CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Software User Forums > OpenFOAM

Simulation seems to converge but crashes suddenly

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   June 27, 2013, 12:28
Default Simulation seems to converge but crashes suddenly
  #1
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Hello,

I'm trying to run a simulation on an unstructured mesh. The simulation seems to converge but, after a thousand of iterations, it suddenly crashes. The strange things is that, if I restart it from just some iterations before the crash and without modify anything, it starts to converge again and crashes further beyond the previous crash point.
The simulation is steady state, MRFSimpleFoam in OF2.1.1.
The case is decomposed in 32 processors.
I use cyclicAMI for periodic faces. RASmodel is kEpsilon.
BCs are :

- inlet : p=zeroGradient U=flowRateInletVelocity
- outlet : p=fixedValue
- all walls : U = fixedValue ( 0 0 0);

I declare inlet, outlet and cyclicAMI as non rotating patches in MRFmodel.

The checkMesh result is

Mesh stats
points: 24362
faces: 232715
internal faces: 211841
cells: 111139
boundary patches: 9
point zones: 0
face zones: 0
cell zones: 1

Overall number of cells of each type:
hexahedra: 0
prisms: 0
wedges: 0
pyramids: 0
tet wedges: 0
tetrahedra: 111139
polyhedra: 0

Checking topology...
Boundary definition OK.
Cell to face addressing OK.
Point usage OK.
Upper triangular ordering OK.
Face vertices OK.
Number of regions: 1 (OK).

Checking patch topology for multiply connected surfaces ...
Patch Faces Points Surface topology
inlet 353 206 ok (non-closed singly connected)
albero 434 250 ok (non-closed singly connected)
wall_inlet 542 305 ok (non-closed singly connected)
ciclico_ps 1878 1059 ok (non-closed singly connected)
dischi 8215 4469 ok (non-closed singly connected)
blade 2916 1606 ok (non-closed singly connected)
ciclico_ss 1928 1084 ok (non-closed singly connected)
top_bott_out 4036 2158 ok (non-closed singly connected)
outlet 572 356 ok (non-closed singly connected)

Checking geometry...
Overall domain bounding box (-0.00825 -0.0229597 -0.0123) (0.0594161 0.055631 0.0032)
Mesh (non-empty, non-wedge) directions (1 1 1)
Mesh (non-empty) directions (1 1 1)
Boundary openness (-9.90405e-17 -1.02286e-16 2.27575e-17) OK.
Max cell openness = 2.85274e-16 OK.
Max aspect ratio = 28.7209 OK.
Minumum face area = 6.8635e-09. Maximum face area = 1.76476e-06. Face area magnitudes OK.
Min volume = 6.18733e-13. Max volume = 6.85671e-10. Total volume = 7.09264e-06. Cell volumes OK.
Mesh non-orthogonality Max: 86.2371 average: 17.0449
*Number of severely non-orthogonal faces: 203.
Non-orthogonality check OK.
<<Writing 203 non-orthogonal faces to set nonOrthoFaces
Face pyramids OK.
Max skewness = 1.07316 OK.
Coupled point location match (average 0) OK.

Mesh OK.

I attached some pictures of residuals.
You can see that residuals keeps steady or falling until they rapidly increase.

When I restart the case from the latest "useful" (=not crashed) time, the simulation proceeds without problems for another thousand of iterations.By the first steps after restart, global results are much different from the ones before the crash. After each crash, results get closer to that calculated for the same geometry with structured mesh (and without any problem!).

After the latest crash, I got almost the same results that I reached with structured mesh. But after some further iteration, the case crashed again.

I would like to ask you if you have any ideas about why the simulation crashes.
I also would like to know if I have to consider correct the results I get in a similar case. I mean that, in my opinion, if simulation crashes it means that results are not reliable. But at the same time, they look so close to the ones by structured mesh, witch I consider to be reliable ( and close to the ones by starCD ).

Please, can someone help me?

Thank in advance.
Attached Images
File Type: jpg residui_1.jpg (27.6 KB, 277 views)
File Type: jpg residui_35000.jpg (42.9 KB, 320 views)
xxxx is offline   Reply With Quote

Old   June 29, 2013, 14:31
Default Interesting....
  #2
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
Hello!

One thing that does not look good is the non-orthogonality of the mesh. Can you post your log file (atleast the residual log of the first 10 iterations) ?
Use gradient limiters ...'cellLimited' for your grad scheme. This should make your simulation more stable.

