Choose a maintenance strategy per failure mode, not per machine
A single asset fails in several ways, and each failure mode deserves its own strategy. Reliability-centred maintenance (RCM, evaluated against SAE JA1011) formalises this; the table below is the short version.
| Strategy | Use it when | Avoid it when |
|---|---|---|
| Run to failure | Failure is cheap, safe, and has no knock-on effects | Failure has safety, environmental, or large production impact |
| Time-based replacement | The failure mode wears out predictably (Weibull β clearly > 1) | Failures are random (β ≈ 1) — replacing early just adds cost and infant mortality |
| Condition monitoring | There is a detectable warning (vibration, temperature, oil debris) with enough lead time | The warning period is shorter than you can inspect and react |
| Redesign | No maintenance task can make the risk acceptable | A task exists that manages the risk at lower cost |
The Weibull shape parameter is the quickest way to tell which world you are in: fit it to the failure history of one failure mode before deciding on scheduled replacement.
Set inspection intervals from the P-F interval
The P-F interval is the time between the point a failure becomes detectable (P) and functional failure (F). An inspection only protects you if it lands inside that window with time left to act.
Inspection interval ≤ P-F interval ÷ 2If bearing damage is detectable by vibration about eight weeks before seizure, monthly readings give at least one — usually two — chances to catch it. Where the P-F interval is uncertain, or failure consequences are severe, inspect more often; where planning and parts take long, subtract that lead time first.
Pick the analysis method that fits the problem
| Method | Best for |
|---|---|
| 5 Whys | A single, fairly simple event where the causal chain is linear |
| Fishbone (Ishikawa) | Brainstorming possible causes across people, method, machine, material, measurement, environment |
| Pareto analysis | Deciding which failure modes or causes to attack first |
| FMEA | Anticipating failure modes in a design or process before they happen |
| Fault tree analysis | Combinations of events leading to one serious top event |
| Weibull analysis | Understanding the life pattern of a repeating failure mode |
A common failure of RCA programmes is stopping at the first physical cause (“the bearing failed”). Keep asking until you reach a cause you can change — lubrication practice, a design margin, a procedure — and record the action and its owner.
Collect failure data you can trust
Every metric on this site is only as good as the work-order history behind it. The essentials:
- Record the failure mode, not just the asset. “Pump 3 failed” cannot be analysed; “Pump 3 — mechanical seal leak” can.
- Separate the times. Waiting for parts, travel, and repair are different problems; lumping them into MTTR hides which one to fix.
- Exclude planned work from MTBF. Scheduled replacements and PMs are not failures.
- Use a fixed taxonomy. ISO 14224 gives a proven structure for equipment, failure modes, and causes.
- Get enough events. Estimates from two or three failures swing widely; treat them as provisional.
Report metrics that drive action
Pair a lagging result with the leading activity that moves it, and always show the trend rather than a single value.
| Lagging (outcome) | Leading (activity) |
|---|---|
| Availability, MTBF | PM compliance, condition-monitoring route completion |
| MTTR | Spares availability, first-time fix rate |
| OEE | Planned vs. reactive maintenance hours |
| Downtime cost | Backlog of approved reliability improvements |
Control charts are the honest way to show whether a change in a metric is a real shift or just noise.
Roll it out in stages
- Baseline. Pick the assets that cost the most downtime; calculate current MTBF, MTTR, availability, and downtime cost.
- Fix the data. Agree a failure-mode taxonomy and make it mandatory on work orders.
- Attack the top of the Pareto. Run RCA on the biggest few failure modes and assign strategy per mode.
- Prove it. Track the same metrics on control charts and show the before/after in money.
- Widen the scope. Extend to the next tier of assets once the loop works.