摘要：

天然气管道失效可能导致人员伤亡和巨大财产损失，运用风险分析能预判管道存在安全隐患，进而通过采取相应措施能降低管道失效发生概率。针对管道事故数据不足，难以精确确定失效概率的问题，本文提出了一种基于排序节点法和层次分析法的输气管道模糊贝叶斯网络风险分析方法。首先，采用模糊综合评判(FCE)实现专家意见量化与系统化分析，计算管道风险因素先验概率；其次，采用层次分析法计算管道基本风险因素权重，采用排序节点法根据权重计算条件概率表(CPT)；最后，将求得的先验概率与节点间条件概率表应用于贝叶斯网络(BN)。从而实现管道事故数据不足、影响因素关系复杂情况下的管道失效概率计算，并通过概率更新，识别关键事件。将该方法应用于某输气管道，根据数据库和现场专家意见，识别潜在风险，以风险因素作为基本节点，综合节点间关系建立贝叶斯网络模型，应用本研究风险分析方法计算先验概率与条件概率，代入贝叶斯模型，求得管道失效概率，确定引起管道失效的关键事件，与现有数据库失效概率对比相近，验证该方法的可行性与合理性。结果表明，该方法能够为管道技术人员提供安全方面的科学指导。

关键词: 贝叶斯网络, 风险分析, 模糊综合评判法, 排序节点法, 层次分析法

Abstract:

The failure of natural gas pipelines may lead to serious casualties and huge property losses, and it also poses a great threat to the surrounding ecological environment and social public safety. Risk analysis is an important technical means to predict potential safety hazards existing in pipelines. By carrying out systematic risk identification, quantitative evaluation and targeted preventive control, relevant departments can accurately grasp the risk level of pipeline operation and take corresponding measures to reduce the occurrence probability of pipeline failure effectively. However, in the actual engineering application of natural gas pipelines, the available historical accident data are often insufficient, and the relationship among various risk factors is complex and uncertain, which brings great difficulties to the accurate determination of pipeline failure probability and the rationality of risk assessment results. To solve this problem, this paper proposes a fuzzy Bayesian network risk analysis method for gas transmission pipelines based on the Ranking Nodes Method(RNM) and the Analytic Hierarchy Process (AHP). Firstly, Fuzzy Comprehensive Evaluation (FCE) is used to realize the quantification and systematic analysis of expert opinions, so as to reduce the subjectivity and fuzziness of expert judgment and calculate the prior probability of each key risk factor of pipelines. Secondly, the weight of each basic risk factor is calculated by using the Analytic Hierarchy Process, so as to reflect the relative importance of different risk factors in the process of pipeline failure. Thirdly, the Conditional Probability Table (CPT) between nodes is calculated according to the weight by using the Ranking Nodes Method, which provides a reliable data basis for the construction of Bayesian network model. Finally, the calculated prior probability and the conditional probability table between nodes are applied to the Bayesian Network (BN), so as to realize the accurate calculation of pipeline failure probability under the conditions of insufficient pipeline accident data and complex relationship of influencing factors. At the same time, through probability updating and reverse reasoning, the critical events that have a significant impact on pipeline failure can be efficiently identified. In order to verify the effectiveness of the method, it is applied to an actual gas transmission pipeline. According to the existing engineering database and on-site expert opinions, various potential risks in the operation process are identified comprehensively. These key risk factors are taken as basic nodes, and the logical relationship and influence mechanism between nodes are comprehensively analyzed to establish a complete Bayesian network model. Then, the prior probability and conditional probability are calculated by using the risk analysis method proposed in this study, and the calculation results are substituted into the Bayesian model to obtain the final pipeline failure probability and determine the critical events leading to pipeline failure. The calculated failure probability is close to the failure probability recorded in the existing database, which fully verifies the feasibility, rationality and accuracy of the method. The results show that this method can overcome the limitation of lack of accident data, realize the scientific and quantitative evaluation of pipeline operation risk, and provide important theoretical support and scientific guidance for pipeline technicians in daily safety management, risk early warning and maintenance decision-making.

Key words: bayesian network, risk analysis, fuzzy comprehensive evaluation method, ranked nodes method, analytical hierarchy process

中图分类号: 

• TE832

• TE88