Crack penetration in cohesive zone models

vinh phu nguyen's picture

Hi all,

I am conducting some cohesive crack simulations within the framework of XFEM. The cohesive model is an initially rigid traction-separation law for mode I.  In order to deal with negative values of the normal sepration/jump, I use a penalty stiffness K.

The example is pretty simple: an un-notched three point bending test where a crack is initiated at the bottom edge of the beam and grow vertically upwards to the upper edge. During the simulation, some negative jumps are observed. The convergence is very much dependent on the penalty stiffness K used. Even for some cases, no convergence obtained at all. I tried K from 10^3 to 10^10.

Could you please share your experiences in dealing with crack penetration in crack grow simulation?

Many thanks,

Phu


Pengfei Liu's picture

http://mypage.zju.edu.cn/per

http://mypage.zju.edu.cn/personnelCard/liupengfei/

The parameter selection in cohesive modeling is a difficult thing in deed and also a defect for promoting the crack propagation simulation. Now their is not a good method to solve this problem. I think you can only try to adjust these parameters such that the theroretical results match the experimental results. In fact, the parameter identification is also a deep topic. 

 


A chinese website?

What do u exactly mean by sending a website in chinese?


Qingda Yang's picture

Re: Crack penetration in cohesive zone models

Phu Nguyen:

I got a message from you through imechanica and it led me here. I will try to email you the paper.

For the problem you discussed here: your observation is true that cohesive zone models do not deal with normal compression (and the resulted friction if it needs to be considered) because of the sensitivity of numerical results to the initial cohesive stiffness especially when it is very highly. A common practice is to define a contact sufrace along the potential cohesive crack plane, whcih can indeed effectively avoid interpenetration. I happened to have some results of this problem in a recent paper:  

http://www.sciencedirect.com/science/article/pii/S0013794413001689