Akshay
Akshay is offline   Reply With Quote

Old   June 29, 2013, 14:44
Default
  #3
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Thank you Akshay for the advise.
Now i cannot post the residual log but I will do it as soon as i can.
Do you think the problem is the mesh? Actually I do think so but many other times meshes with a high number of non-orthogonal cells didn't give me such problems.
Actually, i often find problems when i try to use unstructured mesh.
I use Salome for meshing, but sometimes i would like to have a more powerful tool. May i ask you what you use?
xxxx is offline   Reply With Quote

Old   June 29, 2013, 15:14
Default
  #4
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
OpenFOAM is highly sensitive to the non-orthogonality of meshes. I use snappyHexMesh to create my meshes. It does a great job. Other meshes work as well(even if they are unstructured) as long as orthogonality is maintained.
Akshay is offline   Reply With Quote

Old   June 30, 2013, 12:20
Default
  #5
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Thank you again.
So you think the problem could be orthogonality. Sorry if I bother you, but how can I control non-orthogonality, especially with an unstructured mesh?
Recently I tried to use snappyHexMesh too. Do you use it also for internal fluid dynamics? Because it seems to me that it works better for external fluid dynamics, but I admit I dont know it well. I have tried helixOS interface to simplify the use of snappyHexMesh.
Thank in advance.
xxxx is offline   Reply With Quote

Old   July 1, 2013, 10:19
Default
  #6
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Hello Akshay

I posted the residual log file.

I tried to use cellLimited as grad scheme but OF 211 does not recognize it.
The only limited scheme are:

limitWith
limitedCubic
limitedCubicV
limitedLinear
limitedLinearV


>>>>>>>>> Sorry, I'm not able to have the LOG file uploaded. I post you the first lines:


Create time

Create mesh for time = 34380

Reading field p

Reading field U

Reading/calculating face flux field phi

AMI: Creating addressing and weights between 748 source faces and 758 target faces
AMI: Patch source weights min/max/average = 0.999877, 1.00032, 1.00001
AMI: Patch target weights min/max/average = 0.999917, 1.0003, 1.00001
Selecting incompressible transport model Newtonian
Selecting RAS turbulence model kEpsilon
No field sources present


SIMPLE: convergence criteria
field p tolerance 1e-07
field Urel tolerance 1e-06
field nuTilda tolerance 1e-06
field omega tolerance 1e-06
field k tolerance 1e-06
field epsilon tolerance 1e-06


Starting time loop

Time = 34381

DILUPBiCG: Solving for Ux, Initial residual = 0.000116484, Final residual = 1.95806e-07, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 0.000190872, Final residual = 4.10964e-07, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.00104798, Final residual = 1.49606e-06, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.00671232, Final residual = 6.54732e-06, No Iterations 62
time step continuity errors : sum local = 0.00145536, global = -0.000191862, cumulative = -0.000191862
DILUPBiCG: Solving for epsilon, Initial residual = 3.7014e-06, Final residual = 5.76453e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.98994e-05, Final residual = 2.74708e-08, No Iterations 2
ExecutionTime = 0.49 s ClockTime = 0 s

Time = 34382

DILUPBiCG: Solving for Ux, Initial residual = 6.917e-05, Final residual = 8.73941e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 0.000102618, Final residual = 2.82595e-07, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.000612589, Final residual = 8.99979e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.00271528, Final residual = 2.61552e-06, No Iterations 123
time step continuity errors : sum local = 0.000581223, global = -3.52181e-05, cumulative = -0.00022708
DILUPBiCG: Solving for epsilon, Initial residual = 3.77728e-06, Final residual = 5.73571e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.99839e-05, Final residual = 1.0709e-08, No Iterations 2
ExecutionTime = 0.73 s ClockTime = 0 s

Time = 34383

DILUPBiCG: Solving for Ux, Initial residual = 3.61533e-05, Final residual = 2.63406e-07, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 5.50816e-05, Final residual = 1.01818e-07, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.000340453, Final residual = 5.88818e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.00283995, Final residual = 2.71022e-06, No Iterations 30
time step continuity errors : sum local = 0.000601727, global = 8.04178e-05, cumulative = -0.000146662
DILUPBiCG: Solving for epsilon, Initial residual = 4.42084e-06, Final residual = 6.33828e-08, No Iterations 1
bounding epsilon, min: -0.00650573 max: 1060.77 average: 10.5298
DILUPBiCG: Solving for k, Initial residual = 3.0096e-05, Final residual = 5.41345e-09, No Iterations 2
ExecutionTime = 0.87 s ClockTime = 0 s

