[cnsidero]

Had an offline question from Arvind, that I though would be of benefit for everyone so with his permission, I'm bringing the conversation public.

paraphrasing Arvind:

"With the help of your script i was able to obtain a structured grid for an elliptic cross-section which looked like the first image. I wanted to get an orthogonal grid at the wall and a mapped grid at the center of the cross-section (second image). I tried using extrude option to get an orthogonal grid but some connectors were preventing me from making a mapped face at the center and forming a block.

Is it possible to create and solve for a grid which looks like this in pointwise?"

[cnsidero]

Yes.

Starting with the original surface mesh create on the twisted elliptic surfaces I created from the other thread, select the 4 domains. Create, Extrude, Normal... In the Assemble panel change the type to One Face Per Domain. To do so, at the bottom, click Delete All Faces, toggle the Type to One Face Per Domain, click Assemble and Done to get into the Extrusion settings.

Aside: The assembly type dictates the number of resulting blocks. Type Auto Assemble will figure out the fewest number of blocks possible from the given surface domains. Type One Face Per Domain will create one block per domain. I choose the later because I'll need the topology from the extrusion front to create a block in the core of the pipe. Or if you so choose, you can assemble your own topology.

Go to Boundary Conditions tab, select all the exposed boundaries (ends of pipe), change Type to Constant Z and Set Boundary Condition. Choose the appropriate BC for your geometry. My pipe axis is aligned with the z-direction so Constant Z was the appropriate one.

Go to Attributes tab. Under step size choose appropriate initial step height and growth rate. I used 0.001 and 1.2, respectively. Make sure the normal vectors are pointing in - if not, toggle the direction under Orientation. For this extrusion, no other setting modifications are necessary

Go to Run tab, enter 15 for steps and click Run. Adjust the number of steps for your application. Note: the extrusion at this point is purely temporary so feel free to experiement with the parameters. Click OK to apply and exit extrusion mode.

The last step will be to fill on the inner core. Notice the shape is similar to outer domains so simply follow the same procedure I described previously. Briefly:

- create two structured "H"-domains at open ends of pipe
- select previous 2 domains + 4 inner domains from extrusion and assemble the "H"-block
- if you're OK with the quality of the "H"-block, smooth the 2 end domains created in first step ... but probably not so go to the next step
- create an "OH" topology in the inner core as I described previously using the ButterflyMaker script

As opposed to the topology in the second image in the first post, you'll notice the corner points in mine are not at the major/minor axis locations but mid-way between them. This results in less pinching of the cells around the corner points and will result in a much better mesh quality.

We really need to thank Pointwise for providing the ButterflyMaker script. It makes an otherwise tedious bit of work trivial. Perhaps they can incorporate it directly into the software in the future.

Hope that helps, Chris

[Arvind_CFD]

Thanks for the prompt reply chris.

I tried out the procedure but when i do it for ellipse with a very low aspect ratio i get a very unsymmetrical grid as the corner points are not at the center.Do you think having those points on the axis would make a difference ?

Attached a pic of the case

[cnsidero]

Originally Posted by Arvind_CFD

Thanks for the prompt reply chris.

I tried out the procedure but when i do it for ellipse with a very low aspect ratio i get a very unsymmetrical grid as the corner points are not at the center.Do you think having those points on the axis would make a difference ?

Attached a pic of the case

First, the extrusion looks a bit strange - as in it's extruding further in some locations and less in others. If the cross-section is truly elliptic,l i.e. symmetric about major and minor axis, that would likely be caused by non-uniformity in spacing around the perimeter of the ellipse. Double check that.

What is the aspect ratio of the ellipse? Since it's so narrow some additional tricks may have to be applied to get a good quality grid. I will wait for your reply first.

[cnsidero]

Arvind I guestimated the AR of your ellipse (~ 3:1) and threw together a quick grid.

First, I too see some asymmetry in the extrusion even with uniform spacing around the perimeter. This is because of the 3 dimensionality of geometry (its twisting along the axis). Mine isn't as severe as yours so you might have to use few less layers, use a lower growth rate or a little bit of both. For mine I had to back down the number of layers from 15 to 10. I left the growth alone (1.2).

