[VasCFD]

Hi

I am doing a turbulent flow simulation(K-ephsilon model) of air in a square duct with internal rectangular fins. I previously did the same case in workbench using design modeller and did not get any problem. But now i am doing this in gambit and after a few hundreds of iterations, i am receiving reversed flow in a lot of faces. What should i do??

[QIAN06]

HI,
could you give specific problem？ e.g., give the printscreen of the problem.

[VasCFD]

Originally Posted by QIAN06

HI,
could you give specific problem？ e.g., give the printscreen of the problem.

I have attached geometry as modelled in gambit. I have tried various turbulence models like k-ephsilon, k-omega(So far the best as the reversed flow occurred as late as 890 iterations) and tried to use different schemes for solving momentum equation. But all it did was to delay the problem than solve it.
Actually the problem begins with ever increasing "scaled residuals" and after a point the reversed flow occurs. Please help me out.
Thanks in advance.

[divergence]

It seems to me that your outlet is quite near the last rectangular object. This would imply that the flow would have a hard time getting back on the track and axial instead of swirling in the wake of the last fin.

You could try to extend the square duct in order to give the flow a better chance to stabilize before exiting the domain. If you would still have back flow in the outlet, you could make the outlet somewhat smaller, say 60% of the area of the duct. Mind you that reducing the area is a bad idea if you would do it straight after the last fin.

Hope this helps!

[VasCFD]

Originally Posted by divergence

It seems to me that your outlet is quite near the last rectangular object. This would imply that the flow would have a hard time getting back on the track and axial instead of swirling in the wake of the last fin.

You could try to extend the square duct in order to give the flow a better chance to stabilize before exiting the domain. If you would still have back flow in the outlet, you could make the outlet somewhat smaller, say 60% of the area of the duct. Mind you that reducing the area is a bad idea if you would do it straight after the last fin.

Hope this helps!

Thanks for the idea. I came across this suggestion but will not be able to implement it, as my test case does not allow deviation with the current geometry.
Also i heard through professors that reversed flow does not occur in back flow, but occurs only when the scheme used is not able to solve the governing equations properly and occurs due to generation of absurd values for the parameters.
Thanks for the suggestion though. Please help me with my problem.

[divergence]

So it's more of an academic test you have in your hands there, okay.

I read your first message second time. You mentioned succeeding in your calculations when you created the volume with DM and Meshing (if I understood you right), but got some negative feedback when you created the flow volume and mesh with Gambit.

I assume that the duct itself alongside with the fins are identical in both of the cases as well as the boundary conditions and the discretizations. So actually the only thing that has changed would be the mesh. Therefore, I would proceed with studying the mesh, trying to find some low quality cells that might mess things up for you. Check if there's sudden and drastic changes in the aspect ratio or if the mesh near the fins are inadequate. Additionally, you should check if the under-relaxations are at least moderate if not low.

Kinda sounds weird if you already managed to solve the case and now would have some mysterious problem. I'm sure you'll be victorious sooner than later.

[VasCFD]

Originally Posted by divergence

So it's more of an academic test you have in your hands there, okay.

I read your first message second time. You mentioned succeeding in your calculations when you created the volume with DM and Meshing (if I understood you right), but got some negative feedback when you created the flow volume and mesh with Gambit.

I assume that the duct itself alongside with the fins are identical in both of the cases as well as the boundary conditions and the discretizations. So actually the only thing that has changed would be the mesh. Therefore, I would proceed with studying the mesh, trying to find some low quality cells that might mess things up for you. Check if there's sudden and drastic changes in the aspect ratio or if the mesh near the fins are inadequate. Additionally, you should check if the under-relaxations are at least moderate if not low.

Kinda sounds weird if you already managed to solve the case and now would have some mysterious problem. I'm sure you'll be victorious sooner than later.

Thanks for the suggestion. I have figured out that the mesh would be only difference as i gave assembly mesh in workbench (which gives the best suited mesh for the geometry). I'll definitely review my mesh for error or inconsistencies. But can you suggest me some mesh methods or patterns that would help me solve the problem.
Thanks again.