Pardon my ignorance, I'm quite new to CFD, I'm modeling flow downstream of a moving surface (a 3in diameter speaker). Looking at reynolds numbers it seems that the flow should be laminar (u < 5m/s, l < 0.2m), so I have set up a 2D laminar transient model using CFDesign. However after some number of speaker cycles, the flow sets up a small eddy and the solution diverges. I've tried reducing the mesh size and time step. If I run a trubulent case, the solution converges. Am I making a fundamental error ? What is the reason for my laminar solution diverging ? and can it be fixed ? Thank you for your help.

What is the Reynolds number? If air is the fluid then it could well be Turbulent.

Laminar flow is less diffusive and so less stable therefore it s not surprising that laminar diverges while the turbulent case converges.

Steve

The Reynold's number is right around 60000, using the largest length scale in the problem. So it's kinda bordeline. From what I understand though, the k-e model hugely overpredicts shear at such low Re, so I thought that laminar would be a better model to use.

I'm glad to hear that it makes sense that the laminar flow is less stable. I had thought the same thing for the same reason. Any idea of what mesh size would be required to account for the instablity ? Assuming that the eddy that I currently see forming (before the solution diverges) is about 1inch in diameter, anyone have a guess what element density I would need in that area ? I've been looking for references that might have a suggestion but found nothing. Normally I'd try trial and error but the model already takes over 48 hours to solve, so excessive mesh refinement is not an option.

At Re=60000, I'd say that the flow is at least partially turbulent - at least at the edges of the eddy. Are there boundary layers in the flow - could these be turbulent or transitional? The Reynolds number is probably too high to resolve turbulence by using Navier-Stokes with a very fine grid. Especially if your solution times are 48hrs. However maybe the solution is diverging because your mesh is too fine and you are partially resolving the non-linearities involved with turbulence.

I would suggest trying to simplify the problem and reduce your solution times so that you can work out the parameters that give converged solutions.

As far as element density goes, as you refine the grid the solution will in fact get less stable (but more potentially accurate), due to there being less artificial viscosity. If you want a converged solution then increase the mesh spacing and the solution will converge. However beware, the solution may not be correct and you will struggle to get grid convergence. This is the problem with CFD - the crappier the grid the easier it is to get a solution.

Steve