[Hawxliquid]

Hello.

I am trying to simulate the flow in a nozzle, with an inlet of 34 bar and 3000K, and an outlet at atmospheric condition.
I come across a certain number of problems, first, my solution does not really converge when my mesh is fine, it often stagnate around 1e-2 for energy/mass.

Then, if i use a coarser mesh, I do converge really fast , but something really weird is that when i look at the solution, the pressure at the outlet is not what I defined ! ( Not the 101325 Pa ).

Do you have any idea to help me with one problem or the other ?
Thank you !

[jabier14]

Originally Posted by Hawxliquid

Hello.

I am trying to simulate the flow in a nozzle, with an inlet of 34 bar and 3000K, and an outlet at atmospheric condition.
I come across a certain number of problems, first, my solution does not really converge when my mesh is fine, it often stagnate around 1e-2 for energy/mass.

Then, if i use a coarser mesh, I do converge really fast , but something really weird is that when i look at the solution, the pressure at the outlet is not what I defined ! ( Not the 101325 Pa ).

Do you have any idea to help me with one problem or the other ?
Thank you !

As far as I am aware from reading the users guide, if the flow becomes super sonic the exit conditions are ignored as downstream conditions do not appreciably affect the flow. You might notice some numerical discrepancies around the outlet, but not extensive. This might explain why you are seeing a different exit pressure than the specified outlet.

From the guide:
"Pressure outlet: Exit static pressure (ignored if flow is supersonic at the exit. All the information travels downstream in a supersonic region, hence the pressure at the outlet can be computed by directly extrapolating from the adjacent cell center [ 32]. Therefore, it is not meaningful to use the exit static pressure prescribed in the boundary conditions task page, and the exit static pressure is ignored)."

[Hawxliquid]

Thanks ! That explain indeed the value then !
Do you have any idea for the bad convergence then ?

[jabier14]

Originally Posted by Hawxliquid

Thanks ! That explain indeed the value then !
Do you have any idea for the bad convergence then ?

I might. I recently ran into similar issues with an aerospike nozzle I was analyzing. If you are running a density based transient solution you may see the solver displaying messages about reducing Courant number. This would signify that your time-step may be too large relative to the velocity and mesh size in your model. Basically you will want to look at the cell size and maximum expected velocity (and then apply a reduction factor) to determine your time step as you do not want the fluid to be covering more than one cell per step. You could also try reducing your under-relaxation factors to get a good initial convergence and then incrementally increase them to default (same can be done with Courant numbers).

I also noticed for mine that occasionally reducing the mesh size resulted in a worse quality mesh, which could also be affecting your convergence. Are you using a structured mesh?

[Hawxliquid]

Originally Posted by jabier14

I might. I recently ran into similar issues with an aerospike nozzle I was analyzing. If you are running a density based transient solution you may see the solver displaying messages about reducing Courant number. This would signify that your time-step may be too large relative to the velocity and mesh size in your model. Basically you will want to look at the cell size and maximum expected velocity (and then apply a reduction factor) to determine your time step as you do not want the fluid to be covering more than one cell per step. You could also try reducing your under-relaxation factors to get a good initial convergence and then incrementally increase them to default (same can be done with Courant numbers).

I also noticed for mine that occasionally reducing the mesh size resulted in a worse quality mesh, which could also be affecting your convergence. Are you using a structured mesh?

Okay I will look at this, but this Courant error, I don't see it during my simulation ( Just a problem with temperature gradient, but then it fades away ).
I guess I will play with the mesh and see to what point it still converge.

For the mesh, I try to have the most structured possible, but since I use a bell shape, It's only structured over Area. But the Orthogonal quality is around 0.8 for the minimal, so I think it is quite good ?


EDIT : Seems to converge quite good now.
I want now to get the flow outside the nozzle ( behind it ). So i added a surface, and put atmospheric BC. Now the problem is back again, after 200-300 iterations, it jumps all of sudden to divergence.

EDIT2: Okay, it seems that using k-epsilon resolve everything ? ( If i don't use the temperature, this makes everything diverge )

EDIT3: Nevermind, it worked once and now impossible to get the same result, even though i don't changed anything

[Hawxliquid]

Even if i just put a pressure inlet and remove the nozzle, in ambiant air, i have divergence and messages like " Temperature limited to 5000k " and so.. I can't find what I am doing wrong

[jabier14]

Originally Posted by Hawxliquid

Okay I will look at this, but this Courant error, I don't see it during my simulation ( Just a problem with temperature gradient, but then it fades away ).
I guess I will play with the mesh and see to what point it still converge.

For the mesh, I try to have the most structured possible, but since I use a bell shape, It's only structured over Area. But the Orthogonal quality is around 0.8 for the minimal, so I think it is quite good ?


EDIT : Seems to converge quite good now.
I want now to get the flow outside the nozzle ( behind it ). So i added a surface, and put atmospheric BC. Now the problem is back again, after 200-300 iterations, it jumps all of sudden to divergence.

EDIT2: Okay, it seems that using k-epsilon resolve everything ? ( If i don't use the temperature, this makes everything diverge )

EDIT3: Nevermind, it worked once and now impossible to get the same result, even though i don't changed anything

I am running into a similar problem with my nozzle. In order to try and reduce the complexity of the solution I made an inviscid assumption. My problems seem to be stemming from the Y momentum which gets a little better with smaller courant numbers, but the residuals will increase prior to converging slowly. What time step resolution are you using? Also, I don't know if you ever mentioned, but are you doing a 3-D or 2-D model?