• Registered: 2010-06-13
  • Posts: 1

Topic: Study of Size effect in ATENA 3D

Dear Sir,
I am using ATENA 3D for analysis of size effect in concrete beams. I am finding difficulty in determining mesh density for my models.
I ran simulations for different size of meshes for each beam but every time I got different strength of the same beam with diferent mesh size.
Sometimes the difference is more than half times. I am using 250mm, 500mm, 750mm & 1000mm deep beams with 150mm width & geometrical similarity. The mesh size varying from 40mm to 80mm.
If I reduce the mesh size further more, it will take more RAM & CPU time due to increase in number of elements & DOFs.
It is basic assumption of FEM that accuracy increases with decrease in mesh size. But what is that limit up to which I should decrease it for better accuracy in ATENA ?
In ATENA3D there is only one option of refining the mesh by reducing its size (h-method) if I am not wrong.
Or is it possible in ATENA to optimize mesh by increasing order of element (p-method) because there are only two types of elements Linear or Quadratic in ATENA ?
Error Criterion are norms so its very difficult to find on which side the error is either negative or positive so how can I determine that in ATENA3D ?

Yours Sincerely,
Nitin

Quote Post 1

  • dpryl
  • CC software developer
  • Offline
  • Registered: 2007-07-13
  • Posts: 1,107

Re: Study of Size effect in ATENA 3D

If the largest difference occurs with the 250mm beam, I can only remind you that to capture bending at least qualitatively, at least 4-6 elements per thickness are required. To get the best results for a given total number of elements (or memory consumption), usually a mesh with refinement in the important direction can be recommended (in this case, more elements in vertical direction and coarser in both horizontal directions). Please keep in mind that the element edge length aspect ratio has to stay in reasonable limits (2:1 is OK, 3-4 still works well most times, 10 is definitely too much).

Another frequent error source leading to high mesh sensitivity is applying a node (or line) load into nonlinear material. We always recommend bringing the load (and supports) using elastic loading plates. The load (support) can then be applied to a single point of the plate (which allows easy monitoring).

Yes, only linear and quadratic elements are available in ATENA. Elements of higher order generally do not make much sense with strong material nonlinearities (cracking etc. -> the approximation is frequently not even 2nd order).

A general rule-of-thumb limit for mesh size comes from modeling the heterogeneous concrete material using a continuum with smeared properties - elements smaller than the aggregate size are quite questionable from the physical point of view.

I am not sure if I understand the last question, but maybe displaying the residual internal forces can help you?

Regards,
Dobromil

dpryl's Website

Quote Post 2