Does anyone have a scientific explanation to the fact that why reducing the mesh size in Abaqus/Explicit leads to unstable results and worse agreement with the experimental results?
In this particular case of CFST columns the best fit of the equilibrium path is for the coarse mesh size (best fit for 20 mm).
But if the mesh size decreases to 10 then the stiffness and the load bearing capacity also decreases a little bit. Finaly for the mesh 5 mm the solution becomes wrong. Any leads or papers on the subject?
Maybe you should play with fixed time incrementation in Abaqus/Explicit.
with increased mesh refinement, your stable time decreases (proportionally). As other has said, localized deformation. Another issue you should keep in mind that the more number of increments are involved, the more truncation/roundoff error is accumulated. In terms of element formulation, extra fine mesh leads to shear deformation becoming more dominant than bending. the element size must satisfy the underlying assumptions that were taken in formulation. Lastly, when very fine mesh is used, it's always recommended to run explicit in double precision, and output_precision=full
Check hourglass controls. The default options might not be good enough as you refine the mesh.
The time step is dependant on teh element length for a stable solution.
Additionally, if you have large and small elements next to each other could lead to instability in the numerical process.
The explicit method has to satisfy the CFL stability criterion (Courant-Friedrichs-Levy condition) which states that the time step should be so short that a wave-like disturbance cannot travel through the smallest finite element. You are on the safe side with the longitudinal wave (compressional wave) with speed cp=√(λ+2 μ)/ρ in the case of linear elasticity. The CFL criterion requires smaller time steps as the mesh is refined.
Hi Bartosz
Mr Staat confirm what I wrote earlier and I am in total agreement with him.
Mr Manfred Staat,
Mr Sirous Yasseri
Courant - Friedrichs - Levy condition gives a maximum time step, not the minimum one. And the unstable solution occured for very fine mesh.
Best Regards,
Bartosz Grzeszykowski
Hi Bartosz,
I think you got your answer. The step size depends on the element size.
We should be interested in the maximum step size, since the minimum is meaning less here. We are interested to get answer faster, thus the maximum step size interests us.
- Badges
- Science topic