[kubawlo]

Dear community,

I want to model a flow through an industrial filter which consists of wedge-wire screen. It is composed of tiny triangular wires (1x2mm) arranged radially, and the gap between each is only 30 um. I've made a model in 3D CAD*symmetrical problem) and extracted the volume for Ansys Meshing.

I'm trying to mesh the model, but since the geometry is quite complex, I split the body into 4 and meshed separately. The only problematic geometry is the screen. I tried various methods (automatic, multizone, tets) and none gave satisfactory results. When I set "proximity and curvature" adv. size function, the process lasts forever (left it overnight, it blocks on "preparing to model boundary for part" or on blank "overall process". Intel Core i5, 8GB RAM, Win8.1 64 bit). When I turn "curvature" only, the mesh's quality is very poor (only one cell throughout the big gap).

To simplify the problem, I've prepared a 2D version of the screen, which is a cross-section of the screen (normal to the main axis) and tried to mesh it with proximity and curvature on (automatic, set relevance to 0, relevance center=fine, smoothing=high, span angle=fine, the sizes=default and num cells across the gap=8). The result is much better (80514 elements), there are more cells across the crucial gaps.
Now, since the problem must be solved in 3D (simultaneous radial and axial flow), I would like to transfer the 2D mesh onto the body, for example as a sweep (supposing multizone?).
Any fresh ideas? Ultimately, I'm planning to solve ~3-4 mln elements mesh.

And one more question: If I joined the split body into one in DesignModeller (via Form New Part command), would Meshing automatically "know" that there is full continuity of fluid between the bodies and match the mesh?

Thanks in advance for help (it is a part of my BSc thesis).

I enclose some photos:


EDIT: patch independent mesh generation lasts forever too.

[FrankW]

Hi kubawlo,

in my opinion your mesh of the wedges are to rough. A better way is to simulate 1 wedge and with the simulated parameters would i use a porous domain to emulate the region of wedges.

regards
Frank

[kubawlo]

Originally Posted by FrankW

Hi kubawlo,

in my opinion your mesh of the wedges are to rough. A better way is to simulate 1 wedge and with the simulated parameters would i use a porous domain to emulate the region of wedges.

regards
Frank

Hi, Thanks for your answer. So you suggest that I model a flow through one gap (with several different flow rates) to obtain a pressure drop vs approach velocity? And then i could calculate the pressure drop coefficient ?
Then, I use this coefficient in porous domain? What should be the thickness of the domain? Maybe the same as the height of triangle?

And how about meshing in Icem, maybe It could handle such a specific geometry?

[FrankW]

Hi kubawlo,

So you suggest that I model a flow through one gap (with several different flow rates) to obtain a pressure drop vs approach velocity? And then i could calculate the pressure drop coefficient ?
Then, I use this coefficient in porous domain? What should be the thickness of the domain? Maybe the same as the height of triangle?

Yes, that is my idea to reduce drastically the number of elements. The thickness of the porous domain should be the same as the triangle. Porosity maybe 50%. If you use CFX look at Users Guide 12.4.7 respectively at Theorie Guide 1.8.1.
I think ICEM can handle the geometry but this is not the point. Is your computer able to solve meshes with approx. >5Mio elements? You need at least 5-8 elements across the smallest gap (It depents on your fluid velocity, laminar or turbulent flow)! Maybe you have to take a mesh sensitivity analysis!

regards
Frank

[kubawlo]

Originally Posted by FrankW

Hi kubawlo,


I think ICEM can handle the geometry but this is not the point. Is your computer able to solve meshes with approx. >5Mio elements? You need at least 5-8 elements across the smallest gap (It depents on your fluid velocity, laminar or turbulent flow)! Maybe you have to take a mesh sensitivity analysis!

regards
Frank

Well, I have already made a simulation with >5 mln elements - it took quite a long time, but I'm not in hurry. What is more, I'm planning to use mesh adaptation feature, which will additionally split the cells in gap only. But to do this, first I must have the mesh of acceptable initial quality to avoid such consequences as the residual of continuity of the order of magnitude ~e0025...
But the idea of porosity sounds good too, I have to read more about it.

Cheers!

[mrdelaunay]

Which version of ANSYS are you using?

[kubawlo]

Originally Posted by mrdelaunay

Which version of ANSYS are you using?

The version is 16.0

[Ayoub-moufdi]

Hello sir,

Please could you help me with a tutorial to make a wedge wire screen using solidworks software i'm preparing a master degree in university and i need to elaborate a screen using solidworks to finish my moduls is it possible plz.
Regards.
Chikaouimoufdi@gmail.com.