20 Replies Latest reply on Sep 28, 2010 10:32 PM by Mark Horne

    Predicted vs Tested Deflection 4130 Tubing

    Mark Horne

      I am an engineering student, Journeymen Toolmaker, Machine and Tool designer, and a Kart racer. I have convinced my instructor to let me design a Kart chassis in Solidworks as my semester project. I am using the Student edition at home and the schools computers.

       

      I have two known chassis. Both have known handling characteristics. I modeled both chassis with 3D sketches and using Weldments with the correct tube profiles. My thought was to apply a 100lb load on the front corner of each known chassis and find the displacement. To keep things simple, the connections of the rear axle and front spindles to the frame, I used another tube of heavier wall. The idea is to focus only on how the chassis flexes under load.

       

      So I fixed the end of the left side rear axle and the left side front spindle end. Then applied a 100lb load perpendicular to the top plane.  The displacement on the end of the spindle was recorded. I repeated a study on the rear axle right side end and found the displacement there. This was done to both known chassis.

       

      To verify these numbers where good, I attempted to do a real world test on two samples. A straight tube held in v-blocks. The other a rectangular frame.

       

      SIM TEST 17X19 FRAME-Study 1-Results-Displacement1.jpg  V-BLOCK TEST-Study 2-Results-Displacement1.jpg

      I just cut a triangular hole, leaving the top section, rather than putting clamps in. Not even sure if I needed them.

       

      The results: all tubes 1.125"x.095 wall 4130 n 870c.

      I used a certified weight of 21.96lb with a heavy wire, 90 degree hook to slip into the end of the tube.

      On the frame the predicted displacement was .0112". The dial test indicator showed .034".  (17x19" frame welded after fish mouthing tube ends)

      The straight tube predicted displacement was .012". The dial test indicator showed .016".  (straight tube projected 12" from block face)

       

      I have to show these results in my project to support my design. I used SW materials 4130 N 870c for the tubing material. My instructor is at a loss because he has not done anything with simulation except for the tutorials.

       

      So my question, finally, is: what did I do wrong? Is simulation the correct tool for this? What would be an acceptable deviation from real world to SW?

       

      Below is a sample of how I loaded the chassis. It will be a week or two before we can get the known chassis on the table to test those results to SW.

       

      I was told that even if the numbers were off, the results in SW for both chassis would be relative. Another words, if one is displacing .400" and the other is displacing .600", then my new chassis would have a displacement of .500", it would be right in the middle of the two. This seems too simple to me. What if the error is not linear? It seems non linear with the two simple samples above. Then how would one begin to "guess" what the end design would be. Perhaps I am missing something.

       

      here is the way I did the studies:

      F125 CHASSIS 2 WITH SPINDLES - AXLE-Study 1-Results-Displacement1.jpg

      Any help would be appreciated.

       

      Thanks,

      Mark

        • Re: Predicted vs Tested Deflection 4130 Tubing
          Anthony Botting

          Hi Mark:

          you are not alone. I ran into this very problem while simulating spacecraft bus structures (the frame) for developing new designs. It turns-out, there are two significant effects that are not necessarily taken into account in FEA. 1) component connection stiffness, and 2) Material stiffness. 1) in real life, you typically lose large amounts of stiffness anywhere a load has to pass through a joint. FEA makes all joint connections perfect, whereas in real life this never happens. The second item is 2) a material effect. We found that we had to customize our material model in FEA. We did this by running actual vibration tests in a lab on cantilever beams to identify first mode natural frequency, and backing-out an appropriate modulus, E and density, rho. We plugged into FEA to make a custom material. After we were sure we got the material properties correct, then on to the structure. In some cases we used the elastic foundation at supposedly "rigid" points, or super-stiff springs between part surfaces at connections, and tuned the structure to match known test data for known solutions in bending or vibration. It is quite typical for the need to calibrate the material and the structure to known answers prior to going further. Doing this over a period of several weeks, we had much success with final designs matching within 5 to 10% of simulated models. You are on the right track. Good luck.

            • Re: Predicted vs Tested Deflection 4130 Tubing
              Mark Horne

              Thanks Anthony,

               

              I was hoping it would not be the case to "Find" material properties that reflect the known test, but I suspected it would be. I do not have access to frequency measuring equipment. What would you recommend for steps to follow when adjusting the properties? I am betting just changing a value and seeing what happens will take a long time to hit the correct combination.

               

              Soliworks 4130 normalized 870c material properties:

              Elastic Modulus            = 297322700 LB/in2

              Poisson’s Ratio             = .285

              Shear Modulus            = 11603000 LB.in2

              Density                         = .283599 LB/in3

              Thermal Conductivity = 42.7 W/mk

              Tensile  Strength            = 106023 LB.in2

              Yield Strength            = 66717.3 LB/in2

               

              It seems there should be a strategy when adjusting. If one spec is changed, does another need to be adjusted at the same time? What I am asking is would changing one parameter require another to be adjusted by a ratio or factor?

               

              Is there some reference material that explains the interaction of the different properties and perhaps when and why to adjust? Or is it "seat-of-the-pants".

