3.3 Failure Mode and Effects Analysis (FMEA)

FMEA is not paperwork; it is the mathematical prediction of the future. It forces engineers to stare into the abyss of "What could go wrong?" and build a bridge over it before the first prototype is scrapped. If you treat FMEA as a checkbox exercise to satisfy an auditor, you will find the failure modes in the field, where they cost 100x more to fix.

The Mechanics of Risk (The RPN Engine)

We quantify risk using the Risk Priority Number (RPN) to strip away emotion and focus on data.

Formula: RPN = Severity (S) x Occurrence (O) x Detection (D)

The Variables:

• Severity (S): The impact on the end user.
  • Constraint: You cannot "inspect" Severity down. You can only reduce it by changing the Design (e.g., adding a fuse).
  • 10 = Hazard / Safety Risk without warning.
  • 1 = No discernible effect.
• Occurrence (O): The probability of the cause happening.
  • Driver: Reduced by Process Capability (Cpk) and robust design margins.
• Detection (D): The probability that we catch it before it escapes.
  • Driver: Improved by testing and inspection.
  • 10 = Absolute uncertainty (No check).
  • 1 = Error Proofing (Physics prevents the defect).

Pro-Tip: Never average RPN scores. A single line item with Severity 10 and RPN 90 is infinitely more critical than an item with Severity 3 and RPN 200. Safety trumps statistics.

Execution Strategy: DFMEA vs. PFMEA

Do not confuse the intent. One protects the design; the other protects the build.

DFMEA (Design FMEA):

• Focus: Component physics, circuit topology, geometry, material properties.
• Example Failure: "Capacitor derating insufficient for voltage spike."
• Mitigation: Select higher voltage rating component.

PFMEA (Process FMEA):

• Focus: The machine, the operator, the environment, the method.
• Example Failure: "Operator installs capacitor backwards."
• Mitigation: Add polarity marking to PCB silkscreen + AOI check.

Scoring Calibration:

• IF Severity = 9 or 10 (Safety/Regulatory) -> THEN Action is Mandatory regardless of RPN.
• IF RPN > 100 -> THEN Mitigation plan is required.
• IF Detection = 10 (Visual inspection by human) -> THEN Reject as a primary control for critical features. Humans are only 80% effective.

The AIAG-VDA Harmonization (The New Standard)

Modern Quality Engineering is moving away from pure RPN towards Action Priority (AP) levels (High, Medium, Low) to prevent "gaming the numbers" (e.g., artificially lowering Detection to get RPN < 100).

Logic Flow:

• High Priority: Severity 9-10 with any meaningful Occurrence. -> Action: Review at Management Level.
• Medium Priority: Severity 7-8 with Moderate Occurrence. -> Action: Review at Engineering Level.

Final Checklist