# SolidWorks FEA results questioned - torque on cylinder

Hi everyone,

I have a 100 mm long, solid cylinder (shaft) with a diameter of 20 mm. The base circular surface at one end is fixed and the other end gets a torque of 100Nm.

The SolidWorks simulator (URES) results show that the diameter of this shaft increases by about 0.1 mm due to the applied torque. Similar is the outcome using Autodesk Inventor simulation. My cylinder becomes somewhat conic-like in shape, also not completely simmetrical along its original axis. So far I have only seen books with theories that begin with the assumption that the shaft cross section does not change in diameter.

Could these results be false? Can you people tell me how to get the same results on paper? Name a book that explains these results?

Last edited:

AlephZero
Homework Helper
This is an artefacct of the way the graphics are drawn.

The FE analysis uses small displacement theory, so the results say the displacement of every point on the cylinder is in the tangential direction. That is a good approximation, so long as the deflections are small.

But when you scale up the displacements to make them visible on a plot, it looks like the radius of the cylinder increases.

If this bothers you (or you don't want to put a misleading picture in a report) just plot the stresses on the undeformed structure.

You could fix the problem by doing a large displacement analysis (if that is available in solidworks - I don't use it) which would give the "correct" displacements for an arbitrary rotation (i.e. if you modeled a long cylinder and the torque twisted it through 360 degrees, the displacements at the end would be calculated as zero!). But you can only plot the results of a large displacement analysis correctly at "actual size", not scaled up to make them visible.

If the stresses in your model are sensible (e.g. within the elastic limit of the material) the results from the small displacement analysis will be OK in practice, even if the graphics look strange.

I agree with Aleph, it's probably the graphics on how it's drawn. If that's not the real cause, it's probably due to the mesh size being too large. The equations in finite element theory as formulated such that in the limit as the element size approaches zero, the result becomes more and more accurate. You can't really control the element size too much with COSMOS (or whatever they're calling it nowadays), so your displacements might be off a bit. 0.1mm over a 20mm diameter is 0.5%, so that's not too bad. As well, SolidWorks FEA isn't exactly the most accurate analysis software. It'll get you close, but if you want more exact answers, use Nastran or Abaqus. Also, I think you might be confusing what URES gives you. It will tell you the total displacement of the nodes in those areas, but not necessarily where the nodes are displacing to. I think the 0.1mm is the displacement of the node around the shaft, or in other words the angle of twist times the radius of the shaft.

The main thing you have to keep in mind with FEA is that it only does what you tell it. How you define your constraints, loads, elements, the placement of your nodes, etc. makes certain assumptions. More often than not, it's your fault if it doesn't give you the "right" results because you made some bad assumptions somewhere. I think it's great that you posted this because you're trying to make the results jive with your engineering intuition. It seems more and more people are just blindly trusting results from analysis software and not thinking critically when the results seem counter-intuitive. +1 for that. By the way, a torsion case like in your model doesn't increase the diameter of the rod. You're right about that.

Last edited:
Well even without scaling (on the picture it is about 100x) this 0,1 mm deformation is slightly visible. I was a bit worried about the simulation results vs books telling me there is such no deformation. But now that you have explained what is happening I understand this is not to worry about too much.

Thank you for the answer AlephZero