Bearings are the most replaced component in rotating equipment and the most common single cause of unplanned downtime. They are also one of the best-understood failure mechanisms in maintenance — the failure modes are catalogued, the root causes are known, and the detection technologies (vibration, oil analysis, thermography) are mature. A plant that understands bearing failure has solved most of its rotating-equipment reliability problem.
Here is the field guide: what kills bearings, how each failure looks in the data, and how to read the bearing back to its root cause.
The Failure Modes
Bearings fail in a handful of well-documented ways, and almost every failure traces back to one of four root causes: lubrication, contamination, overload, or misalignment and mounting error. The bearing's surface tells you which.
Fatigue spalling (surface fatigue). The classic end-of-life mode. Material flakes off the raceway in the loaded zone, producing the familiar pitted surface. This is the failure you want — it means the bearing reached its design life under correct conditions. If your bearings fail this way at the expected interval, your program is healthy. The problem is when they fail this way at a quarter of the expected interval, which points at overload or poor lubrication shortening the fatigue life.
Lubrication failure. The most common premature failure mode. Insufficient grease, wrong grease, over-greasing (which overheats the bearing), or relubrication at the wrong interval. The surfaces look polished, glazed, or discoloured (brown or blue from heat), and the cage may be worn. The root cause is a lubrication practice error, and it is fixable with the right lubrication management program.
Contamination. Dirt, water, or abrasive debris ingress through a failed or missing seal, contaminating the lubricant. The raceways show dull grey, satiny wear with embedded particles, and premature spalling. Root cause is a sealing or cleanliness problem, often combined with poor lubrication practices (open grease fittings gathering grit).
False brinelling and true brinelling. True brinelling is denting of the raceway from impact or static overload — dropping a bearing, hammering it on, or shock loads in service — producing evenly spaced indentations at roller spacing. False brinelling happens when a non-rotating bearing vibrates in place (a standby machine near running equipment, or transport vibration), wearing small grooves at the roller positions. The fix for the latter is rotating stored and standby bearings periodically.
Electrical damage (fluting). Variable frequency drives can induce currents through the motor shaft that arc through the bearing, producing fluting (a washboard pattern) on the raceway and accelerated failure. The fix is shaft grounding or insulated bearings.
Misalignment and mounting error. The raceway wear is concentrated on one side or in a diagonal pattern rather than the centre of the load zone. The bearing was installed cocked, the shaft is misaligned, or the housing is distorted. This is an installation problem and it will recur on every bearing in that position until the alignment is corrected. (See laser alignment.)
Cage failure. The cage (separator) cracks or wears. Often a symptom of poor lubrication, vibration, or overspeed rather than a primary mode — the cage is usually the weakest part and fails secondarily.
How Each Mode Looks in the Data
The detection technologies read different modes at different stages.
Vibration is the primary tool. Each bearing fault has a characteristic frequency — ball pass frequency inner race, ball pass frequency outer race, ball spin frequency, and fundamental train frequency — calculated from the bearing geometry and shaft speed. A spike at the outer-race frequency says outer race defect; at the inner-race frequency, inner race. Early-stage faults show as a high-frequency impact signal (kurtosis, crest factor, envelope demodulation) before they show as a discrete frequency. As the fault progresses, the discrete frequency appears, harmonics build, and eventually the signal becomes broadband noise as the bearing self-destructs. (See the vibration monitoring guide.)
Oil analysis catches the wear debris before the vibration signature is clear, especially for oil-lubricated bearings in gearboxes and large machines. A rising iron trend on a gearbox is the first hint of a bearing or gear problem; rising copper suggests the cage. (See oil analysis.)
Thermography catches the heat of friction late in the failure — a bearing running hot is already in trouble. It is a confirmation tool more than an early-warning one, but it is cheap and catches the catastrophic cases. (See thermography.)
Ultrasonic inspection catches the high-frequency impact of early-stage bearing faults before they are visible in the normal vibration spectrum. For plants running a vibration program, ultrasound is the early-warning complement.
Reading the Bearing Back to Root Cause
When a bearing is removed, inspect the races, rollers, and cage before throwing it away. The surface is a forensic record of how it died. A two-minute examination tells you whether you are dealing with lubrication, contamination, fatigue, mounting, or electrical damage, and that tells you whether the failure was random or will repeat on the replacement.
The mistake most teams make is replacing the bearing without reading it. A bearing that failed from misalignment and gets replaced without correcting the alignment will fail again on the same schedule — and the team will call it a bad batch or a bad bearing when the cause was sitting in the coupling the whole time. This is the fail-and-fix loop in miniature, and the way off it is to read every failed bearing, classify the mode, and fix the root cause before the replacement goes in. (This is the defect elimination discipline applied at the component level.)
L10 Life and Why It Matters
Bearings are rated by their L10 life — the life at which 10% of a population will have failed, under a specific load and speed. It is a statistical rating, not a guarantee for any individual bearing. The practical use is in selection: a bearing with an L10 of 100,000 hours for a given duty means 90% survive that long. If your bearings fail at 10,000 hours, they are running at a fraction of their rated life, and the cause is almost always one of the root causes above — lubrication, contamination, overload, or mounting — not the bearing itself. The L10 rating gives you the benchmark against which "premature" is measured.
How OpexMX Handles It
OpexMX stores the bearing data per asset (part number, L10 rating, calculated life for the actual duty), tracks each bearing's failure history and mode classification, and ties vibration, oil, and thermography alerts to the specific bearing fault frequency on that asset. A flagged bearing fault opens a predictive work order with the likely failure mode and root cause pre-filled from the asset's history, so the replacement is paired with the fix rather than just swapped. Failed bearings get a mode tag that rolls up into the bad actor analysis, so the plant can see — across all assets — whether lubrication, contamination, or mounting is the dominant bearing-killer to attack.