Localized Failure

[EddyC]

Hi Folks,

I'm a recent Nastran user. I am analyzing a model that has localized stresses that exceed the material yield stress. Is there a way to allow these particular elements to "drop out" and have a stress redistribution, with the resulting model still stable? Basically I'm trying to determine if the overall global model is still OK with localized failures in it. Thanks.

 

[SWComposites]

Run a nonlinear analysis with a nonlinear material stress-strain curve that represents material yield.

 

[40818]

On a different slant to SWComposites, what would you expect to occur in reality or if you did the analysis by hand?
You could have some sort of contact situation and hertzian stresses far beyond yield, but the whol thing is perfectly acceptable. Or it could be catastrophic failure.

 

[EddyC]

Hi Guys,

Thanks for your input. I think that I'll try to amplify what I'm doing: For my particular model, the loads are statically applied. From past experience, the critical modes of potential failure for my model are yielding or buckling. I have components that have a stress pattern where some elements are below yield and other elements are above yield. This occurs within the same cross-section. If I try to beef up the entire section to reduce the peak stress at a particular point, I'm going to have to add a great deal of material. I'm wondering if I could just let the particular elements that are overstressed deform plastically and transfer their loads to adjacent elements? I've only done linear static analysis in the past, so I don't have any experience with doing non-linear. That is why I posted here. FYI: I'm using Femap/NX Nastran. Thanks.

 

[40818]

If your asking about FE modelling acceptance then its a bit hard from here!
In thelinear FEA your elements will not go plastic and shed load, they will just keep on increasing in stress (if you keep applying greater loads). If you want to use FE and involve plasticity, then it is non-linear i'm afraid. And getting that right can bring a whole new level of complexity to your model. From what your saying your cross section is bending?? If so then you may be able to use hand calcs and basic plasticity in bending. If not then non-linear FE.

 

[ykantor]

If your failure criterion is yield, then your structure fails and there is no need to re-distribute the loads.

I guess you have to clarify what is really your failure criteria.