> no one knows
> something technical on rec.bikes.tech, and no one can even guess
For a mechanical part used in predictable and calculable conditions, a
designer can compute or estimate the percentage of cycles the part will
see each level of fatigue stress. For example, a machine shaft might
have 400 revolutions with low bending stress, then 200 revs at a higher
stress, etc. There's a technique called Miner's Rule which is used to
compute a single equivalent value of reversing stress. That level of
stress can be compared to S-n (fatigue) data for the alloy in question,
to estimate service life.
But it's all pretty complicated and relies heavily on assumptions for
what I'd call fudge factors - for example, assumptions of stress
concentration factors, size factors, material factors, etc. If stresses
are not reversing, but instead varying (a reversing component
superimposed on an average non-zero value) there's a method for
addressing that. It's all a pretty good project even for a device whose
loads are known.
For a device like a bike fork, loads are not really known, so little of
the above applies. I imagine that a large and competent manufacturer
would have prototypes built using some pretty crude assumptions, and
subject them to fatigue testing, with larger loads than would be
expected in any real-life situation. I imagine smaller manufacturers
might have designers who would say "Oh, that's probably good enough.
And look how light it is!" No way to tell where you are in that
spectrum - although most small builders used steel forks instead.
FWIW, our custom-built Reynolds 531 tandem had its steel fork blades
suddenly snap off. Turns out the builder (Jim Bradford, then of the
Atlanta, GA area) had used track gage fork blades, one third the wall
thickness of Reynold's tandem gage blades. Of course, I didn't know
that until I measured the broken blades.
- Frank Krygowski