基于动态贝叶斯网络的深水防喷器系统实时可靠性评估方法研究

深水防喷器系统是海洋油气钻井中一种高自动化和高可靠性的关键井控装备，一旦失效，将引起灾难性后果。本项目提出将动态贝叶斯网络模型应用于深水防喷器系统实时可靠性评估的新方法。建立考虑不完全维修和预防性维护的动态贝叶斯网络结构模型；研究融合多源不完备信息的模块化动态贝叶斯网络参数学习方法；研究多源信息融合的分层理论与动态贝叶斯网络相结合的实时可靠性评估方法；采用动态贝叶斯网络近似推理算法，研究实时可靠性评估模型的快速推理方法；采用敏感性分析方法验证实时可靠性评估模型的正确性；研究不完全维修和预防性维护对深水防喷器系统可用性的影响，研究深水防喷器系统基本部件的重要度序列；开发深水防喷器系统实时可靠性评估软件，并在自主研发的深水防喷器模拟样机上进行测试。本项目的研究对丰富贝叶斯网络的理论和内容，提高海洋石油钻探开发的安全性，具有重要的理论意义和工程应用价值。

Subsea blowout preventer (BOP) system is the key well control equipment with high automation and reliability during offshore oil drilling. It may cause disastrous consequence once it fails. The application process of dynamic Bayesian networks (DBN) in the real-time reliability evaluation of subsea BOP system is presented in this research. The DBN structure model will be established taking account of imperfect repair and preventive maintenance. The modular DBN parameter learning method with multi-source incomplete information will be researched. The real-time reliability evaluation model based on the theory of hierarchies of multi-source information fusion and DBN will be established. The high-speed reasoning algorithm of real-time reliability evaluation model will be researched by using approximate reasoning algorithm of DBN. The real-time reliability evaluation model will be verified by using sensitivity method. The influence of imperfect repair and preventive maintenance on the availability, and the important sequence of basic components of subsea BOP system will be studied. The real-time reliability evaluation software of subsea BOP will be developed and tested in the home-grown subsea BOP prototype. The research will increase the theory and content of Bayesian networks, improve the safety of offshore oil drilling, and have a very important significance in theory and in engineering.