4 Replies Latest reply on Feb 25, 2009 12:24 PM by Chad Schmidt

    Trouble Validating FEA Using Lab Test Results

    Matt Morse
      Hi Everyone,

      I am having trouble making my FEA simulation in Solidworks Simulation match up to lab test data taken on a physical prototype we recently tested. In this case, we're testing for linear geometric behavior of our component, i.e. getting a linear strain response over a given load range. In practice, this component is to be oriented vertically, consisting of 2 horizontal lug holes on either end, 1 to be held fixed on the top end and 1 to hold the load on the bottom end. Between the lugs are some geometric discontinuities as well as the areas of interest for strain measurement. For this test, we take 6 data points in the lab test and I'm currently trying to replicate the strain vs. load relationship in FEA that we saw in the lab. So far I'm not having any luck at all. I've tried making finer meshes, changing aspect ratios, using additional mesh refinements in the areas of interest, and trying a couple different curvature-based meshes with no success. I've also made sure that my restraining bearing load is only holding the upper side of the bearing surface to properly recreate the structural behavior. The load pulling down on the bottom lug hole is also a bearing load which I believe is the most accurate load representation in this case.

      I haven't been able to establish any sort of consistent relationship between the lab data and FEA results at all. My FEA results vary all the way from 1/1000th of the lab test data value at the low end of the load range to 30% above the lab test data value at the high end of the range by testing at the same 6 load values tested in the lab. There are some geometric discontinuities in the solid model I'm testing which I believe are causing some havoc, particularly the large slot-like hole between the two lugs but I'm having some difficulty as to how to best approach this. It should be noted that the applied loads to this component are small enough to be well within the linear elastic range of the component's material so Hooke's law is in effect here. As far as software goes, I'm using Solidworks 2009 Premium with Solidworks Simulation Professional. Anyone have any thoughts?
        • Trouble Validating FEA Using Lab Test Results
          Bill McEachern
          This isn't going to be much help but it should work and the data should line up. I would advise close scrutiny between the test set up and the simulation set up. How the loads get in is no where near as important as how you are holding the part. I would double check the materisl properties and the restraints. Consider whether the test rig itself is adding some flexibility that maybe affecting the experimental results. Sometime you have to add in portions of the rig to the analysis. Also make sure you are looiking at the right results quantites. Sounds like you have been doing all that so like I said I am not sure how else to help with out more info.
          Good luck.
            • Trouble Validating FEA Using Lab Test Results
              Derek Bishop
              I've been doing some simple checks on FEA results and found they correlate well with formulas for stressis in beams. My assessment of the Simulation is that if everything is entered correctly and the pragram doesn't throw a wobbly the results you get are solid.

              Can you post the model. It doesn't sound too complicated and we may be able to confirm the results.
            • Trouble Validating FEA Using Lab Test Results
              Chad Schmidt
              We've had success validating lab results with our micro strain values. Your FEA has to be setup identically to your lab test. I've seen errors in test results but these where attributed to tolerances of the parts, improper bolt fastening and relying on someone else to setup the lab test. If your not setting up the test and checking the parts then it's probably already wrong!
              • Trouble Validating FEA Using Lab Test Results
                Chad Schmidt
                Wanted to add something to this. If you are measuring with strain gages you have to make sure you model a patch where the strain gage is! You have to split the faces to match the exact size and shape of your gage. After doing this create a mesh control on that face with a fine mesh. Then when you "list selected" you get an average of that face (gage). This is assuming that you have a strain gradient across the gage.