Proving Suitable Mesh

[zondashF5M2Q]

Hello all,

I have an issue where my program freezes or crashes when I use convergence. I'm not sure if the model is just too large for my computer to handle or if there is some other issue in my model causing the failure. 

Is it reasonable to recreate the highest stress point in my model (a main 3x3 beam with a plate bonded to the top), apply the load to that beam that has been hand calculated for it, and then use the convergence from that to assume the mesh settings used on the model are suitable? On my simple model I get a convergence rate less than 1% on the first step. My thought process would be that if I am getting consistently good convergence with a simple model with the same mesh settings, I can reasonably assume that the mesh settings in my model are adequate enough to verify that my model is fairly accurate.

Thanks! 

[John_Holtz · Accepted solution]

Hi @zondashF5M2Q 

Let me repeat your problem so that I am sure that I understand.

1. Model 1 is a simplified model. The mesh convergence  tool works properly, so you know what mesh size is needed to get a converged stress solution.
2. Model 2 contains the same parts as model 1 but also has more features (or parts). The mesh convergence does not work with model 2.
3. Since the high stress in model 2 is expected to be in the part that is also in model 1, the mesh size that is suitable for model 1 should also be suitable for model 2.

If my summary is correct, I would agree that the mesh size determine for model 1 should apply to model 2. If the high stress occurs in similar locations in both models.

The other option is to use the Mesh Control to manually refine the mesh in model 2. In other words, you can perform the mesh convergence manually.

John

[zondashF5M2Q]

Yes, you have understood me perfectly. The stress in what you have defined as Model 2 is slightly lower, due to it being part of a modified box-frame assembly. The particular square tube I used for Model 1 carries the vast majority of the load. The stress pattern is identical and matches the neutral axis shift as the distance becomes closer to the parallel plate fixed to the top of the tube.

I have used mesh controls to refine the mesh for that beam and at the location of the stress concentration. Model 1 contains 1.27 million nodes and 653k elements so it may just be too much for the software or my computer to handle. That is why I created my simplified model and I believe that it proves believable accuracy in the full model. I know without a doubt there is a stress concentration here because this particular part has failed in the past multiple times. Just looking for enough accuracy to show the stress that we know is true on the real part, and show how my design changes remedy the issue to a repeatable stress level.