Meshing of a parabolic trough
 

Hello everybody,

I am working on this case now for quite a long time and haven't found the solution for my problem so i thought maybe you can help me out. I am a beginner at ICEM so maybe i just lack of knowledge and my problem isn't that difficult to solve.

We want to create a Hexa mesh around a parabolic trough. I have to make a numerical simulation for three different angles of the trough. You can see two of them in the pictures.

First I tried to create a block structured mesh with an O-Grid around the trough. But since the trough is very close to the bottom i wasn't able to create a high quality mesh (picture 1 shows the trough and the bottom).

After that I tried some tetra meshing using the octree method with prism layer around the mesh and at the bottom of the wind tunnel. This mesh was the best to this point but since my task is to create a hexa mesh I can't use it. It's still not perfect because i quit at some point when i understood how to create a tetra mesh in general.

At the moment i try to create a unstructured hexa mesh using the BFCart mesher. It took me quite some time while testing on how to create an unstructured hexa mesh to realize that the mesh has to be very dense around the trough. The mesh always went through the trough before. But since the trough has a very small thickness even very small hexas around the trough are still to big (see picture 2).

So once again, I am new to ICEM and this Forum so I hope you do understand my problem.

Thanks in advance!

create an ogrid around geometry

Hello Far,

thank you for your answer.

I already tried an O-Grid and it didn't work out that great.
I uploaded a picture of my O-Grid so you can see where i have trouble.
I'd be very happy for any advice to improve this grid :)

edit: Of course this grid needs some refinement etc. but my main problem the transition between the O-Grid and the surrounding mesh especially in the bottom area remains the same I think.

you are almost there. just more splits and move vertices. Also reduce the size of o-block (go to edit block> rescale oblock).

You may need some advance topology strategy. See the video tutorial series (famous 3 part YouTube video on air-foil meshing by Simon) where he has explained how to handle high staggered air-foil with advance blocking.

Thanks for your quick response!
I'll work on this mesh again after watching your suggested series.

edit: The series I found is for a 2D modell. Mine is 3D. Is it the one you were talking about? http://www.youtube.com/watch?v=tYrbScUH9RE

I cannot open link from my office (Youtube is banned). Here is the Google search revealed same videos I was referring http://www.google.com.pk/search?suge...+meshing+3part

http://www.veengle.com/s/ICEM-CFD/2.html

I worked on this basically the whole day and wasn't able to get it done.
I tried the rescaling of the O-Grid and got the error "No rescale direction specified." I don't know where I can specify this and I am not sure how this rescaling option works generally.

I watched the Tutorial. It was very interesting but I couldn't get too much out of it concerning my Problem.

While the mesh was mostly OK my main problem is still the bottom part of the O-Grid where the mesh becomes very dense (as shown in the attached picture). I wasn't able to fix this problem with my little ICEM knowledge.

send files

There you go.

I hope my work is not too bad.

The farfield is huge in both x-directions because we wanna use this mesh to for two calculations to simulate different angles of the trough. Your welcome to comment on this also ;)

Better Topology
 

Your topology is not great...

Instead of thinking of this as some box at an incline, think of it as 1/4 of a circle (imagine the ogrid creating a box around the theoretical center) or as a fillet (imagine the surfaces extending up stream and up towards the top of the box...

Here I will show how to capture the latter...

Start with two splits, one below and one behind the curve...
Attachment 13622

Then put in a quarter Ogrid (one block with 4 faces) to capture the curvature.
Attachment 13623

I rescaled my Ogrid (0.3) to be closer to fitting the model...
Attachment 13624

Then I split out to fit the blade... Split upstream, downstream, upper surface, lower surface to box it out.
Attachment 13625

Other verts could be adjusted also to improve the angles, etc. Then use "Align verts" to get it nice and crisp.


I don't have time to complete it for you, but this should get you started.

