BOEING 737 7H4SOUTHWEST AIRLINES CO

The alternator control gearshaft (L3 gearshaft) failed due to high cycle fatigue intergranular corrosion cracking.  High cycle fatigue is normally vibration driven where fatigue striations are hard to resolve (count) and where the number of cycles far exceeds the flight cycles.  No indications of any abnormal vibration were noted during the event flight and a review of fan and core vibration diagnostic data from the last shop visit in 2017 (the AGB was overhauled at that time) showed all the vibration levels to be within acceptable limits.     A review of the manufacturing records showed that the failed L3 gearshaft was 1 of 32 initially produced from the same production batch.  Due to corrosion issues during the initial manufacturing process, all the gearshafts were subjected to additional manufacturing operations; only 8 were deemed serviceable and put into service, one of which was the event gearshaft.  To understand the impact of the additional manufacturing processing steps performed on the failure L3 gearshaft, a series of chemical etch and black oxide treatments were performed on a sample of the same material.  This sample was compared with the failed L3 gearshaft and revealed that the intergranular attack observed on the failed L3 gearshaft could not be replicated; the sample material showed no intergranular attack even after 5 chemical etchings.  CFMI concluded that the chemical etching alone could not account for the intergranular attack observed on the event L3 gearshaft; instead it is thought that the additional machining process stress combined with the chemical etching needed to remove the corrosion caused the increase in quantity and density of the microcracks along with the grain boundary consumption.  However, this still did not fully account for why the gearshaft failed in high cycle fatigue, which is normally a vibration driven failure mode.  Since the event L3 gearshaft was introduced into service, it had operated almost 20 years; accumulated 63,711 hours time since new and 37,433 cycles since new; and was subjected to in-service inspections at the piece part level on two occasions; once in 2007 and then again in 2017 with no anomalies reported.  Had the additional manufacturing processing been the solely cause of the failure, the L3 gearshaft most likely could not have operated as long as it had, and it would most likely not have failed in a vibratory mode.  Instead the additional manufacturing operations most likely affected the high cycle fatigue capability of the gearshaft, but it wasn't until a change in the vibratory stress that the part was exposed did it fail.  Unfortunately, the engine is not equipped with any vibration sensors on the AGB so detecting any changing vibratory response in the AGB was not possible.  The engine is however equipped with a No. 1 bearing sensor and fan frame compressor case vertical sensor.  Review of the flight data recorder data showed no aircraft vibration monitoring alerts prior to, or after the engine failure event.  Also, a review of fan and core vibration diagnostic data from the last shop visit in 2017, the AGB was overhauled at that time, to the event date showed all the vibration levels to be within acceptable limits.     The transfer shaft locking nut was found loose but was still engaged with the 47 tooth gearshaft and the transfer gearbox.  Review of the maintenance records showed locking nut was last torqued in accordance with the approved procedures during the engine's last shop visit.  The locking nut threads and mating threads on the L3 gearshaft were in good condition, the lock nut retaining ring was present, properly installed, and in good condition, and the transfer shaft was found fully engaged. CFMI's experience with locking nuts becoming untorqued during operation is that the transfer shaft disengages resulting in a subsequent in-flight shutdown; in this case the transfer shaft remained fully engaged.  Therefore, the loose transfer nut did not cause nor a contributing factor to the failure of the L3 gearshaft and was an artifact of the initial failure event.