The instinct after a bearing failure is to get the machine running again. The bearing that stopped it is usually the first thing pulled, wiped down and set on a bench for someone to look at later. Every step in that sequence removes information. A failed bearing is not simply a broken part awaiting diagnosis; it is a physical record of the conditions it ran under, and most of that record sits in material that a rag, a solvent tank or a press removes in seconds and no later examination can restore.
What the assembly is actually recording
Four categories of evidence sit in a failed bearing, and they degrade at different rates. The wear pattern on the raceways records the load path and the alignment. The debris held in the grease records what was circulating and in what form. The surface morphology at the damage origin records the initiating mechanism. The dimensional state of rings, shaft seat and housing bore records the fits the bearing was mounted to.
Only the last is robust. The first three are consumed by ordinary handling, and they are the three that usually decide which mechanism initiated the failure.
As-found documentation precedes everything
Before anything is loosened, the assembly is photographed in place with scale and orientation established, and the angular position of each ring relative to the shaft and housing is marked. That mark is what later allows the loaded arc on the raceway to be related back to the direction of applied load, which is the basis for distinguishing normal loading from misalignment.
ASTM E1188 sets out the practice for collecting and preserving physical items by a technical investigator, and ASTM E860 addresses examining items that may become involved in litigation. Neither is bearing-specific; both are why the sequence is documented rather than improvised.
Cleaning is a one-way operation
Degreasing removes the wear debris that identifies the mechanism, and wire brushing removes the fine surface texture that distinguishes electrical erosion from mechanical wear. Fluting marks and arc-pit morphology are shallow features; abrasive cleaning flattens exactly the detail that separates them from ordinary wear.
Cleaning is a legitimate step in the examination, but it belongs after the as-found record is complete, after the lubricant has been sampled, and under the control of whoever will be asked to explain the result. It is a step in a protocol, not a preliminary tidy-up.
The bearing alone is an incomplete specimen
The shaft journal and the housing bore carry the fit evidence. A ring that turned on its seat leaves fretting and dimensional loss that will not be visible once the shaft is dressed or the housing bored out for a repair. Clearance and roundness measured against ISO 492 tolerance classes and ISO 5753 clearance groups only mean something if the mating surfaces still exist in their as-failed condition. Preserving those sections, along with the seals, the cage and every recovered roller, costs storage space; losing them costs the finding.
Sampling the lubricant properly
The lubricant sample is taken before disassembly, from the bearing itself where possible rather than from a sump or a drain point, and it is kept sealed and unmixed. Grease is scraped and bagged; oil is decanted without filtering. Filters and magnetic plugs are retained whole rather than cleaned.
Ferrographic characterisation under ASTM D7690 and ASTM D7684 depends on debris morphology surviving intact; a sample taken after flushing, or topped up from a fresh drum, answers a different question than the one being asked.
How the bearing comes off matters
Removal introduces damage. Force applied through the rolling elements rather than through the ring being moved produces indentations indistinguishable at a glance from service brinelling, and heat applied for removal alters the temper of a surface under examination.
Where removal must be destructive, cutting the ring away is often preferable to pressing through the raceway, and the method is recorded either way. An undocumented removal invites the argument that the damage relied upon was created after the failure.
Custody and joint examination
Where a supplier, an installer and an owner are all potentially exposed, examination is better conducted jointly, on notice, under an agreed protocol stating what will be sectioned, cleaned or otherwise consumed. Metallographic sectioning is destructive by definition and cannot be repeated on the same material. Agreeing the protocol in advance converts a destructive step from a spoliation argument into an ordinary examination.
The records that expire on their own
Condition-monitoring data is frequently the only account of how the bearing behaved before it failed, and it is held on systems with rolling retention windows. Vibration histories acquired under ISO 20816 methods, temperature trends, motor current signatures and process load data all constrain the failure timeline, and all of them are routinely overwritten within weeks.
Alongside them sit the paper records: the mounting procedure used, the relubrication actually performed, the part number fitted and where it was bought. All are recoverable early and far harder to assemble a year later.
Working with the pressure to restore
Nobody preserving a bearing is arguing that a plant should stay down. The distinction worth drawing is between restoring service and destroying evidence, and the two are separable with a few hours of deliberate sequencing.
Photograph before anything moves, bag the bearing as found, keep the lubricant, retain the shaft and housing sections, and pull the monitoring data the same week. The absence of any of these is itself a finding, and one that speaks to the standard of care as readily as to the mechanism.
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.