[yangboMAE]

Firstly, I generated the volume mesh around a car using the Octree method. Then, I tried to re-generate the mesh using Delaunay method based on existing mesh. But I noticed that the volume mesh sizes around the car were still increased very quickly (not smoothly) in space. So could we generate smooth changing mesh using Delaunay method based on the mesh generated using Octree method in ICEM-CFD? And if we could, how to do that?
Thanks indeed.

[PSYMN]

Delaunay is a more sudden transition than the advancing front method (it started life as an FEA algorithm, where as the advancing front algorithm started life at GE specifically designed for fluids). Personally, I find that method to be too slow to generate.

I prefer the Delaunay method in 12.1 with the TGlib and AF options...

I made a post recently where I showed images of these various methods for comparison.

[yangboMAE]

hi Simon,

Could you please illustrate it in detail? or could you paste the link of your post for me?

Maybe the original Octree volume mesh is not smooth enough for Fluent solution, I guess. So I would like to use Delaunay based on the existing Octree mesh. But I don't find options about how to control the tetra mesh size increasing rate.

What do you recommend? By the way, I am using the ICEM-CFD 12.0.1.
Thanks indeed.

[rwryne]

Originally Posted by yangboMAE

hi Simon,

Could you please illustrate it in detail? or could you paste the link of your post for me?

Maybe the original Octree volume mesh is not smooth enough for Fluent solution, I guess. So I would like to use Delaunay based on the existing Octree mesh. But I don't find options about how to control the tetra mesh size increasing rate.

What do you recommend? By the way, I am using the ICEM-CFD 12.0.1.
Thanks indeed.

http://www.cfd-online.com/Forums/ans...tml#post239296

[yangboMAE]

domain interface?
what does it mean?
Now I set up some density regions around and in the wake of the car to obtain refined Octree volume mesh firstly. Then, re-generate Delaunay volume mesh on the existing Octree volume mesh. It seems better. But I think the volume mesh transition is still not smooth enough. Because the car is a blunt body near road. And tyres contact road. So the Delaunay volume mesh depends on the surface mesh on the road, I think.
Simon, how could we obtain smooth transition surface mesh on the road (or the floor of wind tunnel)??
Thanks indeed.

[BrolY]

You could use some options in the Part Mesh Setup menu such as Tetra Size Ratio and Tetra Width (look at the user manual to explain how it works).

[PSYMN]

You are right in that these bottom up tetra methods are very dependent on the surface mesh. So you should always smooth the surface mesh with the Laplace algorithm before running Delaunay. This algorithm is particularly good at smoothing transitions.



Of course there is also the “Spacing scaling factor” to roughly control the expansion with the Delaunay method. (Width and Tetra Size Ratio are for Octree and don’t yet work with Delaunay (but may at 13.0 )). For Advancing front, you can set the Expansion Factor to control the rate of expansion. There is a similar Expansion Factor for the TGrid method, but I think the TGLib AF method uses the Spacing Scaling factor (I will have to test to confirm).

Another tip would be to "nest" density regions. You can put a size 1 right behind the vehicle, then a larger density box with size 1.5 would extended beyond that one up and down stream, outside of that, you could put another density box with size 2, etc. (the smaller size density box is nested within the larger). Then when you run Delaunay (or Delaunay with TGlib AF) it will respect these and slow down even further.



The final suggestion is to use the Laplace Volume smoother now available in 12.1.