Executive summary
- Use FMEA at design, at process change, at line relocation — not after the defect.
- RPN = severity × occurrence × detection.
- A high severity score must be addressed even when the total RPN is low.
- The usable output is a list of countermeasures with a name and a date on each.
- An FMEA not revisited after a change is an expired document.
FMEA — Failure Mode and Effects Analysis — is a structured way of asking what can fail in a process or a product, what happens when it does, and whether we would notice in time. All of it before the failure actually occurs.
Prevention rather than cure. Done after the event, it is just an incident report with a nicer table.
When to use FMEA
- When designing a new process, product or service.
- When changing an existing process or product, even slightly.
- When taking an existing process into a new environment — relocating a line, opening a new plant.
- When setting the control plan and maintenance plan for new equipment.
- When analysing an existing process that has never been through an FMEA.
Why use it
- It turns risk scattered across several people’s heads into one list that can be ranked.
- It exposes risks that are not severe but are entirely undetectable — usually the more dangerous kind.
- It gives the control plan and the maintenance plan a documented rationale.
- It is organisational memory that survives the person who knew.
- It lowers correction cost: fixing it on paper is always cheaper than fixing it on the floor.
The ten steps of a process FMEA
- Scope it. State where the process starts and ends, and who must be in the room.
- Brainstorm every possible failure mode, without judging likelihood yet.
- State the effects of each failure — on the end user, on the next process, on safety.
- Score severity, based on the worst effect that failure can produce.
- Identify the causes of each failure. One failure mode may have several.
- Score occurrence against real data wherever it exists, not against impressions.
- List the current controls, separating preventive controls from detection controls.
- Score detection. A high score means hard to detect.
- Calculate the RPN as the product of the three scores, and rank.
- Assign countermeasures with an owner and a due date, then rescore once they are done.
The RPN is not the final word
A failure scoring 9 or 10 on severity must be addressed regardless of a low total RPN, because that level of severity means safety or regulatory consequences. Probability does not excuse it.
Linking FMEA to continuous improvement
An FMEA done once and filed expires within a year: the process changes, the material changes, the people change. Linking it to the improvement cycle means three things.
| Link point | What it requires |
|---|---|
| After a real failure | Check whether it appears in the FMEA. If not, add it. If it is there but underscored, correct the score. |
| After a process change | Revisit only the affected rows. A full rewrite is not required. |
| When reviewing the control plan | Every control point should trace to a row in the FMEA. If it cannot, it is being controlled for no stated reason. |
| When reviewing the maintenance plan | A PM task that relates to no failure mode at all is the first candidate for deletion. |
The most common mistakes
- Doing it alone. An engineer filling in the table by himself will miss the failures the operator sees daily.
- Scoring without written criteria, so one line’s scores cannot be compared with another’s.
- Stopping at the RPN. A table with no owner column and no date column will never produce a change.
- Confusing failure mode with cause. Bearing seizure is the failure mode; loss of lubrication is the cause.
- Never rescoring after the countermeasure, so nobody ever learns whether it worked.