Physics-informed dynamic Bayesian networks for time-dependent reliability prediction of subsea wellhead sealing system with multi-states

The Subsea Wellhead Sealing System (SWSS) is crucial for the safety of deepwater operating, yet its reliability assessment faces challenges from harsh environments and multi-factor interactions. This study developed a data-driven, physics-informed reliability assessment method combining Finite Element Analysis (FEA) and Dynamic Bayesian Networks (DBN). An FEA model is established based on metal sealing theory, and a data-driven reliability model is subsequently constructed through sampling analysis, with a numerical-to-state conversion method bridging FEA and DBN. The FEA-DBN approach offers two key advantages: eliminating expert scoring subjectivity through physics-based modeling and effectively capturing multi-factor interactions and time-dependent behaviors. Results show this method can precisely quantify the evolution of SWSS reliability throughout its service lifecycle, with the probability of failure increasing from 0.64 % to 3.38 % over a 30-year service life. Case studies demonstrate its effectiveness for deep-sea equipment assessment, particularly in operating environments where real-time monitoring proves challenging, thereby demonstrating significant engineering application value.