search

KristofferC / Phon

Insert cohesive elements between grains in microstructures
MIT License
29 stars 19 forks source link

Bad quality of cohesive elements when prescribing thickness #14

Open SagarChandra opened 8 years ago

SagarChandra commented 8 years ago

Dear Proff. Kristoffer,

With respect to Issue no. 13 (https://github.com/KristofferC/Phon/issues/13), i visualized the cohesive elements in Abaqus after giving them prescribed thickness of 0.5 mm, in a model of 40mm cube containing 10 grains as a test case.

As you can find in the images attached, the element quality is bad. (distorted elements even when the thickness is reduced to 0.1 mm). Can you please shed some light on this issue. I use Phon 0.4.

Thanking you. Sincerely, Sagar

1 2

KristofferC commented 8 years ago

I am not sure I have even tried this with quad meshes, only tetrahedrons. It looks like something strange happens where the more than two grains intersect. I will not be able to work on this for a long time so you know.

SagarChandra commented 8 years ago

Dear Proff. Kristoffer,

The prime issue seems to be in nodal arrangements and co-ordinates. To test this, i created a simple bicrystal (input file attached) and gave thickness to cohesive elements. When viewed in Abaqus, the following image results:-

image

As you can see, in viewport 1 (left), the boundary elements are Ok, with some finite thickness (as desired). In viewport two (center), the complete model is shown, where it can be noticed that some elements of either of two grains/boundary are overlapping each other. Viewport 3 (right) shows only the grains (cohesive elements hidden). Clearly, the grains have overlapped and the common overlapped region is the grain boundary.

This is the case of a simple bicrystal geometry; which dictates that the issue rests in nodal rearrangements/coordinates.

KristofferC commented 8 years ago

Interesting. Could you plot the normals of the cohseive elements in your picture to the left (and zoom in). There is a display option to enable showing normals. It looks like the elements are connected inside out.

SagarChandra commented 8 years ago

Dear Kristoffer,

So far the only way I could find to visually view any element normals is from Common Plot Options to turn on normals for elements, but this can be done only after running a job, in postprocessing, in Abaqus. However, while trying to create a dummy job for the corresponding input file, the Abaqus throws the error of "The node numbering might not be correct for element 1542 instance tess-1". I tried to renumber the nodes and run the analysis, but same problem occurred.

I couldn't find any other option to view element normals in Abaqus in any module, except in visualization module, which requires an .odb file. And that .odb file will be created only if the job runs successfully. Could this error occur due to element -node connectivity that i pointed originally??

I am attaching the .inp file here.

On Fri, Jun 17, 2016 at 10:42 PM, Kristoffer Carlsson < notifications@github.com> wrote:

Interesting. Could you plot the normals of the cohseive elements in your picture to the left. There is a display option to enable showing normals. It looks like the elements are connected inside out.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/KristofferC/Phon/issues/14#issuecomment-226827116, or mute the thread https://github.com/notifications/unsubscribe/ASXwICNq77Wc0JqvI1lbphtv_IglMUuhks5qMtWZgaJpZM4I2ILe .

SagarChandra commented 8 years ago

Dear Proff. Kristoffer,

Please find attached the image of element normals of the meshed file. The left viewport shows the cohesive elements, while center and right viewport shows the elements in bottom and top grain respectively.

On Mon, Jun 20, 2016 at 12:59 PM, Sagar Chandra <tikku.chandragarg@gmail.com

wrote:

Dear Kristoffer,

So far the only way I could find to visually view any element normals is from Common Plot Options to turn on normals for elements, but this can be done only after running a job, in postprocessing, in Abaqus. However, while trying to create a dummy job for the corresponding input file, the Abaqus throws the error of "The node numbering might not be correct for element 1542 instance tess-1". I tried to renumber the nodes and run the analysis, but same problem occurred.

I couldn't find any other option to view element normals in Abaqus in any module, except in visualization module, which requires an .odb file. And that .odb file will be created only if the job runs successfully. Could this error occur due to element -node connectivity that i pointed originally??

I am attaching the .inp file here.

