Shrink Fit Contact Set Results Interpretation

[morienr]

Hi all,

I have ran a shrink fit analysis of a shaft in a hub with 0.0044" interference. The results of the analysis are attached.

My question is in the interpretation of the results. In the radial stress plot I get high compressive stresses towards the bottom of the fit bore and for the hoop stress plot I get high tensile stresses at the top of the fit bore. This effect is expected as can be seen in the radial and hoop stress distribution for cylindrical fits.

Could someone please explain why these effects repeat every 120 degrees?

Thanks for your help.

 

[salmon2]

This is odd question, all I can think of is change your mesh density and see any difference.

 

[corus]

Could it be that the coordinate system of the stresses is in the global system and not in radial/hoop directions? It's still strange though as you'd expect the results to show similarity at 0 and 180 degrees for instance.

Tata but not yet tara

 

[morienr]

Hi Salmon2 and Corus,

Maybe it would be better if I restate my original question.

I have run a shrink fit analysis of a shaft in a hub with 0.0044" interference. The results of the analysis are attached.

My question is in the interpretation of the results. In the radial stress plot I get high compressive stresses towards the bottom of the fit bore and for the hoop stress plot I get high tensile stresses at the top of the fit bore. These stresses are plotted in their respective coordinate system (radial, and cylindrical).
I think these results are justified by the radial and hoop stress distribution for cylindrical fits in slide 3, but I am not sure.

My question is why these effects repeat every 120 degrees rather than every 180 degrees as in the slide 3?

I also ran a second simulation similar to how these parts would be assembled in the field by leaving the hub at ambient temperature and applying a temperature on the shaft.
(Initially, both bores are size on size and the temperature is applied to the shaft to simulate the press fit).

So, my second question is, which of the results should I use in a report? The thermal differential analysis seems like the proper choice, since this is similar to how the assembly would fit together and the fringe plots are much more uniform as I would expect.

The differences in these two results occur no matter what mesh refinements are applied.


Thanks. again for your help.

 

[corus]

The problem is axisymmetric so you would expect to tee the stress distribution you got from the thermal expansion case.
For the 1st case, instead of running the full 360 degree model, why not run a 90 degree model with symmetry restraints? You should get the same answer as the thermal case but without the unexplained 120 degree peaks.

Tata but not yet tara

 

[kellnerp]

Please create a vector plot showing the first principal stress.

I have seen this before in Cosmos.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[kellnerp]

Here is a similar run done in DesignStar using a vector plot of the first principal stress to show the direction of the tensile stress on the inside of the hoop.

P1 vector plot.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[morienr]

kellnerp,

Sorry for the delay to your request but have been very busy here at work.

The attached vector plot of the first principal stress shows the same story of 3 points of high stress around the diameter.

Thanks

 

[kellnerp]

I've seen that before in Cosmos. Have you been able to inspect the mesh in the areas of high stress to see if something repeats? Does it change with a mesh refinement?

If you cut the model in half or quarters do you still get this effect?

Unfortunately there is little documentation on what Cosmos Simulation is actually doing that you have to sleuth these out by reverse engineering.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[corus]

I notice that there is a difference bwteen the two analyses in that one is contact whereas the other you tie the surfaces together and apply different temperatures.

It could be that the meshes between the two parts, in the contact case, intersect at certain intervals corresponding to the 120 degree angle. With contact problems you can get nodes passing through the surfaces depending on how you've defined contact.

If you've used 4 noded tetra elements, your results will be useless anyway as you need quadrilateral elements. It would be interesting to see a picture of the mesh.

Tara

 

[kellnerp]

Corus

10 node tets are the only other choice (and default) in CosmosWorks. No hex or brick elements at all.

There are two types of contact problem, a node to node element and a surface penetration element. There is no ability in CosmosWorks to see how individual elements or contact sets are setup and there is no documentation on this. Node numbers are not easily determined nor are nodes, especially mid side nodes, easily picked or visualized. That being said, there is a way using probe to plot force or stress at a line of nodes. Perhaps if split face was used for force a nodal boundary the various forces and stresses around the circumference might be visualized.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[corus]

kellnerp,
Quadrilateral elements refer to elements with mid-side nodes, and so refer to 10 noded tets. 4 noded tets give bad results and should be avoided unless used for purely thermal models.

It would be simple to see if penetration of surfaces had occurred by zooming in on the areas concerned and perhaps increasing the deformation scale factor. Generally, though, if the mesh of the two contact surfaces match then there shouldn't be a problem with penetrating surfaces/nodes.

Tara

 

[kellnerp]

I beg to differ. A 10 node tet is an isoparametric element with mid side nodes. Quadrilaterals are nominally rectangular plane or shell elements. They are defined by four distinct corner nodes and any number of midside and inside nodes.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[morienr]

Hello Corus and Kellnerp,

I will work on dividing the part up into 1/3's or 1/4's and apply cyclic symmetry conditions to the analysis over the weekend, but I wonder how this will apply when there are unsymmetrical loadings such as dowel pins in housings, etc.

Corus, I applied a global compatible mesh, meaning that the mesh should be node to node and thus the elements should not intersect into each other. There are two easy ways to see this - in the bonded case look at the normal stress in the bore and should see compresseion on one and tension on the other. The other simple method is to apply no penetration contacts and then plot the contact pressure. You should see contact vectors on both parts.

I also respectfully disagree that tet4 elements are useless since the majority of the work needs to be verified with book equations. It has been my experience that tet4 elements provide an accuracy with hand calcs to within 10 to 20 percent. Agreed, that tet4 are crude, linear elements but are very useful when attempting to optimize a design through several iterations.

Cheers.

 

[kellnerp]

I believe the TET4 is also constant strain so you need to at least do a convergence study to make sure you are getting the right numbers.

Elements with midside nodes do have problems with contact because the reactions at the nodes for constant strain are not equal and not always obvious.

TOP
CSWP, BSSE

http://www.engtran.com

http://www.niswug.org

http://www.linkedin.com/in/engineeringtransport

"Node news is good news."

 

[JacobMcGrew]

Corus,

I believe you are confusing quadralateral elements with quadratic elements. I heard both refered to as quad elements which I think leads to confusing the two.

Quadralateral elements are four sided 2D elements.

Quadratic elements are 2nd order elements (as compared to linear elements). Linear elements don't have midside nodes because you just need two points to define the line. Quadratic elements have midside nodes because they are defined by a quadratic (2nd) order equation.

 

[corus]

You're probably right Jacob. I blame the Greeks for the confusion.

Tara

http://tinyurl.com/4ydjg7m