基于磁力学仿真的油气管道应力预测机器学习模型  

Machine Learning Model for Stress Prediction of Oil and Gas Pipelines Based on Magnetic Simulation

作　　者：邢鑫磊 翁光远 王浩然 刘阳 王兴 赵卓松 XING Xinlei;WENG Guangyuan;WANG Haoran;LIU Yang;WANG Xing;ZHAO Zhuosong(School of Mechanical Engineering,Xi'an Shiyou University,Xi'an,Shaanxi 710065,China)

机构地区：[1]西安石油大学机械工程学院,陕西西安710065

出　　处：《石油工业技术监督》2025年第3期1-5,共5页Technology Supervision in Petroleum Industry

摘　　要：输油气管道结构参数、运行环境、服役荷载是引发应力集中的关键因素,应力集中引发管道变形、开裂,造成管道结构失效。针对管道结构参数、运行环境、服役荷载耦合作用与应力变化的高次非线性、随机性和复杂性,提出一种基于机器学习的输油气管道应力预测方法。利用COMSOL软件进行有限元模拟计算,提取了管道受力关键点的应力分布数据,建立应力数据集并进行数据预处理、特征工程和超参数优化,在相同的数据集下比较不同算法(线性回归、梯度提升树、K最近邻)对输油气管道应力的预测效果,采用决定系数(R2)、平均绝对误差(MAE)和均方根误差(RMSE)分析3种算法的预测性能。结果表明:梯度提升树算法的应力预测效果值为R2=0.8454、RMSE=0.0503、MAE=0.0294,优于其他2种算法,该算法能够准确地预测输油气管道的应力值。The structural parameters,operating environment and service load of oil and gas pipeline are the key factors causing stress concentration,which leads to pipeline deformation,cracking and structural failure.Aiming at the high-order nonlinearity,randomness and complexity of the coupling effect of pipeline structural parameters,operating environment,service load and stress change,a stress prediction method of oil and gas transportation pipeline based on machine learning was proposed.COMSOL software was used for finite element simulation calculation,stress distribution data of the key points of pipeline stress were extracted,stress data set was established,and data preprocessing,feature engineering and overparameter optimization were carried out.The prediction effects of different algorithms(linear regression,gradient lifting tree,and K-nearest neighbor)on the stress of oil and gas pipelines were compared under the same data set.The coefficient of determination(R2),mean absolute error(MAE),and root mean square error(RMSE)were used to analyze the prediction performance of the three algorithms.The results show that the stress prediction effect value of the gradient lifting tree algorithm is R2=0.8454,RMSE=0.0503,MAE=0.0294,which is superior to the other two algorithms,indicating that the algorithm can accurately predict the stress value of oil and gas pipelines.

关 键 词：输油气管道 机器学习 应力 预测 

分 类 号：TE973[石油与天然气工程—石油机械设备] TP181[自动化与计算机技术—控制理论与控制工程]