A question about the using of submodel?

[Lirock]

Hi,everyone:
I got a big model. It's very hard for meshing on my computer so it's out of the question to simulate.
In my opinion, submodel is a method which use the coarse simulation results on the cutting section as the boundary conditions for the submodel simulation.
So I am thinking if I can cut my model into two pieces. First build the first piece and simulate it. Second, build the second piece and simulate it using results of the first piece as a boundary condition.
Anyone pls give me your opinion about this. Am I right?

 

[cbrn]

Hi,
yes, the technique you describe has been used for decades when computers didn't have enough power to handle complex FE models. But it's not sub-modelling, it's sub-structuring. In other terms, you will determine the transfer matrix of a sub-structure, i.e. "how this sub-structure reacts to the external loads / constraints", and then you will "connect" this "super-element" to the second sub-structure. The underlying theory is a bit complex, but there are several books which treat this subject (unfortunately all the ones I know are in Italian). As for the application in ANSYS, yes, it can be done (and perhaps in more than a single way). I suggest to start from the Help manual, both in the application and in the theory parts: ANSYS' help is very well done and extremely rigorous.

As regards sub-modeling, you would do it if you first analyze a model with a coarse mesh and then analyze a small localized part of the same model with a much finer mesh (you can also analyze the coarse model as shell, and the fine one with solids!): the BC of the fine model are all the external loads as in the coarse model, + the "cut-boundary" conditions, i.e. the displacements at the cut-boundaries which derive from the response of the surrounding part to the loads / restraints.

Both methods are a means to analyze only what is needed by taking into acount "everything"; sub-structuring is possible if the analysis of each sub-structure is possible: if the smallest sub-structure is anyway too complex to fit into memory, there is no chance to use this approach; sub-modelling implies that where you put the coarse mesh the results are not so relevant (though the general behaviour must be known in order to correctly simulate the interesting zone), meaning that if all the zones of the model are relevant, or if you can't correctly place the cut-boundaries, you can't use this approach.

Hope this helps...

Regards