[CFDy]

Hi everyone,

I have prepared a very simple case to test the capability of simplefoam to simulate turbulet flow with KOSST turbulence model.

The mesh was created using snappyhexmesh in mm, and it is a simple 20 mm (diameter) pipe. If we solve this without scaling, i.e for a pipe having a diameter of 20 meters, the solution converges for inlet velocity of 10 m/s.

However, once the grid us scaled by 0.001 (in all the directions) no convergence can be achieved, and the velocity develops very high and unreasonable magnitudes.

What could be a problem? How can this be corrected?

This problem does not exist if I run pisoFoam.

Thank you in advance.

[pete20r2]

What are your boundary conditions?

[RobertHB]

I had similar problems when scaling my blockMesh to micrometer scale (1e-6 m). As for my problem, i accounted it to be to many trailing zeros, but maybe i'm wrong.
To offer a solution: You can always make use of dynamic similarity (i.e. Bolster et al., 2011) by scaling your viscosity instead of your mesh size. This will give you the same results as on your smaller scale, but on your larger mesh.

[CFDy]

I don't think that this is a solution. If additional dimensionless numbers must be matched , adjusting the viscosity will not provide accurate and required results.

[RobertHB]

Originally Posted by CFDy

I don't think that this is a solution. If additional dimensionless numbers must be matched , adjusting the viscosity will not provide accurate and required results.

You are right. It doesn't suffice to only scale the Reynolds number. All other dimensionless numbers important to the problem must match on both scales as well.