FMEA: failure mode and effects analysis with example
FMEA stands for failure mode and effects analysis. For each component or process step, you look at what could go wrong (the failure mode), what the consequence is and what causes it. Then you score severity, occurrence and detection, and you tackle the highest risks before they reach the customer.
The general international standard for the method is IEC 60812:2018 . In the automotive industry, the AIAG & VDA FMEA handbook is the standard. In certification, you mainly meet FMEA in IATF 16949, in risk management under ISO 14971 for medical devices, and as a tool for risk management in ISO 9001.
When do you use an FMEA in your management system?
An FMEA is preventive: you look for faults before they happen. That makes it different from a root cause analysis, which looks back at the cause of a fault that has already occurred.
| Standard or sector | Role of FMEA |
|---|---|
| IATF 16949 (automotive) | Standard part of product and process design. Customers usually expect the AIAG & VDA format |
| ISO 13485 / ISO 14971 (medical devices) | Commonly used technique to analyse hazards and failure modes of a device |
| ISO 9001 | Not required, but a good way to apply risk-based thinking to critical processes |
| ISO 22000 / HACCP | Sometimes used alongside the hazard analysis, but the HACCP decision tree remains leading |
For an office or a small service company, an FMEA is usually too much. A simple risk matrix is enough for them.
Types of FMEA: DFMEA, PFMEA and FMEA-MSR
The AIAG & VDA handbook distinguishes three variants. Which one you need depends on what you make.
| Type | Focus | Example question |
|---|---|---|
| Design FMEA (DFMEA) | The product or component | Can this seal leak at high temperature? |
| Process FMEA (PFMEA) | The production or assembly steps | Can an operator fit the wrong bolt? |
| FMEA-MSR | Monitoring and system response during use | Does the system detect a sensor fault and react safely? |
Severity, occurrence and detection: how to score an FMEA
For each failure mode, you give three scores, usually on a scale from 1 to 10. The handbook and IEC 60812 give example scales. Many companies make their own version that fits their product.
| Score | Question | 1 means | 10 means |
|---|---|---|---|
| Severity (S) | How serious is the consequence? | Customer notices nothing | Safety risk or breach of law |
| Occurrence (O) | How often does the cause occur? | Practically impossible | Almost always happens |
| Detection (D) | Do you find the fault before it reaches the customer? | Certain to be found | Certain not to be found |
Note the reverse logic for detection: a high score means poor detection.
RPN or action priority?
For a long time, you multiplied the three scores into a risk priority number: RPN = S × O × D, with a maximum of 1000. The problem: a severity of 9 with a low occurrence can give the same RPN as a harmless fault that happens often.
The AIAG & VDA handbook from 2019 therefore replaces the RPN with action priority (AP). You look up the combination of S, O and D in a table, and that gives high, medium or low. Severity carries the most weight in that table. If you work for the automotive industry, use the AP tables in the handbook. Outside automotive, the RPN is still common, as long as you monitor severity separately.
Never rely on an RPN threshold alone, such as “act above 100”. A failure mode with severity 9 or 10 always needs attention, even if its RPN is low.
Example: PFMEA for filling cleaning product
A manufacturer of cleaning products analyses the step “fill bottle and fit cap”. This is a simplified PFMEA with RPN. You can use the columns as a template.
| Process step | Failure mode | Consequence | Cause | Current control | S | O | D | RPN | Action | S | O | D | RPN after | |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---| | Filling | Too little content | Complaint, legal shortfall | Filling nozzle blocked | Sample weighing once per hour | 6 | 4 | 6 | 144 | Inline scale with rejection | 6 | 4 | 2 | 48 | | Capping | Cap not tight | Leakage, skin contact | Torque too low | Visual check by operator | 8 | 3 | 7 | 168 | Torque monitoring on capping machine | 8 | 3 | 2 | 48 | | Labelling | Wrong label | Hazard symbol missing | Wrong label roll loaded | Check at order changeover | 9 | 2 | 5 | 90 | Barcode scan of label against order | 9 | 2 | 1 | 18 |
Look at the third row. The RPN of 90 is the lowest, but the severity is 9: a missing hazard symbol can harm someone. That is why this failure mode still gets an action. Severity almost never changes through a measure. You mainly lower occurrence and detection.
How to carry out an FMEA step by step
The AIAG & VDA handbook describes seven steps. Outside automotive, this order is also useful.
Planning and preparation
Set the scope, the team and the schedule. An FMEA is done with people from design, production and quality.
Structure analysis
Break the product or process down into components or steps.
Function analysis
Describe what each component or step must do.
Failure analysis
For each function, find the failure modes, consequences and causes.
Risk analysis
Record the current controls and score severity, occurrence and detection.
Optimisation
Choose actions for the highest priorities, with an owner and a date. Score again after implementation.
Documenting the results
Record the outcome and communicate the main risks and actions.
What does an auditor look for in an FMEA?
An auditor in the automotive or medical sector mainly checks whether the FMEA is alive. Typical checks:
- Has the FMEA been updated after a complaint or a process change?
- Do the controls from the FMEA appear in the control plan and work instructions?
- Have high-priority actions actually been carried out and scored again?
- Did a team make the FMEA, or one person?
A complaint about a failure mode that is not in your FMEA is a signal to extend the FMEA. Record that link in your CAPA process or your corrective actions.
Common mistakes with FMEA
- Copying an old FMEA without looking at the new situation.
- Scores that nobody can justify. Record which scale you use.
- Too much detail. Hundreds of lines for a simple process. Start with the critical steps.
- Never updating it. A design-phase FMEA that has not been changed after five years of production.
Further reading
- Creating a risk matrix: the lighter variant for most ISO standards
- Ishikawa diagram: organising causes when something does go wrong
- ISO 14971 and ISO 13485: risk management for medical devices
- IATF 16949 : the quality standard for the automotive industry