Dynamic risk assessment of deepwater drilling using data-based and probabilistic approach

Risks associated with deepwater drilling are dynamic because of various accident-causing factors such as equipment failures, abnormal processes, and operator errors. All these factors should be considered during the implementation of dynamic and quantitative risk assessment (DQRA) for drilling operations. Dynamic Bayesian network (DBN) can graphically present the cause-effect relationships among different types of risk influencing factors (RIFs). Hence, this study developed a four-step DBN model for the DQRA of deepwater drilling. Firstly, multi-type contributing RIFs were identified according to the process flow. Subsequently, a network structure was developed to present the potential accident scenarios and capture the interdependencies among the RIFs. Thereafter, the probabilities of equipment failures, abnormal processes, and operator errors were determined using the probabilistic, data-based, and Standardized Plant Analysis Risk-Human Reliability Analysis (SPAR-H) approach, respectively. Finally, DBN inference was performed to evaluate the probabilistic risk of drilling operations. The model was applied to a case study of DQRA for managed pressure drilling (MPD), where the calculated initial blowout probability was 9.30 x 10-5, whereafter it was updated dynamically. This case study demonstrates the practicability of the proposed approach. This study contributes to a systematic investigation of the role of multisource data in DQRA using a full DBN approach. The assessment results can provide early warnings for practitioners to implement risk elimination or mitigation measures in real time.

Automated summary

This study developed a dynamic Bayesian network (DBN) model for deepwater drilling and demonstrated its feasibility through a case study on managed pressure drilling (MPD). The model identified various risk influencing factors, constructed a network structure, and determined the probabilities of equipment failures, abnormal processes, and operator errors. By evaluating the probabilistic risk of drilling operations, the model provided early warnings for risk elimination or mitigation measures in real time.