Standard Mesher setup at AutoSize level 4 is claiming 80,000,000 cells and counting

DaleKramer · January 18, 2024, 6:54am

I have this Standard Algorithm mesh that grew enormously, what have I done to make it so big?

The 2 refinements I used are the same as I used for my Hex Dominant Parametric meshes which stayed around 15 million cells.

I stopped it at over 80 million cells. Here is the log:

Model import took 20.958646361s.
Maximum precision of model and its entities: 0.001031866166658002 m.
Absolute small feature tolerance: 0.001 m.
Layer refinement was not applied on 8 faces because they were removed by small feature suppression, by the non-manifold boolean for conformal meshing or assigned as start/end faces in a sweep mesh refimenent.
Number of faces after 5m0.046066471s: 986977
Number of faces after 10m0.020021496s: 3740308
Number of faces after 15m0.009116597s: 4715356
Number of faces after 20m0.031016382s: 4715356
Number of faces after 25m0.000201957s: 4715356
Number of faces after 30m0.000317487s: 4715356
Number of faces after 35m0.004106731s: 1674168
Number of faces after 40m0.004760424s: 4232347
Number of faces after 45m0.099892596s: 4743954
Surface meshing took 49m41.95172042s.
Number of cells after 1h14m32.927909123s: 13525231
The mesh may be finer than desired. Please check the model and the mesh settings
Number of cells after 1h39m23.90492571s: 81189113

DaleKramer · January 18, 2024, 7:26pm

Its happening again right now here after I decided to just use a single surface refinement of all faces of my aircraft at a max size of 4 mm, which is an absolute requirement to keep the mesh smooth enough.

Should I cancel it or do the cells get reduced again before completion?

Here is the current meshing log:

Model import took 25.186151475s.

Maximum precision of model and its entities: 0.001031866166658002 m.

Absolute small feature tolerance: 0.001 m.

Layer refinement was not applied on 8 faces because they were removed by small feature suppression, by the non-manifold boolean for conformal meshing or assigned as start/end faces in a sweep mesh refimenent.

Number of faces after 5m0.045202503s: 856136

Number of faces after 10m0.012967398s: 2852486

Number of faces after 15m0.099823476s: 4620310

Number of faces after 20m0.048269625s: 4620310

Number of faces after 25m0.05978723s: 4620310

Number of faces after 30m0.059958988s: 4620310

Number of faces after 35m0.047142006s: 212726

Number of faces after 40m0.049616695s: 1973660

Number of faces after 45m0.056432756s: 4038548

Number of faces after 50m0.00628886s: 4649126

Number of faces after 55m0.020957842s: 4649126

Surface meshing took 58m29.980803959s.

Number of cells after 1h27m45.071278619s: 13532729

The mesh may be finer than desired. Please check the model and the mesh settings

Number of cells after 1h56m59.975564728s: 76996534

DaleKramer · January 18, 2024, 8:07pm

I had to stop it since there was no downward progression to about 15M cells:

Here is where I stopped it:

Model import took 25.186151475s.

Maximum precision of model and its entities: 0.001031866166658002 m.

Absolute small feature tolerance: 0.001 m.

Layer refinement was not applied on 8 faces because they were removed by small feature suppression, by the non-manifold boolean for conformal meshing or assigned as start/end faces in a sweep mesh refimenent.

Number of faces after 5m0.045202503s: 856136

Number of faces after 10m0.012967398s: 2852486

Number of faces after 15m0.099823476s: 4620310

Number of faces after 20m0.048269625s: 4620310

Number of faces after 25m0.05978723s: 4620310

Number of faces after 30m0.059958988s: 4620310

Number of faces after 35m0.047142006s: 212726

Number of faces after 40m0.049616695s: 1973660

Number of faces after 45m0.056432756s: 4038548

Number of faces after 50m0.00628886s: 4649126

Number of faces after 55m0.020957842s: 4649126

Surface meshing took 58m29.980803959s.

Number of cells after 1h27m45.071278619s: 13532729

The mesh may be finer than desired. Please check the model and the mesh settings

Number of cells after 1h56m59.975564728s: 76996534

Number of cells after 2h26m14.95250846s: 72346302

Number of cells after 2h55m29.942915213s: 72478320

And I was really trying to see how well the Standard TET Meshing algorithm was up to my Real World to CFD verification project, I guess not

Without surface refinements I was not able get a mesh that was not bumpy…

igaviano · June 27, 2024, 8:21pm

Hi @DaleKramer, sorry for the very late reply here

Our team was going through some restructuring and only now we’re having the time to answer some posts that were left off from months past!

I’ll do my best to try to answer your questions and provide some insights that might be helpful.

As far as I understand it, tetrahedrical meshes will naturally be larger (when talking about cell count) due to them being composed of tetrahedrons. If we consider for instance an hypothetical regular tetrahedron which side is given by a, it’s volume would be given by

\displaystyle V_{tet} = \frac{a^3}{6 \sqrt{2}} \tag{1}

A cube with same side would have the volume of:

\displaystyle V_{cube} = a^3 \tag{2}

Dividing (2) by (1), we’d get a value of \approx 8.4. Of course, the mesh won’t be exactly that much larger due to the cell’s growth being a bit different depending on the method chosen but we should expect it to be reasonably larger (cellwise).

Since you have a very large domain, it will also be very sensitive to refinements. Your model is composed of faces as small as 0.001m and any refinement that is applied to a region with more resolution than necessary will make the cell count increase exponentially.

As we don’t have access to that large mesh you’ve generated and the inserted parameters, we can only give general advice here! But please, feel free to open a new post if you feel like your question wasn’t fully answered.

Best,
Igor