[Biagio]

Hi, i-m simulating LOx-CH4 Combustion in supercritical condition. I'm using real gas Soave Redlich Kwong equation of state and Eddy Dissipation model dfor the combustion, and i'm struggling with this error: Temperature is below the spinodal point in xxx cells", has anyone ever experienced this? Any solution?

[vinerm]

This should just be a warning and not an error. It may happen if the initialization is not very good. However, if it disappears as the solution progresses, then there is nothing to worry about.

[Biagio]

The problem is that the simulation is going well and the temperature field looks good, but at certain point this message appears and the number of cells in which the temperature is below spinodal point increases till the solution explodes, and i don’t know what to do

[vinerm]

Real gas models require solution steering by user. If the simulation is steady-state, recommended would be to start with first-order schemes and run for at least 1000 iterations. You may have to run for 2000 or more if temperature field is still unstable. Only when it becomes stable, switch to second-order but keep URFs low. Do you have any boundary conditions that are removing the heat from the system or anything that could cause reduction in the pressure? That could bring the system closer to vapor dome, hence, temperature dropping below the spinodal curve. You can also artificially keep a higher pressure in the beginning and slowly bring it down as simulation progresses.

[Biagio]

I'm simulating Lox-CH4 combustion with Methane in supercritical condition (P=5.6 MPa and T=288 K) while LOx is in liquid state(P=5.6 MPa and T=85 K);
My boundary conditions are:
-Mass flow inlet for Methane and LOx inlets, pressure outlet (5.6 Mpa) and stationary walls for walls. k-eps Standard for turbolence and EDM for combustion.
I've tried these method:
-Starting from a solution obtained from non-premixed pdf equilibrium with real gas SRK equation, (in this case first i solved equation for the species, and then enable all the equations);
-Starting from a solution obtained with EDM with ideal gas assumption;
-Starting from a real gas frozen (no reaction) solution;
None of this worked;
I'm using really low URF (order of 0.2 and 0.3).
Do you have any suggestion?

[vinerm]

Since you are using EDM, turbulence model plays a significant role. Though every k- is bad with respect to production of k and over-predicts turbulence, standard performs worst with shear flows that you might have. So, prefer to use either RNG or Realizable with production limiter.

What is missing in the results that you have using Non-premixed model? Or does that not converge as well? Or is it because you have LOx and that is not included in the pdf table?

Are you injecting LOx as Lagrangian particles? Then, you may start with lesser injection and increase the injection rate over iterations to keep energy source smaller.

[Biagio]

The results with the non premixed model converged well. But i'm studying different combustion models for comparing them. So now i need to use EDM and EDC.
For now i'm using an Eulerian approach so i'm injecting Lox as gas, but i'm planning to use the Lagrangian approach later.

[Biagio]

These are the residuals starting form a non-premixed pdf equilibrium SRK and then swithching to an EDM SRK at Iteration 1500.

[Biagio]

Hi, i noticed that i'm getting a strong increase in the pressure at the inlet of LOx, and i don't know why.
My boundary conditions are:
Inlet_LOX: mass flow inlet 0.044 kg/s
Inlet_CH4: Mass flow inlet 0.1431 kg/s
Pressure outlet: 0 Mpa
Operating condition: 5.6 Mpa (That is the working pressure of the chamber).
What could be the cause?

[vinerm]

I suppose you forgot to attach the image.

When you mention Eulerian, do you mean both fluids are gases and you are solving single-phase, multi-species model? That should be easier than when you will use DPM. As far as URF is concerned, 0.2 or even 0.1 could be sometimes high for Real Gas Models. I remember using 0.01 to make it converge though that was with Fluent 6.3. Fluent is certainly better now but 0.2 would still be high with higher order schemes.

[Biagio]

Originally Posted by vinerm

I suppose you forgot to attach the image.

When you mention Eulerian, do you mean both fluids are gases and you are solving single-phase, multi-species model? That should be easier than when you will use DPM. As far as URF is concerned, 0.2 or even 0.1 could be sometimes high for Real Gas Models. I remember using 0.01 to make it converge though that was with Fluent 6.3. Fluent is certainly better now but 0.2 would still be high with higher order schemes.

Yes i'm using single phase models and multi-species. One question, in the operating condition box i have the possibility to set Vapour or Liquid, is it right selecting Liquid for my case (CH4 is supercritical while LOx is at supercritical pressure but subcritical Temperatue = 85 K, so it is in liquid state)?
Surely i'll try 0.01 for the URF's

[vinerm]

If situation is almost always supercritical then that setting has no effect.

[Biagio]

Originally Posted by vinerm

If situation is almost always supercritical then that setting has no effect.

Yes, but O2 is in liquid state at the inlet

[Biagio]

Immagine.png

Immagine1.png

These are the residuals and the pressure at the two inlets and outlet. At Iteration 1500 I switched from A Non-Premixed PDF Equilibrium SRK model to an EDC SRK model.
All my URf's are set to 0.01

[vinerm]

EDC is a rather complicated model but a useful one.

What is p_lox? I assume it is the average pressure at the LOx inlet. Though the system shows divergence, it points towards an increase in the pressure required at the LOx inlet to push it inside. Pressure inside the chamber is common for both oxygen and fuel, so, it should not increase the pressure requirement only at the LOx inlet unless it is density of the LOx incorrectly. That might mean that real gas model for LOx has something wrong; may be a parameter.

[Biagio]

Yes it is the vertex average at the inlet of the oxygen. I'm using an SRK model already implemented in Fluent, and the evaluation of the density is right (1179 kg/m3 for the oxygen at P=5.6 Mpa and T=85 K), so i don't know where the problem could be

[vinerm]

Yeah, you've got a really challenging issue at hand. Did you try other models instead of cubic equation, such as, NIST. But NIST won't allow combustion models; not sure about current versions though.

[Biagio]

I don't think NIST data are available for combustion, and NIST data are only available in a range narrower than the one of interest

[vinerm]

Yes, I never used NIST, myself. Though I remember working with User Defined Real-Gas Model but it was RK and is available in Fluent. I suppose you mentioned that first-order works good. May be you could use higher order only for momentum and energy. Keep first-order for others and try with Linear for pressure. Don't use final results with Linear Pressure but could be used as a stabilizer.

[Biagio]

unfortunately is not working either with first order discretization methods. Using very low Urf's factor (0.005 for all) and switching to second order methods, i was able to get close to convergence and good solution in terms of temperature distribution, but even after 10000 iterations the residual oscillate and never go to fully convergence