[andy_]

Originally Posted by Bob Tipton

I see many posts in the "you can force it to get your job done" category. Perhaps some of you have the luxury of expending hours or days setting up a run.

I am not familiar with cfmesh+/pro but it (now?) comes in a commercial version which will almost certainly inhibit the development of the free version. The website calls it affordable so might be worth an enquiry?

What came out of my brief look 6 months ago was that current freely available meshing software was not appropriate for engineering work except in a few specific cases. I also have some doubts about the appropriateness of some commercial meshing software but that is a separate topic. I briefly looked at working with one or two of the existing open projects but failed to get a positive response (which had also been the case in the past with a number of solvers I wanted to "fix"). The reality of open source seems to be rather different to the general perception of inclusivity though I expect there are some pretty good reasons for it.

My response has been to start working on a mesh generator I would want to use for engineering simulation work in the future. Unfortunately due to ill health progress has been glacial (I am experienced but getting old) and whether production-ready code will ever result is uncertain. It is an interesting topic though considering how best to handle adaptation and optimisation of the geometry and mesh given the range of boundary conforming and nonconforming elements likely to be in use by solvers.

[Bob Tipton]

We're doing some research on an NSF grant and budgets are tight. It's pushing the boundaries of all CFD tools so I've made some requests for "donated" tools. I'm not hopeful.

I'm working on the new mesher now and one question is, is there enough demand to recoup the time?

Our focus is external flow of highly curved shapes with vorticity, slots, ejectors and porosity. I'm not sure how many others will need it.

[arjun]

Originally Posted by Bob Tipton

I was taught that if I have nothing nice to say, I should say nothing at all.


<Long silence goes here>

I am trying to do develop devices relying on passively energized vortices, large vortices. The shapes are complex as is the fluid dynamics. I had expected the engineering challenge to be in developing the shapes and I've run about 100 trade studies. I've done enough to convince myself I have to keep going.

.

What is the thing that is changing from one study to another? Do you have to remesh it every time?

Is there any detail you can share. I just want to see if there is any efficient solution to your problem using Wildkatze. If there is something possible it might save you lots of efforts.

[Bob Tipton]

It's matrix


Angle of attack to the flow, slot size, spacing between sections, curvature of a lofted surface etc. The base model is changing geometry, and the boundary layers have to be rebuilt each time.

[arjun]

Originally Posted by Bob Tipton

It's matrix


Angle of attack to the flow, slot size, spacing between sections, curvature of a lofted surface etc. The base model is changing geometry, and the boundary layers have to be rebuilt each time.

How about if you used immersed boundary method and gave up on meshing in the start.

Then based on the results from immersed boundary you only did final calculations with only few cases with proper boundary layer etc.

[andy_]

Originally Posted by Bob Tipton

We're doing some research on an NSF grant and budgets are tight. It's pushing the boundaries of all CFD tools so I've made some requests for "donated" tools. I'm not hopeful.

When in academia my experience of asking for support for teaching from the commercial CFD providers was wholly positive. I had a minor wrinkle once at a German university when another department was already paying for the CFD software making the company unhappy about supplying it to me for no cost. Rather than press I simply used an alternative commercial CFD package. Perhaps I should add that we used our own CFD software for research and this was only for teaching. The response may not have been as positive for research although if the CFD company saw it as way to get close to some of our industrial sponsors they may have been just as, if not more, supportive.

Originally Posted by Bob Tipton

I'm working on the new mesher now and one question is, is there enough demand to recoup the time?

Demand from whom? Several times over aerodynamic surfaces I have found it more efficient to write a few lines of code to generate the specific grid I wanted with the expansion I wanted rather than wrestling with a more general grid generator. This was only for relatively straightforward shapes and only to satisfy needs within my research group rather than providing software for engineers to use.

Originally Posted by Bob Tipton

Our focus is external flow of highly curved shapes with vorticity, slots, ejectors and porosity. I'm not sure how many others will need it.

Sounds fairly similar to the sorts of things I used to be involved with many decades ago for flows related to gas turbine combustion chambers. When in academia the software was owned by the university and/or industrial sponsors and so was only ever written for the research group to use. Academics sometimes believe they have the same rights to "their" software as they do to "their" publications but this is almost never the case. If you do want your software to be used by others outside your group I would strongly recommend checking your contract first.

[Umut]

Dear tortured souls,

I joined the ranks of digital masochists by trying to generate a mesh for multi-region part with thin bodies. Any update on open-source 3D hybrid mesher?

