7 Replies Latest reply on Mar 7, 2013 6:58 PM by Joseph Brinson

    Hyperelastic Material input tensile data ASTM D638 test

    Joseph Brinson

      Hi,

       

      I am having issues with my nonlinear static simulation. Prior to the simulation, I ran an ASTM D638 tensile test on a hyperelastic material to obtain stretch ratio Vs Stress data (Engineering) and inputted the (Engineering) data into the Ogden hyperelastic material tensile curve properties of the simulation. I defined the Poissons ratio, density of my material, and number of Ogden constants (4) but want the simulation to define them based on my inputted tensile data.

       

      I modeled an ASTM D638 tensile test coupon to try and obtain the same results I reached via bench top. I defined a displaement control at the top of the sample to displace it 7.5 inches as I did on the Instron. And added a fixed restrain at the bottom of the sample. I also defined a distance Vs. time curve in the displacement control. After running the analysis I obtain stresses and strains significantly higher than the bench top test data I have.

       

      Does anyone know what I can do to set this simulation up properly? Or what I am doing wrong?

       

      I show my stress-strain curve from the benc top ASTM test versus what I got in the simulaiton and you can clearly see the difference.

       

      Please reply here or if you'd like to my email below

       

      joseph.brinson@medtronic.com

        • Re: Hyperelastic Material input tensile data ASTM D638 test
          Jared Conway

          Hi Joseph, I sent you a PM.

           

          Is it possible to post the model, setup and material properties?

           

          are there any differences between the way you collected the data and output the data from simulation?

           

          if you compare the actual displacements between physical and simulation do they look ok?

          • Re: Hyperelastic Material input tensile data ASTM D638 test
            Jerry Steiger

            Joseph,

             

            As I recall, engineering data assumes a small displacement and ignores the fact that the sample thins as it stretches, so you don't want to use engineering data for your large strain simulation of a hyperelastic material, you want to use the true stress and strain. This makes gathering the data much harder, as you need to estimate the actual area of your sample. This would, I think, explain part of the difference between your experimental and simulated stress-strain curves, since your experimental values underestimate the actual stresses.

             

            I suspect that most of the difference is because you only used the uniaxial tension test to estimate your material properties. The usual practice is to use several tests, as in the following: http://www.axelproducts.com/downloads/TestingForHyperelastic.pdf. The good news is that you can probably do two of the three tests, simple tension and shear, that Axel uses. The bad news is that the third test, biaxial tension, is a very specialized setup.

             

            It is also likely that your dogbone sample was not long enough to get good uniaxial tension results.

             

            Jerry S.

              • Re: Hyperelastic Material input tensile data ASTM D638 test
                Joseph Brinson

                Hi Jerry,

                 

                I found some resolution to the simulation last night. Here is what I found:

                 

                - I probed the gage area of the sample to plot stress-strain where I noticed the high stress and bad curve overlay relationship to my experimental data.

                - What I did was take the stress values from the probe in Solidworks and converted them to Engineering Stress and this actually fit the curve of my experimental MUCH better. So it seems as if Solidworks plots out True stress Vs Engineering Strain in my results. Not sure why it does this?

                 

                - Do you know how to change the settings so that it changes the output True Stress to Engineering Stress? OR, change the Engineering Strain results to True Stress?

                 

                - Like I mentioned, the curve fits MUCH better now to my experimental data, but still does not fit perfectly as I would expect. This must be something in regards to the algorith with the Ogden curve fitting model to my inputted experimental data??

                 

                - I attached the fit I am now receiving. Let me know what you think?stress strain fit.png

                  • Re: Hyperelastic Material input tensile data ASTM D638 test
                    Jerry Steiger

                    Joseph,

                     

                    SolidWorks shows the true stress because that is the actual stress in the part. Engineering stress and strain values are only approximations; they are only good when the strains are small. I don't know of a way to change the ouput to show engineering stress and strain and I think showing those values would be a mistake.

                     

                    I don't know why the match to the experimental data is not better. If the problem were just a matter of your use of engineering stress and strain values, then I would expect the values to match well close to 0 strain and diverge as the strain increased. The opposite seems to be the case; the values are most different at small strains. Because of that, I suspect that using only the uniaxial tension data to get the Ogden constants is the problem. The dogbone being too short may also be an issue. If you could get a sheet of the material that you could cut into strips with a length to width ratio greater than 10, then you could run the uniaxial tension and shear tests. That might get you a better fit.

                     

                    Jerry S.

                      • Re: Hyperelastic Material input tensile data ASTM D638 test
                        Joseph Brinson

                        Hi Jerry,

                         

                        So I have done more work on this study and here is what I discovered:

                         

                        The subject on Ogden strain density function and how it fits the data is very complex and not only until yesterday was I able to understand what functions and equations it uses to output the stress (Thanks to an expert I reached out to at MIT). Basically it uses the Cauchy True Stress formula, which takes in my inputted stretch ratio's, and (curve fitted Mu's and Alpha's solved by Solidworks) to determine the True stress over each engineering strain step in the analysis.

                         

                        In order to see how well the Ogden curve fits the input tensile data--> I am able to take my experimental uniaxial tension data and input it in the material properties of the simulation for simple tension curve fitting and simulate my analysis. The output .log file gives me the Ogden curve fitting parameters which I export into the Cauchy Stress equation in Excel to see the exact stress-strain curve fitted by Ogdens model in Solidworks. I attached that image here to show you how far off the curve fit is to my experimental data which was produced with 4 material constants (Which is the maximum in Solidworks allows).

                         

                        Therefore, I explored Abaqus to curve fit my data and output me better material Ogden constants for curve fitting. Which it does! I took these new Ogden constants and created a new material in Solidworks with these inputted constants instead of inputting my stretch vs stress data. The output file now follows my experimental data much better, however I encounter singularity stiffness errors after a certain strain (buckling I believe is what I read its called as well). I attached that curve here as well to show you how it follows much better before buckling.

                         

                        Therefore, I don't know how familiar you are with adjusting study options within the simulation to allow for a better convergence or iteration of the data without buckling to occur? I think once I figure this part out my simulation will be complete and 100% accurate.

                        experimental vs input data.jpgExperimental vs abaqus curve fit constants.jpg

                          • Re: Hyperelastic Material input tensile data ASTM D638 test
                            Jerry Steiger

                            Joseph,

                             

                            I'm not familiar enough with SolidWorks Simulation to help with the options. It looks like your Abaqus constants result in the stress-strain curve going horizontal. I think that you may be able to choose how the software works in such cases, but I don't know where the options are set or which one to use.

                             

                            It does seem that the Abaqus constants are not a very good match to the experimental results once you go beyond 50% elongation. Perhaps they will look better once you get past the point where it fails to converge, but I am suspicious.

                             

                            Jerry S.

                              • Re: Hyperelastic Material input tensile data ASTM D638 test
                                Joseph Brinson

                                Jerry,

                                 

                                Sorry I should have expressed a little more. When I take the Abaqus constants and import them into the Cauchy Stress relationship, which is the funciton used to determine true stress from your input engineering stretch data and Ogden constants, I can output the stress-strain curve relationship exactly as I got during bench top testing from an overlay. Therefore, I know the Ogden constants I recieve from Abaqus is a true fit for the model and should produce the same fit in Solidworks. I know I just must be missing something within the Solidworks algorithm or study property/option settings that allows me to produce an accurate convergence during its iteration method.