               

               

              Thanks again,

              Mark

            • Re: Predicted vs Tested Deflection 4130 Tubing
              Roland Schwarz

              What you did right:

              Correlating test results to experimental results.  Benefits of FEA can only be fully realized in a program that consistently compares experimental results to actual results.  The main reason to do this is to verify and refine modelling technique.

               

              What you did wrong?

              Hard to say.  But, there is always something.  Overconstrained or underconstrained joints; material properties; force application; improper constraints, modelling error.  On the physical side, be sure forces and constraints are applied consistent with model.

               

              Material properties?  Almost all steel has the same elastic properties if there is no plastic deformation.

                • Re: Predicted vs Tested Deflection 4130 Tubing
                  Mark Horne

                  Thanks for the help.

                   

                  I did lower it some and got closer to the test displacement earlier this week.

                   

                  I did not lower it as much as you indicated, so I will try reducing it some more and see what I get.

                   

                  Thanks again

                  Mark

                    • Re: Predicted vs Tested Deflection 4130 Tubing
                      Mark Horne

                      I lowered the value of the Elastic Modulus as follows to get the same numbers as the physical tests:

                       

                      On the straight tube: EM was lowered 25% to a value of 22299552

                       

                      On the frame: EM was lowered 81% to a value of 5649219

                       

                      Knowing vary little about this, it seems to me there must be another value which needs adjusting to make both work.

                      Not sure what to do next without just stabbing in the dark.

                       

                       

                      Mark

                        • Re: Predicted vs Tested Deflection 4130 Tubing
                          Mark Horne

                          Now that we finished the racing season last weekend, we will break the cars down to the chassis and fixture them up on the steel table. Then apply the load to the front corner and see what we get there. I will then offset the Elastic Modulus to match what we get on the known chassis and use that number to load the new design.

                          I feel this will get me closer than trying to make the simple samples work to find the EM to use.

                           

                          Any thoughts?

                           

                          Thanks,

                          Mark

                            • Re: Predicted vs Tested Deflection 4130 Tubing
                              Roland Schwarz

                              Since you are in a position to experiment, continue doing so.  Start with some simple stuff, just to get a feel for the level of trust you can put in the math.

                               

                              A simple place to start might be deflection of a single tube.  A fixture or clam, a length of tube, and a weight can be used to conduct quickie experiments and compare to analysis.  Move on to simple structures.

                               

                              There's a lot of funny things about finite element methods that escape the casual user.  It can take multiple iterations to arrive at convergence.  Try doing FEA on a simple model multiple times (bending a bracket with holes is a good one), starting with a very coarse mesh and getting finer and finer.  You want to refine the mesh to the point where reducing element size 1.5x to 2x does not significantly change results in areas with high stress gradients.

                               

                              Also, learn about mesh refinement.  This is where the program automatically refines mesh in response to high strain gradients.  Analysis takes longer but the actual time to get an accurate analysis you can use is reduced (not to mention saving time lost due to bad decisions based on poor analysis).

                            • Re: Predicted vs Tested Deflection 4130 Tubing
                              Roland Schwarz

                              Previously you wrote:

                              elastic modulus = 297322700 LB/in2

                              That's 297e+6 psi

                               

                              Was that maybe a typo?  That's where I got the impression your modulus was off (though that would have made your FEA model stiffer).

                               

                              Elastic modulus for steel (any grade) should be in the neighborhood of 30e+6 psi.

                                • Re: Predicted vs Tested Deflection 4130 Tubing
                                  Mark Horne

                                  297e+6 is the default EM Simulation provides on my system. Keep in mind, I am still just using the Student Edition while I am in school. I intend to buy my own seat early next year.

                                   

                                  I am using Beam Elements as this is the default with Weldments. I did try it as a solid and adjusted the mesh size. The results were still anomalous from our tested simple samples. That is, no single number for EM would satisfy both sample tests. Perhaps I should not expect to have a single number for both.

                                   

                                  As I mentioned above, I will be able to test our reference chassis, hopefully this weekend, and see what I get. I suspect it will be close to the 30e+6 you mention.

                                   

                                  Thanks for the help,

                                  Mark

                          • Re: Predicted vs Tested Deflection 4130 Tubing
                            Mark Horne

                            This weekend we finally had the time to strip the known chassis down and test them. On the first one,we fixed the rear axle on both sides and the front left spindle. We then loaded the front right corner with three different weights and recorded the displacement. I then ran the chassis in Simulation and found that a value for the Elastic Modulus of 11.5e+6 gave me results within .003" of the measured test.

                             

                            So we then predicted the displacement on 2 more known chassis with the EM of 11.5e+6. We then clamped down the other two chassis and found a max resultant deviation of .007" between the two. Confidence level is high that I will get results close to what is predicted. My instructor seems to be satisfied regardless of the differences of the EM values.

                             

                            So why is it 1/3 of the standard EM? I do not know. Perhaps the Student Edition has some thing going on here?

                             

                            Thanks for all the help.

