I'm using the segregated solver to simulate a laminar flat flame in a straight tube. With SIMPLE for pressure-velocity coupling, ideal gas properties and single step chemistry. I'm using a velocity inlet BC and a pressure outlet BC (at atmospheric pressure). The inlet velocity is low at ~3 cm/s. The time step selected is of the order of 10^-7. I can see a wrinkled flame forming after 2/3 time steps (total number of iterations per time step is 2000). The pressure distribution and the velocity profiles look to be okay. Within 2/3 more time steps the solution messes up and I get complete flow reversal. I'm using 0.1 as the relaxation factor for both energy and density, which I increase to 0.2 after 2/3 time steps. For the pressure and velocity relaxation factors I use 0.6 & 0.4 (sum total of 1, as SIMPLE requires). The flame (shown by the high reaction rate zone) dissapears but the reactants-products interface remains. The temperatures remain high of the order of 2500 K.

1. Should I slowly increase the energy and density relaxation factors ? Increasing them abruptly to 1 makes the solution diverge.

2. Should there be any change made to the pressure and velocity relaxation factors ?

3. Grid adaptation in the high reaction rate zones might be another way out.

I must mention here that I'm running an adiabatic case. Does this result in the dissapearance ?

Any suggestions would be of help. I should also mention that the coupled solver failed to give me any kind of a solution with fast divergence and high temperatures (of the order of 5000 K !).

Thanks

Hi,

I think that the coupled solver is the right one being this case a low Mach number flow (uncompressible). Some questions for you: inlet reactants temperature? inlet reatants composition? falme disappears means that reaction rates go to zero?

Let me know.

Cheers

Maurizio

>I think that the coupled solver is the right one being >this case a low Mach number flow (uncompressible). >Some questions for you: inlet reactants temperature? >inlet reatants composition? falme disappears means >that reaction rates go to zero?

The coupled solver doesn't work for me because the residuals refuse to go down, which is because of a fine mesh that I have near and in the honeycomb structure which is present to stabilize the flame. I can not possibly reduce the grid size further, so using the coupled solver is out of question.

Inlet premix temperature = 300 K. Somebody had suggested earlier that keeping a higher inlet temperature inititally and reducing it later would help. I haven't tried that out yet. What do you think ?

Inlet reactants are CH4 and air, stoichiometric mixture.

Flame dissapears means that the reaction rate actually flls to ridiculously low values. It doesn't become zero for me though.

Thanks for the reply

Prateep

(1). Talk to the support engineer to see whether the code can handle the pre-mixed case or not. (2). Or you can try to add a piece of mixing tube, and compute the whole flow field. The best place to ask is the support engineer.

Dr Chien:

Thanks for the reply. I'd like to ask another question related to my computation. I've been successful in getting the unsteady-segregated solver to work. The question is about the time step. I'm using a time step of 10^-6 s. Can I increase it further to lets say 10^-3 s. Of course, the chemical time scales are of the order of 10^-5 s, so capturing the physics would not be possible with higher time steps.

Thanks

Prateep

(1). I normally use the number you mentioned to start the calculation using a different code for non-reacting flows. (2). I had to increase the time step gradually, otherwise the calculation will diverge. (3). I think, the best way to find out is to actually try it. I think, it depends on the problem you are trying to solve and the initial flow field. In my case with the code I am using, it is hard to find a general guideline for convergence. (4). Sometimes, I can use larger time step to force it through the transient phase (non-consistent with the initial flow field), but in most cases, it is safer to increase the time step gradually. (5). As long as the approach is repeatable, you should keep it. If you have time, run some parametric cases.