Flow Divides Into Four Streams
 

Hello everyone.

I am trying to solve this problem since a long time. But I am still getting wired results with SolidWorks FlowSimulation 2019.

It is a simulation with a filter/porous media. The problem is that the filter is perfectly round. The flow is escaping the porous media everywhere in the same ratio. But then the flow is divided into 4 streams in the +/-X and +/-Y coordinate (Pictures).

For me it seems like a wrong result because I do not see any reason why the stream should divide so extremely into four parts.

If anyone knows the reason why I am getting results like in the picture, please tell me.

If you need any further information, please tell me

Thanks a lot.


https://www.bilder-upload.eu/thumb/a...1605602307.png


https://www.bilder-upload.eu/bild-a4...02307.png.html

https://www.bilder-upload.eu/thumb/6...1605602340.png

https://www.bilder-upload.eu/bild-62...02340.png.html

Just for your Information:

I tried many things to get better results:
- Finer Mesh
- Rotating the filter (still the +/-X and +/-Y coordinate)
- Round room
- Different types of filter material and thickness
- Cleaning Parts with simulation cleaning utility
- Turning intern simulation to extern simulation
- Making convergence-tolerance finer
- Trying different inlet/outlet conditions
- ONLY simulating the filter box with top/bottom – part without other parts
- And some more…

In general the simulation is
- Steady state
- Laminar and turbulent
- With air
- Under athmopheric preassure und 20°C

Hi Simon,
I think you forgot to attach a picture as it is really hard to tell what the setup looks like as well as a result image would help and an image of the mesh.

Boris

Hello,

I tried to put the Picture directly in my post. This was not working.
Now you can click on the link and the pictures place and it opens in another tab ( /www.bilder-upload.eu ).

With me its working.


Edit: I copied the link below where the Picture should be.

Yes, this looks strange.
On the first look I would say it is related to the mesh, but I cannot tell without seeing the mesh. Can you show the mesh from the view in the result image you have (full room) and also a more detailed view of the mesh in a closer look of the porous media. So how big the cells are over the thickness of the porous media. As the porous media is primarily responsible for the flow distribution due to the pressure loss it creates.

Hi,

thank you for replying so fast.
You can find the Pictures at the bottom.

I also have a thrid Picture with streamlines close to the porous medium.
For me it is strange that the streamlines are going correctly through the porous medium. Right after it they go together into four streams.

I also was thinking first about an meshing-issue. What I did was to refine the mesh. Even finer than in the "close-view-picture". It was the same result. I wish there is another type of mesh in SolidWorks FlowSimulation (like polyhedral etc...) to see ithe infuence of the mesh type.

The peessure loss is roughly th same than expected.


Thank you Simon

close view:
https://www.bilder-upload.eu/bild-2b...31116.png.html
https://www.bilder-upload.eu/bild-2b...31116.png.html

all view:
https://www.bilder-upload.eu/bild-f9...31139.png.html
https://www.bilder-upload.eu/bild-f9...31139.png.html

streamlines:
https://www.bilder-upload.eu/bild-a5...31149.png.html
https://www.bilder-upload.eu/bild-a5...31149.png.html

Try to run a transient simulation to see if the flow starts correctly initially but then forms to such a pattern.
The environment can have a complex impact on the flow which we naturally would not expect.
I had once a case where you would expect that two jets hitting on the one side of the room would cause two symmetric vortices in the room, but the simulation shows one large vortex which was not symmetric like the conditions were. After a test in reality it turned out exactly as simulated.

Nature might be beautiful, but also unpredictable. So our natural assumptions might be proven wrong in reality.
Your flow field might look completely different when in a round room as the corners are farther away from where the wall is hit perpendicular much earlier and with bigger momentum, the corner is further away from losing momentum which might cause a vortex that pushes the streams towards the corners, more towards these 4 directions.