Time = 34384

DILUPBiCG: Solving for Ux, Initial residual = 2.26295e-05, Final residual = 6.13416e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 3.24352e-05, Final residual = 4.73366e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.00019446, Final residual = 1.83417e-08, No Iterations 3
DICPCG: Solving for p, Initial residual = 0.00185662, Final residual = 1.73255e-06, No Iterations 118
time step continuity errors : sum local = 0.000384432, global = 3.08246e-05, cumulative = -0.000115837
DILUPBiCG: Solving for epsilon, Initial residual = 3.71761e-06, Final residual = 2.17537e-07, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.99508e-05, Final residual = 7.43986e-09, No Iterations 2
ExecutionTime = 1.09 s ClockTime = 1 s

Time = 34385

DILUPBiCG: Solving for Ux, Initial residual = 1.64447e-05, Final residual = 4.3641e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 2.18851e-05, Final residual = 5.3845e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 0.000130193, Final residual = 1.85792e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.0010382, Final residual = 9.90235e-07, No Iterations 114
time step continuity errors : sum local = 0.000219681, global = 3.99351e-05, cumulative = -7.59024e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.57012e-06, Final residual = 8.03344e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.98805e-05, Final residual = 6.51274e-09, No Iterations 2
ExecutionTime = 1.31 s ClockTime = 1 s

Time = 34386

DILUPBiCG: Solving for Ux, Initial residual = 1.18153e-05, Final residual = 3.32574e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 1.49857e-05, Final residual = 8.53877e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 8.9257e-05, Final residual = 2.14161e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.00057172, Final residual = 5.15406e-07, No Iterations 189
time step continuity errors : sum local = 0.00011434, global = 7.86249e-07, cumulative = -7.51161e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.54648e-06, Final residual = 1.1617e-07, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.98313e-05, Final residual = 6.89284e-09, No Iterations 2
ExecutionTime = 1.61 s ClockTime = 1 s

Time = 34387

DILUPBiCG: Solving for Ux, Initial residual = 8.47096e-06, Final residual = 2.02699e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 1.05874e-05, Final residual = 7.0427e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 6.13598e-05, Final residual = 3.56812e-09, No Iterations 3
DICPCG: Solving for p, Initial residual = 0.000366482, Final residual = 3.65723e-07, No Iterations 138
time step continuity errors : sum local = 8.11399e-05, global = 7.36762e-06, cumulative = -6.77485e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.60755e-06, Final residual = 7.63014e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.97936e-05, Final residual = 8.57472e-09, No Iterations 2
ExecutionTime = 1.87 s ClockTime = 1 s

Time = 34388

DILUPBiCG: Solving for Ux, Initial residual = 6.20941e-06, Final residual = 1.95689e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 7.69862e-06, Final residual = 5.75046e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 4.28168e-05, Final residual = 1.70379e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.000244954, Final residual = 2.29733e-07, No Iterations 180
time step continuity errors : sum local = 5.09679e-05, global = -6.05531e-07, cumulative = -6.8354e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.87334e-06, Final residual = 6.84854e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.97638e-05, Final residual = 9.06926e-09, No Iterations 2
ExecutionTime = 2.15 s ClockTime = 2 s

Time = 34389

DILUPBiCG: Solving for Ux, Initial residual = 4.77713e-06, Final residual = 1.43383e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 5.88675e-06, Final residual = 1.4134e-08, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 3.13716e-05, Final residual = 1.14339e-07, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.000170138, Final residual = 1.67529e-07, No Iterations 38
time step continuity errors : sum local = 3.71664e-05, global = -1.77426e-06, cumulative = -7.01283e-05
DILUPBiCG: Solving for epsilon, Initial residual = 6.4676e-06, Final residual = 8.48483e-08, No Iterations 1
bounding epsilon, min: -0.0854313 max: 1060.77 average: 10.5295
DILUPBiCG: Solving for k, Initial residual = 2.981e-05, Final residual = 1.25845e-07, No Iterations 2
ExecutionTime = 2.3 s ClockTime = 2 s

