Progressive Damage Analysis using Continuum Damage Mechanics (CDM)

  • 380 Views
  • Last Post 22 November 2018
obiforeva posted this 12 November 2018

Hi everyone,

Currently, in my PhD thesis, I'm dealing with prediction of burst pressure of composite overwrapped pressure vessels (COPVs). For that I've implemented, Hashin or Puck Failure Criteria for damage initiation law. And for damage evolution "Material Property Degradation" is selected. Results are OK for GF overwrapped steel liners, but not satisfactory for CF overwrapped Al liners. 

So, I decided that I should go for Continuum Damage Mechanics (CDM) for damage evolution. I'm aware that is restricted with Hashin failure criteria but Hashin is OK for me now.

I've selected simpler case for experimenting with CDM. [+-45]ns laminate with applied tensile load (specifically ASTM D3518 test). I'm also interested in Open Hole Tests in similar manner.

I've modelled the [+-45]ns case for non-linear shear behavior of the laminate. I've found the constants needed for CDM. Gft = around 80-130 N/mm (or kJ/m2), Gfc = 80-110 N/mm, Gmt= 0.2-1 N/mm, Gmc=0.5 - 1 N/mm and viscous damping coefficients are from 1e-4 to 0.002 for GFRP or CFRP materials.

I've found several articles about the case and material properties are provided. 

But in the end, my analysis doesn't converge at all, at least for expected loads or displacements. 

I tried several things, lets recap those...

1. Element types: hex20, hex8, quad4, quad8. Best convergence acquired with hex20 but there is still an element violation at about %10-15 of the desired final displacement/force value.

2. Mesh density: Mesh was mapped homogeneously both on D3518 specimen and open hole specimen. Started with 3 mm element size and gone down to 0.5 mm max size. It doesn't help at all.

3. Boundary conditions. I've used both face or edge BCs using ANSYS mechanical interface and also tried "direct FE" option with nodal BCs with symmetry and no symmetry. Result is always the same. I can give more detailed information about my BCs if you want.

4. I've tried some Non-linear diagnostics with newton-rhapson residuals and HDST (high distortion) elements. Generally element violation occurs at near BCs because of Poisson's ratio (I guess). But it happens too quickly!

5. Large Deflections ON. I also turned on Line Search and when I apply force rather than displacement, I also use Newton-Raphson Unsymmetric.  Stabilization option is also tried with constant but didn't go well. But I didn't try too much values on that option.

I always get an element violation error. Sometimes only 1, sometimes several elements. The solid/shell body seems to be distorted around that element. Sometimes that is too much that you cant see the body itself. 

From trial-and-errors that I've conducted, I understand that CDM (at least ANSYS Workbench implemented version) may not be suitable for non-linear shear behavior of composite materials. For that, custom material models might be modelled through USERMAT subroutines. I'll try that way but it seems very complicated for me, at least for now.

Do you have any suggestions about this problem? I can provide my project, BCs, mat properties, schematics, journal articles etc.

Regards...

  • Liked by
  • gmr966
Order By: Standard | Newest | Votes
peteroznewman posted this 12 November 2018

Hi obiforeva,

I can only comment on #3 and hope others will reply to your other questions.

The best BCs to support a material tensile test sample uses a three-plane symmetry model.

Use symmetry at the center of your rectangular coupon in x, y and z planes. On each plane, the symmetry BC constrains the displacement normal to that plane to be zero and leaves all other DOF free.

At the grip end is a face that is at the edge of the grip. It is only a 1/4 of a face of the full cross section due to the symmetry used above. Apply a displacement BC to stretch the material using just the normal displacement, leaving all other DOF free. I understand this doesn't simulate how the material behaves in the grip, but you are not interested in simulating failures at the grip edge.

This setup has the best chance to avoid element distortion errors.

Regards,
Peter

obiforeva posted this 22 November 2018

Hi Peter,

Thank you for your quick response. I had a login problem for ANSYS student community but it is resolved somehow.

About the convergence problem, actually I tried your suggestion about symmetrical Boundary conditions. 

I cut off my model into 4 and 8 parts and applied appropriate BCs as you mentioned. Yeah, it doesn't reflect the exact grip conditions (as there might be some stress concentrations due to gripping behavior) but lots of articles doing the same thing and this greatly decreases convergence problems. 

But no avail, I still get element distortion too early. 

I strongly suspect that damage models available in ANSYS cannot model non-linear shear behavior in composites. Again, during my research, I observed that lots of articles defined custom material & damage models for their simulations while using ANSYS or similar software.

I will try to do the same for my case. I guess I need to learn about UPF and user material subroutines. Thanks for the reply anyway.

Regards,

Serkan

Close