Root Cause Analysis of Bearing Failure

Root cause analysis of a bearing failure is a structured investigation that uses the failed bearing as physical evidence: you preserve the damage, classify the failure mode, trace that mode back to the process condition that caused it, and verify the fix by confirming the failure does not recur.

Most bearings are replaced without anyone asking why they failed. The old one goes in the scrap bin, a new one goes on the shaft, and the plant runs until the same bearing fails again on roughly the same schedule. That is not maintenance; it is a subscription. A bearing carries the evidence of its own death on its raceways, and reading that evidence is one of the highest-return investigations in a plant, because the answer usually names a cheap, permanent fix. This guide applies the general root cause analysis method specifically to bearings, where the evidence lives, how to preserve it, and how to get from a damage pattern to a process change. It is the investigation companion to bearing failure modes which catalogs what the damage patterns look like; here we use those patterns as clues in a case.

What is root cause analysis of a bearing failure?

It is the difference between knowing a bearing failed and knowing why. The failure mode, spalling, fluting, brinelling, corrosion, is the middle of the story, not the end. A root cause is a condition you can control, that the evidence supports, and whose removal stops the failure from recurring. "The bearing spalled" is a mode. "The lube route skips this drive so the film runs thin" is a root cause. RCA is the work of getting from the first statement to the second, and it applies the same three tests as any root cause: the cause must be within your control, verified by evidence rather than merely plausible, and usually found in the process rather than in a person.

Bearings are unusually good candidates for this because the evidence is physical and durable. Unlike a quality defect that vanishes into a finished pallet, a failed bearing sits on the bench with its diagnosis written on it. The job is not to guess; it is to read.

Why is the failed bearing your best evidence?

Because the damage mode and its location narrow the cause faster than any meeting can. A washboard fluting pattern is electrical, chase the shaft grounding and the drive, not the bearing brand. Evenly spaced dents at the rolling-element pitch mean a shock load or debris, not fatigue. A dull frosted band means the oil film thinned and metal touched metal. Rust means water got in. Each mode is a pointer to a class of causes, so classifying the mode correctly eliminates most of the wrong theories before you have said a word.

How do you preserve and read the evidence?

The single biggest mistake in bearing RCA is destroying the evidence before you read it. A technician's reflex is to wipe the bearing clean, and cleaning erases the frosting, discoloration, debris, and grease condition that carry the diagnosis. Treat a failed bearing like a scene, not scrap.

• Record it as removed. Note orientation and which way was up, mark the load zone, and photograph the raceways, rolling elements, cage, and seals before touching them.
• Keep the grease. The old lubricant tells you about contamination, water, and metal wear; sample it before it hits the rag.
• Find the primary mode. Most failed bearings carry more than one mode by the time they reach the bench. Damage cascades, a debris dent seeds a spall, a lost film drives heat until the ring cracks. The primary mode usually sits in the load zone and is more advanced; secondary damage is scattered and shallower. Chase the one that came first, because only it connects to a preventable cause.
• Note the context. Was this a VFD-driven motor? A wet washdown area? A bearing that sat in storage for a year? Context turns a mode into a cause.

How do you work from damage mode back to root cause?

Once the primary mode is fixed, the mode points to a class of causes, and a short 5 whys drives it to something you can change. Fluting: why is there current through the bearing? Because the VFD has no shaft grounding path. Why? Because grounding rings were never specified on this retrofit. That is a controllable, verifiable, process-level cause. For failures with several possible contributors, a bearing that runs hot on a machine with alignment, lubrication, and load questions all open, a fishbone diagram spreads the search across the 6M categories before you narrow, so you do not tunnel on the first plausible story. Verify with the is/is-not boundary: if only the drive-end bearings on VFD motors flute and the line-fed motors never do, the electrical cause explains the pattern and a lubrication theory does not.

A 6-step bearing failure RCA

1. Preserve the evidence first. Photograph and mark the bearing as removed, sample the grease, and note orientation and load zone before any cleaning. Once it is wiped, the diagnosis is gone.
2. Gather the context. Pull the asset history: run hours, drive type, last relube, alignment records, environment, and whether this bearing position has failed before. Repeat failures are the strongest signal that a real cause was never removed.
3. Classify the primary mode. Match the dominant, most-advanced damage in the load zone to one of the six ISO 15243 classes, and separate it from secondary cascade damage.
4. Trace the mode to a controllable cause. Run 5 whys down the single clear chain, or a fishbone when several contributors are open. Keep going past the first person-shaped answer to the process that allowed it.
5. Verify against the evidence. The cause must explain every fact, including the is/is-not boundary, which bearings failed and which identical ones did not. If a theory cannot explain the pattern, it is not the root cause.
6. Change the process and confirm non-recurrence. Implement the fix at the right layer, then keep the bearing's condition metrics on watch through a defined window. Log the mode and cause against the asset so plant-wide patterns emerge, using consistent failure coding.

Why does the same bearing keep failing?

A repeat bearing failure is a finished RCA that reached the wrong conclusion, almost always "the bearing was bad" or "operator error." Those endpoints feel like answers but change nothing, so the failure returns. When a bearing position fails two or three times, stop replacing and start investigating in earnest: the recurrence itself is proof the true cause is still present and still doing its work. This is where bearing RCA graduates into defect elimination, treating the repeat-offender bearing as a bad actor whose cause gets engineered out for good, not nursed. Consistent failure coding is what makes the pattern visible: six flutted bearings across a plant's VFD motors is a grounding design finding, not six unlucky parts, but only if the failures were recorded the same way each time.

How do you make the fix stick?

Non-recurrence is a claim about the future, so it needs a monitored window, not a closed work order. Keep the bearing's vibration and temperature trends on watch after the fix, with an explicit trigger to reopen the investigation if the defect signature reappears. Rank the countermeasure honestly: eliminating the cause (a grounding ring, a corrected lube spec, a heat-mount procedure) beats detection, which beats a reminder to "be careful." And feed the closed case back into your reliability data so the next person facing the same mode inherits your answer instead of starting over. That is the entire point of feeding failures into equipment reliability tracking and your maintenance KPIs: one good investigation should protect every similar bearing in the plant, not just the one on the bench.

What do the standards and numbers say?

• The six in-service damage classes you match against are defined in ISO 15243 Rolling bearings, Damage and failures, Terms, characteristics and causes (ISO 15243). It classifies damage that occurs while the bearing runs, which is exactly the evidence an RCA reads.
• Bearing-manufacturer failure analysis attributes the majority of premature failures to preventable causes, roughly a third to lubrication, with mounting, handling, and contamination behind most of the rest (SKF and bearing-industry failure-analysis training; independent work puts solid-particle contamination alone behind about 14–25% of premature failures). Read these as directional estimates: the through-line is that most bearing deaths have a controllable cause an RCA can find.
• Feeding causes back into reliability data is a discipline with its own standard: ISO 14224 Collection and exchange of reliability and maintenance data for equipment (ISO 14224), which defines consistent failure-mode taxonomies so patterns across assets become visible instead of getting lost in free-text notes.

A bearing RCA only compounds if its conclusions accumulate where the whole team can see them. Harmony pulls maintenance history, failure codes, downtime reasons, and machine signals into one operational data layer, so a recurring bearing mode surfaces as a plant-wide pattern instead of dying in a technician's notebook, and it can draft the corrective work order for a human to approve. It layers onto the CMMS and machines you already run, with no rip-and-replace; see how it works or the CLS case study. For the prevention side of the same problem, see rolling element bearing maintenance and predictive maintenance; motors have their own version of this investigation in motor failure root cause analysis.