar 2009
Posts: 56
Rep Power: 10 
Hoi Frank,
Je pisoloop convergeert niet. Als je kijkt naar de initial residual van p, moet deze minimaal 1 orde lager zijn bij de tweede iteratie in de loop, dit is niet het geval (Hier is die waarde zelfs iets hoger): ICCG: Solving for p, Initial residual = 0.84611, Final residual = 9.43944e07, No Iterations 290 time step continuity errors : sum local = 4.11744e11, global = 1.12633e14, cumulative = 6.75289e10 ICCG: Solving for p, Initial residual = 0.859509, Final residual = 9.95797e07, No Iterations 290 Je kunt proberen om het aantal piso correctors op te voeren of een kleinere tijdsstap nemen. Succes Guido 

February 17, 2006, 06:46 
Ok, more piso correctors is no

#5 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Ok, more piso correctors is not improving the convergence.
On the other hand, a much smaller timestep is actually leading to convergence at the beginning of the motion. After a certain time (which is larger than the previous case) also this solution is diverges. When I turn off the mesh.move() function in my solver no problem occurs. So the problem lies in the piso loop when the mesh moves. Thanks, Frank
__________________
Frank Bos 

February 21, 2006, 07:08 
Could someone please comment o

#6 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Could someone please comment on my solver, shown above.
When I move the mesh, the piso loop diverges, also when I decrease the timestep or increase the number of correctors. When I turn off the mesh motion, the flow solver works fine. In addition when I turn off the flow solver instead of the mesh motion, the moving mesh is calculated well. Do I have to include some extra terms in the piso loop to account for the mesh motion? Thanks in advance. Regards, Frank
__________________
Frank Bos 

February 21, 2006, 12:10 
The solver looks fine (well, a

#7 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
The solver looks fine (well, at least not obviously wrong). However, your residual behaviour indicates a serious problem and my guess is that you have messed up the boundary conditions.
Note that the wall boundary needs to be specified in terms of absolute velocity, so a moving wall with no flux through it will not have fixedValue (0 0 0). If this is the problem use the movingWallVelocity b.c., which corrects the normal component based on the mesh motion. As a test, have a look at the fluxes through your wall boundaries  if all is well, they should be exactly zero. Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

February 22, 2006, 11:21 
As far as I know I already use

#8 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
As far as I know I already used the movingWallVelocity b.c.
My boundary conditions in 0/U looks like (cyl_wall is the moving surface): dimensions [0 1 1 0 0 0 0]; internalField uniform (1 0 0); boundaryField { far_field { type symmetryPlane; } outlet { type zeroGradient; } inlet { type fixedValue; value uniform (1 0 0); } cyl_wall { type movingWallVelocity; value uniform (0 0 0); } frontAndBackPlanes { type empty; } } =========================================== And 0/motionU looks like: dimensions [0 1 1 0 0 0 0]; internalField uniform (0 0 0); boundaryField { far_field { type symmetryPlane; } outlet { type fixedValue; value uniform (0 0 0); } inlet { type fixedValue; value uniform (0 0 0); } cyl_wall { type fixedValue; value uniform (0 0 0); } frontAndBackPlanes { type empty; } } =========================================== When I used movingFlepFvMesh the phi of some surface elements after 1 sec looks like: 1/phi: cyl_wall { type calculated; value nonuniform List<scalar> 280 ( 0 1.73472e18 0 1.73472e18 1.73472e18 0 0 0 0 1.73472e18 0 0 0 1.73472e18 0 0 8.67362e19 8.67362e19 0 8.67362e19 . . . . . } What is wrong with this b.c. implementation. I think I mess thing up with U and motionU........The fixedValue of motionU is overruled by movingFlapFvMesh, right? Regards, Frank
__________________
Frank Bos 

February 23, 2006, 08:09 
Summarising:
What b.c. do y

#9 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Summarising:
What b.c. do you generally have to apply to a moving body? * motionU: fixedValue on the moving surface (this is being overruled by the movingFvMesh class) * U: movingWallVelocity (uniform (0,0,0)) on the moving surface (according to dr Jasak this will correct the boundary flux) Is this correct? or must by movingWallVelocity be applied to both U and motionU? If these settings are correct, my problem of the diverging piso loop has its origins elsewhere. Thanks, Frank
__________________
Frank Bos 

February 23, 2006, 10:29 
motionU: you can do whatever y

#10 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
motionU: you can do whatever you like, this is only used to specify the motion. Typically, if you wish to specify the motion of an object, you would use the fixedValue b.c. on its surface (makes sense).
U: movingWallVelocity. I usually stick to this. It will correct the normal component of the velocity to achieve zero flux through the boundary based on the vertex motion (wherever it may come from). There's not connection between the motionU b.c. and the U b.c.  they are totally independent. I'm not sure what may be going wrong. Try setting very diffusive discretisation and if that still fails you've messed up the boundary conditions or something even more subtly. Have you tried looking at the mesh motion (motion only, no flow), to see if it behaves the way you think it should? This should work without any trouble, Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

February 24, 2006, 08:29 
Thanks for your answers, but m

#11 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Thanks for your answers, but my setup is still not working. I tried central, upwind and blended schemes for the convective flux, but the solution still diverges after 1 second or so.
When I read my ../1/phi file the values at the boundary are of the order 10e18. Does this flux need to be exactly zero or is 10e18 small enough. I also switched off the solver such that the mesh motion can be visualised. In that case the mesh looks very nice at all time up to t=20. No diverges occurs. The solution divergences only when the piso loop is solved. Maybe the movingFlapFvMesh coupling to the fixedValue of motionU is not done correctly.... Frank
__________________
Frank Bos 

February 24, 2006, 09:56 
I'll have a look if you want.

