FMEA (Failure Mode and Effects Analysis)

FMEA is the primary risk management tool in manufacturing quality. Every IATF 16949 facility maintains design FMEAs (DFMEA) and process FMEAs (PFMEA) for critical products. The AIAG-VDA FMEA handbook (2019) replaced the traditional RPN approach with Action Priority (AP) ratings — High, Medium, Low — based on severity, occurrence, and detection combinations.

The "occurrence" rating in FMEA is directly tied to process capability data. A process with Cpk > 1.67 gets a low occurrence rating; Cpk < 1.0 gets a high one. This creates a direct link between statistical analysis quality and risk assessment accuracy. If your capability indices are unreliable due to normality assumptions or small sample sizes, your FMEA occurrence ratings are unreliable too.

The practical impact: an FMEA with inaccurate occurrence ratings misallocates engineering resources. High-risk failure modes get inadequate attention while low-risk ones consume disproportionate effort. Reliable process data is not optional for effective FMEA — it is foundational.

EntropyStat improves FMEA accuracy by providing more reliable process capability data for occurrence ratings. When Cpk is computed using the EGDF instead of assuming normality, the occurrence rating reflects actual process performance rather than a parametric approximation.

This matters most for processes with non-normal distributions. A skewed machining process might show Cpk = 1.45 under the normal assumption (occurrence = Low) but Cpk = 1.05 when the actual distribution shape is considered (occurrence = Medium to High). The entropy-based Cpk changes the FMEA action priority and could shift a failure mode from "monitor" to "immediate corrective action required."

EntropyStat's cluster detection via the ELDF adds another dimension to FMEA. If a process exhibits hidden subpopulations — cavity-to-cavity variation in injection molding, for instance — the ELDF can quantify each cluster's capability separately. This enables more granular FMEA analysis: instead of a single occurrence rating for the process, engineers can assess risk per subpopulation and target corrective actions precisely.