[brcfdstudent]

Hi, I'm working on the simulation of a 3D converging-diverging nozzle. For now I'm running a Euler simulation (no turbulence models, laminar flow) and I'm having trouble achieving convergence.

Geometry (mm): https://imgur.com/a/QE4uky8

Mesh statistics: 3892059 nodes, 3806400 elements

Report mesh quality on fluent:

Minimum Orthogonal Quality = 6.28438e-01
(Orthogonal Quality ranges from 0 to 1, where values close to 0 correspond to low quality.)

Maximum Ortho Skew = 3.71562e-01
(Ortho Skew ranges from 0 to 1, where values close to 1 correspond to low quality.)

Maximum Aspect Ratio = 7.03460e+01

Apart from the residuals I've set two variable monitors to help me judge convergence: mass weighted average outlet temperature and mass weighted average velocity magnitude on outlet. Here's a link to the plots after ~700 iterations: https://imgur.com/a/u67DBij. There's a spike in residuals after ~300 iterations because I went from a first order discretization method to a second order discretization method.

As you can see I don't see the residuals going down anymore. I wonder if there's anything I can do to better judge convergence. Should I just leave it running for a couple more hours and then come back and check my plots? Are outlet static temperature and outlet velocity magnitude good variables to keep track and judge convergence?

Also, what can I take away from Report Quality results? Is the mesh good enough?

Thank you very much. Any help is kindly appreciated, I'm kinda lost.

[RaiderDoctor]

Hey brcfdstudent,


I'm assuming this is a steady-state simulation? What are your boundary conditions?

[brcfdstudent]

Yes, I'm running a steady state simulation. The BC are based on an article called Numerical simulation of near-critical fluid by Arina, R. I got pressure inlet and pressure outlet: https://imgur.com/a/AGj6L8i. Also worth mentioning that I've set the operating pressure to 0.

Oh, by the way, I'm still running it up to ~1300 and there are no signs of the residuals going down or my monitors getting stable.

Thank you in advance.

[LuckyTran]

Yeah keep cranking (iterating).

[brcfdstudent]

My simulation has just diverged. AMG Solver: temperature

[RaiderDoctor]

Out of curiosity, can you post a pic of your mesh?

[brcfdstudent]

Originally Posted by RaiderDoctor

Out of curiosity, can you post a pic of your mesh?

Sure I can, here you go: https://imgur.com/a/vXdvl2n

I sincerely don't know if this mesh is good enough. I've added 20 inflation layers and 3 edge sizing. Any hep is highly appreciated.

[RaiderDoctor]

I'm going to go out on a limb here and say that it's your mesh. Although you have really good resolution along the boundary layers, your cells in the middle of your domain are enormous. Mesh defeaturing (where the cells are finer closer to the boundary and larger towards the center) is perfectly acceptable, and even recommended to help save on computation time. But, I think the solver is having a hard time trying to resolve all of the numerical errors created when the fluid is transferring from the "good" section of the mesh to the "bad" section.
More to the point, though, I gotta ask; why is this 3D? It appears as though your case is axisymmetric. If this is true, then you could run a 2D simulation and have a solution in a lot less time.

[brcfdstudent]

Originally Posted by RaiderDoctor

I'm going to go out on a limb here and say that it's your mesh. Although you have really good resolution along the boundary layers, your cells in the middle of your domain are enormous. Mesh defeaturing (where the cells are finer closer to the boundary and larger towards the center) is perfectly acceptable, and even recommended to help save on computation time. But, I think the solver is having a hard time trying to resolve all of the numerical errors created when the fluid is transferring from the "good" section of the mesh to the "bad" section.
More to the point, though, I gotta ask; why is this 3D? It appears as though your case is axisymmetric. If this is true, then you could run a 2D simulation and have a solution in a lot less time.

I'll take a look at the mesh. Any practical recomendations on how to improve it? Maybe just lowering the max tet size until I get better cells size towards the middle? Is it also ok if i decrease the number of inflation layers in order to have a lower cell count and save computation time?

I have already ran 2D simulations. I'm now requested to run it 3D.