Knurled Interference Fit Analytical Analysis

[amtrak17]

thread404-213125
I am looking for a way to analytically determine stresses on a knurled interference fit. The referenced thread above was seeking the same, but unfortunately no one actually answered the original question. I am not interested in whether or not a knurled press fit is a good idea. This is a mature production part with no history of failure. I simply need to perform a safety analysis for one part of a certification. Is there any analytical approach to this problem that would avoid a full FEA type of analysis?

 

[Tmoose]

Before attempting any analysis I would dissect some new assemblies and some old assemblies and new and NOS shaft and hubs. This would be to understand the range of knurl ODs, and what happens to the knurl and bore when assembled.
I envision the knurl tips may be being sheared off, flattened to some degree, or are cutting perfectly into the bub bore.

What is the load in service ? Torque, axial withdrawal, some bending ?

 

[amtrak17]

Those are good ideas, and I think your understanding of the mechanics of the fit are correct. The operational loading is purely torque. I am performing a Monte-Carlo type analysis to simulate the range of shaft/bore diameters and come up with a statistical range of interference at that press-fit interface. Again this is a mature product with 10s of thousands of assemblies and no issues. My goal is not to identify some type of existing failure, simply to perform an analysis on the stress at the interface. A dissection is not going to give me the stress values I need to put into the report.
Do you know of an analytical way to evaluate stress at a knurled press fit?

 

[geesamand]

It seems like an exceedingly complex phenomena to model analytically. Heck, basic friction calcs are just rough estimates (pun intended) but if you work in plastic deformation, material adhesion, and manufacturing tolerance, it would be a WAG on the best day. So the validity of the calculation would be far inferior to empirical production experience and/or testing. And with volumes of production examples at your disposal, you have something to test.

David

 

[amtrak17]

Thanks to all for the comments and answers. This is resolved for me now. Looks like testing will be the only way to have a reliable estimate.