I have simulated a concrete cube under tensile loading. For this purpose, I have activate element removal option from mesh module and I input 0.95 for max degradation( corresponding to max damage parameter that is specified in material property), But When I run my model and check the results, I notice that elements have not removed from the mesh.Also, I have compared the resulted stress-strain curve by experimental curves and there is considerable difference between them in softening branch (its figure is attached ) and I think it is due to element removal.
What is your opinion?
Dear Mirshad,
Your stress strain results show mesh dependency. If you are using local damage model then you will have this problem of pathalogical mesh dependence.
Hi Muhammad,
I have run my model by element length 1 and 2 mm. But the result did not change very much and resulted curve does not have any change specially in softening branch .Can you explain more?
In the figure you have attached the energy dissipation is zero as soon as softening begins. In local damage models, damage and strain accumulates in one row of elements so the energy dissipation in local deformation band becomes mesh dependent. Perhaps 1mm or 2mm are very small compared to the actual (experimental) local deformation band so you dont see a big difference. Use significantly larger elements and you may see the difference.
To resolve this issue you either need to use a nonlocal formulation or any other regularization technique which is more suitable for your application.
See chapter 4. Everything is explained and you will find a lot of useful references.
http://doc.utwente.nl/82517/
Dear Muhammad,
Thank you for your help. I tried by element length 5 mm too recently, but the softening branch did not have any significant change. Since, the cube has dimension 25*25*25 mm, I think increasing element length more than 5 mm is not rational. On other hand,I saw In a paper, they had used Concrete-Damage-Plasticity model and they had got proper stress-strain curve by element length 0.5 mm for cube 25*25*25 mm but with a different, the mesh shape in paper was not same mine, I mean the partitioning shape of cube for meshing was different.
Can you show the damage and strain distribution in the mesh? Zoom into the damage region.
I have attached 3 figures from each side of cube(L1,L2,L3,R1,R2,R3) in different frames. As you see, damage in each frame than previous frame is increased. I have 500 frame in my analysis, but here, I killed the analysis at frame 162. It sould be noted that I have used irregular mesh too, but reulted curve did not have significant change and the different was only in location of damage band.
Assuming that red color represents max damage and blue color represent 0 damage, you can see that one row of elements have maximum damage and the neighbouring rows have 0 damage.
Assuming that the element size in these figures is 2mm, how does it compare with damage distribution when element size is 1mm? Do you see the red color in two rows in that case or in one row? If damage accumulates always in one row when you change mesh size then it shows mesh dependency.
Apart from the mesh dependency, you can also play with the damage model parameters. Set the damage parameters such that damage grows very slowly then check how the stress strain curve look like.
See the attachment
- Badges
- Science topic