First, I really admire your work. It is indeed a hard job to such geometry using blockMesh.

For complex cases, it is really easy to increase the number of points along the Z axis but for X and Y, It is tricky. I believe you got a lot of errors like "Inconsistent number of faces between block pair". With 287 blocks it is really hard, even to spot the problem.

My advice to maintain the quality of your mesh is to use refineMesh without refineMeshDict or simply run refineMesh -all. I tried with your coarse mesh (24400 cells), it gave me final mesh with 195200 cells, pure hexahedra. Indeed, the number of non-orthogonal faces will increase but it is typical for such case.

If you want to refine a certain volume of your mesh, not the complete domain, you have to use refineMesh with refineMeshDict. But first, you have to define a cell set using topoSet utility. Based on my experience, sometimes, the interface cells quality between the refined and non-refined area became very bad. You can test it for your case and see what will happen.

Thank you again, your work really proves that how blockMesh is a very good tool.

Dear Eduardo, try having a look here

https://openfoamwiki.net/index.php/RefineMesh

Can you post the error you get from the checkMesh?

Franco

Hassan I. Kassem · Thanks for your recommendations I'm trying right now refineMesh with -all and is working very well, for now. I also have to change my blocks definition because the refineMesh seems to split the existing blocks into more pieces, but remaining its original form, and for a curved edge when the mesh is coarse, the elements are lines, and the split mesh didn’t solve this. But redefining the size of the big blocks with curved edges and with your recommendations I improve the solution. Right now I'm having a finer mesh and only one error:

***Max skewness = 11.4477, 5 highly skew faces detected which may impair the quality of the results

I’m going to continue my test to see how to change this

thanks again.

Franco Concli · Hi thanks for the link I saw this documentation about the refineMeshDict, my question about this s how to create the set of cells to apply the refinement. But after reed the recomendation of Hassan I. Kassem about the quality between the refined and non-refined area I'm gone to keep trying to refine all the mesh, not locals sets.

about the checkMesh, when I run with coarse  blocks in the x,y (5x5xZ) direction the result of checkMesh is:

Overall domain bounding box (-1 -0.5 0.2) (1 0.5 6.29)

Mesh has 3 geometric (non-empty/wedge) directions (1 1 1)

Mesh has 3 solution (non-empty) directions (1 1 1)

Boundary openness (6.39383e-16 -1.03849e-17 -2.49558e-17) OK.

Max cell openness = 4.86847e-16 OK.

Max aspect ratio = 63.3181 OK.

Minimum face area = 1.31814e-05. Maximum face area = 0.0204992. Face area magnitudes OK.

Min volume = 9.27453e-07. Max volume = 0.00184706. Total volume = 11.9841. Cell volumes OK.

Mesh non-orthogonality Max: 83.8537 average: 16.5052

*Number of severely non-orthogonal (> 70 degrees) faces: 256.

Non-orthogonality check OK.

***Max skewness = 11.4477, 5 highly skew faces detected which may impair the quality of the results

Try to check where are these faces in ParaView first and I believe with some tweeking you will be able to avoid them.

For the second checkMesh report after changing x,y, it shows really big problems. Mainly due to the extermally high aspect ratio.

For non-orthogonality and other mesh realated problems, I would recommend to read Hrvoje Jasak in UK user meeting last week. His presentations were really useful (Day1).Links below.

http://ukri-openfoam.ex.ac.uk/

http://ukri-openfoam.ex.ac.uk/docs/Monday/trainings/DiscretisationBestPractice_2014.pdf

Hello again. Thank  you very much Hassan, for the information presented in these links, there are many presentation with very good recommendations not only about the mesh, also in the configuration of the case.

Tnahk you very much again I will be publishing progress.

Hi, i'm glad to finally show my first approach of a full mesh domain of a rotor NREL 5MW using structured mesh. Now I need some advices on how to build the hub part because its very ugly and not aerodynamic in the current version. I think that using snappyHexMesh just for this part may be the solution, but snappyHexMesh didn't work on meshes created with stitchMesh, maybe.

I'm glad to hear that.

I worked on a case before in which one part of the mesh was created using blockMesh and the second part was created using snappyHexMesh and it worked without problems. First you shall need to merge then stitch. Hopefully it will work for your case.

Bw,

Hassan

Hi, Hassan, thanks again for your advices.

In my case, I tried to do it in that way. First run the snappy on the close_hub and blade_domain to extract the hub.stl geometry for each case, then merge the generated meshes and after that use the stitch to eliminate the patch. But the stitch didn't work, it always fails, even if I use it with -partial option, or changing the tolerances to a very hight values.

To me it looks like snapping problem from snappyHexMesh (green part). Sure you need partial option and the slave patch should be the smaller one.

I'm confused about your statement "First run the snappy on the close_hub and blade_domain to extract the hub.stl geometry". What do you mean?