The transient simulation should show the flow development. Some flow simply is not symmetric, ideal and steady state and transient and non symmetric considerations might show the reason.

Hello, sorry for answering late.

I started the transient simulation. After a while the follow is going through the filter media. At first it flowing constantly around the object but when the flow is getting stronger the 4 streams are forming. From time to time they change the direction slightly but at the end they stay constantly.

I am getting the same flow by only simulating the "Filtration-part" as a cylinder (Steady state).

Another thing is that if I
- change the porous media to a solid
- move the inlet from the inner part of the Geometry to the outside of the solid
- and fit the inlet conditions
The flow is (in my oppinon) correct. But ist not the thing I wanted to simulate…….



Thank you Simon

Sorry for getting back to you so late, I was too busy to look after the posts here.

I think the problem you have is not as easy as "it should be completely circular as the model is. The room is not circular and the setup is not easy. You are blowing out flow at the top, suck it in at the bottom, and worst case, the fan also has a swirl. The room walls will cause the flow to spread out once it hits them and the more you get to the corner, the flow will hit the other flow from the other wall which in return will cause a vortex or move of the flow to the upper part of the room or lower and on the lower you suck in the air again so it is not as "simple" as you might think.
That's the trick with designing things, they look plain simple and should behave as we think, but in reality, they don't.

Split the problem into smaller portions and you will see it will change drastically as you increase complexity. I setup your model very simple with an inflow in the inner cylinder and a porous media in an external environment.
Result is an equal outflow, although diagonal upwards (depends on the porous media and inlet BC), and it looks all as it should be.

Now you can add a room to see the impact of the walls on the flow and then also some intake at the bottom where the fan is sucking the air in again etc. Step by step increasing the model complexity will show you when and why the flow might be impacted this way.
In my case, the walls are not there and no suction, the flow can simply leave the domain.

Regards,
Boris

Tank you a lot for answering and for the effort!


i was doing another project during the past weeks. Now I had time to follow your instructions (some of them I also was doing previously). The result is really not changing as soon as the flow is going through the porous media.

I had the chance to try another CFD-Program where I was simulating the exact same geometry. In this program it was fine like expected.

Can you tell me which program you are using? With me it was not fine with SolidWorks FlowSimulation but with AltAir Simulation it was correct.


Regards,
Simon

I'm using Simcenter FLOEFD which is the same code as SolidWorks Flow Simulation.

Have you tried a step-by-step approach with your model?
Start simple and see what the results do as you add one uncertainty/complexity after the other. I used a simple cylinder with the upper ring part being a porous media and had a simple inflow BC, so no fan that sucks in at the bottom and blows out at the top.
You can see, it is a perfectly round outflow. It was done in an external simulation without the room walls.
Go from there, add the room, then add the suction at the bottom and see how suddenly things will change as they interact with each other.

Comparing apples with oranges won't help. You are using 2 CFD codes which both should give the same results in an ideal case. The question is, was it ideal?
Did you use the same turbulence models, the same BCs, the same environment etc.
Even then you don't know which code is correct and which is wrong.
If you want to know if the results are accurate, measure it in reality and model it in CFD as close as you can to reality. And make perfectly sure that there is nothing you have forgotten, no simplification you have done that could have an impact. Don't think A has a bigger name than B so A is correct. I have seen this several times and it turned out A was wrong.

You can also try another thing, rotate the room by 45° so the mesh orientation is not aligned with the room and see if it changes.

I'm not saying the software can't be wrong, every software can be wrong. But make sure it really is the software by eliminating every possible error, even own errors in setting it up.

Boris

Yes, this looks strange.

On the first look I would say it is related to the mesh, but I cannot tell without seeing the mesh.

Can you show the mesh from the view in the result image you have (full room) and also a more detailed view of the mesh in a closer look of the porous media. So how big the cells are over the thickness of the porous media.

As the porous media is primarily responsible for the flow distribution due to the pressure loss it creates.