Also, here is a 30 second start on the other option, the circle option... Here is the basic blocking after 4 splits and an Ogrid thru the model (one block, 2 faces). From here, you would split out the airfoil, etc.
Attachment 13626

Here is what i got... i don't know why winrar couldn't compress it in zip format, so i made it in my dropbox, here is the link:
https://dl.dropbox.com/u/35161486/parabolic.rar



https://dl.dropbox.com/u/35161486/cfdonline.JPG

picture is huge :o sorry about that

Yes, now the question is if that resolution is sufficient... (clearly it is not yet, but I assume you plan to keep working on it).

I hadn't really taken a close look at the shape of your airfoil before... Quality on the curved "corners" won't be ideal. You could probably also place a small Ogrid around the airfoil blocks for an extra refined boundary layer...

true... because refining on a laptop requires a bit more of memory... but i added and o-grid around. hope it's correct
here is new link :
https://dl.dropbox.com/u/35161486/barabolic2.rar

I don't know why this pictures are so big !!!

https://dl.dropbox.com/u/35161486/cfdonline2.JPG

https://dl.dropbox.com/u/35161486/cfdonline3.JPG

of course quality can be improved more when playing with vertices...i'll play with it more in lab. PCs are faster there.
thanks Simon for this blocking strategy and the advices .

On the right track...
 

Yea, I think you are on the right track now...

The topology is done, and you just need to spend some time fine-tuning the vertex placement and working out the edge distributions...

Make sure you learn how to use "Align Vertex" and "Set Location", both of these are under "Move Vertex" and can really help you align everything for maximum quality... You control the extent of their influence with the index control, so hopefully you have figured out how to use that also.

Wow, these are great ideas.

Thank you very much everybody.

I'll try to do it in a similar way on my own and post again when I am done.

I have managed to make blocking as Simon suggested. Still internal o-block is remaining and also some splits on wing or whatever ;)

Horizontal split is necessary?

Should we crate the y-blcok at the corner where two splits meet?


Which option is better, this one or quarter Ogrid?

@ Far

The horizontal split assumes that there is a floor in this model... I assumed there is or the model would have centered the scoop. If there is a floor and we are considering viscous effects... It is nice to keep that boundary layer clean. Without that lower horizontal split, the boundary layer would peel away from the wall and shoot straight up, wasting mesh in the volume, creating volume jump problems and leaving no boundary layer for downstream...

No idea what you mean about the Y block... But I don't think anything else is necessary. Simpler is usually better, even if the quality is just a little worse. This blocking suggestion should give more than adequate quality.

That depends on what the flow looks like (or what is expected). If the flow is fast and this scoop sends a large portion upward for a long distance, the Quarter Ogrid will be good because you can get higher resolution in that area above (downstream) of the scoop. You could even come back and adjust the blocking slightly to align the vertical portion of the Ogrid with the flow and get better resolution.

But if the flow is lower speed and not much happens between the scoop and the top of the airfoil, then that mesh refinement is wasted in the quarter Ogrid and the circle method may be better... The circle method also has the advantage of being somewhat self contained... In other words, it doesn't propagate and cause difficulty elsewhere in the model. That is not a concern for this particular case, but imagine if we had many such scoops and all the quarter ogrids started intersecting, etc. It could become a mess.

Here is the final blocking with finishing touches. It took less than a hour to make the blocking and another hour to fine-tune it.


But took hours to rectify some duplicate blocks which were never created intentionally and min quality was ~-0.9 . These were so thin, when blocks were activated they were not visible with naked eye. So when moved some vertices, then I realized double vertices hence duplicate blocks. So moved them and merged them with internal vertices (on inside Oblock in the wing or scoop).

After removing them, now min quality is higher than 0.4. Also during this process (above paragraph), I was trying different processes to improve quality, one of them was to increase the no. of nodes on edges. Therefore mesh size was increased to around 4 million, now you can reduce the size without affecting quality much.

I am attaching both files (with problems of duplicate blocks and final good quality blocking), so this may help others how the problem can popup from nowhere and how to work around in ICEM to deal with them.

