my question is, should this sudden jump have any major effect on results? assume that the larger cell size is enough to resolve the flow

That jump in cell size is always bad regardless of the type of simulations you are using

The transition should be always smooth. From you geometry it doesn't look that hard to make a smooth transition between cells

well the core jump is pretty easy to fix but under the vortex finder I'm at a loss for a solution

I guess I have to make the whole mesh around that constraint... because the thickness is 2mm

so I need to have an integer number of cells there...

I do not get what you're trying to explain

but always try your best to create a good mesh

it's okay, nevermind. thank you for your inputs!

Is it possible for Spalart-Allmaras turbulence model to converge even if the first layer is at 4<y+<5? Or it need to be at y+<1?

within y+ < 5 the flow is still laminar, so it shouldn't be much different. It can affect prediction of separation and therefore results, but my guess is it shouldn't affect convergence

I do not understand why it couldn’t be turbulent in this region, could you substantiate this a bit?

To my understanding your second statement is true to, you will possibly solve you boundary layer wrong, so your solution could convergence, but it is still wrong in such case.

At the wall, v=0, so for a tiny distance far from the wall there is a laminar flow. Below y+ < 5 that is. Or what do you mean @J4NN3S™ ?

So even after re=(turbulent value for chord length) in the lower region of the boundary layer the flow is laminar?

it doesn't matter what Re is, y+ is non dimensional. The y (the absolute distance) will simply get smaller with bigger Re.

Re is also non dimensional

Because I want to test the case, I can safely said that I'm not too concern about the prediction of seperation and the results in general for the moment. Once I can confirm that all the schemes, the BC and the IC are wright, I'll switch to a refine mesh. But right now, what it does on the residuals it decrease and then increase a bit for finally getting a 'plateau/stagnation' where the residuals stay basically the same iteration after iteration. The reason why I was asking if it could converge even though 4<y+<5 is that, I use the wall function 'nutUSpaldingWallFunction' for nut but didnt fix any 'maxIter' nor 'tolerance' parameters yet.

Because I want to test the case, I can safely said that I'm not too concern about the prediction of seperation and the results in general, for the moment. Once I can confirm that all the schemes, the BC and the IC are wright, I'll proceed with a refine mesh. But right now, what it does on the residuals it decrease and then increase a bit for finally getting a 'plateau/stagnation' where the residuals stay basically the same iteration after iteration. The reason why I was asking if it could converge even though 4<y+<5 is that, I use the wall function 'nutUSpaldingWallFunction' for nut but didnt fix any 'maxIter' nor 'tolerance' parameters yet.

I'm pretty confident that the solvers are good, but the grad, the sngrad & laplacianSchemes could maybe needed some adjustement. I've a mesh with a non-orthogonality less then 60, but if I recall, greather then 40. I'm maybe wrong by assuming that the issue could likely reside in the BC and/or IC....

I’ll dig into it a little more someday

Since y+ is a function of y(wall disntance) and Re a function of c(chord length) I don’t think your statement is always the case (the bigger Re, the bigger y+

Or charasteristic length

I might be wrong tho

I just want to understand

The reference length used to calculate Re doesn't matter. What matters is your velocity profile close to the wall. It gets more and more steep with a more turbulent flow. So you need finer and finer cells to capture this gradient accurately. The 'viscous sublayer', the region that is so close to the wall that viscous effects dominate gets smaller. The viscous sublayer is still in the region 0 < y+ < 5 though

Inherently transient cases will rarely converge in steady-state, so taking into account most flows in real life are transient, you have quite a slim chance of getting anything to converge neatly... You puck most probably drags huge separated vortexes behind it so steady-state will behave exactly like you described

The effect that you mentionned is probably accentuate with a 3D mesh also. Obvisouly, my strategy is to create IC with a steady-state turbulence model with an appropriate mesh and going forward in unsteady state afterward. But in this case, like you mentioned, what could be my approach then? Going straight with a unsteady case can take a long time before I can validate the mesh, the drag coefficient and all of the other parameters.....

I wasn’t sure y+ was a function of Re as well, but as the skin friction coefficient used to estimate the wall shear stress is a function of Re it is actually connected to the reynolds number