CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   how to improve the convergence of chemical reac.? (

lily April 12, 2006 07:15

how to improve the convergence of chemical reac.?
I have 14 species and 31 related reactions. Can anyone tell me if there is really chance that i would get an converged result? Can anyone give me some suggestions that i can follow to improve the convergence? Thanks in advance.

Allan Walsh April 12, 2006 13:32

Re: how to improve the convergence of chemical rea
It may work. I'm modeling 12 to 15 species, with 10 to 12 separate reactions and convergence is not a problem. I'm modeling combustion cases with a modest number of cells (1/2 million) and enthalpy and radiation are the last to converge (and first to diverge).

I never bother with decreasing under-relaxation to less than 1 for species.

lily April 12, 2006 22:42

Re: how to improve the convergence of chemical rea
thank you for your reply, allan. but now i am facing the problem of divergence already with only one reaction activated. can you give me some tips that how did you make it converge? thanks.

AkA April 13, 2006 00:16

Re: how to improve the convergence of chemical rea
can u specify the type of model u have selected for ur reactions?

before activating the reaction(if it is combustion) try to run the case in cold flow , once this is converged activate the reaction.

Allan Walsh April 13, 2006 16:47

Re: how to improve the convergence of chemical rea
lily - not sure about your problem, but we are modeling combustion in industrial boilers. So, our velocities are relatively low and pressure gradients small, but temperatures above 2000K exist. Divergence of multiple finite-rate/eddy dissipation reactions has never been a problem. Our grids our coarse - say 0.1 m - but the gradients in species concentration can be large.

We use the basic k-epsilon model for turbulence - are you using something more complicated? Does your case converge without the reactions? Or is heat generated so that the solution is trivial without the reaction?

You could set up a box or small 2-d case to check your reactions. In order to test sub-models, we have looked at cases with half-a-dozen species and a few reactions, to confirm that the finite rate/eddy dissipation model in Fluent matches up with laboratory data.

Good luck.

lily April 13, 2006 17:10

Thanks. Allan, my case is like this
the very opposite to your case, my domain is very very small, a small nozzle, the grid is like 1e-7m, i have very high velocity and also very high temperature, the gas so is compressible. every gradient is huge, T, P. I didn't do detailed calculation to see how big the reaction would affect the whole calculation, but it's for sure not very convincing without it, the initial reactant has a mass ratio of 38% in total, and to break the bonds need energy. if the temperature field got affected by the reaction, so will the momentums.

lily April 13, 2006 17:15

thanks, AkA, i use finite-rate
thanks, aka. i want to pull it out by modeling Species Transport and Finite-Rate Chemistry. I started from modeling only all the species transport and it converged well. but as soon as i added a reaction, it went wild.

murdock88 March 15, 2013 12:41

Do not forget to include radiation at combustion, because 500^4 small with respect to convective heat flux, but 1800^4 is high enough to decrease temperature of flame zones.

All times are GMT -4. The time now is 07:33.