Hi,
I've been researching on how to deal with stresss concentrations and singularities in FEA models. My coworkers tell me to input a bilinear kinematic model and check that the plastic strain is less than 5%.
First of all, I think this might be valid at stress concentrations such as fillets, where it is possible to converge on stress and strain with a fine enough mesh. In this case, FEA can give us the stress concentration factor (= strain concentration factor for perfectly elastic material) by comparing the notched and un-notched part.
However in the second case, I question whether this would be a good approach at singularities such as right angle intersections of plates, sharp corners, or interfacial stresses at corners between materials. These places are singularities, where convergence is not possible no matter how fine the mesh is. IIRC, theoretically there is no SCF here, and both stress and strain will always increase as more elements are added. In my experience true singularities are much more of a problem since they cannot converge.
Also, in the second case, it doesn't seem possbile to get the SCF for a joint anyway, because you can't model the part without the joint like you can model a bar with and without a notch.
Is this correct? How would you use Neuber's rule on singularities, or otherwise determine if a joint containing high stress can be ignored?
I've been researching on how to deal with stresss concentrations and singularities in FEA models. My coworkers tell me to input a bilinear kinematic model and check that the plastic strain is less than 5%.
First of all, I think this might be valid at stress concentrations such as fillets, where it is possible to converge on stress and strain with a fine enough mesh. In this case, FEA can give us the stress concentration factor (= strain concentration factor for perfectly elastic material) by comparing the notched and un-notched part.
However in the second case, I question whether this would be a good approach at singularities such as right angle intersections of plates, sharp corners, or interfacial stresses at corners between materials. These places are singularities, where convergence is not possible no matter how fine the mesh is. IIRC, theoretically there is no SCF here, and both stress and strain will always increase as more elements are added. In my experience true singularities are much more of a problem since they cannot converge.
Also, in the second case, it doesn't seem possbile to get the SCF for a joint anyway, because you can't model the part without the joint like you can model a bar with and without a notch.
Is this correct? How would you use Neuber's rule on singularities, or otherwise determine if a joint containing high stress can be ignored?
