3 Replies Latest reply on Jun 19, 2013 7:23 PM by Serhey Moroz

    Local over-stress

    Serhey Moroz

      Hello colleagues

      Is there any one from structural or mechanical professional who can sugest on my little concern.

      I am testing coal bin for stress. It's 1/4 thick shell and a lot of stiffeners everywhere. Stiffeners are 1/2 thick and in general look OK, except a few local overstresses at corners. Please look at picture attached.

      We all know that it's right location for maximum stress at that case, but do we really should be concerned on that? It's about 95 KSI at the corner which is 3 times over stress for A36. At the same time just a little out of corner I have only 5 KSI, which is 6 time under stress. Is that really the case when we have to over-kill our stiffener thickness because of a little local stresses? Even though, displacement at corner is only 2.5 mm, which is nothing comparing to bin sizes. I tried to round of corner, it's getting even worse, for some reason.

      Well, I kind of spent a lot of thinking on this, it looks safe to me for sure, not going to fall appart. But is there any practical advice can anyone provide me? In terms of what kind of problem should I expect and what the practical way to judge or avoid local overstresses on FEA models?

      Thank you

        • Re: Local over-stress
          Mike Pogue



          This comes up a lot. FEA cannot interpret stress at a sharp interior corner, because the correct stress is proportional to 1/r^n. I don't really kow what n is, it depends on other geometry. In your model, r is zero, so, the solution converges to infinity. The good news is, there is always some non-zero radius, even the corner of an end mill has a radius of around .0002in. The bad news is, the actual radius may be impractically small to model, like .0002 in. In your case, the effective radius has to do with the weld leg, I'm guessing.


          • Some solutions are
          • Modelling the welds--remember, the welds don't go under the stiffeners, only around them.
          • Ignoring the local stress and looking at deformation--this requires some intuition about what is acceptable.
          • Verifying the stress is not real by using mesh controls to make the mesh around the stress very small. You can trust stress values that converge, i.e., stop getting bigger as the mesh gets smaller. you can effectively drie the stress into the corner and ignore it, as you see  that the convergent stress very close to the corner is acceptable.

          There is a lot more to this, but you can google it if you want a more detailed explanation.