You have flow separation because you have adverse pressure gradient (decelerating the flow). Simple!

Hi professor,

Thanks

Yes. but Is existence of  reverse flow logical after iteration of 6000 or 10000 for example?

In other words, Shouldn't flow form(without reverse flow) finally? ie Is always there these reverse flow in this simulation until solution converges? According my experience, I thought that flow should form finally after longer iteration!

It depends on the size of the wake and also on the 3D effects on the boundary layer.

Can you show also the mesh and a contour plot of the last iteration ?

Thanks every body.

Abdul aziz Jaafar, reversed flow occurred on pressure outlet. Now, with K-W SSt, there is also this problem and reversed flow occurs on pressure outlet.

Is there any problem that I place outlet far away?

stream lines:

http://upsara.com/images/c1we_capture.jpg

pressure contour plot:

http://upsara.com/images/wjnj_capture22.jpg

mesh:

http://www.upsara.com/images/vwc2_capture333.jpg

http://www.upsara.com/images/4dkh_captur5777e.jpg

http://www.upsara.com/images/lp35_capture343434.jpg

http://www.upsara.com/images/6jt3_captur554e.jpg

http://www.upsara.com/images/o4ox_capture34345554.jpg

http://upsara.com/images/lysj_captu5545re.jpg

http://upsara.com/images/64lv_captu465656re.jpg

I think that your first problem is that your boundaries are way to close to your airfoil.

Whatever condition you put on the outlet you will have problems: look at the pressure for instance. Being so close to you airfoil you experience pressure non- uniformities. However, you are imposing constant pressure outlet, right?

I would put the outlet AT LEAST 3 times the chord further downstream. AT LEAST!

Yes.

The computational domain is limited to four and eight blade chord lengths upstream and downstream from the blade row, respectively.

http://upsara.com/images/u6cy_capture66666.png

However, there is still the reversed flow.

No I was talking about the trailing edge, not suction/pressure sides.

I am not a turbo-machinist, but I would leave more space between the trailing edge and the left boundary. As for the upstream/downstream, 4 and 8 chords in your last sketch look much more than what we see in the contour/streamline plot. Is it just a zoom issue?

To me the reverse flow would occur because of the pressure field you have and because of the unsteadiness that your simulation is shown (from the mass flow residuals) , combined with having the left boundary too close to the airfoil.

First , check if you have any benefit in putting the left boundary further. If not, the problem is purely on your numerical settings, and we will try to understand what is going on

I do not think your problem is in the flow separation

Thanks for your answer.

Yes. it is zoomed.

What is your exact meaning of "left boundary"?

I am grateful that guide me according the following figures:

http://upsara.com/images/0z2f_capture34345554_-_copy.jpg

http://upsara.com/images/ydi4_capture333_-_copy.jpg

I mean the left edge of your domain from your picture

Therefore your means is space between the leading edge and the left boundary.

In Fact, Can this effect on outlet results and convergence?

Yes, the space from the leading (or the trailing edge, whatever) should be increased.

Something like in the sketch. This is my first , coarse guess: you are asking the software to put a constant pressure in that outlet, whereas you can see from your simulation that this is not physical, since the boundary has still strong non uniformities.

As I said, I do not know how turbomachinist would pose the problem, but a constant pressure outlet there, especially if you are so close to your object, is a problem.

Could you plot the pressure from that boundary, just to check ?

Yes, You are right.

Outlet pressure is set 0 pascal(static pressure=gauge pressure) in Fluent.

Yes, I plotted pressure from that boundary. outlet pressure usually is close to 100 pascal!!!

Should be  this value close to zero?

In fact you say that if I place outlet boundary  to ten blade chord lengths downstream from the blade row, outlet results and convergence become better. right?

It might not solve the problem but it should help: at the moment your simulation is not able to respect the BC that you introduced. That's a big, big, problem.

Now if the BC condition are the good ones, that's another story. But at least the solution should respect them, no :D?

Thanks Miguel Alfonso Mendez .

I will investigate your statement.

Can you tell us if the separation is at the hub or at the shroud. If it is at the hub you may have a simple explanation for the phenomenon.

Hi,

What is off-design condition??!!!

It means as after stall? In fact,  I want to say that stall condition is off-design condition.

because there is not design operating range after stall.  Am I right?

Generally,  if there is an oblique body through the duct, the viscosity force cause the swirl or Recirculations near the boundary can make 'Reverse Flow' problems. As recommended, it can disappear after many iterations and then the reversed flow will disappear. The solution of this case can be listed below:

1. Increase the mesh number and quality.

2. Increase the duct length after the oblique body about 11 times the diameter of the oblique.

3. This problem appears in ANSYS FLUENT and the solution is to switch the outlet boundary condition from 'Outlet Pressure' to ' Outflow'.

4.If the length of the duct is short, I recommend using ANSYS CFX and make the outlet condition as 'opening' boundary condition.

Dear Ali Sam,

Did you solved this issue?

Thanks,

Eyub

Reversed flow can occur when pressure gradients due to geometry mean that the flow wants to re-enter the domain at the location where you have placed an outlet. Some flow reversal can be tolerated, but will have an effect on mass-conservation and thus convergence

Hi, Ali. I currently have thesame issue with my meshing. Please, how did you resolve this ?

How can this reversed flow issue be solved? It shows in previous iteration and after that it dissapears. Has it been solved later or need to re run the simualtion?