crack propagation

[D-Nguyen]

Hello,

I try to reproduce your three point bending test (http://mofem.eng.gla.ac.uk/mofem/html/frac_mech_simple_examples.html).

- I do not use cubit but gmsh to generate the mesh.

- The crack is initiallized using 

-cut_mesh 1

-cut_surface_side_set 400

-edges_block_set 100

-vertex_block_set 101

with 400 holds the cut surface, 100 keeps  all the coorner edges, 101 keeps all corner nodes. All options, I think, are correctly defined so that the crack propagates until last failure.

The problem I do not understand is: the initial crack is planar, but during crack propagation, the mesh nodes at this initial crack moves so that this part is no longer planar (as shown in attached figures). It means that there is something still missing since the sliding contraints should be applied on this initial crack path as this part is not affected by crack propagation. Do you have any idea about this problem or I missed some options?

Best regards,

Van Dung

[Lukasz Kaczmraczyk]

Hello,

In code, it is specified how large is an area where sliding constraints apply, beyond that nodes on the crack surface are fixed in material space. If surface makes large kinks, that can happen at the start of the analysis, sliding surface constrains can give such behaviour in this specific case, but this should be limited to some small area. To avoid such problems, in keys of initially kinking crack, you can create an initial physical crack, then add a cutting surface. 

Alternatively, this could be a problem with cutting algorithm. That mesh cutting algorithm magnifies oscillations each time cut is done and has nothing to do with surface sliding constrains. In the version on web pages, crack front changes were done only at the crack front. However, we changed this in a later version, to allow users to stop analysis and change surface, add another surface, etc. Consequently,  cut all mesh at each cut step. That could have an undesired effect when kink is created.

One of the students working on this problem at the moment, working with Salome, and version v0.9.51. I will have results to compare with you and give you precise explanation.  I am going on holidays, and I could advise you better after the 22nd of July. 

BTW We work on the new release of fracture module to target that issue in cutting algorithm and improve robustness & efficiency. We improved mesh refinement algorithm at the crack front, so for cases like this one will be managed much better.  

Kind regards,

Lukasz

[D-Nguyen]

Thanks for your response.

Best regards,

Va

[Hoang Nguyen]

Hi,

Can you please upload all the input files (mesh, param_file.petsc) and the command line that you used to run the analysis. I will try to reproduce and hopefully fix the issue that you are having. BTW, can you also let me know which version of the fracture module that you are running?

Kind regards,

Hoang

[D-Nguyen]

Hello,

Here my inputs: mesh by gmsh, configure file, parameters file, and h5m file. 

I used add_meshsets -my_file model.msh -meshsets_config model.cfg to create the h5m file.

I used master branch of current version of mofem.

I also attached the solution of last failure stage (volume + crack surface). 

Best regards,

Van Dung

[Hoang Nguyen]

Hi,

Thank you for sending the files.

With some tweaks, the crack was able to propagate smoothly after about 30 steps before the structure was fully fractured and the part for the initial crack remained planar as you can see in the gif image attached.

The main changes that I made compared to your inputs include the magnitude of Young's modulus and the param_file.petsc

Regarding the Young's modulus, as you used g_c 0.352 in your param file, I supposed you used the units of Newton for force and millimetre for length, if so the magnitude of Young's modulus (2800 MPa) in config model.cfg file should also be 2800 rather than the one you set.

Regarding the param_file.petsc, I attached the new file that I used for this analysis. This file is tailored to work more effectively with new updates of the fracture module 0.9.52 which is the one that you have if you work on the current master branch.

There are some parameters you can play with when you first run the crack propagation. For example, -arc_s, -adapt_step_min_s, -adapt_step_max_s which are expected, min, max values of the increment of the crack surface in each step, respectively. Based on the initial crack surface area (in the log, it is 'Crack surface area =  2.3598e+02'), you can set those parameters appropriately to have fewer steps to run before the structure is fully fractured, of course not too big as it causes divergence. I guess you had to wait more than 100 steps because -arc_s was quite small.

Additionally, you can use the parameter -ref_before_cut 1 (or 2 or 3, etc.) to locally refine the mesh at the vicinity of the crack surface. With this parameter, you do not have to manually refine the mesh around the crack in advance. We will update the tutorial so the next users can run the problem easier.

Hope this is helpful and please let me know if you can reproduce the results.

Kind regards,

Hoang

[Hoang Nguyen]

For some reason, gif file does not work when uploaded to the group. Please see enclosed the crack surface at the last step.

[D-Nguyen]

Thanks alot. Your explanation and correction are helpful for me. 

It seems that the initial part is still deformed but the finner mesh can reduce this effect. 

Best regards,

Van Dung

 

[Lukasz Kaczmraczyk]

This a case when crack start propagate in mixed mode-I and mode-II, in that case crack is kinking, i.e. crack surface is not smooth. Mesh refinement is a way to capture this. 

However we working now on improvements for crack population, where some input parameters will be removed to improve robustness and accuracy. As I write before, it will be a new version in beginning of august which will fix some problems, with additional features for crack propagation in heterogeneous materials. We will let you know when it will be ready.