Time = 34390

DILUPBiCG: Solving for Ux, Initial residual = 3.80104e-06, Final residual = 1.01539e-08, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 4.65179e-06, Final residual = 5.97925e-09, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 2.38548e-05, Final residual = 4.96177e-08, No Iterations 2
DICPCG: Solving for p, Initial residual = 0.000118929, Final residual = 1.15022e-07, No Iterations 108
time step continuity errors : sum local = 2.55176e-05, global = 3.39004e-06, cumulative = -6.67383e-05
DILUPBiCG: Solving for epsilon, Initial residual = 4.35129e-06, Final residual = 6.37832e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.97125e-05, Final residual = 6.79241e-09, No Iterations 2
ExecutionTime = 2.51 s ClockTime = 2 s

Time = 34391

DILUPBiCG: Solving for Ux, Initial residual = 3.12327e-06, Final residual = 8.26615e-09, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 3.79e-06, Final residual = 5.2095e-09, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 1.87191e-05, Final residual = 2.82144e-08, No Iterations 2
DICPCG: Solving for p, Initial residual = 8.62629e-05, Final residual = 9.76075e-08, No Iterations 113
time step continuity errors : sum local = 2.16542e-05, global = 1.94692e-06, cumulative = -6.47914e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.68029e-06, Final residual = 6.11616e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.96539e-05, Final residual = 7.78769e-09, No Iterations 2
ExecutionTime = 2.73 s ClockTime = 2 s

Time = 34392

DILUPBiCG: Solving for Ux, Initial residual = 2.63464e-06, Final residual = 6.09912e-09, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 3.18131e-06, Final residual = 4.0811e-09, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 1.50655e-05, Final residual = 5.82442e-08, No Iterations 2
DICPCG: Solving for p, Initial residual = 6.69605e-05, Final residual = 9.75085e-08, No Iterations 17
time step continuity errors : sum local = 2.16325e-05, global = 5.92714e-07, cumulative = -6.41986e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.46206e-06, Final residual = 5.47954e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.96205e-05, Final residual = 8.97014e-09, No Iterations 2
ExecutionTime = 2.86 s ClockTime = 2 s

Time = 34393

DILUPBiCG: Solving for Ux, Initial residual = 2.259e-06, Final residual = 5.01807e-09, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 2.72394e-06, Final residual = 5.76879e-09, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 1.23467e-05, Final residual = 4.15094e-08, No Iterations 2
DICPCG: Solving for p, Initial residual = 5.2557e-05, Final residual = 9.71596e-08, No Iterations 178
time step continuity errors : sum local = 2.15551e-05, global = -7.09067e-07, cumulative = -6.49077e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.35479e-06, Final residual = 5.28543e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.95863e-05, Final residual = 8.85696e-09, No Iterations 2
ExecutionTime = 3.14 s ClockTime = 3 s

Time = 34394

DILUPBiCG: Solving for Ux, Initial residual = 1.96087e-06, Final residual = 3.74087e-09, No Iterations 2
DILUPBiCG: Solving for Uy, Initial residual = 2.36285e-06, Final residual = 3.33676e-09, No Iterations 2
DILUPBiCG: Solving for Uz, Initial residual = 1.0291e-05, Final residual = 2.50191e-08, No Iterations 2
DICPCG: Solving for p, Initial residual = 4.31965e-05, Final residual = 9.70765e-08, No Iterations 19
time step continuity errors : sum local = 2.15365e-05, global = -3.91339e-06, cumulative = -6.88211e-05
DILUPBiCG: Solving for epsilon, Initial residual = 3.29362e-06, Final residual = 5.36464e-08, No Iterations 1
DILUPBiCG: Solving for k, Initial residual = 2.95555e-05, Final residual = 8.68175e-09, No Iterations 2
ExecutionTime = 3.27 s ClockTime = 3 s



limiterBlended

What scheme did you mean? Is it in OF 220?

Last edited by xxxx; July 1, 2013 at 10:24. Reason: sorry, i had some problems with file uploader
xxxx is offline   Reply With Quote

Old   July 1, 2013, 13:34
Default
  #7
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
Hey

try this...

gradSchemes
{
default cellLimited leastSquares 1.0;
}


And...
You did mention you were using an unstructured grid right?
Akshay is offline   Reply With Quote

Old   July 2, 2013, 02:02
Default
  #8
Senior Member
 
T. Chourushi
Join Date: Jul 2009
Posts: 321
Blog Entries: 1
Rep Power: 18
Tushar@cfd is on a distinguished road
Quote:
Originally Posted by xxxx View Post
Thank you again.
So you think the problem could be orthogonality. Sorry if I bother you, but how can I control non-orthogonality, especially with an unstructured mesh?
Recently I tried to use snappyHexMesh too. Do you use it also for internal fluid dynamics? Because it seems to me that it works better for external fluid dynamics, but I admit I dont know it well. I have tried helixOS interface to simplify the use of snappyHexMesh.
Thank in advance.
Hello,

1. For the fvScheme file in your case check the following post:
http://www.cfd-online.com/Forums/ope...orrectors.html
I hope that will solve your problem regarding scheme implementation.

