A failed gearbox arrives as a problem to be solved, and the first instinct is to open it and look. That instinct destroys more gear evidence than the failure itself. The relative position of the gears, the distribution of debris in the sump, the band the mesh left on the flanks and the condition of the oil are all recoverable on the day and largely unrecoverable a week later. Almost none of it is lost deliberately. It is lost to ordinary, well-intentioned troubleshooting, carried out by people whose job is to get the machine running again.
The assembly is part of the evidence
A gear set does not fail in isolation from the housing that locates it. Shaft position, bearing seating, backlash measured at several angular positions and the relationship of each gear to its mate exist only while the box is together. Once the shafts are pulled, those measurements are gone and can afterwards only be argued about from drawings.
Where misalignment or housing distortion is a candidate explanation — and it usually is — the pre-teardown measurements are what allow that candidate to be tested rather than merely asserted.
Document before anything moves
Photographs of the box as found, covers off but nothing disturbed, cost an hour. They should record every flank from more than one angle and under raking light, which reveals surface texture that flat lighting washes out. Scale references and a tooth index matter, because a report referring to the damaged tooth on a set of forty is not usable later.
Marking the gears with a permanent index before removal preserves the mesh relationship — which tooth ran against which — that debris patterns and progressive damage sequences depend on.
The oil is a sample, not waste
Draining into a waste drum discards the most information-dense item in the failure. The oil carries viscosity and additive state, water and contaminant loading, and a wear-debris population whose particle morphology distinguishes fatigue spall from abrasive cutting.
A sample drawn before the drain, from a circulating system while it can still circulate, is more representative than one scraped off the bottom afterwards. Where the box has already been drained, the drum, the filter element and the magnetic plug all still hold usable material and should not be cleaned or discarded.
Debris location, not just debris
Where the debris sits is evidence in its own right. Material packed against one side of the case, concentrated at one bearing, or absent from the sump entirely tells a different story from an even loading of fine wear particles. Sweeping everything into one container answers the composition question and destroys the distribution question at the same time.
Cleaning is the most common single loss
Solvent scrubbing, wire brushing and abrasive blasting are standard practice for making a part presentable, and each removes the fine surface detail that identifies the mechanism. Micropitting texture, transferred material from scuffing, oxide and temper colours around a hot contact, and the fine markings on a fracture surface are all lost this way.
Where cleaning is needed to see anything at all, the order is: document, then clean by the least aggressive method that works, then document again. Fracture surfaces in particular should not be cleaned by anyone other than the examiner.
Do not fit the broken tooth back on
The instinct to hold a liberated tooth against its stump to see how it fits is understandable, and it damages both mating faces at precisely the location carrying the origin information. Matching fracture surfaces should be kept from contact with each other, packaged separately and handled at the edges.
Corrosion is the other clock running. A fracture surface left in a humid shop for weeks loses detail that was readable on day one. Light oil or a desiccant pack is a small precaution against a permanent loss.
Keep the undamaged parts too
There is a strong pull to scrap everything that is not obviously broken. Undamaged teeth from the same set establish the baseline geometry, case depth and surface finish that the failed teeth are measured against, and the mating gear carries the other half of the contact record. Discarding either reduces the analysis to a description of damage with nothing to compare it to.
Records have retention clocks
Vibration monitoring history, oil-analysis trend reports, drive torque and speed logs, alarm and trip records, maintenance work orders, and the original gear drawing with its rating calculation all exist somewhere at the moment of failure. Several of them overwrite themselves on a rolling schedule measured in weeks.
Requesting them early is cheap. Reconstructing a duty cycle afterwards from recollection is not, and it is the part of the analysis most exposed on challenge.
Sequencing restoration against preservation
The pressure to get a line running again is real and legitimate. The distinction worth drawing is between restoring production and destroying evidence, and the two are usually separable: fit the spare drive, crate the failed one intact, take the oil sample, and leave the teardown to a documented examination.
Where a box must be opened on site, the marginal cost of an hour of photography, an oil sample and an index mark is trivial against the cost of an analysis that can only report what somebody remembers seeing.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.