Learning Forum

[J.LoongHee]

What is best practice in monitoring Combustion & Heat Transfer Model?

Hi all,

I have a few questions to ask, now I am doing combustion modelling in which I have activated (species transport, radiation, energy equation and of course the standard turbulence model to properly model the mixing processes). 

Geometry: furnace with burner and outlet (3D, one fuel inlet, one air inlet and an outlet). The real question is how do I properly monitor the steady-state solution / final converged solution of the combustion model (in the sense of combustion and heat transfer)?

I have created a plane across the burner + furnace where I have a monitoring probe that measures the velocity magnitude variable area-weighted average over iteration. That is straightfoward and I can see that the velocity in the enclosed space eventually reaches a steady-state final value with minimal oscillation. 

The tricky question is the temperature / combustion modelling / radiation model (or the thermal conduction of 1 layer material at the boundaries) where from my simulation I think they are lagging behind velocity by some margin in terms of convergence. I have used a low URF for energy & radiation. I can't afford to use high default URF which will result in divergence. 

Because I was told that the total sensible heat transfer (i.e. NET) needs to be as low as 2%. And everytime I am changing radiation model (either P1 or Discrete Ordinates or changing URF by 0.5 to 0.55 or 0.65), the whole total sensible heat transfer at the report changes by several KiloWatt / significantly.

Could someone guide me if I should keep the settings all constant and run on HPC for days, will it eventually goes down to close to several hundred watt out of say 60kW? I was trying the approach of tunning the URF by 0.05 every 1000 iteration but that has significant impact on the result which means it will be forever not reaching a converge final solution even after I iterate for 10000 iterations.

Sorry, hope my question make sense if not I should reword in a right way. 

Thanks
JL

 

[Rob]

If you drop energy URF from default (usually 1.0 but 0.75 for PBCS) you may find you need a LOT more iterations to reach the solution. It's OK to decrease the URF to get a model going, but you need to get back to the default by the end of the run. 

Re radiation & combustion models. These depend on what you're modelling. DO will cover most of the P1 capabilities, but P1 is no where near as comprehensive as DO. You may find S2S or Monte Carlo are better suited to your needs: again you MUST understand their use and limitations. The User's Guide and Theory Manual are fairly good in this respect. 

Area averaged values are good to get an idea on convergence, but will also mask moving jets and the like: ie the mean is always more-or-less the same but the jet wanders around. Use points and save images every some iterations to see what's going on. 

Have a look through the tutorials and videos in the Fluent Help system.