Sensitivity Analysis with FMEA — SA-Integrated Failure Mode and Effects Analysis

Standard FMEA computes a single RPN for each failure mode based on fixed expert scores. SA-FMEA treats those scores as uncertain inputs and systematically varies them to test how sensitive the resulting RPN rankings are to scoring assumptions. The core FMEA structure is unchanged; sensitivity analysis is layered on top to audit which inputs drive the conclusions.

Simple one-at-a-time sensitivity analysis can be performed in a spreadsheet by varying each of S, O, and D across their plausible range while holding the others fixed and plotting the resulting RPN. Global sensitivity methods such as Morris screening or Sobol indices require more computation and are typically implemented in R (sensitivity package) or Python (SALib library). The appropriate method depends on team capability and the number of failure modes under analysis.

For most FMEA teams, one-at-a-time (OAT) analysis or tornado charts are sufficient and interpretable. When interactions between S, O, and D are suspected to be important, Morris screening provides a practical middle ground. Sobol indices are most informative for complex weighted-RPN variants but require considerably more evaluations. Start with OAT and escalate only if interactions appear to matter.

Yes. Many organizations use weighted or alternative RPN formulas (e.g., assigning different exponents to S, O, D, or using additive combinations). Sensitivity analysis applies equally to these variants — the analyst simply defines the weights or formula parameters as additional uncertain inputs alongside the S, O, D ratings, and the sensitivity analysis reveals which weights most influence the resulting priorities.

Fault tree analysis (FTA) is appropriate when you need to model the logical combination of component failures that lead to a system-level top event, particularly in safety-critical systems where probability calculations are required. SA-FMEA is better suited when you are working at the failure-mode level and the primary question is how uncertain scoring affects prioritization rather than how failure causes combine probabilistically. Many reliability programs use both: FMEA to enumerate failure modes and SA to audit priorities, FTA to model system-level causality.