Introduction
Failure Mode and Effects Analysis (FMEA) has a reputation problem. In many organizations it’s treated as a documentation exercise — something the quality team generates to satisfy customer or audit requirements, with designers providing input reluctantly during a compressed review cycle. The result is a FMEA that is technically compliant but functionally useless: it doesn’t change designs, it doesn’t prevent failures, and no one reads it after the program is approved.
Done correctly, FMEA is one of the most powerful design tools available. It structures a systematic search for failure modes before they occur in hardware, forces multidisciplinary discussion of risk, and produces prioritized action items that actually improve the design. This article explains the FMEA methodology at the level a design engineer needs to use it effectively — not just contribute to a completed form.
What FMEA Is (and Isn’t)
FMEA is a structured, team-based analysis that:
- Identifies the ways each component or function can fail (failure modes)
- Assesses the effect of each failure mode on the system and end user
- Identifies the potential causes of each failure mode
- Evaluates the current controls that prevent or detect the failure
- Calculates a Risk Priority Number (RPN) to prioritize action
- Drives design or process changes to reduce the highest-priority risks
FMEA is not a root cause analysis tool (that’s for post-failure investigation), not a replacement for testing, and not a guarantee that a product won’t fail. It is a systematic risk reduction tool that improves the probability of success before hardware is built.
Design FMEA vs. Process FMEA
Two types of FMEA are relevant to mechanical engineers:
- Design FMEA (DFMEA): Analyzes failure modes inherent to the design — features, tolerances, material choices, interface definitions. Owned by the design team. Conducted before design is frozen.
- Process FMEA (PFMEA): Analyzes failure modes introduced by the manufacturing process — machining errors, assembly errors, inspection misses. Owned by manufacturing or process engineering, but design input is required where design choices affect process risk.
This article focuses on DFMEA, which is the primary FMEA responsibility of the design engineer.
The FMEA Worksheet: What Each Column Means
A standard DFMEA worksheet has the following key columns:
Item / Function
Identify the component and what it is supposed to do. Be specific — “provide axial preload to bearing assembly, holding inner ring against shoulder during thermal cycling” is more useful than “bearing housing.” Vague function statements produce vague failure mode identification.
Failure Mode
The way in which the item could fail to perform its function. For a bearing housing: fatigue crack at stress concentration; corrosion-induced dimensional change; insufficient surface hardness causing fretting; interference fit slip under temperature. For each function, there are typically 2–6 plausible failure modes.
Effect of Failure
What happens to the system and the end user when this failure mode occurs. Effects cascade: a bearing housing crack (local effect) leads to bearing misalignment (next level effect) which leads to premature bearing failure (system effect) which leads to machine downtime and potential injury (end user effect). Document all levels.
Severity (S)
A numerical rating (typically 1–10) of how serious the end user effect is if the failure occurs. The scale is standardized in the AIAG/VDA FMEA handbook (the current industry reference). Key thresholds:
- S = 9–10: Safety-critical failure — injury or regulatory non-compliance
- S = 7–8: Major product dysfunction — loss of primary function
- S = 5–6: Degraded performance — reduced function, customer dissatisfaction
- S = 1–4: Minor effects — cosmetic, minor inconvenience
Severity is a property of the effect, not the design. You can change occurrence and detection; you cannot change severity except by redesigning to a less severe failure mode or adding a fail-safe.
Cause
The mechanism by which the failure mode occurs. Good cause statements are specific and technical: “stress at re-entrant corner exceeds material endurance limit due to combination of design load and stress concentration factor of 2.3” rather than “overstress.” Specific causes lead to specific corrective actions; vague causes lead to vague corrective actions.
Occurrence (O)
A numerical rating (1–10) of how likely the cause is to occur during the product’s intended life. Ratings are tied to estimated failure rates or probabilities from historical data, analysis, or engineering judgment. O is reduced by design changes that eliminate or reduce the cause.
Current Controls — Prevention and Detection
Prevention controls reduce occurrence; detection controls find the failure (or its indicator) before it reaches the customer. Both types are important. A failure with no prevention or detection controls is a significant risk regardless of the O rating.
Detection (D)
A numerical rating (1–10) of how well current controls detect the failure mode or its cause. D = 1 means virtually certain detection; D = 10 means no detection capability. Detection ratings are frequently underestimated — engineers tend to credit controls with better detection capability than they actually provide in practice.
Risk Priority Number (RPN)
RPN = S × O × D. The product ranges from 1 to 1000. High RPNs — and particularly high Severity scores regardless of RPN — should drive corrective actions. The limitation of RPN as the sole prioritization criterion: a failure with S=10, O=1, D=10 has the same RPN as S=2, O=5, D=10, but the first is far more dangerous. Always give priority attention to high-Severity failure modes regardless of their RPN.
FMEA Rating Scale Reference
| Rating | Severity | Occurrence | Detection |
|---|---|---|---|
| 10 | Hazardous without warning | Failure almost certain | No detection control exists |
| 8–9 | Loss of primary function | Repeated failures | Very low detection probability |
| 5–7 | Degraded performance | Occasional failures | Moderate detection probability |
| 2–4 | Minor effect on function | Relatively few failures | High detection probability |
| 1 | No effect | Failure unlikely | Detection almost certain |
Making FMEA Actionable: The Part That Usually Gets Skipped
The output of an FMEA is not a completed form — it is a list of actions that reduce risk. For every failure mode with an unacceptable RPN (or an unacceptable Severity regardless of RPN), there should be a specific action item with an owner and a due date.
Actions take three forms:
- Design changes: Modify geometry, material, tolerance, or configuration to reduce occurrence or severity
- Additional controls: Add detection methods (inspections, tests, interlocks) to improve detection
- Analysis/testing: Conduct specific analysis or test to better characterize a risk that is currently assessed by judgment
After actions are implemented, update the FMEA with revised ratings and verify that the RPN has been reduced to an acceptable level. An FMEA that is never updated after corrective actions is evidence that the analysis was never taken seriously.
FAQ
Q: When in the design process should FMEA be conducted?
A: DFMEA should begin as soon as the design concept is defined enough to identify components and functions — typically after concept selection, before detailed design begins. Starting FMEA late (after drawings are complete) reduces its value because design changes become expensive. The FMEA should be updated at each major design review as the design evolves and new failure modes become identifiable. A final FMEA review should occur before production release.
Q: How do I handle failure modes where severity is very high (S=9 or 10) but occurrence is extremely low?
A: High-severity failure modes warrant special treatment regardless of RPN. For S=9 or S=10 failures, the standard approach is: first attempt to redesign to achieve a less severe failure effect (a design that fails safely rather than catastrophically). If the severe failure mode cannot be designed out, add detection and prevention controls to the maximum practical extent, document the residual risk explicitly, and obtain formal sign-off from a designated risk owner. Do not allow high-severity failures to be de-prioritized based on low RPN.
Q: How granular should the FMEA be — should I analyze every sub-component?
A: The appropriate level of granularity depends on the criticality of the assembly and the development stage. For safety-critical assemblies, every component and its failure modes should be analyzed. For non-critical standard components procured to industry standards, a higher-level analysis covering the component’s function (rather than internal failure mechanisms) is often sufficient. As a practical guide: new designs, novel materials, and safety-critical functions deserve detailed analysis; re-used proven designs with established reliability data can be handled at a higher level.
コメント