Investigating human and organizational risk propagation in ship foundering: A data-driven HFACS–Apriori–DEMATEL–FCM framework

Maritime foundering is a low-frequency but high-consequence accident mode associated with multiple human, technical, supervisory, and organizational vulnerabilities. This study develops a data-driven framework integrating an adapted Human Factors Analysis and Classification System (HFACS), Apriori association rule mining, the Decision-Making Trial and Evaluation Laboratory (DEMATEL), and a Fuzzy Cognitive Map (FCM). A database of 155 foundering investigation reports from the China Maritime Safety Administration was coded into 50 risk factors. Apriori association mining identified empirical conditional associations in the coded reports, and their confidence values were used to parameterize the subsequent DEMATEL and FCM analyses. The modeled results distinguish highly activated terminal vulnerability accumulators from upstream intervention factors. Under the baseline FCM settings, idealized suppression of substandard physical and mental condition (P10), a lagging emergency response mechanism (O11), and the absence of an emergency preparedness plan (S8) yielded the largest reductions in aggregate activation across the remaining factors, with Global Sensitivity Index magnitudes of approximately 0.124–0.136. The findings support cross-level safety interventions that target latent organizational, supervisory, and preconditional vulnerabilities before they propagate into foundering.