摘要
为实现含局部减薄缺陷油气管道爆破压力的精确预测,将神经网络算法应用于油气管道爆破压力的预测研究,建立了含局部减薄缺陷油气管道爆破压力预测的径向基函数(RBF)、广义回归(GR)、极限学习机(ELM)和误差反向传播(BP)4种神经网络模型。分析对比了4种模型用于X46、X52、X60、X65、X80材质含局部减薄缺陷油气管道爆破压力的预测误差,结果表明:4种神经网络模型的预测误差均小于ASME B31G,修正ASME B31G,DNVRP-F101三种通用规范的预测误差;BP神经网络模型的预测误差最小,预测误差为-1.10%~4.70%;神经网络预测模型用于油气管道爆破压力预测时具有操作简单,适用范围广,便于工程实用等特点。
In order to accurately predict the burst pressure of pipelines,the algorithms of artificial neural networks are applied to the prediction burst pressure of oil and gas pipeline with uniform corrosion defects.RBF neural network model,GR neural network model,ELM neural network and BP neural network model are established to predict the burst pressure of oil and gas pipelines with uniform corrosion defects.The prediction errors of four models used in X46,X52,X60,X65 and X80 oil and gas pipelines with uniform corrosion defects were analyzed and compared.The results show that the prediction errors of the four neural network models are obviously smaller than those of the three general codes such as ASME B31G,modified ASME B31G and DNV RP-F101.The calculation errors of BP neural network model is between-1.10%~4.70%.The neural network prediction model has the advantages of simple operation,wide application range and good engineering practicability when it is used to predict the burst pressure of oil and gas pipelines.
作者
王昱
贾思奇
李乃文
郭涛
Wang Yu;Jia Siqi;Li Naiwen;Guo Tao(Hebei Institute of Special Equipment Supervision and Inspection,Shijiazhuang,Hebei 050061,China;State Key Laboratory of Market Regulation-Safety Evaluation of Steel Pipes and Fittings,Shijiazhuang,Hebei 050061,China;School of Mechanical Engineering,Hebei University of Technology,Tianjin 300401,China)
出处
《绿色科技》
2023年第18期111-115,共5页
Journal of Green Science and Technology
基金
河北省市场监督管理局科技计划项目(编号:2018ZD13,2020ZC26)
河北省特种设备监督检验研究院科技计划项目(编号:HBTJ2023CY001,HBTJ2023CY003,HBTJ2023CY004)。
关键词
局部减薄
管道
爆破压力
神经网络
local wall-thinning
pipeline
burst pressures
neural network
