Sub-structuring
Sub-structuring
(OP)
I am trying to model part of a very large bridge structure that has been analysed before using a coarse mesh. So I know the support reactions and externally applied forces around what is now my local area of interest. Now I wish to analyse this local area with much finer mesh and incorporate holes that were not part of the original coarse mesh analysis. I wish to cut off and isolate this part of structure for analysis purpose. I am not sure how far away these cut lines should be and more importantly, what boundary conditions should be applied to these cut lines. Do I need to obtain stresses as well as deflections at these cut surfaces and incorporate them in my new model?
I use a program called Lusas. I understand what I am trying to do is called 'sub-structuring'. But cannot find any references to it. Can anyone help please.
Thanks in advance.
normm
I use a program called Lusas. I understand what I am trying to do is called 'sub-structuring'. But cannot find any references to it. Can anyone help please.
Thanks in advance.
normm





RE: Sub-structuring
what you need to do is not "sub-structuring" but "sub-modelling". The difference has been already explained elsewhere in this forum or perhaps in the Ansys forum.
- "how far" the cut-boundaries? "far enough" as not to influence the local results. The results on the cut-boundaries of the sub-model may have artifacts due to the interpolation routines, but in a general section of both the coarse and the fine models, where you know that there are no "stress raisers", the stress states must be very similar if not <almost> identical.
- "how to apply the BCs?" normally if the software is "sub-modelling - ready", there should be routines to perform the cut-boundaries interpolations for you. Otherwise, it's a bit difficult, because you have to retrieve the deformation results from the coarse model, on the nodes of the cut-boundaries in the coarse model; then, interpolate these results for the nodes of the same cut-boundaries but in the fine model; then, set the "initial state" of the cut-boundaries as imposed displacements on the cut-boundaries nodes of the fine model. Loads and BCs which are not on the cut-boundaries must be integrally re-produced in the fine model.
Hope this helps...
Regards
RE: Sub-structuring
The new substructure was modelled/refined/etc... and then the field applied to all nodes common to the cut planes. Extrenally applied loads/displacments/pressures/etc... were then applied to the features which are common to the substructure.
The substructured model was run, and post-processed. A check of the sub vs the global model was always done with a simple loadcase to verify that the BCs were applied in a similar fashion/correctly.
jetmaker
RE: Sub-structuring
Gurmeet
RE: Sub-structuring
sub-structuring (or super-elements in NASTRAN-speak) divides the model into "bit-size" pieces. the benefit of sub-structuring is that you don't have to re-run the entire model ... each component is reduced to a stiffness matrix in terms of the boundary grids, re-run what you have to, re-use what you can !
actually you can do both the way way ... create a sub-structure in your coarse grid model which will be your area of interest, re-mesh this region to your heart's content (or your computer's capacity), re-run this component (which has the same mesh at it's boundary as the old coarse-mesh model), and re-run the global (combined components).
if you wanted to you could refine the boundary mesh, and MPC it to the original coarse mesh.
i think, like gurmeet notes, that this too probably isn't required with today's computing capacities ... a single structure would probably work just as well.
you could always take this to the extreme, an refine a single coarse-model element
RE: Sub-structuring
A.A.Y.
RE: Sub-structuring
Would agree that with today's CPUs most applications can be done using single models and bypass the substructuring all together. However, there are still many practical reasons for it, and I just got done doing one which I'll share.
I was given a model from a designer that was found to be structurally acceptable at the time of analysis. However, certain features were later found to be incompatible with the design, and minor detail changes were needed (spotfacing, additional csk details, tighter fillet radius). Alot of these could be cleared using classical methods, but some needed the FEM to be re-run to show good. It was a complicated model, and to re-mesh the entire structure would have taken the better part of a day, and the same on the solving time. By sub-modeling, I was able to concentrate on the areas that needed it, and had results by day's end.
Althought it still took a fair bit of time to do, it was faster than re-meshing the new model.
jetmaker
RE: Sub-structuring
I disagree both with not doing submodeling anymore and with not doing substructuring anymore.
A computer can very powerful nowadays, but it will still be too slow and have too little memory for some problems.
Regards