Notch Sensitivity "q"

[imcjoek]

We have a stepped shaft in the field that has failed (rather clearly through fatigue due to bending).

I am trying to determine what should be done on the replacement shaft. It has failed right at a fillet, so obviously I will make that fillet larger (since the step cannot be eliminated).

I am also looking at material and heat treat selection, but have hit a problem. I need to use Kf (fatigue stress concentration factor), but that requires a "q" value (notch sensitivity). I cannot find any data on this (other than generic "here is some trend for steels"). Original shaft was 4140 (normalized), and I would like to see what going to Q&T 4340 would do. I expect that will increase "q" and I am concerned that that would just negate any benefit of the increased strength.

So: where do I find this data? (or similar. Nuebert?)

 

[metengr]

imcjoek;
Increasing the fillet radius to an elliptical profile may be all you need to do to reduce stress concentration.

Was the fatigue failure high or low cycle and what was the application? Can you post any pictures that show the location and appearance of the fracture surface on the shaft?

 

[imcjoek]

It is a calendering roll. My guess would be high-cycle, since it lasted for about 8 years. No can do on the pictures, company policy.

Currently I have increased the fillet from R0.063 (on a 2.0-2.5in diameter step!!) to R0.250 for a reduction of Kt from 2.3 to ~1.4.

Definite improvement, that. Still wondering on the notch sensitivity value thing, largely out of my own curiosity.

 

[TVP]

Have the fillet rolled, not just machined (turned, etc.). Ecoroll and Cogsdill make roller burnishing tools that can roll fillets and induce residual compressive stresses, thereby increasing the fatigue strength. Commonly used on crankshafts and underhead fillets for high strength aerospace fasteners.

Values for q will be hard to find, since it is not really a material property. Have you use Google Scholar to search scientific jornals and other technical references?

 

[imcjoek]

We seem to have this handled. At least for now anyways. The powers that be seem to be content with the fillet size change.

If values of q are difficult to find or unreliable, why do all machinery design books reference it in their chapters on fatigue life? What is the alternative? Surely (don't call me Surely), aerospace must have this fatigue business pretty well figured out. What do they use for references? What I'm asking, I guess, is what is a good method of sizing/materials selection of cyclic stressed components other than "hey, that's how we always dun it. And it only fell apart once!".

Thank you everyone for your responses!

 

[redpicker]

I would expect the value of "q" to be inversly proportional to the CVN impact strength. See if you can test the failed shaft to determine the impact strength and make sure your 4340 QT has significantly higher impact strength. This should inusre the 4340QT has a lower value of "q".

Also, if you can look at the fracture surface, estimate how much results from fatigue (for rotational bending fatigue, the fatigue portion will be flat and smooth while the final fracture will rough, either ductile or brittle). IMO, if the fatigue portion covers more than 25% of the fracture surface, increasing the toughness will make little improvement in the fatigue life. If, on the other hand, the fatigue portion is less than 5% of the fracture area, then increasing the toughness may help the fatigue life.

rp

 

[CoryPad]

Notch sensitivity q for different materials and treatments is documented in Stress Concentration Factors by Peterson, and also the newer revision entitled Peterson's Stress Concentration Factors by Pilkey.

 

[imcjoek]

IMO, if the fatigue portion covers more than 25% of the fracture surface, increasing the toughness will make little improvement in the fatigue life. If, on the other hand, the fatigue portion is less than 5% of the fracture area, then increasing the toughness may help the fatigue life.

The fatigue bit looks like about 85%+ of the surface area. On a 2" diameter shaft, there was only a 3/4" or so nub left for the final failure.

If I follow your line of reasoning:
The nominal stress was low enough that the 3/4" area could hang on by itself. So the failure was induced by some crack, the fillet, or corrosion causing fatigue cracks to begin. A higher UTS material cannot prevent this and the problem must be fixed geometrically and through maintenance.
True?

Notch sensitivity q for different materials and treatments is documented in Stress Concentration Factors by Peterson, and also the newer revision entitled Peterson's Stress Concentration Factors by Pilkey.

I was looking for something like this. I will check these out.

Thank you, everyone.

 

[redpicker]

If I follow your line of reasoning:
The nominal stress was low enough that the 3/4" area could hang on by itself. So the failure was induced by some crack, the fillet, or corrosion causing fatigue cracks to begin. A higher UTS material cannot prevent this and the problem must be fixed geometrically and through maintenance.
True?

Not quite. The reasoning is that if a extremely deep crack does not result in a brittle faliure, then the toughness of the material is adequate for the application.

Increasing the material strength (UTS) _might_ improve the fatigue resistance, but if crack initiation is due to other factors (corrosion fatigue, for example) increasing the UTS may have little effect.

From what you have posted here, I'd agree that geometry change, cold rolling, and improved periodic inspections (maintenance) will have a larger effect than a change in material.

rp

 

[metman]

imcjoek,

In lieu of q values

Values for q will be hard to find

if not found in Peterson as suggested by CoryPad or CVN testing, another approach is to compare area under Stress/Strain curve of Normalized 4140 vs Q&T 4340. Area under Stress/Strain curve is a good indication of toughness.

Also you should compare the transition temperature of the two different alloys having different heat treatments.

 

[metman]

I meant to say..."or in lieu of CVN testing..."