Mesh 3D LEF problem

[FRANCESCA PISTORIO]

Dear all,

 

I hope I am not  disturbing you. I am modeling a semi-elliptical crack on a spherical surface and I have some problems with the mesh.

Since the model is 3D, I have created a sort of “torus” along the crack front. Then, since I can not use in 3D the command KSCON to create quarter-points singular elements, I have meshed the circular faces of the torus using mesh200 elements (and Kscon command) and then I swept the mesh along the crack front. I also tried two different combinations:

 

1. 3-D quadrilater with 4 nodes for mesh200 elements and solid185 for the swept volume;
2. 3-D quadrilater with 8 nodes for mesh200 elements and solid186 for the swept volume.

 

I compute the J-integral using the CINT command. The first combination give me a good results: the J-integral is path-indipendent and it converges to a practically constant values as the number of DOF increases (mesh convergence). Since  I wanted to further validate this results I tried to apply the displacement extrapolation methods to compute the mode I SIF  value  along  the crack front. However, the result differs greatly from that obtained by using the J-integral.

 

I tried to use SOLID186 elements along the crack front to improve the accuracy but when I swept the mesh200 elements I obtained very degenerated elements along the crack front. The J-integral computed with the CINT command  is completely wrong (it seems to fluctuates).

 

Is there a way to avoid the degeneration of the SOLID186 elements along the crack front? Why the SIF I computed using the displacement estrapolation method is different from the value obtained sing the J-integral?

 

Thank you for your help and time.

 

Best regards,

 

Francesca Pistorio

Student at Politecnico di Torino