A mechanics-data-driven methodology for dynamic risk evaluation of riser and new hang-off system

Typhoons pose substantial reliability challenges to drilling operations, particularly when risers operate in hang-off modes. Extreme hydrodynamic loads and platform motions may trigger nonlinear responses and cascading failures, while the coupling of a new hang-off system with a hydraulic compensation mechanism complicates quantitative risk evaluation. Therefore, a mechanics-data-driven risk assessment model is developed to quantify the dynamic catastrophe risk of the riser and hang-off system under severe marine conditions. The model integrates fault tree analysis, machine learning, and a dynamic catastrophe model, incorporating environmental and equipment uncertainties and time-varying behavior. A dynamic catastrophe fault tree identifies natural and equipment-related failure sources, where environmental variables follow Weibull distributions and equipment faults are modeled as uniform processes. A particle swarm optimization–deep neural network surrogate model is trained on simulation data to predict key structural responses, and Latin hypercube sampling estimates failure probabilities under different fault scenarios. The dynamic positioning system failure probability from literature is incorporated via the total probability theorem, while a dynamic Bayesian network captures temporal dependencies and mode transitions. Application to riser system under typhoon conditions verifies that the model effectively characterizes nonlinear coupling and time-dependent risk variation, providing a reliable basis for improving riser safety.