Skip to content
TimeQeeper

FMEA analysis: how to do an FMEA, with RPN, AP and a template

Reliability11 min readUpdated 15 November 2026By TimeQeeper

An FMEA (failure mode and effects analysis) is a team review that records, for each component or process step, how it can fail, what happens when it does and why. Each failure mode gets three ratings from 1 to 10: severity (S), occurrence (O) and detection (D). Take an idler on a belt conveyor that seizes: the belt rubs across it, with a risk of fire (severity 9). That happens a few times a year (occurrence 6), and the monthly inspection round only sometimes catches it in time (detection 6). Multiplied together, that gives a Risk Priority Number (RPN) of 9 × 6 × 6 = 324.

What is an FMEA?

An FMEA looks ahead: you search for failure modes before they happen and choose measures that help reduce the likelihood of failure or limit its consequences. Every row is a chain with three links:

  • Cause: why it goes wrong, such as a bearing worn by dust and moisture.
  • Failure mode: how a component stops performing its function, such as an idler that seizes.
  • Effect: what the plant or the people around it notice, such as a belt rubbing across the roller.

Its origin is usually traced to MIL-P-1629, a US military procedure from 1949. Today you will mostly come across two documents: the general standard IEC 60812, in its third edition since 2018, and the FMEA Handbook published in June 2019 by AIAG and VDA, the North American and German automotive industry associations.

What types of FMEA are there?

The method is always the same; the subject changes.

TypeWhat you analyseWhenExample failure mode
Design FMEA (DFMEA)A product or part as drawnDuring designA shaft breaks under load
Process FMEA (PFMEA)The steps of a manufacturing or assembly processBefore production starts and when it changesA bolt is not tightened enough
Asset FMEA, for maintenanceAn installation in operationFor critical or new assets and after failuresAn idler seizes

Design and process FMEAs were developed mainly in the automotive industry, where suppliers are usually bound to them by contract. For maintenance, the asset FMEA is the most useful: you analyse an installation that is running and has a failure history, and the measures are mostly maintenance tasks. In RCM (Reliability-Centered Maintenance), this analysis is the basis for choosing the task for each failure mode.

How do you carry out an FMEA? The seven steps

The AIAG & VDA handbook divides an FMEA into seven steps: steps 1 to 3 form the system analysis, steps 4 to 6 the failure analysis and risk mitigation, and step 7 the risk communication. For a belt conveyor:

  1. Planning and preparation: what you analyse, with whom, by when and on which scales. Start with the most critical installations, which an asset criticality analysis identifies.
  2. Structure analysis: break the installation down into drive, belt, pulleys, idlers and safety devices.
  3. Function analysis: what each component has to do, ideally with a figure, such as "carry the belt at 30 tonnes of waste an hour". Include protective functions too, such as the pull-cord emergency stop.
  4. Failure analysis: the failure modes for each function, with the effect at the level above and the cause at the level below.
  5. Risk analysis: record the current prevention and detection, give the ratings and set the priority.
  6. Optimisation: measures with an owner and a date; once they are in place, check that they work and rate again.
  7. Results documentation: summarise what was found, what has been done and what is still open. This step and step 1 are new in the 2019 handbook; the earlier VDA manual only had the five steps in between.

The classic approach of the earlier AIAG manuals (the last edition dates from 2008) centres on the worksheet, which you fill in from left to right. If you set out the structure and the functions first, you overlook fewer failure modes.

How do you rate severity, occurrence and detection?

The AIAG & VDA handbook has rating tables written for the automotive industry; for an asset FMEA you usually draw up your own scale from 1 to 10, such as this one:

  • Severity (S), for the most serious effect: 1 = no noticeable effect; 2–3 = a minor nuisance; 4–6 = reduced capacity, extra work or a stop of less than a shift; 7–8 = standing still for a shift or longer; 9–10 = danger to people or the environment, or a breach of the law or a permit, with 10 if someone could be injured straight away.
  • Occurrence (O), how often the cause occurs given current prevention: 1 = practically ruled out; 2–3 = once every few years; 4–5 = about once a year; 6–7 = a few times a year; 8–10 = monthly or more often.
  • Detection (D), how well the current controls spot it in time, so 1 is good and 10 is bad: 1 = an automatic protection intervenes; 2–4 = a fixed test, measurement or inspection spots it in time; 5–6 = an inspection round sometimes spots it in time; 7–10 = you only notice when it goes wrong.

Agree the scale beforehand and note the assumption behind every rating. Base occurrence on data where you can: with consistently recorded failure codes, you can count how often a failure mode actually happened.

How do you calculate the RPN, and what are its drawbacks?

