Learning Forum

[ae22b001]

Hello all, The inlet of my geometry has a supersonic nozzle. I am trying to simulate flow through it in steady state initially and then later in 

unsteady conditions.

With the density based solver I was able to get a decent steady state solution. When I started to run the transient solution , my results start diverging near the wall, which I think is caused by high accept ratio cells near the wall/inflation layers and sharp corners. Without the inflation layers, it is working.

 

However, when I use the pressure based solver  I am able to get really good results with inflation layers as well. I want to recreate the results using density based solver as Mach number in my flow exceeds 3.

Please let me know how to fix this issue.

I will attach pictures

Solver settings for steady state simulation

Courant no - 5
when I start the transient sim, after some iterations:

1. How do I prevent this from diverging as shown above in transient sim?
Thanks

[Federico]

How are you initializing your solution for the transient case?

Also, what happens if you reduce the Courant number to 1?

[Matt Jones]

Hello any ideas on how to fix, thanks

[Matt Jones]

 

 

 

Hello, Thanks for replying.
Im using the steady state solution to start the transient solution. I was able to obtain the steady state solution by slowly increasing the courant no from 5 to 15. 

 

 

[Matt Jones]

 

after obtaining the steady state solution, without using any udf (so nothing is changing and everything shld instantly converge), I tried out a transient simulation to make sure everything was working, but even with a courant number of 1, the residuals shot up.

 

[Matt Jones]

Hello any ideas on how to fix?

[Federico]

You mentioned the aspect ratio of your inflation layers. How large is this ratio? What is your y+?

[Matt Jones]

y+ is less than 2 on the walls. Ascept ratio is less than 100.

[Federico]

Looks good to me. 

What are your solver settings for Transient case?

[ae22b001]

Same as steady state but with timestep of 1e-6, second order timestepping and courant no 5.

 

[Federico]

I don't see any issues with the information that you are giving me. You can try reducing Turbulent URF and/or the time step to see if this helps with stabilization.

Side note: the pressure-based solver nowadays may be just as accurate even for flows M ~ 3.

[Matt Jones]

 

I tried using a timestep of 1e-7, and it seemed to be working better. The only problem is that I’ve seen multiple papers use a timestep of 1e-5 and 1e-6. So it seems weird that this problem is happening to me and that it is diverging from the inflation layers.

The bigger problem is that after I get the steady state solution, I simply changed it to transient with a timestep of 1e-5 (nothing else is changing). Ideally it shld converge instantly, but after a few iterations it started diverging.

I saved the residuals for postprocessing and observed where the residuals were high. and it was really high in some parts of the inflation layers.
This is confusing because none of these problems happen when using pressure based solver

[Matt Jones]

https://www.researchgate.net/publication/366732294_Effects_of_isolator_length_on_pseudo-shock_wave_in_a_rectangular_duct

for your reference, this is the paper i am talking about 

[Federico]

time step size will depend on the mesh, so unless you have identical mesh size, the time step size may not be comparable.

[ae22b001]

Yes but The bigger problem is that after I get the steady state solution, I simply changed it to transient with a timestep of 1e-5 (nothing else is changing). Ideally it shld converge instantly at every timestep, but after a few iterations it started slowly diverging.

[Federico]

Not sure what you mean by "it should converge instantly at every time step".

As I said, time step size is related to the mesh size. The fact that you see improvement when reducing dt and the fact that the solution is stable when removing inflation layers (which are your smallest cells), really points to the time step being too large for me.

[ae22b001]

Understood, Thank you for your help