If you are not getting proper convergence try this one too.
2. Since, your mesh is "severely non-orthogonal faces: 203". I would rather run a transient run instead of SIMPLE and with transient case try increasing the nCorrectors and nNonOrthogonalCorrectors in fvSolution file.

Do, try these steps and do let us know.
Tushar@cfd is offline   Reply With Quote

Old   July 2, 2013, 02:59
Default
  #9
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
xxxx: You've got all Tets and a total count of 111139. You are decomposing this size into 32 partitions!! that is an overkill!!

Tushar: I don't think he needs to bring in more complexities through a transient run. MRFSimpleFoam should do the job. Non-Ortho of 86 can be handled through some stable solver settings. I've run cases with non-ortho of 87 successfully.

And...

xxxxx: Please use another name here and not 'xxxxx'
Akshay is offline   Reply With Quote

Old   July 2, 2013, 08:19
Default
  #10
Senior Member
 
T. Chourushi
Join Date: Jul 2009
Posts: 321
Blog Entries: 1
Rep Power: 18
Tushar@cfd is on a distinguished road
Hello,

One more reason why the solver had convergence problem. If you see the checkMesh output the max skewness is:

Code:
Max skewness = 1.07316 OK.
It is actually higher.

Tushar@cfd is offline   Reply With Quote

Old   July 2, 2013, 11:06
Default
  #11
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
Tushar: Skewness of 1.07 is as low as it can get. Above 5 and i'll call it a MINOR problem
Akshay is offline   Reply With Quote

Old   July 2, 2013, 16:55
Default
  #12
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Hello everybody.

Akshay: thank you. I will try cellLimited soon. The case was partitioned between 32 processors to see how it works ( I have just started to use multi-processors and I still don't know exactly how it works).. You can call me 'zzzzz' if you prefer :-) .
In the meantime I talked with a person more expert than me in CFD, but not in OF. He supposes that, since crash occurs in only one step, this could be caused by errors in reading/writing in RAM.

Tushar: I'm not an expert but skewness as low as 1.07 should not be a problem.
xxxx is offline   Reply With Quote

Old   July 3, 2013, 02:00
Default
  #13
Senior Member
 
T. Chourushi
Join Date: Jul 2009
Posts: 321
Blog Entries: 1
Rep Power: 18
Tushar@cfd is on a distinguished road
Well,

Akshay
Quote:
Tushar: Skewness of 1.07 is as low as it can get. Above 5 and i'll call it a MINOR problem
And, xxxx
Quote:
Hello everybody.

Akshay: thank you. I will try cellLimited soon. The case was partitioned between 32 processors to see how it works ( I have just started to use multi-processors and I still don't know exactly how it works).. You can call me 'zzzzz' if you prefer :-) .
In the meantime I talked with a person more expert than me in CFD, but not in OF. He supposes that, since crash occurs in only one step, this could be caused by errors in reading/writing in RAM.

Tushar: I'm not an expert but skewness as low as 1.07 should not be a problem.
If you happen to see the definition of skewness as mentioned in the following web-links:

http://www.tchpc.tcd.ie/fluent/Unpac...ug/node318.htm

http://www.cfd-online.com/Forums/flu...an-0-98-a.html

http://en.wikipedia.org/wiki/Types_o...uiangular_skew

You, will realize the actual problem comes when skewness exceeds 1 (>1).
Also, when mesh is distorted then the error induced by convection term increases significantly which causes unboundedness. Now, if suitable scheme is not used then the solver crashes.

I hope you guys got my point.
Tushar@cfd is offline   Reply With Quote

Old   July 3, 2013, 02:10
Default
  #14
Member
 
Akshay Kumar
Join Date: Aug 2010
Location: India
Posts: 84
Rep Power: 16
Akshay is on a distinguished road
Tushar:
OpenFOAM has its very own definition of skewness. What it shows in checkMesh is based on its definition of skewness. In OF skewness above 4 is usually called high skewness. Please don't confuse fluent with openfoam. They are very different!
Akshay is offline   Reply With Quote

Old   July 3, 2013, 02:33
Default
  #15
Senior Member
 
T. Chourushi
Join Date: Jul 2009
Posts: 321
Blog Entries: 1
Rep Power: 18
Tushar@cfd is on a distinguished road
Quote:
Originally Posted by Akshay View Post
Tushar:
OpenFOAM has its very own definition of skewness. What it shows in checkMesh is based on its definition of skewness. In OF skewness above 4 is usually called high skewness. Please don't confuse fluent with openfoam. They are very different!
Can you write the definition of skewness with respect to OpenFOAM?

