Discussion Forum

Hello Markus,

I know nothing about crack modeling. What I do know is that COMSOL is capable of modeling any system that can be described by equations; whether they be PDE, ODE, or algebraic, or any combination. Further, COMSOL gives you all the tools by which you could generate inputs, geometries, meshes, physics, solutions, and examine results. So, if you yourself do understand the mathematical description of the cracks you want to model, then you could then implement them into COMSOL and gain a simulated solution to that problem. Perhaps you should give that a try instead of waiting for so long for someone else to do it. Then, if you gain a following of other COMSOL users who also want to model cracks, then you might develop a business in crack modeling, and then you would have the crack modeling business for yourself or even a new COMSOL module. Who knows, you might even be able to become the COMSOL crack module leader. That sounds exciting !

Hello James,

thank you for the kind reply! Looks like your pushing me into getting Santa Clause myself.

I am quite aware of the Comsol capabilities and the methods to control and get access to the deep levels of the computations in Comsol via Matlab. But this has some limitations - due to usability and due to the lack of my leisure time. There have been some attempts to couple other computational approaches to implement crack modeling with Comsol (see e.g. A SYNERGISTIC APPROACH TO MODELING CRACK PROPAGATION IN NANOREINFORCED POLYMER COMPOSITES by P. McCarron). There are other work-arounds to the problem that were presented at Comsol conferences.

But - as mentioned - these have limitations concerning their applicabiltiy on a daily basis. And if I would have an urgent need about it, I would try to implement something. But for the time being it is simply easier to use other software products.

But I believe in the Comsol idea, and I am just wondering, why noone pushes the implementation of any of the commonly known approaches.

Dear Markus, I was not thinking about using a the programmable interface via MATLAB or JAVA. However, that could also be done, but is more difficult. Rather, I was thinking of the PDE mode of COMSOL as regular input (or equation-based modeling). On the newer versions, the COMSOL base module comes with the "physics builder" which provides an improved GUI interface for using the PDE mode. Are you aware of the physics builder ? This feature is what really sets COMSOL apart from other codes. There is no end of possible equations that could be modeled. There is an interesting new article in the COMSOL blog by David Kan that helps to introduce to this feature.

www.comsol.com/blogs/modeling-viscous-fingering-equation-based-modeling/

Dear James,
thank you for this hint. I found this useful for another problem I had. But the approaches mentioned before require more than just different PDEs. I.e. the ability to split nodes or to allow for interaction of different meshes. I am almost sure, there is currently no way to implement this in a straightforward fashion.

Also, since I do not see any other response than yours, I also allow to conclude, that there is currently no convinient way to do this.
Again, Comsol officials, no interest in this business?

Markus

Hi,


I am certainly not an expert in this area...but...I did some preliminary reading and analysis on this.

If you want to include continuum damage mechanics, this is possible to some extent. There is already a paper in COMSOL published by Dawei Hu et al. on chemo mechanical damage.

I also saw an example by Henrik Sonnerlind on cohesive zone model where a traction separation law is used to demonstrate tearing (he has attached an example).

The main problem I found during the implementation is that we have to use steady state with continuation technique which means that the load step is fixed by the user and there is no automatic stepping as in nonlinear analysis possible, at least I did not try. Of course, when I say load step, I mean displacement step assuming there is no snap back in post failure. If there is snap back then you have to use some arc control technique which I think should be possible.

The other main problem is on pathological mesh sensitivity but this can be worked around via the use of implicit gradient (Helmholtz equation) but convergence was very difficult and I did not bother further with it.

But if you are talking about discrete crack modelling I have no knowledge other than the J-integral type example given in the COMSOL library.

Frankly, I do not like a set up where one has to go out of comsol to matlab or java as this can be very time consuming especially if multiphysics is involved, which I experienced in other couplings.

Some two years ago I had discussed this with COMSOL support, but they did not support crack modelling.

Like you, I was guessing and thought it is the right time that COMSOL introduced this feature at least in version 4.4.


Suresh

I also saw an example by Henrik Sonnerlind on cohesive zone model where a traction separation law is used to demonstrate tearing (he has attached an example).

Suresh, can you be more precise about the model mentioned above? I was not very successful in finding it...

I also saw an example by Henrik Sonnerlind on cohesive zone model where a traction separation law is used to demonstrate tearing (he has attached an example).

Suresh, can you be more precise about the model mentioned above? I was not very successful in finding it...

It is in this thread: www.comsol.com/community/forums/general/thread/18335

We will include a version of it in the Model Library in the near future.

Regards,
Henrik

Dear Henrik,

this is a really nice example for implementing crack growth. Thank you for sharing and I am also looking forward to include this in the model library!

I was thinking about this option as well, but I evaluated only quickly with symmetric models, i.e. I tried only on outer surfaces. Did not work with the thin elastic layer, so I skipped to follow its capabilities.

You do not accidently know, wether there are plans to include X-FEM approaches as well? I am just asking because of the restriction to predefined crack paths as being used with cohesive zone modeling or VCCT.

Thanks again for the model!