[RagingPotato]

Hi good people of CFD online. I'm currently trying to create a structured mesh around a 3D Windsor body, an automotive reference model similar to the Ahmed body. While I've done a fair amount of unstructured meshing in ICEM I'm a total novice when it comes to structured meshing (as will no doubt be clear from the images below!).

The images attached show some of the problems I'm currently facing, with overlapping grids on surfaces, and areas of mesh within the body which shouldn't be there. Now i'm fairly confident these issues are due to the blocks I've used, but despite starting from scratch many many times I always end up with similar problems.

So my question: is anyone able to look at these images and provide some insight into where they think I've gone wrong, or what I should be looking to change. I appreciate it's probably not easy to make things out from these images, so I can upload the mesh files if anybody is willing to take a look.

Any feedback very gratefully received, constructive or otherwise!

[RagingPotato]

Mesh files are here: https://skydrive.live.com/redir?resi...AMVUMebfVuLayU

[Far]

http://www.cfd-online.com/Forums/ans...tml#post363879



http://www.cfd-online.com/Forums/ans...tml#post363879

[diamondx]

if it can't find a surface to project to, your mesh will go overlapping somewhere else.
check the square next to surface under the geometry tree.
Then click with right button and select solid, so you can see the surfaces in your geometry. I bet you don't have a surface there, or you forgot associations. create a surface, associate, then premesh again.

[RagingPotato]

Thanks Far, I found your Ahmed body mesh files which are very interesting. The issue I'm having though is that the Windsor body is more complex than the Ahmed body, I was originally following an Ahmed body tutorial, but the Windsor body doesn't just have constant radius features at the front, and trying to split blocks to allow me to capture these features is whats got me into the mess I'm currently in.

diamondx, thanks! I'll check that over now.

[RagingPotato]

It's doesn't seem to be a surface issue, they're all in place. As for associations, all the edges which should be associated to curves appear to be, should I be associating faces to surfaces too or is that not necessary? Thanks for your help!

[Far]

Originally Posted by diamondx

if it can't find a surface to project to, your mesh will go overlapping somewhere else.
check the square next to surface under the geometry tree.
Then click with right button and select solid, so you can see the surfaces in your geometry. I bet you don't have a surface there, or you forgot associations. create a surface, associate, then premesh again.

Geometry has lot of problem. Even spaceclaim is not able to rectify these problems.

[PSYMN]

Based only on the images...

The symmetry plane on the inside of the body doesn't appear to have a good surface to project to. You can either create a surface for it or just set the association for those faces to "none".

The blocking topology is very poor on that front lower corner... not sure what you are doing there, but it looks very bad.

I will download and give it a try later if I have time...

[PSYMN]

I suddenly had 10 minutes to take a look...

The problem was you were trying to capture the surface patches instead of the actual shape... The surface patches are complicated, the actual shape is not.

I just wanted to mess with the blocking, so I left the geometry alone, but if I had more time, I would create isoparametric curves along the center of each fillet so I could properly associate edges... Instead, I just moved the verts to the middle of each fillet and captured things that way.

The shape is just a box with some round corners... So block it like that... The boundary layer ogrid will solve any problems you would have had at the corners...

WindsorBody.jpg

I am attaching my rough 10 minute version. I have not gone thru and aligned verts to be straight, setup edge distributions, etc. Perhaps you want to split off a block behind the body and put another ogrid in there to capture the re-circulation, etc. To open it, just load your project and then load my blocking (replace).

Have fun.

[maalan]

Simon, would you mind tell me the best association strategy to mesh the three-dimensional ahmed body nose with ICEM?? I have a rectangular domain and the simplest block distribution with an O-grid around the body.

I would like to have something as the windsor body mesh you post long time ago.

I will be very gratefull!!
Thanks.

Antonio

[PSYMN]

With ICEM CFD Hexa, half the skill is just viewing the geometry in its simplest topological form...

The Ahmed body is basically a block with a shallow champher on the back (backlite angle) and fillets around the front.

The easiest way to block it is to add an isoparametric curve to the fillet so you get a line where the imaginary corner would be... Then block it like you are cutting a box out of the windtunnel. Add a split for the angle change at the back. Associate with all the curves that line up with edges (including the isoparametric ones for the front edges) so it is basically a box that tapers at the back. Then select the two blocks inside the vehicle and put them in a new part (SOLID). Then select those Solid blocks for adding the OGrid to. You don't need to select any faces, but make sure to check the "Around Blocks" option before hitting apply.

Apply. Now you have your Ogrid which should help with your viscous boundary layer.

If your version has the little pegs under it, you would split upstream and down stream and on the side of each so they end up in a box about twice their diameter. (If you have multiple pegs, one split may work for more than one) Then Ogrid and select the box (or all the boxes at once if you like). Add a face(s) where the box(es) meet(s) the symmetry plane. Make sure to select the block(s) inside the car directly above each peg also (or select the face between the under-body and the car solid which will do the same thing). Make sure to turn off "Around Blocks" since we split this one larger than the peg. Apply. Associate the inside of this block with the curves of the little pegs.

For extra marks, you could have split off a block behind the vehicle and either put an Ogrid in it (re-circulation zone) or a CGrid all the way back to the outlet. This will help better resolve your wake... Actually, there are a lot of fancy ways you can refine the wake with the blocking structure, but many Solvers these days have a way to automatically refine the mesh as needed...

[maalan]

Thank you for your fast reply, Simon!! I got it!

[neutron]

I have trying to do blocking for a windsor body. However there seem to be automatic association being created from the domain upon the geometry. I have tried manually deletingt the associations but the mesh is still strange.

I have tried associating edge to surface as done in the previous sample but ti was unsuccessful.

Please suggest how to go forward? need help

[neutron]

These are the files. Any help would be great.