CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   OpenFOAM (http://www.cfd-online.com/Forums/openfoam/)
-   -   (chtMultiRegionFoam) reducing gravity increase the simulation time ! (http://www.cfd-online.com/Forums/openfoam/72488-chtmultiregionfoam-reducing-gravity-increase-simulation-time.html)

 openfoam1 February 8, 2010 02:00

(chtMultiRegionFoam) reducing gravity increase the simulation time !

Hi foamers;

when i simulate a simple conjugate heat transfer problem using chtMultiRegionFoam solver and i don't want any buoyancy effects appear (plug flow) ,, i changed the g file from (0 0 -9.81) to (0 0 0) , when i did this the simulation time increase horribly

i don't know what is the reason why g effects the simulation greatly ,, and i tried it with other simulation and also the tutorial case and did the same thing (great increase in the simulation time for every step ! )

any one can help

thanks very much ..

 openfoam1 February 9, 2010 17:10

can any one help me to get red of that bad behavior of chtMultiRegionFoam solver
(it may be a bug)

 Bene85 February 12, 2010 03:42

effects of g=0

Hi openofam1, hi all

trying to run a heat transfer simulation with chtMultiRegionFoam takes very long till I reach steady conditions (if I reach it!?). Then I read your post and tried to run the simulation faster without any gravity.
Analysing the written log.file I had to realize that there is nearly no temperature change but a significant change in execution time as you already mentioned.

Here you can see the difference:
Temperature at the beginning: 298,15K
As a heat source I use q=606[W/(m*m)] in changeDictionaryDict. According to the used area it matches P=1,5W

without g:
Solving for solid region cube
DICPCG: Solving for T, Initial residual = 0.009375165, Final residual = 8.318404e-07, No Iterations 24
DICPCG: Solving for T, Initial residual = 0.001351302, Final residual = 7.010276e-07, No Iterations 17
Min/max T:min(T) [0 0 0 1 0 0 0] 298.2413 max(T) [0 0 0 1 0 0 0] 298.3844

ExecutionTime = 1849.95 s ClockTime = 2964 s

Region: domain0 Courant Number mean: 3.232212e-05 max: 0.2909719
deltaT = 1.428571
Time = 150.1

Moreover after 2522sec it only reaches max(T) [0 0 0 1 0 0 0] 299.2234

with g:

Solving for solid region cube
DICPCG: Solving for T, Initial residual = 0.00038313, Final residual = 3.640629e-08, No Iterations 2
DICPCG: Solving for T, Initial residual = 2.75657e-06, Final residual = 4.424209e-08, No Iterations 1
Min/max T:min(T) [0 0 0 1 0 0 0] 299.3279 max(T) [0 0 0 1 0 0 0] 299.4493

ExecutionTime = 27342.29 s ClockTime = 40150 s

Region: domain0 Courant Number mean: 0.001197062 max: 0.2999681
deltaT = 0.002828811
Time = 150.445

-> more than 1K difference and nearly 15 times faster!

In my Opinion without g the convection is not considered and thats the reason for the lower executionTime

Does anybody know whether it is possible to run the case faster and how??
I can't imagine that it takes that long but I do not know what's to change!?

overall the domain0 has 92520 cells
cube has 15424 cells
plate 1540 cells

but I think this number is necessary because a coarsen mesh has bad effects on the stl-files.

If anybody knows a solution I would be very grateful!

best regards
Bene

 kawuppdich February 12, 2010 08:55

Hi Bene85,
how did you use the heat source flux? What have i to type in the changeDictionaryDict file to use it?
It would be very helpful for me.

 Bene85 March 6, 2010 04:44

Hi kawuppdich,

sorry for the late answer, but I have been quite busy.
Using the heat-flux I used the following in the changeDictionaryDict:

cube_to_plate
{
type solidWallHeatFluxTemperature;
K K;
q uniform 6302; // the value depends on Power and face: here e.g. P=15,6W and face=0,05m x 0,04951m -> q=6302W/(m*m)
value uniform 298.15;
}
}This entry has to be made in both changeDictionaryDicts of cube (cube_to_plate) and plate (plate_to_cube) to specify both sides, but one has to be 0 that the heat flux only takes place in one direction. I hope you understood what I meant.

 kawuppdich March 10, 2010 05:06

thanks for reply, it works fine

 Bene85 March 10, 2010 07:40

Hi kawuppdich and OpenFOAMers,

What can you say about your ExecutionTime? Does it take very long? Do you run your simulation in parallel?

I have another question. Maybe you can help me!?
OpenFOAM has libraries for so called RAS turbulence models. Do you know whether RAS is conform to the abbreviation RANS?

Would be very grateful for any help, but it's quite urgent.

Thanks

Bene85

 kawuppdich March 10, 2010 10:41

in my case there is no change of calculation time.
I run my testcase in parallel on 2 cpus

RAS stands for Reynold Averaged Stress and RANS for Reynold Averaged Navier Stokes and thats all what i know about it. I´m just a beginner in cfd so sorry.

greetings

 All times are GMT -4. The time now is 05:36.