Not sure how to proceed with a series of Quasi-static experiments

[g.alshamsi]

Hello all, so here's my issue: I have a cold-formed steel shear wall made up of a frame and a thin sheathing (0.8 mm) subjected to a quasi-static cyclic loading protocol. Let's call this experiment one.

After the experiment is complete, the sheathing was manually cutout from the frame as shown in the following figures; Then the frame was subjected to another cyclic loading protocol (let's call this experiment 2).

I was able to simulate experiment one with acceptable results using s4r elements, linear kinematic hardening and the explicit solver. My issue is on how to proceed from here to simulate experiment 2. The main challenge is to model the inelastic deformations/strains that were generated from the cyclic loading from experiment 1. I'm thinking of building a model where the sheathing is cutout from the centre of the frame and subjecting it to two quasi-static displacement load protocols using two separate steps. I'm uncertain however if this method would give acceptable results or even work the way I want it to. If anyone has any recommendations or have previous experience with something similar, feel free to add your own thought/recommendations.

Thanks for your time, cheers!

 

[FEA way]

Did you try solving this using other analysis procedure (step type) than explicit dynamics ? Abaqus has special analysis types for cyclic loading / low-cycle fatigue. I would also experiment with different types of elements and material models.

Of course, you can divide the analysis into several steps with different loading schemes that will represent the whole physical test. When the steps are nonlinear, the model’s state is updated with each step and plastic strain and damage accumulate.

 

[g.alshamsi]

Yup, I've experimented extensively with the solvers and explicit was basically my last resort option. The only way I could get the implicit solvers to converge was by using a large dissipation factor which changes the physics of the problem.

Same problem with the elements/material models, initially I wanted to use higher order elements (S8R or S95R) and perhaps a multi-linear kinematic model; again I ran into convergence issues. The assembly is comprised of thin steel sheets which are very prone to buckling and softening. Add to that contact interactions between 5 sheets and load reversals due to cyclic loading protocol and you have a problem that runs into multiple convergence problems.

 

[FEA way]

Right, explicit might be the best choice in such a case. Maybe you could even simulate the whole process, including removal of the sheathing from the frame (element deletion functionality can help with that).

 

[g.alshamsi]

Interesting, the element deletion functionality didn't occur to me. I'll give that shot.

Cheers!

 

[FEA way]

It's essential e.g. in ballistic studies but here it could let you remove part of the mesh from the model and continue the analysis without it. You just have to set a proper failure criterion.

Unfortunately, the *Model change functionality, which is more suitable for such situations, can't be used in Abaqus/Explicit.