[skris2009]

Hi
I'm looking to create a completely structured (preferably body aligned) grid for a circular-to-rectangular transition duct, i.e. a duct with a circular inlet, undergoes a shape transition to end in a rectangular outlet. I'm looking at high speed flow (Re~400,000) of air through this duct.

I tried using Gambit, but I'm not satisfied with the grid, as I need to use a 'pave' mesh for the circular end.

In Gridgen I can create a polar mesh for the circular end. But as I progress towards the rectangular end, I want to be able to use a Cartesian mesh. Is there any way I can do this? Currently, I am starting at the polar mesh end, creating many small blocks, until I get to the rectangular end, but this yields a very poor mesh in the transition region (which is the main area of interest.) I have also tried starting at the middle and proceeding towards either end, but the grid is not body aligned.

Bottom line: Is there any way in Gridgen, I can give a polar mesh at one end, select the faces (or domains) that form the volume, and allow Gridgen to extrude the mesh from polar to cartesian?

I will appreciate any ideas that come up. Many thanks in advance.

[rmatus]

Would a central H-block surrounded by an O-topology work for you? If so, check out the transition duct tutorial in Pointwise, which uses this topology. It extrudes the O-blocks from the walls and fills in the middle with an H-block.

If you have a current Gridgen license, you can download and run Pointwise with the same license.

[skris2009]

Thank you very much for your reply. I knew I had seen a similar geometry tutorial somewhere (Gambit, Gridgen, Hypermesh or SnappyHex) , but I couldnt put my finger on where. I will try this one out and get back to you if I have trouble.

Thanks again. I appreciate your help.

[skris2009]

Hi. I tried what you said, and it works perfectly for a symmetric (half) duct (as in the tutorial). I'm currently running a very high speed flow, flow which I want to pack in more grid points, while not compromising on solution time. Hence, I'm doing a quarter of the circular-to-rectangular duct. And I am able to get a good grid for this as well.

Any thoughts on how I could get a polar grid at the circular end, a cartesian grid at the rectangular end, with some sort of transition in between? Basically, I want to get to a structured, orthogonal and body-aligned grid. I was hoping there might be some extrusion process, starting from circular end) that I can constrict using domains.

Thanks again for your help.

[jorkolino]

You have the surface of the block defined already, right? Did you try to mesh both ends (i.e. the circle and the square) of it with structured grid? Then applying block command? I have managed decent structured grid on a circle, so this should merge smoothly into the rectanglular section.

By the way, I have a question concerning your first post. As you mention correctly, you can use pave grid with Gambit (which I have done extensively). I wonder, can I also use it with Gridgen too? Still unstructured grid, but with quads, instead of tets. I can nowhere find option in Gridgen to make him use this particular element type.. Anybody fluent in Gridgen can cast light?

ALso, I read there is software named Pointwise. What it differs from Gridgen? Any particular reason they develop two mesh generators?

[rmatus]

George:

Pointwise does not make unstructured quads, but you can make multi-domain, structured quad grids. The domains can fit together in an unstructured manner to make the desired shape, and they can then be extruded simultaneously to create a multi-block, structured, hexahedral grid system.

Pointwise is the current mesher from Pointwise, Inc. It takes the same meshing techniques used in Gridgen and puts them into a more modern (easier to use) graphical interface and a more modern software architecture that makes things like multiple-level undo and redo available. The same license file lets you run both Gridgen and Pointwise, so you can choose which one you want to use for a particular problem.

Hope this helps.