[MattDuff99]

Hello,

I am currently attempting to simulate a pressure-swirl atomizer. I am using a simplified 2D axisymmetric model with the 'Axisymmetric Swirl' model included to simulate the swirling flow. I have calculated the radial and tangential velocities required of an equivalent annunar inlet to the flow through conservation of mass and angular momentum, and I am confident that I have these values correct.

My issue is that the swirling flow causes my residuals to diverge and eventually reach a floating point error, of which it says the specific cause cannot be identified. This was initially the case when I was modelling the full multiphase flow with the implicit unsteady solver. I thought it may have been a meshing or time step issue, but I was unable to solve it that way. I have also tried to use just water as the fluid in both implicit unsteady and steady state but even then the results diverge and eventually crash regardless of models, mesh, or time steps.

I have tried to follow all advice on other problems suffering from floating point error's but I believe mine is directly relating to the swirl and I have not been able to find any advice on this matter.

Here is a google drive link to a steady-state attempt at simulating this swirl:

https://drive.google.com/file/d/1gJT...ew?usp=sharing

If anyone could offer any assistance it would be greatly appreciated.

Thanks,

Matthew

[MattDuff99]

Just some further information as I realise I may not have provided enough previously.

Boundary Conditions:
Inlet: Velocity Inlet with velocity components [0.0, -4.0971, 4.2259]
Outlet: Pressure Outlet with pressure = 101325pa
Axis: x-axis
The rest are walls with no-slip condition
Initial condition is specified as air with water flowing from inlet

Models:
Axisymmetric
Axisymmetric swirl
Cell quality remediation
Implicit Unsteady
K-omega turbulence
Multiphase
Segregated flow
SST
Turbulent
Volume of fluid

Multiphase interactions:
H20-Air - Surface Tension Modelling
H20-H20(vapour) - Cavitation (Schnerr-Sauer)

Mesh:
Polygonal Mesher
Prism Layer Mesher
Approximately 80,000 cells with local refinement particularly around the nozzle. I have attached a picture of the mesh at the nozzle to give an idea of the mesh.


Using a time-step of 5e-5s to test the flow but this diverges every time. I have also used the adaptive time-step model to keep the CFL numebr below 2 but this also diverged after around 3 hours of running and 40,000 iterations.

I am at a loss for where to go from here, I am not sure if the issue is related to the mesh or under-relaxation factors but any assistance would be appreciated.

[LexMedina]

Matt, have you found a solution? I am facing similar problems.