[rewol]

Hello everyone,

I am having a trouble initiating combustion in my case. I will try to give as much detail about the setup and mesh as possible.

Case Setup :
-Viscous, Laminar Flow
- DENSITY-BASED Solver !
- Species Transport Model, Methane-Air 1 step Mixture, (Non-premixed)
- Axisymmetric geometry, where fuel and Oxidizer enter the domain from different inlets @300K
- High Pressure Combustion Chamber ( Operating Pressure is 5 bar), so Real-Gas model is used for density.
- Both inlets are mass-flow-inlet, outlet is pressure outlet, a symetty axis and wall on top with no heat flux.

PROBLEM :

* First, i simulate a cold flow that converges relatively well, as instructed in User Manual. ( Steady State, Pseudo Transient, Second Order Upwind, NO REACTION)
* I patch the temperature in an area, as shown in third picture.
* I turn on volumetric reactions, use stiff chemistry solver, and start the simulation but it diverges after 100-200 iterations where resiuals go sky-up at first (understandable, due to high gradients and density change) but does not converge. (Again, second order upwind and ROE scheme, pseudo transient)

ATTEMPTS :
* I tried nearly all possible patching areas for temp even though it shouldnt matter in steady-state in theory. No luck there.

* I have used coarser mesh, more dense mesh, all uniform mesh but no luck there.

* Instead of patch, i stopped the solution in cold flow and turned on REACTIONS and change the OXIDEZER BC to blow hot air (2000K) in order to initiate combustion and slowly lowered it back to 300 K, no luck there.

* Tried with / without inlet / thermal diffusion, no luck there.

* First order upwind scheme does not yield any better results either.

* PATCHING TEPERATURE at inlets causes reversed flow because density and momentum balance goes crazy, making it really hard for velocity field to converge, in my opinion.

Basically at this point, i would take any suggestions since i ran out of opinions.

I would appriciate any tips / guides to achieve convergence.

Regards.


PIC_1_COLD_FLOW_SIMULATION_RESULTS.jpg

PIC_2_MESHING_AT_INLET.jpg

PIC_3_PATCHING_AREA_FOR_TEMPERATURE.PNG

PIC_4_RESIDUAL_BEHAVIOUR_AFTER_REACTIONS ARE ON.jpg

PIC_5_CASE_SETUP.PNG

[rewol]

OK, here is the updated information. I was able to progress a bit more :

Changes :

* I did not like the mesh at fuel enterance so i edited it a little bit. Cells are more "square like" at fuel inlet and i canceled mesh bias. Picture is added. Also added the geometry too.

* Switched back to full steady-state (no pseudo transient !) with Courant Number set as 5.

* No cold flow, straight to temperature patch as shown in upper post and REACTIONS ON and calculation.

Results :

- With this mesh and setup , flow field and temperature converged (order of 1e-6) but species are on the order of 1e-1, which is not nice. I added the temperature field though, behaviour is very much in agreement with results with academical paper(s).

- Combustion product species residuals are very high ! In attachment, results for first 1500 iterations can be found. There is a great oscillation at first. Any comments or suggestions on this ?

- I tried pseudo with same initialization after steady-state. Flow field converges the same with SS but residuals for species are a bit worse than steady-state.

Question: How do i make the residuals for species smaller and get rid of oscillations ? Would playing with the Courant number help ?

In my previous experiences, increasing the courant number would make it more unstable, i reckon.

Looking forward to all the help and guidance i can get.

Thanks !

PIC_6_GEOMETRY.PNG

PIC_7_REFINED_MESH_AT_FUEL_INLET.jpg

PIC_8_solution_after_1500_iteration_and_residual_behaviour.jpg

PIC_9_Converged_Temperature_Field.jpg

[rewol]

Hello,

Attempts & Results :

- Reducing the COURANT number and increasing number of iteration solves the oscillation in residuals but does not yield a better convergence, as expected.

- Flow field, energy converges but "combustion products" do not. Allthough the final look seems logical, prodcut species just stuck on 1e1 order residuals.

- Combustion occurs in both real-gas and ideal gas, with the same behaviour. ( Thermodynamic property values are different of course, as expected)

What can i try to lower the species convergence ?

Since the equivalence ratio at the inlet can be decided from fuel and ox inlets, i calculated the "lean combustion product mole fractions" analytically and place the results on the outlet. Didn't work.

Thanks in advance.

Alper

[rewol]

OK, i solved the problem and solution is converged finally.

Problem was the initial conditions. In reacting flows (or combusting flows), species field would have to be initialized very accurately if a " steady-state " approach is employed. Closer the initial solution to final solution, better the convergence.

I did not use previously completed simulation results as IC's, just guessed better values for species during initialization and it worked. But using previous solution as IC would probably yield better results.

Best Regards.