Flame Lenght Overpredicted - Methane / Air Combustion
 

Hello every one,

I hope someone could help me.

I´m trying to simulate a Laminar Difusive Non-premixed Methane/Air Combustion in CFX.

The aim of this project is compare the results for the flame lenght varying the velocity of the flow with those experimental results encountered by Burke-Shunnemam and Rope.

The point is that the flame obtained is aproximatelly 3x bigger than the experimental data, and I´m having difficulty to discover what is the problem.


> Is a laminar flow (v=0,1m/s, T= 298K) in a cilindrical domain. Cilindrical methane inlet at the centeline, and a coflow air inlet;
> Finite Rate model is used in a 2 Steps reactions (oxidation of methane and oxidation of monoxide); NASA format;
> Radiation is desconsidered;
> Expert parameter Stiff Chemistry is used;
> Initial Domain is a 1400K with a small mass fraction of the elements.


I will be greatfull if someone could gave some advices, to solve this issue.

Hello,

Why are you not considering radiation ?
Flames tend to emit a lot of energy by radiation. The heat release by radiation decrease the flame length.

Regards,
Possa

Have you covered the basics? http://www.cfd-online.com/Wiki/Ansys..._inaccurate.3F

Hello Possa

Yes, your right the radiation decrease the flame length;
And my next step in the project is consider radiation to make a comparison.

In the most cases the maximum temperature, with radiation, decrease 150K.
I can be complete wrong, but I can't understand that involving radiation will change the flame to 1/3 of its length.

It should be a very simple simulation, in a very refined structured mesh, all the physic suppose to be right, convergence obtained is in order of 10^-4 or 10^-5.

Hello Cristian,

You're right. Considering radiation won't decrease flame length to 1/3.
As I see it, flame length is not easy to accurate predict in CFD due to the simplified combustion models. As far as I know you would need a higher steps reaction mechanism to reach a good result when compared to experimental data.
So, I believe you will only reach a result near your real flame length in CFX if you use more accurate models for both combustion and radiation. For CFX, 3 times bigger is not that bad.
For flame length and other flames variables in simple reactor problems the software CHEMKIN can solve mechanism with more steps. But then you will only be able to solve reactor types of problems, and not 3D CFD.

If you intend to really use CFX for that purpose I recommend two things:

1- Reach a better convergence. For complex problems where you are interested in a variable like flame length I would recommed a convergence of MAX error of 10E-5 at least (not RMS error).
2 - More complex models for combustion ( I don't know how far CFX goes in combustion models), and to take into account all the physics of your problem.

Maybe that can help you reach a better result, but as I said before keep in mind that for combustion problems CFX does not gives good results for combustion specific variables (Due to the simplified combustion models). It gives good results for global problems involving combustion system.

Regards,
Possa.

Oh I forgot...

you can always try to refine your mesh and do a sensitivity analysis.

Regards,
Possa.

You get sick of writing this after a while. That's why I just refer people to the FAQ. Feel free to update the FAQ if you have some hints or tricks of your own.

Assuming the basics are working properly (you have done convergence and mesh sensitivity studies and everything else in the FAQ) then I would consider the flame model you are using. The flamelet libraries from the old CFX-RIF which are now part of the core solver should help accuracy as they include more accurate models of complex combustion reactions.

Hello Cristian,
i suggest you to check the overall diffusivity of your reactants. You have laminar flow, turbulence has no effect on mixing hence your reactions are influenced by pure diffusion. And since you have only 2-step reaction mechanism, this may have big impact on your results as you are neglecting many intermediate species.

regards,

Peter