The inner core block requires a little tweak of the default parameters in the ButterflyMaker script. First ensure the direction you want the OH topology to propagate matches the radio button selected. Second, I changed the H Region 3D Scalar parameters to 0.7 0.4 0.5 (in fact the last one has no influence since we're only propagating it one direction). You may have to experiment a little - use the Preview Topology feature to do so. Finally I used 4 Grid Points on Ogrid Ribs (in retrospect I should have used more).

The resulting mesh quality is not as good as I showed before but still acceptable. FYI, the lowest min angle I allow is 20. Meshes with a minimum min angle greater than 40-45 I call good. So this one definitely passes. Mind you I spent about 5 mins on this. If I investigated a little more time I'm sure I could get it up another 5 or so but there's probably an upper bound due to high aspect ratio shape.

-Chris

[Arvind_CFD]

Originally Posted by cnsidero

First, the extrusion looks a bit strange - as in it's extruding further in some locations and less in others. If the cross-section is truly elliptic,l i.e. symmetric about major and minor axis, that would likely be caused by non-uniformity in spacing around the perimeter of the ellipse. Double check that.

What is the aspect ratio of the ellipse? Since it's so narrow some additional tricks may have to be applied to get a good quality grid. I will wait for your reply first.

the aspect ratio is 0.3. I changed the extrusion method from hyperbolic to algebraic and it looks a little better but there is still some asymmetry. In your pics since the aspect ratio wasn't that low the difference in extrusion was not obvious but it can be observed when you have a closer look at them.

[Arvind_CFD]

Originally Posted by cnsidero

Arvind I guestimated the AR of your ellipse (~ 3:1) and threw together a quick grid.

First, I too see some asymmetry in the extrusion even with uniform spacing around the perimeter. This is because of the 3 dimensionality of geometry (its twisting along the axis). Mine isn't as severe as yours so you might have to use few less layers, use a lower growth rate or a little bit of both. For mine I had to back down the number of layers from 15 to 10. I left the growth alone (1.2).

The inner core block requires a little tweak of the default parameters in the ButterflyMaker script. First ensure the direction you want the OH topology to propagate matches the radio button selected. Second, I changed the H Region 3D Scalar parameters to 0.7 0.4 0.5 (in fact the last one has no influence since we're only propagating it one direction). You may have to experiment a little - use the Preview Topology feature to do so. Finally I used 4 Grid Points on Ogrid Ribs (in retrospect I should have used more).

The resulting mesh quality is not as good as I showed before but still acceptable. FYI, the lowest min angle I allow is 20. Meshes with a minimum min angle greater than 40-45 I call good. So this one definitely passes. Mind you I spent about 5 mins on this. If I investigated a little more time I'm sure I could get it up another 5 or so but there's probably an upper bound due to high aspect ratio shape.

-Chris

Thanks a lot Chris..the grid should probably do the job when used with fluent. I'll get back to you in case i have problems getting a good solution with the grid.

[Arvind_CFD]

Hi,

With the inputs from this thread I was able to create a structured grid for hellically twisted elliptic pipe. I was not able to use the extrusion feature for creating layers of mesh close to the wall surface as suggested because the twist in longitudinal axis of the geometry was causing a problem preventing uniform extrusion(I used 120 grid points along the length of the twisted pipe). I used the spacing constraint(5.0e-5) on connectors in the cross-section to get the points closer to the wall which worked in cases where the aspect ratio of the ellipse was larger (i.e 0.5 or 0.7) but the simulations results were quite off when I tried the same grid structure for a smaller aspect ratio of 0.3 . Would be great if you could suggest ways to improve the mesh quality at the boundary surface by tweaking extrusion parameters(step size, growth rate?) or should i try unstructured grid with T-rex ? Attaching some pictures and the pipe surface's .glf file for 0.3 aspect ratio in case you want to give it a try when you have time.

Thanks,
Arvind