基于数据驱动的管道环焊缝风险预测模型研究
Research on data-driven risk prediction model for pipeline girth welds
浏览(649) 下载(24)
- 引用格式:
-
杨仪,孙晁,王婷,帅健,陈健,李云涛.基于数据驱动的管道环焊缝风险预测模型研究[J].天然气与石油,2023,41(6):147-154.doi:10.3969/j.issn.1006-5539.2023.06.021
YANG Yi, SUN Chao, WANG Ting, SHUAI Jian, CHEN Jian, LI Yuntao.Research on data-driven risk prediction model for pipeline girth welds[J].Natural Gas and Oil,2023,41(6):147-154.doi:10.3969/j.issn.1006-5539.2023.06.021
- DOI:
- 10.3969/j.issn.1006-5539.2023.06.021
- 作者:
- 杨 仪1,2 孙 晁3 王 婷3 帅 健1,2 陈 健3 李云涛1,2
YANG Yi1,2, SUN Chao3, WANG Ting3, SHUAI Jian1,2, CHEN Jian3, LI Yuntao1,2
- 作者单位:
- 1. 中国石油大学(北京)安全与海洋工程学院, 北京 102249; 2. 应急管理部油气生产安全与应急技术重点实验室, 北京 102249; 3. 国家管网集团北方管道有限责任公司, 河北 廊坊 065000
1. College of Safety and Ocean Engineering, China University of Petroleum(Beijing), Beijing, 102249, China; 2. Key Laboratory of Oil and Gas Safety and Emergency Technology, Ministry of Emergency Management, Beijing, 102249, China; 3. Northern Pipeline Company of PipeChina, Langfang, Hebei, 065000, China
- 关键词:
- 油气管道;环焊缝;风险预测;机器学习;风险因素重要度;管道安全
Oil and gas pipeline; Girth weld; Risk prediction; Machine learning; Importance of risk factors; Pipeline safety
- 摘要:
为完善当前油气管道环焊缝失效风险判定方法,提高风险焊口开挖选点的有效性,依据油气管道环焊缝排查数据统计分析和环焊缝失效机理研究相关成果,建立了管道环焊缝全生命周期的失效风险因素识别体系,并基于集成算法建立环焊缝风险预测模型,所建立的预测模型与使用其他机器学习算法建立的模型相比,具有较高预测准确度和泛化性,模型整体预测准确率达86.44%,结果表明,模型对不合格焊口的识别精确度较好,可应用于工程中指导未开挖风险焊口的科学选点工作。基于所建立的预测模型,还提出了一种分析风险因素局部重要度的方法,从数据建模的角度出发,得到风险指标的因子权重,权重结果可为环焊缝的定量风险评价提供参考,有效提高管道环焊缝数字化管理水平。
In order to enhance the current methodology for failure risk determination of oil and gas pipelines girth welds and to improve the effectiveness in the selection of excavation points of risk girth welds for inspection, this paper proposes a comprehensive index system of risk factors of pipelines’ girth weld. This system, covering the entire lifecycle of pipeline girth welds, is established based on statistical analysis of oil and gas pipeline girth welds investigation data and pertinent research findings on the failure mechanism of pipeline girth welds. Additionally, a risk prediction model is developed based on the ensemble learning algorithm. This model, when compared to those models established using other machine learning algorithms, demonstrates superior prediction accuracy and versatility. The overall prediction accuracy of this model is reported to be 86.44%. The results signify that this model accurately identifies girth weld failures and serves as a beneficial guide for scientifically selecting unexcavated risk girth welds in engineering construction projects. Building upon the established prediction model, this paper introduces a method for analyzing the local importance of risk factors. This approach, obtaining the risk index factor weighting from a data modeling perspective, can serve as a reference point for quantitative risk assessments of girth welds, thereby significantly enhancing the digital management capability of pipeline girth welds.
