Just to confirm that compared to your original mesh, all you’ve done is rotate it and scale it down?
If so, this makes me think that you are just hitting into numerical precision issues. What is wrong with runing the scaled up version if all the non-dimensional groups have the same value?
Have you also changed your initial condition? Otherwise, your BC is now not compatible with your initial condition which will result in a similar inconsistency.
Regards, Freddie.
Thank you for your answer Freddie, Can you clarify a bit for me? The initial velocity for is not compatible with all four boundary conditions. However I feel that this is inevitable and occurs in any simulation, otherwise there would be no need to solve anything. For example in the “couette_flow_2d” example case the initial velocity profile is incompatible with the no-slip boundary condition at bcwalllower.
I guess I misunderstood you somehow or I am missing some nuance.
There is an appreciable different between wall boundaries and inlet/outlet/far-field boundaries. This is why when starting a real simulation we always try to make the IC compatible with the boundary conditions (typically by making the ICs equal to the free-stream conditions or a minor perturbation thereof). This is fine as it is the walls (more precisely their locations) which drive the physics.
Case in point: a vortex coming towards a wall is typically a non-issue, a vortex coming towards a far-field condition is very often problematic and may result in reflections or otherwise destabilise the flow (hence why it is common to find things such as sponge regions near outflows).
Regards, Freddie.
Thank you Freddie and will, after accepting both your suggestions- running the scaled system and making all the B.Cs compatible with the we were finally able to make things run without diverging
Many thanks
Nathan