On Fri, Jun 17, 2016 at 10:42 PM, Kristoffer Carlsson < notifications@github.com> wrote:

Interesting. Could you plot the normals of the cohseive elements in your picture to the left. There is a display option to enable showing normals. It looks like the elements are connected inside out.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/KristofferC/Phon/issues/14#issuecomment-226827116, or mute the thread https://github.com/notifications/unsubscribe/ASXwICNq77Wc0JqvI1lbphtv_IglMUuhks5qMtWZgaJpZM4I2ILe .

SagarChandra commented 8 years ago

Please find attached the image of element normals of the meshed file. The left viewport shows the cohesive elements, while center and right viewport shows the elements in bottom and top grain respectively.

normals

SagarChandra commented 8 years ago

Dear Proff. Kistoffer,

Is the issue of element thickness for quad meshes resolved now?

On Wed, Jun 22, 2016 at 10:36 AM, Sagar Chandra <tikku.chandragarg@gmail.com

wrote:

Dear Proff. Kristoffer,

Please find attached the image of element normals of the meshed file. The left viewport shows the cohesive elements, while center and right viewport shows the elements in bottom and top grain respectively.

On Mon, Jun 20, 2016 at 12:59 PM, Sagar Chandra < tikku.chandragarg@gmail.com> wrote:

Dear Kristoffer,

So far the only way I could find to visually view any element normals is from Common Plot Options to turn on normals for elements, but this can be done only after running a job, in postprocessing, in Abaqus. However, while trying to create a dummy job for the corresponding input file, the Abaqus throws the error of "The node numbering might not be correct for element 1542 instance tess-1". I tried to renumber the nodes and run the analysis, but same problem occurred.

I couldn't find any other option to view element normals in Abaqus in any module, except in visualization module, which requires an .odb file. And that .odb file will be created only if the job runs successfully. Could this error occur due to element -node connectivity that i pointed originally??

I am attaching the .inp file here.

On Fri, Jun 17, 2016 at 10:42 PM, Kristoffer Carlsson < notifications@github.com> wrote:

Interesting. Could you plot the normals of the cohseive elements in your picture to the left. There is a display option to enable showing normals. It looks like the elements are connected inside out.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/KristofferC/Phon/issues/14#issuecomment-226827116, or mute the thread https://github.com/notifications/unsubscribe/ASXwICNq77Wc0JqvI1lbphtv_IglMUuhks5qMtWZgaJpZM4I2ILe .

KristofferC commented 8 years ago

Sorry no, I have not have time to work on it. Are you sure this is only a problem when there is finite thickness

One possible thing could be that I am moving them in the wrong direction for quads. Try to change this line: https://github.com/KristofferC/Phon/blob/653aec2ec14eb3dda0ebbc18e81d86e7abbca2ee/phon/mesh_tools/create_matrix.py#L92 to use a minus and see if it helps.

SagarChandra commented 8 years ago

Dear Proff. Kristoffer,

The issue of elements connected 'inside-out' seems to be resolved by changing node.c += r to node.c -= r in create_matrix.py. Thank you very much for the suggestion.

Irregular elements also seem to be removed.

[image: Inline image 1]

However, few cohesive elements have variable thickness; i.e. some edges have either greater thickness than prescribed, while some have no thickness at all. An example of zero thickness at an edge is shown below:-

[image: Inline image 3]

I am attaching the neper output inp file, and also the phon output file for you to have a closer look, where the thickness of 8 microns was given to cohesive elements in a 500 microns cube.

On Thu, Sep 1, 2016 at 1:14 PM, Kristoffer Carlsson < notifications@github.com> wrote:

Sorry no, I have not have time to work on it. Are you sure this is only a problem when there is finite thickness

One possible thing could be that I am moving them in the wrong direction for quads. Try to change this line: https://github.com/ KristofferC/Phon/blob/653aec2ec14eb3dda0ebbc18e81d86 e7abbca2ee/phon/mesh_tools/create_matrix.py#L92 to use a minus and see if it helps.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/KristofferC/Phon/issues/14#issuecomment-244001703, or mute the thread https://github.com/notifications/unsubscribe-auth/ASXwIHYL8fD8NekWz4u2LRVCHkQJF9Y3ks5qloJegaJpZM4I2ILe .