An Uncertainty-Aware Framework for Runway Excursion Risk Assessment Using SHELLO and Set Pair Analysis

Runway excursion (RE) risk during aircraft landing emerges from the coupling of technical, environmental, human, information-related, and organizational factors, and is characterized by the coexistence of deterministic deviations and uncertain intermediate states. This study proposes an uncertainty-aware RE risk assessment framework by integrating the SHELLO model with Set Pair Analysis. A six-dimensional indicator system covering Software, Hardware, Environment, Liveware, Liveware–Organization, and Organization was established based on accident reports, regulatory and operational documents, 94 valid pilot questionnaires, and expert consultation. Indicator weights were determined through structured expert judgment, and grading thresholds were established by integrating industry standards, operational evidence, expert judgment, and simulation-supported information. A discrepancy-refined ternary SPA formulation was employed to characterize identity, discrepancy, and contrary relationships and to further resolve the discrepancy component into identity-side, residual, and contrary-side components. The principal value of connection degree and set pair potential were then used to characterize the separation of the evaluated state from the neutral position and the relative dominance between identity and contrary components. A case study of an A320-landing RE scenario indicated that the evaluated system fell within the Grade II intermediate diagnostic region. Hardware and Environment characterized the background risk conditions, whereas Software-related information deviations and Liveware–Organization coordination deficiencies showed the most prominent negative diagnostic contributions. Four additional anonymized A320-landing incident cases exhibited distinct dimensional risk profiles. Discrepancy refinement changed three-dimensional classifications relative to conventional SPA, illustrating its effect on boundary-sensitive risk diagnosis.