Dynamic failure-risk assessment of a rotary self-priming ECD-reduction tool under complex deepwater drilling conditions using a multi-fidelity dynamic Bayesian network

Offshore drilling under complex deepwater conditions is highly sensitive to equivalent circulating density (ECD) fluctuations, where narrow pressure margins and coupled hydraulic-mechanical loads impose stringent reliability requirements on downhole pressure-control equipment. To assess the failure risk of rotary self-priming ECD-reduction tools under complex operating conditions, a hierarchical multi-fidelity dynamic Bayesian network (H-MF-DBN) framework is proposed. The framework integrates time-slice state transitions, confidence-weighted fusion of low-fidelity simulation and high-fidelity field evidence, weighted Noisy-OR parameterization, and trapezoidal fuzzy evaluation to characterize degradation evolution and quantify failure probabilities under data-scarce conditions. Application to the JJSY-2H engineering test well shows that the baseline failure probability is 30%, with stator elastomer damage and insufficient turbine power identified as the dominant risk drivers. After sensitivity-guided optimization, the predicted failure probability decreases to 19%. Field testing and post-test inspection confirm stable operation and satisfactory structural integrity of the optimized tool. The proposed framework provides an interpretable reliability-assessment approach in which incoming operational evidence recursively updates node-state and system-failure probabilities through Bayesian posterior inference, thereby supporting engineering optimization of ECD-reduction tools under complex downhole operating conditions.