Static FEA Shaft Analysis: ignore stress riser effects? 1

[blademaker]

Hi All--this is my first time posting on this site. I need some advice on shaft design & interpretation of FEA results.

I'm working an linear static FEA model of a small wind turbine shaft (AIS 1026 steel) using COSMOS Adv. Professional 2008. The shaft is subject to a bending moment, axial & torsional loading--see attached photo. It also has a couple of step featues, but the small step feature (45.9 mm to 45.01 mm with a 1 mm radius fillet) is causing me trouble. I am using 3 mm solid elements with much finer mesh around the fillets (I have varied the mesh in this area from 1 mm to .1 mm). The mesh aspect ratio is quite good, as well.

This feature is causing very high stresses in the area of the 1mm fillet. I have performed a series of successively finer mesh scenarios around this feature, and the stress continues to rise. The stresses in the rest of the shaft are well below yielding, though. My first inclination was to do a quick hand calculation to check the results, but all of the S.C.F. formulae & charts for shafts with a filleted shoulder do not apply to such a small step.

It seems to me that the mesh is well-refined in the area of the fillet. My questions are as follows: first, how should I interpret the FE model? Refine more? Refine less? Second, any thoughts on the proper K_t for hand calc verification?

Thanks very much for your help!

 

[SWComposites]

Suggest you correlate an FE model of a simplified shaft with a fillet configuration for which you have a SCF formula. Determine the required mesh density, mesh size to radius ratio and post-processing method to get the FE result to match the SCF result. Then use the same modelling approach for your actual shaft.

Or just test the shaft and correlate the results to the FE model.

 

[EdR]

Looks to me like the model is giving good results. You have essentially a 90 degree corner (don't remember what the SCF is for this situation but you should be able to find it in a reference book) and the mesh will continue to give high stresses in the corner regardless of the mesh as this is the actual situation.....You need to put a reasonable fillet on the corner ..... probably best done with hand calcs.....A mechanics of materials book gives a SCF of about 4 for a sharp radius and a handbook will probably give even better estimates....

Ed.R.

 

[corus]

Generally on radii it's recommended to have at least one element every 15 degrees, or 6 elements for 90 degrees. You could check out your level of refinement by using a 2D axisymmetric model with the axial loading. Also make sure you have no direct point loads or restraints at the maximum stress as those will give you infinite stress as you refine the mesh more.

corus

 

[blademaker]

All--thank you for your responses...

For reference, I didn't want to change the fillet radius value, since this part is being tested for increased loads "as-is." In other words, this shaft is already in production and we'd like to use it with increased loads w/o modification. I may end up changing a few values to ensure my model is reasonable...

I'm in the process of simplifying the model even further so that I can actually use a real SCF value to compare the values. Corus, thanks for the elements versus degrees recommendation....I may also modify the way I applied my loads & restraints.

 

[Stringmaker]

Well the best way would be to submodel the area in question. I'm not certain if Cosmos has that capability. Also, it would help to show the stress rather than the factor of safety in your plot.

Peterson's book is the best source for SCF info. If your model predicts close to a value that you get from there I'd say you have fairly good convergence. That being the case your FS < 1 which might present a bit of a problem.

Good luck.

 

[corus]

I'm not sure if the factor of safety in Cosmos is simply based upon yield. In this case you'll be looking at fatigue failure of a rotating shaft and you should be looking at the rotating bending fatigue limit of the material, factoring in surface finish, size of specimin, etc. In addition you'd need to make sure the stress was below the UTS of the material. If the stress doesn't satisfy the fatigue limit it will mean that the shaft will have a finite life, and not infinite.

corus

 

[rb1957]

isn't the model showing localised plasticity due to the fillet ? i suspect that if you take a section thru the tangent point of the fillet that there is only a sliver of the surface that has FoS < 1. i'd be more concerned about the fatigue life ... is this high strength steel ? do you have a fatigue spectrum ? i understand you're doing a static test of the part, fatigue test too ?

 

[blademaker]

All,thanks again for your helpful posts....

My goals for the analysis of this shaft were twofold: first, I wanted do a simple linear static analysis to ensure the shaft could handle the new loads. Second, if the static analysis looked OK, I would proceed to fatigue analysis, which is the more relevant one for rotating shafts.

The show stopper occurred when the linear static showed the FS < 1. I want to be sure that those peak stresses were correct, and not just an artifact of improper FE modeling. Is the localized yielding (plasticity) around the small fillet indeed a show stopper (see photo)? As rb1957 mentioned, there are only a few areas of apparent yielding. Perhaps I should move on to fatigue analysis?

 

[rb1957]

if these are linear stresses, then they're definitely not "right" ... do a quick material non-linear run to see the difference (the "real" stresses will be significantly lower as the material yields and redistributes away from the peak).

can you section the model (to see the stress distribution thru the thickness) ?

if your original post is still good (FoSmin = 0.1) i'd be nervous about the fatigue results.

you mentioned this to allow for increased loads ... would you be able to spon peen the fillet ? this'll help your fatigue life significantly.

 

[blademaker]

rb1957: my gut instinct tells me that those results are a bit odd. I'm running a non-linear analysis to verify peak stresses right now.

Did you mean shot peen the surface? Yes....that would be an excellent option for increasing the fatigue life.

Thanks!