Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
A Swept Quad Mesh Within Free Tet - Quality Concerns
Posted 18.03.2017, 01:00 GMT-4 Heat Transfer & Phase Change, Geometry, Mesh, Structural Mechanics Version 5.2 6 Replies
Please login with a confirmed email address before reporting spam
Hi everyone,
I'm looking for some advice on a meshing problem that I have - I have a heat load imposed on a 3D rectangular beam embedded within a larger beam of the same material; the two beams form a union. The area where the heat load is applied is very finely meshed with a mapped distribution so that I can control the exact number of elements in my area of interest.
This is great but my problem is that in order to cut down on computation time/memory cost - I am converting the boundaries of this swept block (or another swept block around it) to tri's and then putting in free tetrahedral mesh for the large remainder of the geometry. This has the unfortunate effect of creating a region of poor mesh element quality around the swept block where I need my mesh to be the most accurate.
I have attached an image of what the mesh element quality looks like in one iteration of this model. It seems that every time I try a slightly different meshing scheme - my results are significantly different and I have tried many variations. If anyone can suggest a better alternative to my current meshing scheme, or point me in the right direction of increasing the mesh quality, I would greatly appreciate it.
Finally, my heat loads are asymmetric and so unfortunately I cannot exploit any symmetry to cut down on the computational cost.
I'm looking for some advice on a meshing problem that I have - I have a heat load imposed on a 3D rectangular beam embedded within a larger beam of the same material; the two beams form a union. The area where the heat load is applied is very finely meshed with a mapped distribution so that I can control the exact number of elements in my area of interest.
This is great but my problem is that in order to cut down on computation time/memory cost - I am converting the boundaries of this swept block (or another swept block around it) to tri's and then putting in free tetrahedral mesh for the large remainder of the geometry. This has the unfortunate effect of creating a region of poor mesh element quality around the swept block where I need my mesh to be the most accurate.
I have attached an image of what the mesh element quality looks like in one iteration of this model. It seems that every time I try a slightly different meshing scheme - my results are significantly different and I have tried many variations. If anyone can suggest a better alternative to my current meshing scheme, or point me in the right direction of increasing the mesh quality, I would greatly appreciate it.
Finally, my heat loads are asymmetric and so unfortunately I cannot exploit any symmetry to cut down on the computational cost.
6 Replies Last Post 24.03.2017, 10:12 GMT-4
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
I wanted to update my post and bump it back up since I posted it on a Friday and it may have gotten lost over the weekend - I have not been able to resolve the issue of a low mesh quality at the border of the converted quad mesh and the free tetrahedral. Someone suggest that I boolean the parts, but I think that this just means forming a union which I'm certain has already happened.
I'm hoping that there may be some master meshers around that can give me a clue on how to improve this mesh quality problem. Thanks again.
Cheers
I'm hoping that there may be some master meshers around that can give me a clue on how to improve this mesh quality problem. Thanks again.
Cheers
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Nicholas-
Interesting challenge.
Without having access to the actual mph file, my initial suggestion would be switching to a Free Tetrahedral meshing algorithm and controlling the maximum element size of the domain/boundaries of interest. The COMSOL free meshers have some automatic options for improving element quality that may be of benefit. Is there any specific reason you need to control the exact number of elements? Seems to me that the maximum element size may be a better parameter to "control" when it comes to studying mesh dependencies in regions of interest.
--
Best regards,
Josh Thomas
AltaSim Technologies
Interesting challenge.
Without having access to the actual mph file, my initial suggestion would be switching to a Free Tetrahedral meshing algorithm and controlling the maximum element size of the domain/boundaries of interest. The COMSOL free meshers have some automatic options for improving element quality that may be of benefit. Is there any specific reason you need to control the exact number of elements? Seems to me that the maximum element size may be a better parameter to "control" when it comes to studying mesh dependencies in regions of interest.
--
Best regards,
Josh Thomas
AltaSim Technologies
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Nicholas-
Interesting challenge.
Without having access to the actual mph file, my initial suggestion would be switching to a Free Tetrahedral meshing algorithm and controlling the maximum element size of the domain/boundaries of interest. The COMSOL free meshers have some automatic options for improving element quality that may be of benefit. Is there any specific reason you need to control the exact number of elements? Seems to me that the maximum element size may be a better parameter to "control" when it comes to studying mesh dependencies in regions of interest.
--
Best regards,
Josh Thomas
AltaSim Technologies
Hi Josh,
Thank you very much for your response! I'll give this a try and update the post with my results - the reason that I would like to control the number of elements is to allow the user of my application to pick how many elements they would like in a regular grid along the surface of the heat load. I then extract the subsequent heat load data as a corresponding regular grid. I suppose that I could determine a relationship between maximum element size and number of elements?
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Hi guys, I attempted to solve my mesh quality problem by creating successive layers of blocks with incrementally lower and lower maximum mesh size but it appears that mesh quality is now lowest in the area where I want it to be highest. I have attached a screenshot of a mesh volume quality plot where I tried a free boundary triangular mesh on the top center block and made the max element size very low - but the quality is still poor and strange. I also attached a mesh quality plot of a cross-section of the same model and basically identical mesh. I can't figure out why the quality would be so poor in the center where the element size is set to be smallest. Do I really need to make the elements that much smaller?
Please let me know if you have any suggestions. Thanks very much!
Please let me know if you have any suggestions. Thanks very much!
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Nicholas-
Mesh quality is a measure of the stretched nature of an element not of size. You can have large elements with poor quality or small elements with poor quality. If elements have high aspect ratio they have a lower quality. It is hard to see from your images, but perhaps your z-height of the domain of interest is very shallow so the elements with low quality are small yet the z-height is very, very small causing the stretching of the tetrahedral elements.
Also note that having some lower quality elements doesn't necessarily mean an unacceptable solution. Most meshes are going to have at least some elements with element quality on the lower range of the spectrum. If you are running into convergence difficulties or your results are mesh dependent then you will want to consider adjusting mesh resolution and improving element quality of the worst offenders.
--
Best regards,
Josh Thomas
AltaSim Technologies
Mesh quality is a measure of the stretched nature of an element not of size. You can have large elements with poor quality or small elements with poor quality. If elements have high aspect ratio they have a lower quality. It is hard to see from your images, but perhaps your z-height of the domain of interest is very shallow so the elements with low quality are small yet the z-height is very, very small causing the stretching of the tetrahedral elements.
Also note that having some lower quality elements doesn't necessarily mean an unacceptable solution. Most meshes are going to have at least some elements with element quality on the lower range of the spectrum. If you are running into convergence difficulties or your results are mesh dependent then you will want to consider adjusting mesh resolution and improving element quality of the worst offenders.
--
Best regards,
Josh Thomas
AltaSim Technologies
Please login with a confirmed email address before reporting spam
Posted:
7 years ago
Thanks again for the sage advice Josh; I am beginning to understand that and I actually found the explicit expressions by which quality is calculated. I'm going to have another look at the model today and hopefully resolve the issue.
Also - I think you hit the nail on the head about the z height of the domain - perhaps I can change the mesh scale in that direction?
Best regards,
Nick
Also - I think you hit the nail on the head about the z height of the domain - perhaps I can change the mesh scale in that direction?
Best regards,
Nick