A knowledge transfer-based method for risk analysis and procedure optimization of emergency schemes

Abstract

Due to insufficient knowledge of rare accidents, it is essential to transfer knowledge from source domain with sufficient cases (e.g., onshore blowouts) into target domain with limited cases (e.g., deepwater blowouts). In this paper, a knowledge transfer-based method is proposed to evaluate the performance of emergency schemes in presence of limited accident cases. By considering dynamic evolution and operational characteristics of the accident, to evaluate emergency schemes from the perspectives of emergency risk and timeliness, a hybrid model integrating dynamic Bayesian networks (DBN) and program evaluation and review technique (PERT) is applied. In the integrated model, we transferred graph structures and parameters to obtain emergency schemes based on the similarities between the source systems and target systems. On the one hand, to evaluate and control emergency risk, DBN is utilized for dynamic risk analysis of emergency operations. On the other hand, to judge the response-time requirement, an optimized PERT is applied to enhance the completion probabilities of the emergency procedures. A case study on the installation of onshore blowout preventer and offshore capping stack is applied to illustrate the applicability of the proposed methodology. The results show that equipment failure and human error are crucial risk-influencing factors leading to emergency failure. Activities duration and completion probability of emergency schemes are optimized by emergency timeliness analysis.