#12 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
I'll have a look if you want. Could you please pack the code (ready to run, if the mesh is made with blockMesh, please delete unnecessary files, with timestep, discretisation, initial fields etc set up) and the case (clean up the object files and similar garbage + please include everything I need to compile it up).
Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

February 24, 2006, 13:30 
Thanks,
I just sent you th

#13 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Thanks,
I just sent you the case and solver files. Both are ready to run/compile. Frank
__________________
Frank Bos 

February 27, 2006, 18:50 
Very easy: it does not work be

#14 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
Very easy: it does not work because you've set the time step to give the Courant number of 3.5, which is jolly high. I have dropped the time step by a factor of 4 and with the Co number of around 1 it works beautifully for the first second or so and then the Co number goes higher again which makes it blow up. Nothing wrong with the solution, just the normal flow development on a newrrt fine mesh.
Then, I have added automatic timestep control and it works beautifully. Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

February 28, 2006, 08:00 
Thanks again for your time :)

#15 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Thanks again for your time :).
In my first case (dt=0.02) also the Co of around 3.5 gives me a good solution for the first second. After that the Co always blows up which is rather independent of the chosen timestep. I really like a constant time step. Since I need to simulate for 100 seconds or so, this is easy to plan and estimate calculation times. Though, this seems to be impossible in my case. Frank
__________________
Frank Bos 

March 4, 2006, 16:22 
I did the first 5 seconds and

#16 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
I did the first 5 seconds and made a little animation for you (it is ugly but proves the point). The automatic timestep control is wellbehaved and there's no problem with the simulation (I am using 1 nonorthogonal corrector). If you want the regular output dumps with automatic timestep control, e.g. every 1 second or similar, use adjustableRunTime writeControl option.
cylinder animation Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

March 6, 2006, 06:54 
Allright, thanks. Is there act

#17 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Allright, thanks. Is there actually some sort of guideline how many nonorthogonal correctors are needed? Is the accuracy or the efficiency affected by this choice, or not?
Frank
__________________
Frank Bos 

March 6, 2006, 07:12 
You have several options. I s

#18 
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
You have several options. I suspect the problem you are getting is that the mesh deformation may become quite sensitive to the convergence of the motion solver. The issue is that you've got a really big mesh qith a little bit of motion in the middle. Ideally, you would want to subset a piece of the mesh around the cylinder which is big enough to accomodate the motion but smaller than the complete domain. This would make yor simulation much faster (less work in mesh motion) and probably cause less problems. However, it is a bit more difficult to set up (you will be dealing with 2 meshes, where the motion mesh is a subset of the first one).
If you want to see exactly what is going on, try running the nonrothogonality check from time to time during the mesh motion. You will see that the nonorthogonality changes and this influences the convergence behaviour. You can tackle it in a number of different ways, but basically they fall into the two categories you mention  a less accurate but more robust discretisation for the laplacian: limited Gauss. You lose accuracy but it does not affect the speed  nonorthogonal correctors: (typically, 1 is plenty). Increases the cost of simulation but is more accurate than the limited snGrad stuff from above. Enjoy, Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

November 18, 2006, 09:43 
Hi everybody,
In my simulat

#19 
Senior Member
Frank Bos
Join Date: Mar 2009
Location: The Netherlands
Posts: 340
Rep Power: 11 
Hi everybody,
In my simulations concerning moving meshes, I need a very small timestep in OpenFOAM compared to other CFD solvers (e.g. Fluent). In OpenFOAM I use secondorder CrankNicholson, so in fact a smaller timestep than Fluent (which is first order in time for moving meshes) should work. The mesh motion is OK for larger timesteps, but for convergence of the icoDyMFoam solver a much much smaller timestep is necessary. Otherwise the Courant number increases till the solution diverges. Why is CrankNicholson in OpenFOAM not unconditionally stable, as proved in many textbooks (e.g Ferziger, Peric or Hirsch). Regards, Frank
__________________
Frank Bos 

November 18, 2006, 11:55 
Why is CrankNicholson in Open

#20  
Senior Member
Hrvoje Jasak
Join Date: Mar 2009
Location: London, England
Posts: 1,800
Rep Power: 24 
Quote:
I have shown in my PhD that CrankNicolson is not unconditionally unstable (based on numerical diffusion) so what you say is definitely news to me. Would you clarify please. Hrv
__________________
Hrvoje Jasak Providing commercial FOAM/OpenFOAM and CFD Consulting: http://wikki.co.uk 

Thread Tools  
Display Modes  


Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Moving mesh  Martijn  CFX  3  November 26, 2008 22:46 
moving mesh  sadid  Main CFD Forum  1  May 4, 2008 05:33 
Moving Reference frame  UDF  Moving mesh  modisa  FLUENT  0  April 18, 2008 13:31 
moving mesh  sulfi  FLUENT  7  April 13, 2005 23:52 
Moving Mesh & Not Rotating Mesh  AB  CDadapco  1  October 25, 2004 03:10 
Sponsored Links 