[S2.Mike]

Hello FOAMers,

I asked the same question in Running and Solving forum but I'm not sure if it was the right forum for my thread. I'm new in OpenFoam and i have a big problem.
I'm writing my bachelor thesis in a company which want to use OpenFOAM instead of Fluent in future. It works fine, but OpenFoam is much slower than Fluent to find the solution. The calculation is running on a 64 core cluster and I'm using chtMultiRegionSimpleFoam.

So i decided to implement a fuzzy algorithm control for the relaxation factors to get faster convergence and more stability. For the fuzzy algorithm it is necessary to write every timestep new relaxationfactors. But now I've a problems. The problem is that OpenFoam don't read the fvSolution every timestep which is necessary for my relax control algorithm. In the global controlDict i use timeStepMaster for fileModificationChecking.

How can I ensure that OpenFoam reads the fvSolution every timestep? Please can anybody help me with my problem?

thanks and best regards, Mike

[olivierG]

hello,

In order to have run time modifiable fvSolution, just put "runTimeModifiable yes;" in your controlDict.

However, if you want to speed up your case with chtMultiRegionSimpleFoam, you might try also to change solver for T (try GAMG if it work), and run first thermoFoam for a good initial T field. And this is weak coupling here, so for faster convergence, try congugateHeatSimpleFoam (from foam-extend) instead.

regards,
olivier

[S2.Mike]

Originally Posted by olivierG

hello,

In order to have run time modifiable fvSolution, just put "runTimeModifiable yes;" in your controlDict.

However, if you want to speed up your case with chtMultiRegionSimpleFoam, you might try also to change solver for T (try GAMG if it work), and run first thermoFoam for a good initial T field. And this is weak coupling here, so for faster convergence, try congugateHeatSimpleFoam (from foam-extend) instead.

regards,
olivier

hello oliver,

thanks for your fast reply. I already have runtimeModifiable yes in my controlDict. OF reads the fvSolutions at runtime, but i think not before every timestep.

My logFile looks like:

Time = 1258


Solving for fluid region Air
DILUPBiCG:  Solving for Ux, Initial residual = 0.001473127, Final residual = 6.388003e-06, No Iterations 1
DILUPBiCG:  Solving for Uy, Initial residual = 0.001077828, Final residual = 4.460612e-06, No Iterations 1
DILUPBiCG:  Solving for Uz, Initial residual = 0.0007397743, Final residual = 3.149015e-06, No Iterations 1
DILUPBiCG:  Solving for h, Initial residual = 0.00128943, Final residual = 8.301581e-05, No Iterations 14
Min/max T:362.9948 649.5947
DICPCG:  Solving for p_rgh, Initial residual = 0.002547044, Final residual = 9.128748e-05, No Iterations 6
time step continuity errors : sum local = 0.0002680669, global = -3.173526e-05, cumulative = -0.01920438
Min/max rho:0.7 0.9575195

Solving for solid region LED_osram108_1b
GAMG:  Solving for h, Initial residual = 0.0002935136, Final residual = 1.034723e-11, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 406.7548 max(T) [0 0 0 1 0 0 0] 422.999

Solving for solid region Print
GAMG:  Solving for h, Initial residual = 0.0002210894, Final residual = 8.670698e-11, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 402.9372 max(T) [0 0 0 1 0 0 0] 407.0176

Solving for solid region Kuehlkoerper
GAMG:  Solving for h, Initial residual = 0.0004534505, Final residual = 1.119512e-10, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 402.3478 max(T) [0 0 0 1 0 0 0] 403.8735

Solving for solid region LED_osram108_1a
GAMG:  Solving for h, Initial residual = 0.0002873201, Final residual = 1.146439e-10, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 403.2385 max(T) [0 0 0 1 0 0 0] 418.1941
ExecutionTime = 647.37 s  ClockTime = 665 s

Time = 1259


Solving for fluid region Air
DILUPBiCG:  Solving for Ux, Initial residual = 0.001474693, Final residual = 6.37829e-06, No Iterations 1
DILUPBiCG:  Solving for Uy, Initial residual = 0.001076896, Final residual = 4.449169e-06, No Iterations 1
DILUPBiCG:  Solving for Uz, Initial residual = 0.0007395095, Final residual = 3.140754e-06, No Iterations 1
DILUPBiCG:  Solving for h, Initial residual = 0.001269364, Final residual = 8.975471e-05, No Iterations 19
Min/max T:362.9948 649.9584
DICPCG:  Solving for p_rgh, Initial residual = 0.002549795, Final residual = 8.721853e-05, No Iterations 12
time step continuity errors : sum local = 0.0002561782, global = -2.981797e-05, cumulative = -0.01923419
Min/max rho:0.7 0.9575195