The Risk Priority Number (RPN) is the product of the three ratings:

The RPN runs from 1 to 1000; the higher it is, the sooner the failure mode should be dealt with. Only compare RPNs within the same analysis and scale. The number does have well-known weaknesses:

  • Same number, different risk. An RPN of 120 can be reached in 24 ways: 10 × 4 × 3 can injure someone, 5 × 8 × 3 often causes extra work and 2 × 6 × 10 is a minor nuisance that nobody spots in time.
  • Multiplying ranks. A detection rating of 6 is not twice as bad as a 3, yet the RPN doubles.
  • Gaps in the range. Only 120 of the numbers up to 1000 can come out as an RPN, and just 17 of those lie above 500.
  • Thresholds. With a limit such as RPN 100, 501 of the 1000 combinations land above it, while a failure mode with severity 10 and an RPN of 80 falls below. And nudging a rating down by one point can take a failure mode below the limit without anything actually changing.

Action Priority: the successor to the RPN

The AIAG & VDA handbook replaces the RPN with Action Priority (AP). You look up the combination of S, O and D in a table that gives the most weight to severity, then to occurrence and only then to detection. There are three outcomes:

  • H (high): identify a measure that improves prevention or detection, or justify and document why the current controls are adequate.
  • M (medium): identify a measure; justifying why none is needed is allowed at the company's discretion.
  • L (low): a measure is optional.

So AP ranks not the risk but the need to act. The handbook recommends that failure modes with severity 9 or 10 and an AP of high or medium are, at a minimum, reviewed by management. Two rules from the table, using the bands of the scale above:

  • Severity 9–10 with occurrence 4 or higher: high, except occurrence 4–5 with detection 1 (medium).
  • Severity 9–10 with occurrence 2–3: high for detection 7–10, medium for 5–6, low for 4 or better.

That gives the three ways of reaching 120 a different AP each: 10 × 4 × 3 high, 5 × 8 × 3 medium and 2 × 6 × 10 low. Outside the automotive industry the RPN is still common; since 2018, IEC 60812 also describes other ways of calculating the RPN and a method based on a criticality matrix.

FMEA template: the columns of the worksheet

A worksheet has one row for each combination of failure mode and cause, because occurrence and detection differ from cause to cause. These columns follow the classic layout and work in any spreadsheet; the AIAG & VDA form arranges them by steps 2 to 6, with AP instead of the RPN.

ColumnWhat you enterExample: belt conveyor
ComponentFrom the structure analysisIdlers
FunctionWhat it has to do, with a figureCarry the belt at 30 tonnes of waste an hour
Failure modeHow the function failsIdler seizes
EffectWhat happens nextBelt rubs across the roller, which heats up: risk of fire
Severity (S)For the most serious effect9
CauseWhy it happensBearing worn by dust and moisture
Current preventionWhat works against the cause todayNone; a roller is replaced once it has failed
Occurrence (O)Given current prevention6
Current detectionHow you spot it todayMonthly round: look and listen
Detection (D)10 means hard to detect6
RPN and APS × O × D and the AP table324, AP high
ActionWhat you do, and which rating it targetsRollers with better-sealed bearings (O), a thermal imaging round every week (D)
Owner and dateWho, and by whenMaintenance planner, before the next shutdown
ResultThe new ratings9 × 3 × 3 = 81, AP low

The diagram below takes the same row apart.

One row of the conveyor's FMEA worksheet. Each rating belongs to a column: severity to the effect, occurrence to the cause, detection to the current check. 9 × 6 × 6 gives an RPN of 324 and AP high; after the action, 9 × 3 × 3 = 81, AP low.Idlers · function: carry the belt at 30 tonnes of waste an hourONE ROW OF THE WORKSHEETFAILURE MODEIdler seizesEFFECTBelt rubs across theroller, which heatsup: risk of fireSCAUSEBearing worn bydust and moistureOCURRENT DETECTIONMonthly round:look and listenDRPN966324Severity 9Danger to people:risk of fireOccurrence 6A few times a year;current prevention: noneDetection 6The round sometimesspots it in timeRPN= S × O × D966××=324HAction Priority: high. In the AP table, severity 9–10 with occurrence 6–7 always gives H, whatever the detection.ACTIONAFTER THE ACTIONRollers with better-sealed bearingsoccurrence 6 → 3A thermal imaging round every weekdetection 6 → 39 × 3 × 3 = 81LAction Priority: low
One row of the conveyor's FMEA worksheet. Each rating belongs to a column: severity to the effect, occurrence to the cause, detection to the current check. 9 × 6 × 6 gives an RPN of 324 and AP high; after the action, 9 × 3 × 3 = 81, AP low.

