Sheared hub shaft, check the photo!

The loads are actually pretty easy to calculate depending on how accurate you want to get. The easiest way is to simply say that on race rubber 1.3g is about the limit of traction, there is a 100% transfer to the outside wheels and 60% in the rear. The car weights 3200lbs, so the load where the tire meets the ground is 2880lb and work from there. It will be a little less braking in to the turn and a little more accelerating out of the corner but it’s close and I would want to safe a safety factor of at least 1.5 which will wash away whatever error there was in the load assumptions.

To your point any kind of defect won’t make things worse, whether its tool marks, rust pitting, or previous impact. My only point was that the failure is right where I would expect to see a failure. The model I ran is just a shaft that looks about right with loads in the right directions to confirm that the assumption that cornering loads should fatigue the shaft at the base of the radius and clearly they do. This is the third first hand account of a shaft 3with this exact failure in track use I’ve heard so I have to believe any analysis that says the shaft won’t break is wrong.

Since I already knew the shafts are prone to fail at the base of the radius I didn’t bother with exactness. I think I would want a more exact analysis before blessing a redesign. I’d like to see the large OD extended over to the small bearing….or the large OD extended 10mm then tapered down to the small OD and maybe a nice big hole drilled to about 2/3 of the way through the outer bearing section and a small hole the rest of the way through the shaft and have a shaft that is both significantly stronger and a bit lighter than stock.

 

I gave it a quick tune-up and re-ran the analysis. I made the large OD a bit longer and tapered it down to the small OD and bored out the center. The shaft is 10% lighter and 25% lower stress. This is just a generic design to show the form of the solution as I have not measured an actual shaft to generate an exact drawing/solution.
Image Unavailable, Please Login
Image Unavailable, Please Login

 

Hmmmmm , the axle does not operate in a void , what happens when the bearings and axle housing are added to the equation ?

Why is the axle draawn as bent somewhat . . the CV joint on the end of the axle does not support / force the end of the axle in any particular direction.

How great is the "bend" ion the axle as shown in the first illustration ? (in thousandths of inches please . I'm old !)

 

Think free body diagram. The other part only matter where they can impart a force on the shaft so I set constraints where the bearings contact the shaft then loads where the wheel bolts put the wheel loads into the shaft. Everything that needs to be in the equation should be.

The axle is drawn bent because I set the analysis to show the deformed shape....or the shape the parts takes under load. The software multiples the deformation to make it easy to see. I didn't look at the actual deofrmation number, I tjust looked at stress distribution. I'll take a look in the morning, I'd expect it to be around .001" I think.

 

In the model I made the OD of the flange moves .005" and the shaft just inside the outer bearing moves .002".

Remember though that this is a rough model I made in about 10 minutes and I didn't have the correct dimension info to plug in. Its a model that just looks about like the real part so it will act about like the real part but the actual numeric values don't really mean very much, I was simply looking to comfirm that the point of highest stress is the location the shaft failed.

 

Just getting onto the subject of the computer simulation only for a moment ... (intentionally getting away from the axle itself)

Is it possible to design something and then add a specific "damage" to the item ? I am guessing if that is possible you could test all kinds of materials for suitability if you hadn't decided upon one already ...

 

I respectfully disagree. If the shaft is completely constrained in the model at both bearings then there can not be deflection at those locations. Therefore there is no deflection between the bearings. This can be visualized or tested by clamping a metal rod or a piece of wood at two locations along its length. Then pull up on one end of the metal/wood and see where the bend is. It will be outboard of the clamps not inbetween.

The only way for the shaft to bend between the bearings would require that one or both bearings to be crushed or the bearing bores in the housing to yield significantly. Neither of these senarios is likely.

 

If that were the case it implies that one or both bearing would be out of spec, or that spec is not enough for the (racing)application. I don't think that the bearing needs to be crushed.
The bearings may 'feel' good, but are they really within spec? I'd wager that if not, this could introduce a deflection of the shaft between the bearings and the subsequent damage.

 

Tony,
Getting the constains correct is by far the hardest part of an FEA analysis and where most peolpe go astray for exactly the reasons you state. The classic quote is "FEA makes a good engineer fast and everyone else dangerous". Garbage in, garbage out also applies nicely.

The models I posted show the deformed shape and you should be able to see that the parts are not only bent between the bearing locations, but the peak stress (and therefore any failure) would occur between the bearings. If I had simply grabbed the bearing sufaces and locked them I would have gotten exactly the result you describe, so that is not what I did. I grabbed the surface just inside the OD bearing where the daimeter changes and locked y, z translations only and lock on rotations(x is the axis of the shaft) so it can't move up/down or front/back but can move side/side, spin along the X and also flop along the y and z just like it could inside the bearing. Then I grabbed the end of the shaft which would be the outside of the inner bearing in a more complete model and locked x, y, z translations and rotation about the x. This provides the required 6 degrees of freedom constains (x, y, x translation and rotations - remember locking translations at 2 point locks rotation). On a real analysis I wold be more careful about making sure the constrains act act the center of the bearing and things like that, but for the purpose of a quick look I didn't bother with that level of accuracy.

Mark

 

A single row ball bearing will flop around quite a bit even when new. they are design to support the shaft, not prevent rotation is any direction. To lock a rotation in 2 or the 3 directions you need to use 2 bearing spaced some distance apart and doing so causes the shaft to bend if any load ia applied in that direction.

As far as a bad bearing causing a bending load, a little maybe but not really. The bearing will fry and fail completely and take the other bearing with it long before it could cause a bending load in the shaft.

 

I thought when we discussed this before the failures were happening on the outer flange?

This is the first time I've seen the failure at this location.


cheers,

Sean

 

I don't know....I thought the others were inside...I really don't recall now that you ask the question.

 

Sean, you are correct; we discussed mostly axles which broke at the outer flange. One of the examples you can find here:
First picture in post 15.

http://www.ferrarichat.com/forum/showthread.php?t=119568&highlight=mondial+stub+axle



Best Regards from Germany

Martin