Dimension-specific aggregation of FMEA expert scores for marine gas turbine risk assessment: A critic–topsis approach

Failure Mode and Effects Analysis (FMEA) is widely applied for reliability assessment, safety analysis, and maintenance planning of complex engineering systems. In expert-based FMEA, judgments are aggregated into Occurrence (O), Severity (S), and Detection (D) values before calculating the Risk Priority Number (RPN). However, the aggregation method is often selected without systematic justification, although it can influence failure-mode prioritization. This study proposes a dimension-specific procedure for selecting expert-score aggregation operators rather than applying a single default rule. Eight candidate methods are evaluated separately for O, S, and D using four criteria: Breakdown Point, Spearman rank correlation with Borda consensus, Jackknife standard deviation, and Bootstrap RMSE. CRITIC is used to derive criterion weights, while TOPSIS ranks the alternatives. The framework is applied to a marine gas turbine FMEA case involving 48 failure modes and 10 experts. Results indicate that the median is preferred for Occurrence and Severity, whereas the arithmetic mean is selected for Detection, yielding RPN* = O(Median) × S(Median) × D(AM). The resulting ranking remains strongly correlated with the conventional AM-based RPN (ρ = 0.9451), while individual failure modes shift by up to 11 positions. Extensive robustness and sensitivity analyses further support the dimension-specific nature of aggregation selection.