CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   Phoenics (https://www.cfd-online.com/Forums/phoenics/)
-   -   CONVERGANCE IN IMMERSOL RADIATION MODEL (https://www.cfd-online.com/Forums/phoenics/51990-convergance-immersol-radiation-model.html)

anjani January 3, 2006 02:03

CONVERGANCE IN IMMERSOL RADIATION MODEL
 
HI

i m trying to simulate the heating cycles of an anealing furnace, the problem is that when i m using immersol model for radiation the results are not converging (% error oscillating around E05) but when this immersol model is removed the convergance is met.

so anyone please help me to decide the right convegance parameters like dtfals or any other way to meet the convergance.

regards

Mike January 3, 2006 05:39

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
Try linear under-relaxation factors of say 0.3 on both TEM1 and T3. Also set VARMIN and VARMAX on TEM1 and T3 so as to prevent non-physial values occuring during the course of the simulation.

mick January 6, 2006 18:15

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
It could well be that your model set up is not correct. Remove Temperature and immersol and make sure that you can get a converged solution for flow only first!


anjani January 7, 2006 02:07

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
thanks mick for ur response.

but the problem is that ,i am having only the time temperature characteristics with physical model details and properties like pressure and velocity inside the domain are not known . Thus i m trying to get the temperature charteristics with the help of inlet volume flow rate and velocity profile . so if i simulate the problem for flow only, i can only judge this with desired flow profile ,which i m able to get in my current simulation . The possible reasons for large error may be i was using a pressure relief object in place of outlet to the domain , this arrangement may be causing mass accumulation and large pressure inside the domain. now i m again using outlets to the domain in place of pressure relief object and pressure is within the desired limits.

mick January 9, 2006 09:14

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
Yes, if you do not have an outlet your vessel will become pressurised and it could become challemging to converge.

anjani January 13, 2006 02:55

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
hi mick

As you suggested , to solve the case for steady state case only without any heat source and provison for oultet ( to avoid mass accumulation), I am trying to solve the case with follwing differences from previous case considered:-

1. only presuure and velocity is being solved

2. heat sources and radiation model are removed.

4. The outlet is made and placed properly in the model.

5. Grids are arranged more suitably.

But still with above improvements , the error is of the order of E04 (10000).

The chief objetive of the case being studied is to get the temperature build up within the blockage object (with maximum and minimum temperature location) in course of time.

Therefore ,please suugest what should i try to meet the convergance in the steady state flow simulation case. i m trying with -.03 DTFALS values for pressure.

i can mail u the q1 file ,if i may get ur email address.

Regards & Thanks in advance

Anjani Kr Dubey


Mick February 16, 2006 04:44

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
Sorry, I am too busy with consultancy and paying customers, but maybe someone else can take a look, any volumteers out there come on!

1/ Check the grid have equal spacings where possible, especially if axisymmetric. For example you can't just have 3 cells in the axisymmetric for the circumferential direction X if thats the case.

Also carefully align the objects are they really inj the correct position, Zoom in and check the position and size.

Thin/small objects may well cause convergence difficulties.

2/ If you have very thin objects this will give cells with extreme aspect ratio

3/ try CONWIZ the automatic relaxation convergence setting, but set the increments to be small say 1.0 or 0.1 rather then 100 for velocities.

4/ Try and get mass balance if you have this and errors are O(E+4), then things may not be as bad as you think as these errors are calculated in a relative type way.

5/ If spot values are not changing and residuals have come down a couple orders of magnitude


Mick February 16, 2006 05:25

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
Apologies my answer was a bit stunted: Also I am sorry that I do not have time to examine your Q1 at the moment.

1/ Check the grid have equal spacings where possible, especially if axisymmetric. For example you can't just have 3 cells in the axisymmetric for the circumferential direction X for 360 degrees if thats the case. Also you will need xcycle=T if 360 deg circumferential and polar.

5/ If spot values are not changing and residuals have come down a couple orders of magnitude then you may have a reasonable solution as the errors are divided by a RESREF value that may be extremely small.


Anjani Kr. dubey February 17, 2006 02:15

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
Thanks mick for your suggestions.

pl. tell me where to set xcycle=T in the q1 file as i am using 360 deg polar cylidrcal co-ordinate system and objects.

Also pl suggest me how to get the time taken by a trasient heat trasnfer to become steady in phoenics i.e how can i know the steady state settling time in phoenics .

Mick February 21, 2006 08:45

Re: CONVERGANCE IN IMMERSOL RADIATION MODEL
 
XCYCLE can be set through -> VRMenu-> Sources -> Cyclic Boundary Conditions.

The transition time to steady state conditions is of course problem dependent. If you have some physical idea of what this might be then use this as a first-stab bounding condition. I dont think this is something that you can calculate easily unless the problem is quite simple. I suspect people generally use intuition on this or alternatively dont know, so it may be a question that the CFD is being used to answer.

What a CFD code will require is a descretisation of that time into suitable time steps. If inexperienced I guess the best way to proceed might be to make a simpler copy of your case (If slow to run) to experiment on (keep the basic design) but have a coarser cell distribution (for turnover speed). Then try this discretisation on the original q1.

Otherwise just experiment with your proper case.

Remember (generally speaking) that there is something called the CFL criterion that you might have to keep your eye on to a more or less extent depending on the type of set-up you are using.

But I think for your ConjHeatTrans case then this is not raelly so important and you should be able to get away with quite coarse time intervals.

last word is that you may wish to expand the time intervals over time geometrically, so that the time steps grow over time. This can be done with the POWER LAW button and can help towards reaching steady-state for obvious reasons.

good luck


All times are GMT -4. The time now is 00:46.