Hope you like my little effort (credit goes to the creater of this blocking, Simon)

Pics
 

I forgot to place the horizontal split. But it is important to do so

Pics:
http://img13.imageshack.us/img13/8550/58896525.png
http://img11.imageshack.us/img11/5889/57865644.png
http://img42.imageshack.us/img42/5211/63858281.png
http://img9.imageshack.us/img9/5406/46283969.png
http://img163.imageshack.us/img163/5043/14158267.png
http://img72.imageshack.us/img72/7579/73831113.png
http://img717.imageshack.us/img717/2430/37566679.png

So I tried to create a mesh with your ideas but I have problems to find the reason why the mesh is very bad around the trough.I spent a lot of hours figuring it out and trying a lot of things to get this mesh better but failed.

I hope you can help me out once again.

I attached some pictures (third pictures shows all cells <0.4 quality) and a zip file so that you can see what my problem is.


Thanks again for your great help!

Align edges with curves. For this use edit edge menu.

Thanks for your answer.

When i go to Blocking -> Edit Edges I have the options Split Edge, Unsplit Edge, Link Edge and Unlink Edge.

I find the align edges with curves option under Geometry->Create Faceted->Align edges to curve.

I tried both the Link Edge tool and the Align edge to curve tool this morning but couldn't eliminate the bad cells.

Do you have thin blocks (as discussed above). I believe it is not the problem with your blocking but you can recheck. Lets say it has 10% priority in our list.

Now the major problem is the edge to curve association. By default edge will follow the nearest curve (This is serious problem when you selected all edges and all curves while making edge-curve association) . In this case you must apply the spline or automatic linear (use this command after setting the edge mesh parameters with enough no of nodes : at-least 50-80) on green edges. Once this is done then go to 3rd button > link edge shape and select the appropriate green edge and then black edge. Repeat this procedure for both sides of wing or trough. Also for the edges on the outer boundary.

If you compare your blocking with mine, you will notice extra split on the air-foil trailing edge. It is necessary for same reasons as discussed above (Fig. 4 & 5).

I already looked for thin blocks yesterday and could not find any. I moved all the vertices around the trough but there were no double vertices. But the bad cells which are shown in my third picture above have no thickness at all. This would be the result of very thin blocks, right? I just might just have not found the thin blocks? Could there be another reason why these cells have not thickness?


I added a picture which shows my associated edges to curves. Its basically the inner part of the O-Grid which is associated to the curves of the trough on both sides.

I'll try your suggestions of linking the edges. After that I'll try your suggested blocking on the trailing edge.

I dont think so it is the case with your blocking.


Yes, association is correct. But as I described already it will create problem. See this thread http://www.cfd-online.com/Forums/ans...ng-issues.html

I tried this in different ways and i still have those bad cells.
I splitted the green associated edged with automatic linear after I increased the number of nodes to 80. The edges "jumped" on the curves of the trough.
After that I linked the blue edges around the trough to the green ones and they were shaped like the trough after that. I also tried to cut the edges of the trailing edge before and after this procedure but it didn't solve my problem.

I also forgot to mention that i splitted the blocks in the field one more time to be able to create a nearfield later (see pictures). I hope this doesn't cause these bad cells.

Once again thanks for your help, Far. I appreciate it A LOT!

I also noticed that I have an extra split in the middle of the trough. There seem to be to edges which are almost identical but there is a small gap between them. See pictures (from old a blocking version) to compare your splitting to mine.

edit: I merge some blocks there and this problem seems to be fixed now
edit2: Maybe this solved the problem. My Laptop ran out of battery and I have no power supply for it here but with his last breath I created a pre mesh where the worst quality was -0.1. That is the best I had to this point so I hope I will create a good mesh on the weekend.

Alright, I still have some bad cells left which have a quality of ~-0.1.

I might found the reason but wasn't be able to fix it. The boundary edges aren't split like the edges directly at the trough. The splits from the O-Grid around the trough do not go through the domain. I tried to split the blocks afterwards but could not find the right tool to do so. To explain my problem I added two pictures. First the boundary edges which are nicely splitted (Fars work) and picture two shows my boundary edges which have not the splits of the O-Grid.

edit: Ok, solved this problem also. Now my worst cell has a quality of ~0.35 without finetuning the mesh. I hope this will work now.

Once again, thanks everybody. I really appreciate your help.

Great....We have also learned new topology tips from Simon while working on this meshing.