For a single installation a spreadsheet is enough; with more assets, software that records the analysis per asset helps keep scales and versions consistent.

Example: an FMEA of a belt conveyor

A belt conveyor carries waste from the bunker to the shredder at 30 tonnes an hour, with no second conveyor. The team looks at two functions: carrying the waste, and stopping the belt when someone pulls an emergency cord. Besides the idler from the template, it finds these failure modes, among others:

  • Belt runs off-centre: waste falls off the side almost every week, because material builds up on the pulleys. The operator notices quickly, and a misalignment switch stops the belt if it wanders too far.
  • Belt splice tears: the belt parts and a new splice takes one to two days; it happens once every few years. The splice is checked during the weekly stop.
  • Pull-cord switch fails to trip: in an accident, nobody can stop the belt. Dust or corrosion jams a switch about once a year; the switches are tested monthly.
Failure modeS × O × D = RPNAPActionAfter the action
Idler seizes9 × 6 × 6 = 324HRollers with better-sealed bearings, a thermal imaging round every week9 × 3 × 3 = 81, AP L
Belt runs off-centre5 × 8 × 3 = 120MAdjust the belt scrapers, clean the pulleys every week5 × 6 × 3 = 90, AP M
Pull-cord switch fails to trip10 × 4 × 2 = 80HSwitches with better sealing against dust10 × 2 × 2 = 40, AP L
Belt splice tears8 × 2 × 4 = 64LNone; the check during the weekly stop stays8 × 2 × 4 = 64, AP L

Two lessons:

  • A threshold would have missed the emergency stop. With a limit of 100, only the idlers and the belt tracking would have been tackled, while the pull-cord switch, at an RPN of 80, has the most serious effect. AP rates it high.
  • Looking harder is not enough. With the thermal imaging round alone, the idler's RPN falls to 9 × 6 × 3 = 162, but at severity 9 and occurrence 6 AP stays high, whatever the detection. Only once occurrence falls too does AP drop to low.

The belt tracking stays at medium; the team records why that is acceptable: the effect is cleaning work, not danger. The new ratings are an expectation, to be checked against the failure history. The thermal imaging round and the monthly test become tasks in the maintenance plan; a round like that is condition-based maintenance, driven by the state of the roller rather than by the calendar.

Common FMEA mistakes

  • Mixing up cause, failure mode and effect. Wear is a cause, lost production an effect. Test every row: "because the bearing is worn (cause), the idler seizes (failure mode) and the belt rubs across the roller (effect)".
  • Forgetting protective functions. Emergency stops and standby units often fail unnoticed, until you need them. Give them a function of their own, like the pull cord.
  • Only inspecting more. That improves detection, but the cause remains. With high severity and high occurrence, AP then usually stays high.
  • Not rating again. A measure without an owner, a date and a check afterwards remains an intention. If an unknown failure turns up, add the failure mode; a root cause analysis helps find the cause, for example with an Ishikawa diagram.

How FMEA differs from FMECA

FMECA, with the C for criticality, also ranks the failure modes by criticality. If you want a well-founded ranking per asset to base your maintenance on, read what FMECA is, with a worked example for a cooling water pump.

Frequently asked questions

What does FMEA mean?

FMEA stands for failure mode and effects analysis; in German it is Fehlermöglichkeits- und Einflussanalyse. It is a method for working out in advance how a product, process or installation can fail, what the consequences would be and which measures are needed first.

What is the difference between a DFMEA and a PFMEA?

A DFMEA (design FMEA) analyses the design: can the product fail as drawn? A PFMEA (process FMEA) analyses how it is made: can a process step produce a faulty product, even when the design is sound?

What is a good RPN?

There is no limit that applies to every analysis. The RPN depends on your scales and mainly gives you the order within a single FMEA. Also look separately at every failure mode with severity 9 or 10, even if its RPN is low.

Who should be on an FMEA team?

People who know the installation from different sides, such as a technician, an operator, the maintenance planner or a reliability engineer, and someone from safety. One person leads the session and guards the scale.

Ready for more control over your maintenance?

Book a tailored demo. We'll show how TimeQeeper fits your machines, way of working and team, and we'll get in touch quickly.

We use functional storage and cookieless statistics. With your consent, LinkedIn measures visits and demo requests and helps us show relevant ads. LinkedIn may link this data to your profile.

Functional storage and cookieless statistics
Always available
LinkedIn marketing
Disabled

Your choice lasts 180 days. Declining does not affect the website or your demo request. You can change your choice here at any time.

Privacy statementCookie statement