8 Replies Latest reply on Jun 13, 2014 5:09 AM by Mark Scholten

    local stress peaks

    Mark Scholten

      Hi everyone,


      i am still very new to FEA and continuing to ask questions. i am wondering about the following:

      when simulating an almost square cabinet, lifted by two points on the diagonal, i find some very localized stress peaks. filleting sharp reentrant corners does not take away the stress peaks, but adding structure does.


      even though the stresses are beyond the yield stress, the equivalent strains are quite small. is it safe to assume that, as long as the stress state is not hydrostatic, the material will only deform plastically in a very small volume and NOT rupture?


      thanks, Mark

        • Re: local stress peaks
          Mike Pogue

          Hello Mark. You should post screenshots or a model--there are some experts here who can give you really good advice about your specific situation. In general though, if you don't have a fatigue problem and your material is ductile, very small volumes of yield are probably acceptable. What is very small is open to interpretation.

            • Re: local stress peaks
              Mark Scholten

              Hi Mike,


              thanks for your reply. I hope the picture is clear enough (I often find that i need to rotate back and forth the structure on screen to get a clear sense of the geometry).


              It is a sectioned von Mises stress plot. The structure that is pictured basically is a corner post with two beams meeting perpendicularly at the top and a diagonal beam in between. The load case considered here is lifting the structure by lugs mounted on the tops of the OTHER two diagonally opposed corner posts (so NOT by the one pictured here).


              The region at the red dot where the lower flange of the diagonal top beam meets the lower flange of the side beam is highly stressed at 777[MPa]. I know that this is not realistic as this value is beyond the yield stress, and that plastic deformation will cause the peak stress to lower while the less heavily stressed regions will attract more load. However I do not have access to non-linear analysis.


              As can be seen from the picture the mesh around the stress peak is quite fine. I have checked convergence there, so the stress peak is actually going to be there.


              The structure will only be lifted two times (once during load testing by two lugs and once to be placed at its operational location by all 4 lugs). It will be subjected to significant wind load, but those will not cause the same stress peak to occur. so fatigue will not be an issue in this region in my opinion.



              The equivalent strain is about 0.003, and the components are of similar magnitude. so quite small, the material's breaking strain is 0.5.


              Considering that no fatigue problems are expected and only two lift situations will be experienced, the remaining concern is that the plastic deformation will cause visibly noticable warping after the lifting load has been removed.


              Do you agree with this chain of thought?

                • Re: local stress peaks
                  Mike Pogue

                  I tend to agree. If the stress is convergent, and the volume in yield is really small, I doubt there is likely to be a failure in this load case. Can you show a factor of safety plot iso-clipped to FS=1 to show the volume in yield?

                  • Re: local stress peaks
                    Jared Conway

                    hi mark, do some searching on singularities and convergence


                    if displacements are converging but you have a high stress > very likely a singularity which is numerical that you will have little control over other than improving the mesh in that area and making it smaller

                    • Re: local stress peaks
                      Mike Pogue



                      It's not possible to see in the picture, but you did write that you filleted the area of concern, correct?


                      I'm assuming that this is in fact welded. Welds present a an analysis challenge, I think, because the stress concentration depends on the quality of the weld. This is not a big deal for a one-time load. The welded area is also not homogeneous, because of the heat treatment and the filler material.


                      It all gets to be a little beyond my skill level. I'd love to know how better engineers deal with welds.


                      You always have the option to add a gusset to unload that section.

                        • Re: local stress peaks
                          Mark Scholten

                          Thanks Jared and Mike!


                          Mike: the region in yield was about about 5x5x5mm. You're right about it being a region which includes a weld. The weld expert here agrees that it should be verified that the stress state is not hydrostatic. I have plotted the principal stresses, which turned out not to be equal: P1 is positive, P2 very small and P3 negative (about same magnitude as P1). In other words: no hydrostatic stress state.


                          Jared: I had already added fillets and the stress convergence plot i made had flattened out nicely (i have a question about convergence, but i'll ask that in a separate thread when i have time).


                          The client's lead engineer also agrees, he has given OK for production so it is out of my hands now.