INS on hold with mesh converted to second order #23570
-
Hi everyone, I have been working on an application to study the viscous deformation of a material with INS. Beforehand, I was working with elastic deformation and only needed a first-order mesh with HEX8 elements. Now, with velocity and pressure at different orders, I would need the same mesh but converted to second order. I came across the Running my input file with a simple generated mesh of elements e.g. HEX20 (commented in the input file) works just fine. Any idea where that can come from? It seems like something is off with the second-order conversion... I could convert my mesh to second order with some lower-level libmesh code but this is what MOOSE can do, so I would like to stick to it! Thank you a lot in advance for your help! |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 20 replies
-
Hello So your preconditioning is failing. By commenting it out, I get some progress.
With the equivalent MOOSE-made mesh in terms of elements (x50 more elements than the one you have checked in though), it s faster and converges. But then when you make it the same dimension as your simulation mesh (1k m dimension roughly), you get similar bad convergence. So this is not really a MOOSE mesh vs external mesh problem or a second order mesh problem. I would try to see if a smaller dimension system helps (using a TransformGenerator to scale the mesh for example). Then we ll have to figure out a working preconditioner, possibly using a field split. Guillaume |
Beta Was this translation helpful? Give feedback.
we can help you figure out what to do for proper slip conditions for the channel "walls" (left, right, bottom, top), but you will need to figure out a boundary condition for velocity for at least one end of the channel. I think that the 0 Dirichlet BCs you have on your "upstream" boundary are a great first start. Then a natural boundary condition on the "downstream" boundary might work.
For doing slip BCs, for boundaries that are collinear with a mesh axis I would apply only a 0 DirichletBC for the collinear component. For boundaries that are not collinear with a mesh axis, I would suggest trying the
INSFEMomentumFreeSlipBC
... but of course now that I'm looking at the code I'm seeing tha…