A DBN-GO approach for success probability prediction of drilling riser emergency disconnect in deepwater

Abstract

Drilling risers are crucial connections of subsea wellhead and floating drilling vessel. Riser emergency disconnect (RED) is the most effective protective measure to secure risers and wellhead in case of emergency. There are many different stages for RED depending on riser-connected operation. A novel approach based on Dynamic Bayesian Network and GO (DBN-GO) model is proposed to dynamically analyze the variation of RED risk over time in different stages. The GO model was built by translating the operation flow chart of RED and then mapped into BN. A DBN model was established considering the dynamic reliability of equipment. The cognitive reliability and error analysis method (CREAM) and improved analytic hierarchy process (IAHP) were used to determine and modify the probabilities of human factors in the DBN-GO model. The practical application of the developed model was demonstrated through a case study. The results showed that DBN-GO approach could be used to predict the dynamic success probability of RED under different stages during a drilling duration. Additionally, the key steps corresponding to each stage were identified, and some preventive measures to mitigate the failure risk of RED were proposed.