Solving for solid region LED_osram108_1b
GAMG:  Solving for h, Initial residual = 0.0002932964, Final residual = 1.033519e-11, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 406.7556 max(T) [0 0 0 1 0 0 0] 422.9998

Solving for solid region Print
GAMG:  Solving for h, Initial residual = 0.0002209446, Final residual = 8.658666e-11, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 402.9372 max(T) [0 0 0 1 0 0 0] 407.0184

Solving for solid region Kuehlkoerper
GAMG:  Solving for h, Initial residual = 0.000453174, Final residual = 1.116956e-10, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 402.3475 max(T) [0 0 0 1 0 0 0] 403.874

Solving for solid region LED_osram108_1a
GAMG:  Solving for h, Initial residual = 0.0002870728, Final residual = 1.144746e-10, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 403.2387 max(T) [0 0 0 1 0 0 0] 418.196
ExecutionTime = 647.52 s  ClockTime = 665 s

regIOobject::readIfModified() : 
    Re-reading object fvSolution from file "/home/attrmi/Desktop/Simulation/processor0/../system/Air/fvSolution"
regIOobject::readIfModified() : 
    Re-reading object fvSolution from file "/home/attrmi/Desktop/Simulation/processor0/../system/LED_osram108_1a/fvSolution"
regIOobject::readIfModified() : 
    Re-reading object fvSolution from file "/home/attrmi/Desktop/Simulation/processor0/../system/Print/fvSolution"
regIOobject::readIfModified() : 
    Re-reading object fvSolution from file "/home/attrmi/Desktop/Simulation/processor0/../system/LED_osram108_1b/fvSolution"
regIOobject::readIfModified() : 
    Re-reading object fvSolution from file "/home/attrmi/Desktop/Simulation/processor0/../system/Kuehlkoerper/fvSolution"
Time = 1260

Before Time 1260 OF reads the fvSolutions. But not before Time 1259 for example. Do you know how I can ensure that before every timestep the fvSolutions has been read. I tried with fileModificationChecking inotifyMaster but i got an error.

best regards Mike

[jherb]

If you are running the code on a cluster this might be a problem of the file system. Are you sure, that the fvSolution file has been written (modified) before the next time step started?

[jhoepken]

I would suggest to write a function object that gets the fvSolution dictionary from the mesh (or runTime) and alters it there. Hence you don't have to take that detour over the filesystem.

An other option would be to change the datastructure that actually takes care of storing and returning these relaxation factors, which would be the more elegant way. Look for solutionControl and it's derivatives.

Jens

[S2.Mike]

Originally Posted by jherb

If you are running the code on a cluster this might be a problem of the file system. Are you sure, that the fvSolution file has been written (modified) before the next time step started?

I also tried to run on a 4 core desktop pc in parallel and it didn´t work. I read the fvSolution from a template and overwrite the old fvSolution using python. I checked the new fvSolutions by writing the new files every timestep to txt-files too. I think maybe OpenFOAM don´t get the signal the fvSolution is modified. I tried to change in source code of OF the file IOOutputFilter.H from "MUST_READ_IF_MODIFIED" to "MUST_READ" and it works, but i think thats not a "clean" solution for my problem.

regards, Mike

[S2.Mike]

Originally Posted by jhoepken

I would suggest to write a function object that gets the fvSolution dictionary from the mesh (or runTime) and alters it there. Hence you don't have to take that detour over the filesystem.

An other option would be to change the datastructure that actually takes care of storing and returning these relaxation factors, which would be the more elegant way. Look for solutionControl and it's derivatives.

Jens

Hello Jens,

do you mean something like that http://foam.sourceforge.net/docs/cpp...0.html#details ? Do i have to write the new relaxFactors in the lib.so file?

The second option sounds better but I'm not sure if I have enought know how of the OF structure to do that. I think it takes too much time for me to until i know what's to do. I have to finish my thesis until august otherwise i can't start with my master in september.

regards,
Mike

[sap_92]

Originally Posted by S2.Mike

I also tried to run on a 4 core desktop pc in parallel and it didn´t work. I read the fvSolution from a template and overwrite the old fvSolution using python. I checked the new fvSolutions by writing the new files every timestep to txt-files too. I think maybe OpenFOAM don´t get the signal the fvSolution is modified. I tried to change in source code of OF the file IOOutputFilter.H from "MUST_READ_IF_MODIFIED" to "MUST_READ" and it works, but i think thats not a "clean" solution for my problem.

regards, Mike

I am also doing the similar thing of altering fvSolution dictionary using Python. Currently, I am changing fvSolution dictionary after every 50 time steps. However, I would like to do it in real-time at every time step. How do your python scripts know that the OpenFoam simulation for one-time step has been completed? How are you communicating between Python and OpenFoam?
Thank you.