摘要
针对船舶风险侧重于计算发生概率,其管控和干预效果难以评估及推理难题,提出了基于do算子的船舶风险干预及推理方法。首先利用先验知识建立因果关系网络;其次引入Leaky Noisy-Or假设和隐变量节点,建立基于结构性孪生因果网络模型;最后根据do算子干预进行反事实推理,利用引起故障的充分因和必要因来确定干预措施的有效性。实例分析表明:所提方法可以有效衡量风险管控措施的有效性,判断干预的行为对危险源排序的影响,具有一定的稳定性和优越性。
Aiming at the ship risk focusing on calculating the occurrence probability,its control and intervention effect is difficult to assess and reasoning difficulties,the do operator-based ship risk intervention and reasoning method was proposed.Firstly,the causal network was established using a priori knowledge.Secondly,the Leaky Noisy-Or hypothesis and hidden variable nodes were introduced to establish a structural twin-based causal network model.Finally,counterfactual inference was performed based on do operator interventions,and the effectiveness of interventions was determined using sufficient and necessary causes that cause failures.The case analysis shows that the proposed method can effectively measure the effectiveness of risk control measures and determine the impact of the intervention's behavior on the hazard source ranking,with certain stability and superiority.
作者
王学峰
苏高辉
谢旭阳
曾国庆
王劲韬
陈国兵
杨自春
WANG Xue-feng;SU Gao-hui;XIE Xu-yang;ZENG Guo-qing;WANG Jin-tao;CHEN Guo-bing;YANG Zi-chun(School of Power Engineering,Naval University of Engineering,Wuhan 430000,China;Qingdao Equipment Technical Quality Monitoring Station,Qingdao 266000,China)
出处
《科学技术与工程》
北大核心
2023年第12期5332-5339,共8页
Science Technology and Engineering
基金
国家自然科学基金(51802347)。
关键词
船舶风险评估
孪生网络模型
do算子
反事实推理
ship risk assessment
twin network model
do operator
counterfactual reasoning
