Abstract: Ice-induced vibration is a significant risk factor for the operational safety and structural stability of offshore oil platforms. If the hazards posed by sea ice, currents, and other environmental loads are not properly assessed, offshore platforms may suffer catastrophic consequences such as collapse, oil spills, and casualties. Due to the uncertainty of ice load and ice-induced vibration, traditional numerical analysis and statistical prediction methods struggle to simultaneously meet the requirements for high-precision early warning and timeliness in ice-induced vibration risk assessment. Therefore, based on the measured winter sea environment data of the JZ20-2MUQ platform in the Bohai Sea, this paper uses the interpretive structural model (ISM) and the expectation maximization (EM) algorithm to construct a Bayesian network model. Quantitative analysis indicates that the main risk factors affecting the vibration of the offshore oil platform in the Bayesian network are ice thickness, ice velocity, ice direction, wind speed, and tidal flow velocity and direction. Furthermore, the accuracy and validity of the model are verified. The average accuracy of the model is over 0.75, and the accuracy of the target node is above 0.9. Verification results confirm that the model has high accuracy and can provide reliable data support for risk prediction.

Key words: offshore oil platform, ice-induced vibration, risk level prediction, Bayesian network

摘要： 冰激振动是引起海上石油平台结构失稳的重要风险因素之一。若管理者对海洋平台所承受的海冰、海流等环境荷载风险评估错误,则海洋平台可能发生垮塌、溢油、人员伤亡等事故。由于冰荷载与冰激振动存在不确定性,传统的数值分析与统计预测方法在冰振风险等级评估上难以同时满足高精度的预警需求和高时效要求。该文以渤海JZ20-2MUQ平台冬季海域环境实测数据为依托,利用解释结构模型和最大期望算法搭建贝叶斯网络模型。定量分析结果证明：贝叶斯网络海上石油平台模型振动的主要风险因素为冰厚、冰速、冰向、风速、潮流流速和流向。该研究对模型进行了准确性和有效性验证,其平均准确度超0.75,目标节点的精度在0.9以上;准确性和有效性验证结果均表明该模型性能良好,能够为风险等级预测提供有力的数据支持。

关键词: 海上石油平台, 冰激振动, 风险等级预测, 贝叶斯网络