So far, I tried usual suspects of SHM, cfmesh, salome, gmsh. Here is how each of them fails (by my experience):

SHM: Insists on uniform hexas, resulting in a mesh several orders of magnitude larger than necessary.

CfMesh: Can’t generate multiRegion mesh. (I doubt even paid version would handle thin bodies gracefully)

Salome: Basically Gmsh with (arguably) a better UI but consistently worse results. (Geometry creation is way better for sure)

Gmsh: Almost there, works for 2D or embarrassingly simple 3D cases. But for geometries slightly passing simple, cells in the vicinity of pyramids are beyond fixable/usable quality. (Please check: Hybrid Mesh Generation in Gmsh) (I must admit I didn't try this enough perhaps)

I’ve given each tool my time, patience, a non-trivial portion of my soul. I couldn’t find a reliable way for hybrid 3D mesh in open-source.

If anyone has stumbled upon a hidden gem, a secret workflow, or even a mildly effective ritual sacrifice to the meshing gods—please share. I’m all ears (with rapidly declining sanity).

[Bob Tipton]

I've been doing high end FEA/CFD/CAD/CAE work since 1984 for big outfits like McDonnell Douglas and DOD. The secret sauce is often writing your own custom mesh generator on a case by case basis. That's how most 2d airfoil meshes are made.

Meshes like this are not generated by a general purpose tool.

I'm working some PITA 3D cases where the meshers fail miserably. I've got a lot of C++ code in place already. Enough to layout the block mesh, view it, edit it, write the OF polymesh files and define boundaries. The plan is to grow that to a new tool which is way in the future, but if you're up for writing your own code - it's better than starting from nothing.

[arjun]

Originally Posted by Umut

Dear tortured souls,

CfMesh: Can’t generate multiRegion mesh. (I doubt even paid version would handle thin bodies gracefully)

I recently did a calculation that had 14 regions and many mesh to mesh interfaces in Wildkatze. However the mesh was generated by CfMesh or by Snappy (I do not remember because i got it in openfoam format).



What i did was converted the mesh for each region using openfoam to fluent converters and then combined them all into one mesh using Wildkatze and did my simulation.



Doing this in openfoam would be really painful so, company just decided to use Wildkatze for this purpose otherwise they mostly use openfoam.

[Umut]

Originally Posted by Bob Tipton

The secret sauce is often writing your own custom mesh generator on a case by case basis. That's how most 2d airfoil meshes are made.

Dear Bob,

Thank you for your reply, I truly appreciate the insight you shared from hand-on experience.

However, I’d argue that 2D bluff bodies are hardly complex geometries. In my view, there are scenarios where unstructured meshes are preferred and others where they are absolutely required. Allow me to elaborate:

• Some cases require tetrahedra. Certain software tools only allow remeshing (not smoothing) on tetrahedral cells. An argument can be made that after generating a structured mesh, hexahedral cells can be subdivided into tetrahedra. However, this approach is often overkill for a tetra-dominant mesh.

• Structured meshes can force unnecessary refinement. In some scenarios, using a structured mesh results in an unnecessarily high number of cells in regions where such resolution isn’t needed—consider, for instance, a large backward-facing step domain.

• Complex 3D internal flow geometries from customers. Sometimes, customers provide a series of intricate 3D internal flow geometries (e.g., STEP files) and expect results quickly.

• Parametric design surveys benefit from structured meshes. If I’m allowed to generate the geometry and plan for a design survey with many similar geometries, it’s worth the effort to create a block-structured mesh. I once ran 1,200 different 2D car side silhouettes for a generative design study, writing scripts to generate 1,200 unique geometries.

I want to use non-commercial tools for industrial applications. Therefore I want to prepare streamlined workflows for various mesh scenarios. I believe a reliable hybrid mesh generating open-source tool would be invaluable. (Yet I failed to find one )

[Umut]

Originally Posted by arjun

I recently did a calculation that had 14 regions and many mesh to mesh interfaces in Wildkatze. However the mesh was generated by CfMesh or by Snappy (I do not remember because i got it in openfoam format).

Hello Arjun,

That is an interesting workflow, thank you for sharing it. I’m not sure if I fully understand it because I never used Wildkatse.

So, there was already a polyMesh folder with 14 cellZones (or regions) with acceptable quality. Was the mesh at the region-to-region interfaces were not conformal? Why required interfaces?

This may not be a remedy for my situation at hand though. I’m having trouble to generate a mesh with acceptable quality. Imagine there is a thin and wide body in the computational domain where the whole geometry is too complex to generate block-structured mesh.