Thanks for that detailed explanation Dan. Much appreciated.

In my simulations I'm solving for steady gas phase combustion and then using these results to update solid phase consumption, using a partially decoupled algorithm thru user scalars inside Fluent. So this method may be difficult to incorporate within this framework.

I have recently noted that in at least one past model they have found convergence difficulties too, and simply omitted the reaction, chosing to include part of the effects in the gas-solid reaction. For the time being I'm taking this approach and will have to evaluate how good an approximation it is.

I really appreciate your thorough explanation Dan.

Thanks

Greg

Great thread guys!

I would add that LSODE has been superseded by VODE, so if you don't know how to use either, you should use VODE, not LSODE.

I don't think it's a good idea to try to use a commercial CFD code to solve laminar premixed flames, as of now commercial codes are unsuitable for stiff chemistry, and you really need good mesh adaption to solve these sorts of problems. My dissertation was on the unsteady simulation of combustion in porous media, and I wrote my own code using VODE and chemkin to solve the problem. Now I use Fluent for non-premixed turbulent combustion and there's no way I'd try to do a laminar flame with Fluent.

Star-CD is in the process of implementing a stiff solver in their code. You'd still have to trick it out to do the solid-gas heat transfer, but at least it should have the numerical capability to solve stiff equations. If you try stiff chemistry in a non-stiff solver, like Fluent, you'll get unpredictably wrong results. You'd also have to figure out how to do the radiation and dispersion effects. That'd be tough, but maybe possible.

As to which code is "best" for combustion, I'd say Fluent if you limit the discussion to industrial combustion (I don't think laminar flames in porous media is industrial combustion). This conclusion would be just based on a feature list, not experience (I only have experience with Fluent). Fluent supports unstructured meshes and parallel calculation for most if not all combustion models. It has both equilibrium and flamelet pdf models. It has DO radiation with WSGGM, and has had it for several years, so it's mature. It is starting to support LES for reacting flows. Fluent has recently added a Zimont model for premixed turbulent flows.

CFX and Star are probably both a couple years behind Fluent in combustion models. I have a friend at CFX who is trying to get their combustion models ported to an unstructured solver. Industrial applications generally need unstructured meshes, so I have trouble being objective about CFX. It just doesn't contend because it can't do the geometries I want to simulate. We're going to buy Star at work soon because they are definitely moving in the right direction. I'm very anxious to find out where they are now.

My opinions only, YMMV Mike

Hi Mike, where can I find more about your implementation?, How did you solved the advection-diffusion problem? As I know LSODE or VODE and Chemkin only solves the chemical reaction. Thanks

I used a method of lines algorithm to solve advection-diffusion. You're correct that chemkin doesn't do that for you. I added a P-3 radiation solver also but that calculation was lagged, so not really solved using method of lines.

If you want to learn about my implementation, there's an article in Combustion and Flame, Vol. 117, No. 4, 1999 (I don't have the page numbers). The gory details are in my dissertation which you can get from UMI (http://www.umi.com/hp/Products/DisExpress.html) Their number for it is 9837980 and the title is "Simulation of transient combustion within porous inert media." Or e-mail me and I'll send try to find a postscript file to send you.

Hello everybody !

After an year's work trying to simulate laminar premixed combustion in a Rijke tube using Fluent 5.1 to 5.5 ....... we (the group) has decided to abandon using commercial codes ...... we need to work with turbulent high swirl rigs now - both premixed gaseous and liquid sprays. And the kind of difficulties I have had with getting a decent solution of the laminar problem using Fluent, it gives me the shivers to think what will happen if we continue using it !

I've been able to get a solution for phi=1 case and am simulating a phi=0.6 case nowadays. Real low time-step (1e-5 s) even with 8 processors on an SGI Origin2000 resulted in high computation time. The phi=1 case has been able to predict a lot of the characteristics of the flat flame and the match with experimental results is pretty good, of course, the limit cycle frequency is off by few Hz !

Prateep

(1). As I have said before, a commercial code is a commercial code. That's all. (2). So, don't expect too much from a black box, unless it matches your need exactly.

Consider checking CFD-ACE+, it has probably best features for gaseous and surface reactions, plus interface with Sandia's CHEMKIN, thus can read chemkin files directly and use chemkin database. Stiff solver is available. Besides it handles many plasma reaction applications with very fast reactions.

Some experience that i had with this class of problems suggests need for very large linear relaxation, possibly down to 0.01 level. Startup solutions do become critical, as well as rates manipulation (ie slower rate first and then slowly increase), it can make a difference how you initiate the flame and grid resolution you utilize. Very fast reactions require (by physics) very small timesteps and in steady-state formulation this leads to relaxation. Assume you have rate of 10E10 for one reaction and 10E5 to another. So if you use relaxation of 0.9 for 10E5 it is logical to assume that you need relaxation of 1E-5 for another, just to be physically consistent.

I am interested to see what happens (with FLUENT) if you go for 0.01 to 0.001 linear relaxation coefficients for T, species, viscosity and other scalars (P etc).

I would not try to generalize about commercial codes applicability for laminar combustion and pre-mixed flames, but for very complex physics there are no UNIVERSAL BLACK-BOX solutions. Unfortunately, when using commercial packages people often utilize fire-and-forget approach instead of think-think-think.

vvk@cfdcanada.com