2 Replies Latest reply on Nov 17, 2007 8:01 PM by Vince Adams

    Interpreting study results

    Aaron Hayden
      Hi all,

      I know this topic has come up more than once in the past. When one gets results from a study how does one weigh theoretical results against actual field conditions (without actually being able to test the part)? I understand that the study is only as accurate as the data being fed into it. I usually try to keep the model as intact as possible to achieve the most accurate testing scenario, however performance is typically inversely proportional. One thing we don't do is insert weld beads into the models. We run our studys with global bonding for weldments as well as bolted joints as it is the fastest study. Any guesses as to how much error is introduced when using bonded for a bolted joint rather than simulating the fasteners (I'm sure it's a case by case deal)? Then there is the hot-spot issue. Anyone have any first hand experience running a study on weldments with and without the weld beads? The hardest part about running FEA for me is determining how critical the individual hot-spots are (second hardest is by far meshing). Will the material yield in those high stress areas which will then allow the stress to redistribute? Will a crack develop and propagate? Does the hot-spot actually exist? These are seemingly difficult questions to answer.

      I've attached some JPEGS of a typical model showing hot-spots. The plot is Design Check (FOS) to Yield. The scale is 0-3. Anyone care to comment on this model regarding hot-spots? I'm certainly not expecting someone to "ok" the design, just your thoughs based on your experiences.

      Thanks in advance!
        • Interpreting study results
          genexxer genexxer
          I bang my head then do some rudimentary studies. Back to basics. Cantilevered prismatic beam with a single point load at the free end. The other end is the interesting end. There, try various boundary conditions and the results should give you some knack for the proportional exagerration you get from using bonded vs fastened joints. These studies shouldn't get you too steep on the inverse proportion curve since they are pretty simple.
          • Interpreting study results
            Vince Adams
            Hi Aaron, you hit the proverbial nail on the head when you said you were sure it was a "case by case deal." Here are a few observations...
            1) There is a saying that "Red spots are red flags" They are indicating something to pay attention to, not something to ignore. Only after examining them and establishing a likely cause should dismissing them even be an option. That said...
            2) "Hot spots" are often caused by near-singularities where a finite load is being carried by a very small area, thus approaching a "divide by zero" (F/A) situation. In the physical world, local deformation and/or yielding will always increase the area until failure or equilibrium is reached.
            3) Hot spots at the edges of restraints typically indicate that the model goes from elastic to perfectly rigid too rapidly. You should consider revising your model, even to the point of adding the parts represented by the restraint, to make sure the hot spot goes away.
            4) Hot spots at sharp inside corners, especially in bonded plates that would otherwise be welded are very common and no amount of convergence will ever fix them. Use Free Body Forces to estimate the load being carried to size the weld and then add an appropriate safety factor. There is a whole COSMOS Companion session on this. (www.solidworks.com/Companion)
            5) Bolt can be modeled many ways depending on the goals of the analysis and the distance of the bolt from areas of concern. Be very careful of over-bonding a bolted connection. Tensile load is ONLY carried by the bolt, not the entire area that would otherwise carry compression.
            6) Individual hot spots that you can't dismiss for some clear justifiable reason must be addressed as potential failure areas. Make sure your loads, restraints, contact, and geometry are as good as possible in these locations before trying to make a failure decision. If you are looking at a single loading condition (or only a couple) local (point) yielding may never turn into anything significant. However, a nonlinear analysis may be required to clarify since once an area crosses that threshold, local results become meaningless since stiffness redistribution is not necessarily predictable. If you are looking at manyloadings, cyclic or on-off, these local stress risers become much more ominous. Fatigue failures can occur at areas of stress well below yield so the presence of localized stress exceeding yield is almost an invitation to a fatigue crack. You should consider local geometry changes or, better yet, re-routing your load path to relieve these regions.

            After this... it is a case-by-case thing...

            Good luck!!