[enginpower]

Hello,
I'm simulating a simple water flow through a control valve. So the flow is incompressible and steady. Max reached internal velocity is less than 100m/s.
I'm using ACE+ as software, of ESI Group (the same now owns OpenFOAM) which uses the SIMPLE algorithm.

As boundary conditions, I set pressure, normal velocity, turbulence intensity, and the inlet diameter for the dissipation rate calculation in inlet; while in the outlet I set the fixed pressure, kinetic energy and again the outlet (=inlet) diameter for the dissipation rate calculation :

• inlet: p=1psi, 1.5 m/s, T=300K, I=0.05, d=0.25m
• outlet: p=0psi, T=300K, k=?, d=0.25

The resulting kinetic energy, using the same inlet velocity, is very low (k<0.01m^2/s^2). Using this value the solver diverges soon, and I thought that the ouput flow is very turbulent due to internal geometry of the valve. So usually I set an upper value for k (for example k=1).


It seems very converging in the beginning but then it does not converge. I attached a picture imageACE+.jpg of the normalized residual plot and the mass imbalance.


If I let the solver continue to simulate then the residuals will oscillate but never converge, and sometimes they are very oscillating after 2000 iterations, and globally all normalized residuals never go down 0.1 value.



What do you suggest to do to get convergence? Is it a good setting? What do you suggest in general to do this type of simulations?
Even if you don't know the software, how would you set this type of problems?


Thanks in advance for any useful advice!

[FMDenaro]

Have you tried also different RANS turbulence model? What about the grid resolution? The geometry of the valve is axisymmetric?

[enginpower]

Originally Posted by FMDenaro

Have you tried also different RANS turbulence model?

I tried k-epsilon model and SST k-omega model: always the same behaviour

Originally Posted by FMDenaro

What about the grid resolution?

The grid is very dense, let's say about 7M cells, obviously more dense where flow goes through holes of the internal parts of the valve, and less dense in the large duct of the bell close to the inlet and outlet section.
I created the grid with the Esi software for mesh generation, that is VisCart. About generated mesh, it passes all validation checks

Originally Posted by FMDenaro

The geometry of the valve is axisymmetric?

Yes, the valve is symmetric, so I'm simulating just a half