code_saturne User's Forum

Hello,
I am using CS 8.03 in Ubuntu 22.

I am modeling MHD driven flow in a closed loop. This loop has a small region where the MHD pumping effects are simulated by using a momentum term. In the image, the red region is the pump region, and the blue is the rest of the loop pipe.
https://en.wikipedia.org/wiki/Electromagnetic_pump

I am having a difficult time getting convergence. From my time using OpenFoam and Comsol for this type of simulation, a pressure point constraint was helpful for convergence. Since there is no inlet or outlet, there is no fixed pressure BC. described here: https://www.comsol.com/blogs/how-to-sol ... ty-problem
Is this a potential issue, and how would I set a single node to a zero reference pressure?

Other things I've tried:
-auto vs green-gause gradient reconstruction
-hexa vs tetra meshes. Mesh size, boundary layer sizes. Currently using ~200,000 volumes in a tet mesh, with boundary layers
-fixed vs adjustable time steps. short and shorter time steps.
-dropping the drive body force down towards zero

Thank you for any guidance.

After looking further at different types of mesh, it looks like with improved mesh quality I can get it to not diverge. It oscillates after residuals drop about 1.5 decades.. Perhaps pinning a point at a fixed reference pressure will allow further stability and convergence(?).

Hello,

I'm not sure how you could force a pressure at a given point with no outlets using the current code options, and am not sure it would help (since normally things depend on the pressure gradient but unless you have physical quantities depending on pressure, there should be no influence). I any case if could probably be done by modifying the code itself but I doublt it would help.

Mesh quality is the most important thing here. With tetrahedra, using an extended neighborhood for gradient reconstruction can help, but given that the geometry seem not seem to complex, using hexahedra in the full mesh (or in as many parts of the mesh as possible if you eventually add some more complex details will lead to better convergence and precision for a given cell count than with tetrahedra.

If the flow is steady, using a pseudo-steady time scheme (local in space and time) may help with convergence speed, but you still need to provide a well-chosen reference time step.

Best regards,

Yvan