4 Replies Latest reply on Jul 19, 2015 1:06 AM by Michael Oswald

    problem defining fixture for static analysis of cylinders

    Michael Oswald

      I have two hollow cylinders standing on top of each other axially.  The faces in contact with each other have a very shallow, very wide chamfer.  The bottom cylinder is sitting on the ground, and I apply a 2000# force to the top cylinder (axially downwards).  I want to see if/how much the chamfers between the cylinders deform.  I think this should be really straight-forward, but when I try to run the analysis I get an error stating that "The solver has numerical difficulties.  Model may not have adequate fixtures".


      My only physical constraint is the floor that the bottom cylinder is sitting on.  I've tried to model this with:

      • a roller/slider fixture on the bottom face
      • a fixed geometry fixture on the bottom face
      • an advanced/flat face on the bottom face with the normal translation set to 0
      • an advanced/reference geometry on the bottom face with an axis to define the direction and the translation set to 0
      • by creating a coincident reference plane on the bottom face and setting it as a virtual wall. 


      I get the same error for all but the fixed geometry.  For that I get incorrect results - deformation in a circular motion, mainly to the positive-x direction.   I also have the global contact set to no penetration instead of bonded, because that's how the parts really are.  But I've also done it bonded, and gotten the same error.


      I'm sure this is just a stupid error on my part.  If anyone has any suggestions, I'd really appreciate the help.  Thanks so much.

        • Re: problem defining fixture for static analysis of cylinders
          Richard Wehmeyer

          There are a few things you need to do to make this work.  First, it sounds like you have a point/line contact between two parts.  You cannot reasonably do this in FEA.  If you must make a sharp edge contact another body, add a small flat edge to cut off the sharp point.


          If the two cylinders are identical (the area of interest anyway) then only model one cylinder against a solid wall (for a linear analysis this will do the same thing.


          Assuming the identical cylinders you can now take a single model, cut a flat spot on the chamfer, use a spring constraint on the cylinder wall (value can be quite large like 100-1000lb/in2) then make a sliding contact on the flat face and force load on the top/bottom of cylinder.  This will give you the most accurate result acievable with a linear static analysis.



          Edit:  I created this part doing what i think you are trying to do although to get the results you are looking for I reversed the load and contact.  This makes no difference for the study since a linear analysis will follow the rules of statics to the "t".

            • Re: problem defining fixture for static analysis of cylinders
              Michael Oswald

              Thanks so much for your help Richard.  I should have given more details in the original post.  The chamfer takes up about 1/2 of the area of contact between the 2 cylinders, so there is a flat there.  The part you made is very similar to what I've actually got, the main difference (like you said) is that I'm applying the load on the top and the fixture on the bottom whereas you did it the other way.


              I don't really understand the elastic support fixture on the circumference because there isn't anything around my parts to push against, but when I added that to my model it was at least able to converge on a solution.  I've got to fool around with the parameters as it still seems odd (e.g. there's uneven radial displacement - more in one direction than another) but I couldn't even converge on a solution before.


              One thing I did want to mention is that I took two of your parts and created an assembly with them, mating the small-diameter faces together.  I then added the force on the large face of one part and the rolling/sliding fixture on the other.  I got the same error I was getting before (model may not have adequate fixtures...) until I changed the global contact from 'no penetration' (which is what I think it should be because the parts are just sitting on top of each other) to 'bonded'.  Once bonded, it came up with a solution real quick.  But I don't really want to bond those faces together because in reality they could slide.  Is there any other way to define that type of contact?

                • Re: problem defining fixture for static analysis of cylinders
                  Richard Wehmeyer

                  All bodies in an FEA static must be properly restrained.  Think of all parts as being coated in thin oil.  If you restrain the lower body and press the upper body down on it a tiny truncation in the calculation can cause it to try to jet out.


                  The spring constraint does not act on another body, it is like a soft support.  It acts on the coordinate system.  if the body moves to the left, it gets a force to the right.  The further it moves the more it gets pushed back.


                  The bonding of the parts gives the top body the left-right-front-back support, just like the spring will.  Again, since the parts are identical bonding will create the exact same effect as if it slid.  Remember: this is a linear analysis.  slide, contact, bond, invisible wall: the result will be the same unless you INTRODUCE an anomaly.


                  You have only a downward force on the top body and contact on the bottom.  What restrains it front to back?  This scenario requires a little help to keep it from popping out, hence the spring to stabilize it.  Use the spring constraint on the top body to get your desired results from the assembly.  In the part I sent the spring constraint was not necessary because I had fixed the base.  Had I used a sliding contact then it would have been necessary to stabilize the body.

                    • Re: problem defining fixture for static analysis of cylinders
                      Michael Oswald

                      Thanks for the explanation Richard, I appreciate your help.  I understand what you're getting at, and I do indeed see the bodies 'popping out' when I don't include the spring fixture (the few times it i\was able to converge in that circumstance).  I was trying to get a more 'realistic' answer by using sliding contacts instead of bonding, but in that case I should have included things like friction as well to take care of those tiny truncations you mention.