Originally Posted by
riff_raff
noooby,
Your point about a sector of the planet gear bearing inner race (which is fixed wrt radial loads) being subject to most fatigue cycles is correct. And normally one would expect this to be the most likely location for contact fatigue spalling to occur. But there can be other problems that would cause the planet gear to suffer the type of radial fracture propagating outward from a race surface spall like this example.
First, it is not possible to pre-load the type of spherical roller bearing used in this case. So there will normally be some amount of radial clearance present. If this radial clearance is excessive, the gear rim will be subject to greater reverse cyclic bending loads than it was designed for. Think of a rotating cylinder subject to opposing local inward radial loads as it gets squeezed between the ring and sun gear. The planet gear race surface will experience a tension/compression load cycle every 180 degrees of rotation, adding to the normal contact stresses. If the fatigue/fracture analysis of the planet gear did not consider a condition with excessive radial clearance in the bearing, this problem would not have been exposed. Normally, the analysis assumes the bearing is manufactured within certain tolerances.
A good clear description. If I may add a little:
There are further complexities that would make analysis of stresses in these gears a difficult 4D problem when considering fatigue. The squeezing described above is not equal but rather the 'cylinder' is wrapped around 3 quarters of the bearing. All the clearance bunches up behind the bearing (in advance of the bearing as it orbits the sun gear) which will happen once in 360 degrees. So you have effects once per rev and effects twice per rev of the planet.
The 2 races of rollers per gear are inclined and bear against the concave inner surface of the gear (the outer raceway of the bearing). As well as radial forces, these also impart axial tensile forces into the body of the gear around the loaded sector creating axial tensile stresses that vary with the thickness of the 'cylinder'. The opposite effect occurs on the inner race which is axially compressed.
As was revealed in the G-REDL accident report, the barrel shaped rollers only truly 'roll' at one position over their length, with some of the contact under loading conditions involving some sliding and frictional force components that had not originally been considered in the design and which can lead to a crack radiating outwards.
Whilst we can envisage these forces and stresses, actually calculating and combining them all accurately requires sophisticated software tools. Even then there have to be certain assumptions made regarding utilisation load spectra, load sharing, temperature effects etc. and the validity of assumed boundary conditions in any model. Then you have to consider material and manufacturing consistency that the whole model is based on. Factors that affect the various heat treatments have previously been discussed.