[Esamor]

Hi,

I try to model an axial-symmetric turbulent Bunsen flames as described in this dissertation http://multiphase.asu.edu/paper/itm_2001.pdf Page 110ff. Therefore i am initializing a bunsen burner test case with two inlets. One smaller main inlet 1 with unburned methane-air (G=-1, T=300K, velocity = 65m/s) and another support inlet 2 with burned methane-air (G = 1, T = 2000K, velocity = 1,5 m/s).

-I am using the k-epsilon model for the turbulences.

-For the Reaction i use the implemented Methane Air Flamelet Library, Single Step.

-So far i successfully simulated a similar test case (Slot Burner), therefore i know my boundary conditions are not completely bad.

-I am simulating steady state for initial values of the transient simulation. My aim is not a converged solution, but a initial mean flame front for a faster solution, as described in the dissertation.
Question:
- Is there any other way to initiate a flame front in CFX, not using the spark ignition model? (but that's not my real problem)


My real problem:
- As soon as i switch from steady state simulation to transient simulation the pressure near the flame front drastically rises. This accelerates the Fluid to Mach 3 and more, which is completely unrealistic. I expect maximum values of around 100m/s.
- Also i observe random flame arising at Inlet 2

Here some pictures:
-Picture of steady state before transient simulation (Temperature). Not converged, but good enough so far
https://ibb.co/cDvxCz

-Picture of relative pressure. Too high at inlet1
https://ibb.co/jnCNee

-Picture of random unburned pockets at inlet 2 (G-Variance is set to zero)
https://ibb.co/mjUPsz

-Picture of velocity, especially to high at inlet (rises with simulation duration to over 1000m/s)
https://ibb.co/nKhAXz

I read about time dependent equations which are turned on in transient mode. Is there a documentation which equations are these exactly? Maybe some important pressure equations are not used for combustion simulation in steady mode?

My following solutions did not work:
- changing Inlet boundary condition from "normal speed" to "static pressure" or "mass flow rate"
- reducing time step
- reducing mesh length at inlet
- clip pressure minimum in advanced solver control (i get negative relative pressure of 1 bar and positive relative pressure of 20 bar and more)
- Tried Upwind and High resolution advection scheme
- tried first order and second order euler transient scheme
- reduced turbulence numerics from high resolutions to first order


I hope anyone has seen similar things or can explain these effects.

Greetings

[Esamor]

Hey,

Since i get no answers maybe i have to formulate my problem differently:

1. I model two flows of burnt methaine-air in inlet 1 with T = 300K and in inlet 2 burnt methaine-air with T = 1400K, so there is no combustion, but i can see the mixing of the two different inlets, because of the temperature mixing. I use the steady state solver and let it converge to RMS 1e-5. No problem, realitic solution. See pictureTemperature Mixing no combustion.jpg

2. I "turn on" combustion by setting inlet 1 to unburnt --> G = -1 and i use the solution of the first simulation for intitial values.
Then i solve two timesteps of 1e-6 seconds. I can see the (green) flame front of G = 0 and get realistic pressures. Lets have a look on the pressure and zoom in: relative pressure steady .jpg
Flame front in green, pressure varies a little, but nothing abnormal. But i need a transient solution, because of the expected flame front fluctuation.

3. I turn on transient mode, still have inlet 1 at G = -1 and solve two timesteps of 1e-6, still use the first simulation as initial values
Results:
The relative pressure drops around -0.5 bar in the whole domain and rises more than 0.5 bar at inlet 1 --> accelerates fluid over Ma = 1 and no convergence, solver ends with error, when modeling more timesteps. The absolute pressure is 1 bar when initializing and should not vary more than 1000-2000 Pascal. This is also important because the used flamelet library only has informations for p=1 bar and the more i turn away from these values, the more i get a wrong solution.
Here a picture
pressure transient.jpg

Why varies the transient solution that much after just 2e-6 seconds? What is the difference in the used equations? And how can i work around this problem?

Have a nice day

[ghorrocks]

Combustion modelling is highly numerically unstable due to the rapid changes in properties and temperature across the flame front. So you should expect difficulties in convergence.

To counter this, first do the basics for numerical stability: make sure your mesh quality is as good as possible, double precision numerics and the best initial condition possible.

I would also look at the CFX tutorials for how they do combustion modelling. Also ask ANSYS support as they have many more examples and that will help a lot I suspect.