Beam element meshing 1

[skr3178]

Hello,
I am working on a model with beam elements on wires.
My concern is that the results/analysis have huge variation depending upon the mesh size I use. As I recollect mesh sensitivity analysis for beam elements is not yet permitted in abaqus. With this view, how can I determine what is the best size of the mesh I should select in order to have "accurate" results.

The overall size of the bounding box is about 70x20*4mm.
I have chosen approximate element sizes of 0.15 and 0.5mm. These provide me highly contrasting results. Both have identical boundary conditions and loads

Thanks and sincerely,
skr3178

 

[FEA way]

Do a classical convergence study - run the same analysis on several mesh sizes (with increasing density) and check when the results converge.

 

[skr3178]

Hello,

I ran some simulation in order to better understand the effect of mesh size.
Two identical studies with one done with 12K nodes and second with 75K nodes.
Here are my observations. The less dense mesh(12K) seems to be performing better in terms of not penetrating through the beam elements for the same displacement. A hard contact plus tangential penalty friction was used.
Can anyone explain this? Is it because less mesh density causes some sort of less probability to pass through(assuming the penetration is taking place after a certain time through the nodes). I have attached the images here for reference.
Also could FEAway explain what it means to check when results converge? Does it mean the total number of increments taken or the total time taken for the dynamic explicit analysis to complete? I would be glad for the clarification.
Thanks,
skr3178

 

[FEA way]

In some cases coarser mesh can be better in case of analyses with beam contact because Abaqus assumes circular beam cross-section for contact and there might be a problem with intersections of contact edges.

What I meant is that you should check how the results change between subsequent analyses. If the difference between old and new mesh is significant, run the simulation again with refined mesh and check the results. If the difference is small, you can assume that results are converged.

 

[skr3178]

hello,

I did some further investigations into the effect of mesh size on the final analysis.
For a general overview, the overall model dimensions are 10mm diameter.
The study is an explicit dynamic type with thermal loading on isotropic material. Below is an overview of the analysis:
[ul]
[li]Bilayer structure with different thermal material properties.[/li]
[li]Undergoes thermal loading(temperature increase). No dissipation of heat considered. Friction considered.[/li]
[li]On loading the 2 layer transform into a strain minimized deformed structure.[/li]
[/ul]
Can we explain lack of strain minimized structure stable structure when using high mesh density?

Sincerely and thanks,
skr3178

 

[skr3178]

Hi,
I realized that the total time step has an effect on the final structure.
In all the static thermal analysis, I increased the time step from 1.5 to 60s. The results are all in agreement and with the final deformed shape. I would be glad if someone could help me understand how does the time step affect the analysis in explicit. I have read that inertia has a role to play in explicit analysis with deformation due to change from strain energy to kinetic energy.

The loads in the model are in the following order:
1. Initial temperature of 293 K
2. Final Temperature of 693 K.

Both are applied to the same time step. Perhaps, my previous post was not well describing the situation.
Thanks,
skr3178

 

[FEA way]

You should use the time step that corresponds to the duration of the simulated process. However, it’s often necessary to reduce time step (speed up the even artificially) so that the analysis completes faster. This approach is not perfect since inertia may affect the results.

 

[skr3178]

Thank you FEAway for your advise.
My efforts are focused on working through the abaqus manual for verification of my model.
My model seems to not follow the guideline"ETotal should be constant or close to constant". I am assuming that since I am placing it in an oven which heats from 293K to 693K, it does not have to follow this guideline(energy is added to the system by increasing the temperature). I would appreciate any comments on this.
-Thanks
Skr3178