基于大数据和神经网络的管道完整性预测方法

Pipeline integrity prediction method based on big data and neutral network

为了全面、客观地评价油气管道完整状态,预测管道潜在威胁,提高油气管道安全管理水平,通过采用人工神经网络机器学习理论,建立了对管道潜在威胁和管道状态进行学习和预测的框架与方法。该方法根据管道潜在威胁类型和管道状态定义,充分利用管道建造、施工、操作、运行、失效、检测等各类数据,作为潜在影响因素,以模拟人脑思维和学习的方式,对管道威胁和管道状态进行客观、有效的学习,给出潜在影响因素的重要度排序,利用获取的知识,对管道潜在威胁和管道状态进行客观预测,并将预测结果用于评价管道风险和制定管道检测周期。该方法避免了以往分析方法依赖专家意见、主观性强的缺点,全面利用各类管道数据集,客观地对管道完整性进行预测,对提高管道安全水平和决策效率有重要作用。

In order to comprehensively and objectively evaluate the integrity of oil and gas pipeline, predict the potential threats and intensifies the safety management of pipelines, this paper proposes the framework and method to learn and predict the potential threats and pipeline conditions by adopting artificial neutral network machine learning theory. According to the definition of potential threat types and pipeline conditions, this method simulates the human thinking and learning, with pipeline construction, operation, failure and detection as the potential influential factors. It can learn the pipeline threats and conditions objectively and effectively, and provide the order of importance of potential influential factors. With the information obtained, it makes objective prediction on the pipeline threats and conditions, and then applies the prediction results to evaluate pipeline risks and define pipeline detection period. Unlike previous analysis methods that rely on experts＇ opinions and feature strong subjectivity, this method can make full use of various pipeline data sets to predict the pipeline integrity objectively, which is of great importance to improving pipeline safety and efficiency of decision-making. （2 Figures, 2 Tables, 12 References）