The event
At 06:41 on a winter morning, a take-up bearing on a plant conveyor at a diamond concentrator in Lesotho seized in service. The grease in the overheated housing ignited. The fire was extinguished, the bearing was replaced, and the conveyor was back in production within hours. The site’s safety flash recorded the event cleanly — basic cause: “bearing seized on the take-up pulley”; action taken: “the fire was extinguished and the bearing was replaced.” On paper, the incident was closed.
What the record missed
The record stopped at the part, not the cause. “Bearing seized” is a failure mode, not a root cause — it describes what happened, not why. Nothing asked the next question: why did a bearing that had run for years suddenly lose its lubrication film, overheat and burn? A bearing does not seize for no reason, and grease does not ignite on its own. With the cause untouched, the same failure mode was free to return — on this bearing, or on any of its identical siblings across the plant.
The investigation that should have followed
Run as a structured Root Cause Failure Analysis, the event opens into a clear chain. The honest output is not a single confident answer but a ranked set of causes and a short list to verify on the asset:
| # | The question | What the evidence points to |
|---|---|---|
| 1 | Why the fire? | A seized bearing overheated; the churned and purged grease ignited. |
| 2 | Why did it seize? | Loss of the lubrication film → metal-to-metal contact → heat. |
| 3 | Why was lubrication lost? | Candidates: over-greasing (excess churns, heats, blows the seal); contamination at an exposed, low take-up; wrong or mixed grease; misalignment overload. |
| 4 | Why did that go unchecked? | No lubrication standard — no grease spec, measured quantity or set interval — and no temperature or vibration monitoring on the bearing. |
| 5 | Why no standard or monitoring? | The take-up bearing had never been ranked in criticality or FMEA, so it defaulted to run-to-failure. |
The three roots
A complete RCFA drives to three levels of root. The original investigation reached none of them:
- Physical (most probable). A lubrication-management failure: loss of the lubrication film through over-greasing and/or contamination ingress. Never formally confirmed on the asset at the time — which is itself part of the lesson.
- Human. Greasing carried out without a defined specification, quantity or interval.
- Latent / system. Critical conveyor bearings were never ranked, so no maintenance tactic, lubrication standard or condition monitoring was ever assigned. This is the root that allowed the failure — and guarantees its return until it is corrected.
The fix — a strategy, not a spare
The corrective action is not a better bearing; it is a maintenance strategy for the asset:
- Rank it. Add the take-up bearing to the criticality / FMEA register and assign a deliberate maintenance tactic instead of run-to-failure. (Pillar II — Asset Criticality)
- Standardise the lubrication. Create a lubrication PM task with the correct grease specification, a measured quantity and a set interval; fit a single-point automatic luber to remove human over-greasing variability. (Pillars III & IV)
- Watch it. Add the bearing to a condition-monitoring route — temperature and vibration, or a thermal trip — turning an invisible asset into a measured one. (Pillar VI)
- Protect it. Verify the seal arrangement and contamination guarding at the exposed take-up. (Pillar III)
- Spread it. Apply the same standard to every take-up and idler bearing on site — one investigation, fleet-wide elimination. (Pillar VIII)
Where it fits in the framework
This case lives at the heart of the Defect Elimination pillar and reaches across the framework: criticality (II) decides the asset deserves a tactic; maintenance tactics (III) and work management (IV) build and execute the lubrication standard; performance management (VI) keeps it under watch; and integration & sustainment (VIII) spreads the fix across the fleet. Anyone can replace a part. The discipline — and the difference between a plant that firefights and one that runs — is replacing the reason it failed.
From the real file: this case study is built on a genuine safety flash from a Southern African diamond operation. The original record closed at “bearing replaced.” Everything after that is the analysis the event deserved.