Comparing Stress Values with Non-Uniform Mesh Sizes

YunMiao_1997

Hello, everyone

I have two models. One, derived from experimental data, uses a 0.1mm mesh for static analysis. Stress singularities arise near the concentrated load application point. My current benchmark is the stress value three rows of elements away from the singularity. I want to utilize this value as a reference for future simulations, to evaluate potential risks in a second model. However, the second model contains small geometrical features (less than 0.1mm), requiring a finer mesh (smaller than 0.1mm) in those areas. How should I select a comparable stress value from the second model’s analysis to compare with my established benchmark? My key question is, how to select meaningful stress values for comparison when mesh sizes differ significantly due to localized geometrical details in a model?

Any experience or suggestions would be greatly appreciated.

Adriano_Koga

as you've mentioned, these sharp corners will present non-convergency after all, so it is nice to defie some point at a controlled distance, that is not affected by the singularity. For real applications also, submodelling approach could be applied, so you run first a global model, with coarse mesh, and later on, you run a finer mesh inside a submodel (except singularities).

loistf

Tough question! I am not sure if there is a single best response for this. But mesh convergence is always a good path. (Mesh-converged stress results are strictly the only thing you can compare).

For this, two step modelling is a nice tool (instead of remeshing the full part)
https://help.altair.com/hwsolvers/os/topics/solvers/os/subcase_specific_global_local_modeling_r.htm#subcase_specific_global_local_two_step_modeling_r

YunMiao_1997

Thank you for your thorough response. I agree that mesh convergence analysis is indeed the best approach. However, for some components, such as adhesives in product designs, CAD models often have sharp corners, and it’s difficult to determine the exact radius of the fillets. This leads to stress singularities, making mesh convergence difficult to achieve. Do you have any other suggestions on how to address this issue?

Adriano_Koga

as you've mentioned, these sharp corners will present non-convergency after all, so it is nice to defie some point at a controlled distance, that is not affected by the singularity. For real applications also, submodelling approach could be applied, so you run first a global model, with coarse mesh, and later on, you run a finer mesh inside a submodel (except singularities).