flow field phenomenon
 

Hi there,
I hope that someone here can explain to me a phenomenon that I haven't been able to explain until now. Note that I have encountered this phenomenon while using different software, advection scemes, boundary conditions and meshes, so I think that it is a general "problem" not specific to any software.
I'm simulating a symetrical profil consisting out of two eliptical halfs and a straight part in between. It is a two dimensional simulation. What I do not understand is, why the velocity is higher in the region of the two elliptical parts and is lower in the straight section (I have attached a picture to show what I meen). I would have expected that the velocity is highest at the thickest point and stays the same for the straight part, or something similar to that. Can anyone give me a physical explanation why the velocity is higher than slows and is higher again or is this some numerical effect like numerical dispersion? If anyone can help me with that I would be very thankful.

Which way is the flow? left|right or down|up?

You make the assumption of 1D flow to arrive at this expectation probably. Try plotting only the streamwise component of velocity instead of full velocity magnitude. It might help.

Because the body cannot negotiate with the flow infinitely far away. Only the flow closest to the body feels the influence of the body while having to navigate its way around the body, causing a local acceleration. It's actually a lack of dispersion (both physical and numerical) that gives this result.

The flow is left/right.
I have already tried to only display the velocity in flow direction. That flow field has pretty much the same features. There is still an accelaration, decellaratin and then accelaration again.
Could you maybe explain your last statement a bit closer? I don't see yet why the thinning body should cause an accelaration in the flow field and the connection to the missing dispersion.

I don't understand the model you are using, is it a 2d geometry extended in the third direction for a small lenght? Is it a viscous flow model? the flow is uniform at zero AOA going from left to right

The picture is not so clear ...

Yes in this case it is an 2D Model extended in the third direction for a small bit. The flow is inviscid (I also did it with viscous flow, pretty much the same result only with a boundary layer, if it comes to the phenomenon I'm asking about).
Uniform flow with zero angle of attack left/right.

For inviscid flow, you must consider that the generation of a lift (what you are observing) in case of totally simmetric condition is determined by the setting of the Kutta condition. Just think about the non-zero circulation over a cylinder. Likely, your grid and numerical method implicitly set this condition. Are you sure that the solution reached a correct convergent solution? Are you prescribing periodic condition in the third direction?

This can't be a correct solution for an inviscid 0 aoa setup. It isn't symmetryc. You are either messing up with the bc or you are far from convergence

Ok just to explain things a bit further. Yes I do know that this picture looks a bit weird and yes I think it is because it might not really be a converged solution and the mesh is rather crappy too.
But as I said before I have done this simulation with a different software too, which I do know better. There the solution is totally symetrical and everything is just fine ecept for those two regions where the velocity is higher. But I don't have any pictures of that test case.
So I am really interested in why you can see this phenomenon in every software (I have used), with every boundary condition (and I have tried a lot) and no matter how fine I mesh this thing.
Because I have tried so many things I do not think that it is anything to do with the setup but rather with a physical or numerical effect I can't explain. Which doesn't mean that the setup of the case isn't responsible for the other problems in this picture, but as I have already corrected that in the other software I'm only interested in an explanation for this specific phenomenon.

You can't claim that It is a common feature but just show the one where some stuff is probably messed up.

For what I know, you might have messed up in all the remaining n-1 attempts with other codes.

Also, you need to tell Us your full setting to actually have some help. This includes bc, numerical method, order of residuals, etc.

Immagine to solve the equation for the potential Lap phi =0 with the BC dphi/dn=0. This way you ensure only the condition on the normal component of the velocity nothing being prescribed for the tangential component (and hence on the circulation). The numerical solution can be affected by a lot of things (grid and discretization, as well as the convergence of iterative solvers) and the line integral of the tangential component can be not zero. That means a non specular velocity field as in your case.
As Paolo addressed, you should provide all the details of the setting in your flow model.