I know one can solve with higher skewness in OpenFOAM. But, for that you need to find the correct scheme. May be, some more efforts to skewcorrect the solution. That I think was a reason behind the divergence of the mentioned solution.

What do you think solving is more important ? or, solution accuracy? with respect to the actual results. When the user uses higher skewness it incorporates more chances of errors.
Tushar@cfd is offline   Reply With Quote

Old   July 3, 2013, 17:09
Default
  #16
Member
 
xxxxx
Join Date: Feb 2013
Posts: 34
Rep Power: 13
xxxx is on a distinguished road
Akshay, may I ask you another thing?
Following your advise I started to learn how to use snappyHexMesh. I still have some doubts.
First of all, I saw that when I use snappyHexMesh (with snappyHex disctionary taken from simpleFoam/turbine tutorial and modified to adapt to a simple geometry) OF creates to folders, named 1 and 2, each containig a polyMesh folder. I think this are, let me say, a first trial mesh and a refined one, rigth? Than only the mesh in folder "2" is copied to constant. If I perform checkMesh I see that mesh at time 1 seems to be better that at time 2 ( smaller skewness, and so on..). If I copy polyMesh folder from time 1 to constant and open paraview, the mesh doesn't look much different from the one at time 2. Am I wrong? In this case, should I use the mesh of folder 1?
Another question is, can you please explain briefly what the two numbers in

" levels ((2 4)); "

in snappyHexMeshDict mean? I know they are related to mesh levels and using them one can control the resulting mesh, but how do they actually work?

The last question is about mesh quality. It seems to me that snappyHexMesh produces some holes in the surface. I highlighted some of these in the attached picture.You can also see there's a sort of bubble in the foreground (pointed by an arrow). The colours are cell normals, so different colours mean in neighbour cells mean steeply variations in orientation. Can this problem cause the crash of the case? How can I remove it?

Thank in advance.
Attached Images
File Type: jpg snappy1.jpg (92.1 KB, 155 views)
Attached Files
File Type: txt checkMesh.txt (3.2 KB, 13 views)
xxxx is offline   Reply With Quote

Old   September 12, 2014, 09:07
Default
  #17
Senior Member
 
T. Chourushi
Join Date: Jul 2009
Posts: 321
Blog Entries: 1
Rep Power: 18
Tushar@cfd is on a distinguished road
Quote:
Originally Posted by Akshay View Post
Tushar:
OpenFOAM has its very own definition of skewness. What it shows in checkMesh is based on its definition of skewness. In OF skewness above 4 is usually called high skewness. Please don't confuse fluent with openfoam. They are very different!
First of all I owe Apology to Akshay,
Please, excuse me for my earlier comment it was too rude.

For few cases skewness of 1-4 are considered OK with OpenFOAM.

Quote:
Hello,

I'm trying to run a simulation on an unstructured mesh. The simulation seems to converge but, after a thousand of iterations, it suddenly crashes. The strange things is that, if I restart it from just some iterations before the crash and without modify anything, it starts to converge again and crashes further beyond the previous crash point.
The simulation is steady state, MRFSimpleFoam in OF2.1.1.
Yes xxx, your case is of turbine. So, you must re-mesh as handling of that bad mesh in OpenFOAM will be very difficult with the mentioned MRFSimpleFoam solver.

-
Best Regards!
Tushar@cfd is offline   Reply With Quote

Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Problem in initializing transient simulation with a finer mesh sidd CFX 8 April 29, 2016 03:25
Dynamic mesh simulation (pimpleDyMFoam) problem sidlof OpenFOAM Running, Solving & CFD 6 April 12, 2013 12:24
T reactor simulation - Dont converge - Help plz Dinocrack FLUENT 0 September 14, 2010 18:15
Replacing mesh while running a simulation akultane CFX 1 November 15, 2009 14:46
3-D Contaminant Dispersal Simulation Apple L S Chan Main CFD Forum 1 December 23, 1998 11:06


All times are GMT -4. The time now is 11:02.