@rmerlob wrote:
... and even though theory tells me that some localized yielding can be acceptable on ducile materials (again, keywords: strain hardening and stress redistribution) theres no guidelines on how much of it is acceptable.
This leaves no more options other that testing, even for a case that I consider could be the simplest real life application of finite element analisys.
The reason for this thread is that I DO NOT want to believe this is the case and I think there is somebody out there with more experience and training than myself that can say: this local yielding is acceptable or not because x and y reasons.
Firstly numerical analysis is always and approximation of the real world and is only as accurate as the simplifications and assumptions made creating the 'model'. It is never intended to be used as replacement for load testing, rather to reduce the number of tests required when developing prototypes. Obviously in many industries conditions are well known and FEA is used as a calculation method only where physical testing is not required. I believe the latter would apply in your case too since a clevis would usually always have been used in similar successful circumstances.
Regarding local yielding, the structural codes don't cover this well but is well documented in the pressure equipment industry. It is all just basic stress analysis and failure theories so can be applied to structural too. First it helps if you understand stress classification so I will only give the short version. Basically stresses can be classified a primary and secondary, whereby secondary stresses are self limiting by nature i.e. displacement limited where local yielding will redistribute the load. Primary stresses are not self limiting and will cause failure or gross distortion if the load is applied past yield.
Primary stress can be sub-divided into general and local membrane, and bending. General membrane, i.e. direct stress that has not been affected by stress concentrations at discontinuities should be limited to your allowable stress per your design code, you don't want this near yield. If you have pure bending or combined stress such as membrane and bending, like in most FEA then you can go up to about yield (I try to use 90% for structural work, 1.5*0.6=0.9). If your stress result also includes secondary stresses (self-limiting) then you can exceed yield because the stress will 'shake down' on repeated reapplication of the load. What the PV codes use as the check for protection against ratcheting (low cycle fatigue), in linear analysis is an allowable of twice yield in local areas. Obviously twice yield is not attainable in real life but in linear analysis any stress beyond yield is fictitious anyway and only indicates a lower bound for shake-down. This can be checked in more detail with non-linear analysis if required.
There are also lots of cases where the stress at the surface show excess of yield but the through thickness stress distribution drops off rapidly, in these cases local yielding is not always likely to occur. Also, when using stress classification, how you judge whether stress is considered local for complex geometry is not clear cut (for vessels it is based on thin shell theory). So it comes down to engineering judgement and experience when deciding if the stress above yield on a case by case basis is acceptable.
In some cases it might not be a strength limit state but a serviceability limit state where deformation would prevent it performing as intended, just as the clevis becoming out of round to the extend where it is deemed unacceptable (which JD has already discussed).
If you are interested Google 'stress classification' for more information.
Hope it helps.
Please mark as "Accept as Solution" if it answers your question or "Kudos" if you found it useful.
---------------------------------------------------------------------------------------------------------------------
Stew, AICP
Inventor Professional 2013, Autodesk Simulation Multiphysics 2013
Windows 7 x64 Core i7 32GB Ram FX2000