                            Mark

                              • Re: Predicted vs Tested Deflection 4130 Tubing
                                Bill McEachern

                                You seem to have forgotten why you are changing the material stiffness. From a brief read of the posts it appears you are changing the modulus as it is the only thing available to you in the model you have and the program you are using. You are correcting for things that at not in your model and maybe somewhat difficult to model like joint stiffness and restraint stiffnesses, which, it appears, are perfectly rigid in your model. There is nothing wrong with the program you are just getting out what you put in. And your solving problem with the tools at hand to a known level of uncertainty which you appear to now have quantified..

                                  • Re: Predicted vs Tested Deflection 4130 Tubing
                                    Mark Horne

                                    Thanks Bill,

                                    This is my first attempt at using Simulation on something I can test/prove and I had some assumptions which are apparently wrong. You are probably correct with joint stiffness. The welded joints must be providing less resistance when the loads are applied on the real world chassis. Can a joint be welded in SW and that joint stiffness be reflected in Simulation? If so, can the filler material also be specified? I am looking at the program now and do not see anything obvious. Perhaps it is uses nomenclature I am not familiar with.

                                     

                                    The tutorials in SW lead one to believe the output with the default material properties will be very close to a tested reality. At least that was the notion bouncing around in my head. After this experience I now understand it depends on the model, restraints and load application.

                                     

                                    I have two SDC Publication books on Simulation. Both do not mention verifying results and adjusting joints, restraint stiffness etc. What would be a good source for this material?

                                     

                                     

                                    Thanks again,

                                    Mark

                                      • Re: Predicted vs Tested Deflection 4130 Tubing
                                        Bill McEachern

                                        for welded joints I would guess adjusting the material modulus is probably your simplest option. You would need to be careful about the "span" of the connection and making sure that however it is in real life it is in the model. On welded tube to tube - it matches up pretty well but if you have plates welded to the tube end and  then the plate is welded to say some flat face - the flexibility of the plate may come into play. To illustrate imagine that the plate is 10 x the diameter of the tube and is thin and you only weld around the perimeter and then you put bending across the joint - it will be really flexible but if you use default bonded in the analysis that effect won't show up and the joint will be quite stiff ans the whole face of the plate is attached and not just the perimeter. You need to look really closely at how the thing is put together and how it behaves under the applied loads. If you want to match reality you have to model reality.

                                         

                                        The bigger uncertainty might be at the restraints.

                                         

                                        The other thing is that most linear analysis is a less accurate estimate than what you would get with a non-linear analysis, though most of the time this limitation is not an issue but then again it depends on the subtleties in the model.

                                          • Re: Predicted vs Tested Deflection 4130 Tubing
                                            Mark Horne

                                            Thank you,

                                             

                                            It is good to get real and honest feedback.

                                             

                                             

                                            Mark

                                              • Re: Predicted vs Tested Deflection 4130 Tubing
                                                RICH PAUZA

                                                VERY interesting stuff.

                                                 

                                                Over the last 10 years - 4130N tube - commercially available, has had some consistency problems.

                                                 

                                                The ( accepted tolerence ) is pretty wide.

                                                 

                                                Based on my experiance, the tube you can buy today is not the same quaility as the tube you could buy 10 years ago.

                                                 

                                                You can see this very clearly in tube bending - how much spring back today's material has compared to old stock bent the same way.

                                                 

                                                You can also see the differance in the cross section of the tube after bending, the way it machines & the way it welds.

                                                 

                                                It is possible the sample tubes you used were not up to the specs SW has locked in for the material.

                                                 

                                                Not sure if this is information you want to here, but it is a fact.

                                                 

                                                 

                                                All things being the same - if my computer results did not match my real world testing results, I would use the computer results for comparison - and I would DESIGN based on my real world test results.

                                                 

                                                 

                                                Rich Pauza

                                                  • Re: Predicted vs Tested Deflection 4130 Tubing
                                                    Mark Horne

                                                    Thanks Rich,

                                                    We were having that very concern about the material not matching the specs when nothing was even close to what SWS predicted.

                                                    One of the chassis was manufactured in California back in circa 1995. The second one was built by me back in 2003. The third one was built by my racing buddy last year. All the material for the 2 newer chassis was sourced from different vendors. Hard to say if any of it was from the same heat or not. We did not record that. We have no idea of the material source or heat from the Cal chassis. I guess a good thing to do is make sure all the new tubing is from the an identical heat. That would at least provide consistent material for the new project.

                                                     

                                                    What ever the differences are, they seem to be minor when compared to each other, at least using the EM of 11.5e+6 value and tested chassis. Is it possible to buy tubing that is to spec? The stuff we have is marked with a matching mil-spec number. I do not have that in front of me right now. (Mil 6732B seems to come to mind. Could be different) One of our sources is a local steel retailer the other is a local aircraft supplier. It would be interesting to test something at nominal spec material against what we have.

                                                     

                                                     

                                                    As far not wanting to hear something, I am not afraid of the truth. I would rather get honesty over BS any day. :-)

                                                     

                                                     

                                                    Thanks again,

                                                    Mark