If you accept Murphy’s Law as a non-negotiable boundary condition of system design, you acknowledge that your product or manufacturing line is going to fail. But acknowledging failure is not enough. You must anticipate the exact mechanism of that failure before the design leaves the building.
In amateur engineering environments, risk analysis consists of a few engineers looking at a CAD model and asking, "Do we think this is strong enough?"
In high-reliability environments—such as aerospace, medical device manufacturing, and elite automotive design—teams do not rely on gut feelings. They use a ruthless, quantitative framework developed by the military in the 1940s to systematically strip risk out of a design: Failure Mode and Effects Analysis (FMEA).
Simple Definition of FMEA
FMEA is a structured, step-by-step approach for identifying all possible failures in a design, a manufacturing process, or a product or service.
A "Failure Mode" is the specific way in which something might fail. An "Effect Analysis" involves studying the consequences of those failures. By breaking a complex machine down to its individual components and interrogating every single part, engineers can stop reacting to unexpected catastrophes and start engineering them out of existence.
Why Engineers Trust FMEA
The true power of an FMEA lies in its ability to remove subjectivity from design reviews. Instead of arguing over which failure is "worst," the team calculates a Risk Priority Number (RPN) for every single failure mode using three variables (each scored on a scale of 1 to 10):
- Severity (1-10): How catastrophic is the failure? (1 = no noticeable effect; 10 = critical safety hazard without warning).
- Occurrence (1-10): How likely is this specific failure to happen? (1 = highly unlikely; 10 = practically guaranteed).
- Detection (1-10): If the failure happens, how likely are our current controls to catch it before it reaches the customer? (1 = certain to detect; 10 = completely invisible to current tests).
If an O-ring failing causes a machine fire (Severity = 9), happens somewhat frequently (Occurrence = 5), and is completely hidden from inspection (Detection = 10), your RPN is 450 out of 1000. That is an unacceptable risk that requires immediate design action.
It is important to note that RPN is a prioritization tool, not an absolute measure of risk. Different combinations of Severity, Occurrence, and Detection can produce identical RPN values, which is why high-severity failure modes must always be addressed regardless of the final score.
“A high RPN is not a warning. It is a leading indicator of future downtime.”
The Contrast Insight: The Detection Blind Spot
Most engineering teams naturally focus on Severity and Occurrence. They want to make the product stronger or use better materials. But the most cost-effective way to drastically reduce an RPN is often to improve Detection.
In many mature systems, Detection becomes the dominant lever for risk reduction because it can be improved without fundamental design changes, making it the fastest path to lowering system-level risk. You cannot always stop a tool from breaking (Occurrence) or change the fact that a broken tool ruins the part (Severity). But if you install a sensor that instantly stops the machine the millisecond the tool breaks (Detection score drops from 9 to 2), your RPN plummets. This is how the Swiss Cheese Model is successfully implemented: by building detection layers that catch the failure before it escapes.
Engineering Controls for Running an FMEA
An FMEA is useless if it is treated as a paperwork exercise to satisfy an ISO auditor. To make it a functional engineering tool, enforce these controls:
- Cross-Functional Execution: Never let a design engineer do an FMEA alone. They suffer from creator's bias. You must put manufacturing, quality, and maintenance engineers in the room. The people who assemble the product know exactly how it will break.
- Set an RPN Threshold: Establish a hard limit (e.g., any RPN over 100). If a failure mode scores above that threshold, the product cannot be released to manufacturing until an engineering control is added—often via error-proofing methods like poka-yoke—to lower the score.
- Prioritize High Severity: A failure mode with a Severity of 10 (life-threatening) and an RPN of 90 must always be addressed before a nuisance failure with an RPN of 150. Severity always overrides the raw score.
Quick Self-Check: Is Your FMEA Process Broken?
- Is your FMEA filled out after the design is finished, just to satisfy a checklist?
- Do teams artificially lower the Detection or Occurrence numbers just to get the RPN under the threshold?
- Are field failures regularly occurring that were never listed as a possible Failure Mode on the original FMEA?
- Is the document ignored once production begins, rather than treated as a living "lessons learned" database?
Frequently Asked Questions (FAQ)
What is the difference between DFMEA and PFMEA?
DFMEA (Design FMEA) focuses on how the product itself could fail due to design flaws (e.g., a bracket snapping under load). PFMEA (Process FMEA) focuses on how the manufacturing or assembly process could fail (e.g., an operator using the wrong torque spec during assembly).
Why is the RPN detection score counter-intuitive?
Because a higher number is worse. If you have an excellent, foolproof detection system, your score is a 1. If the failure is completely invisible and you have no way to detect it, your score is a 10.
When should an FMEA be started?
As early as possible in the conceptual phase. The cost of mitigating a failure mode is practically zero when it is just a CAD model, but it can cost millions if the failure is discovered after tooling is cut and production begins.
The Framework for Reliability Engineering
Engineering is not about hoping for the best. It is about mathematically predicting the worst and designing a system that survives it. If you aren't calculating your risks, you are just waiting to be surprised by them.
Do not wait for a failure to tell you what went wrong. Calculate it before it happens.
To master the definitive industry framework for risk analysis and learn how to execute a flawless Failure Mode and Effects Analysis, explore the authoritative guide by reliability expert Carl Carlson, Effective FMEA.
Comments
Post a Comment