气体管道运行状态特征提取与状态识别  

Feature extraction and state recognition of gas pipeline in operating state

作　　者：阚玲玲[1] 叶蕾 高丙坤[1] 梁洪卫[1] 路敬祎[1] KAN Lingling;YE Lei;GAO Bingkun;LIANG Hongwei;LU Jingyi(College of Electrical and Information Engineering,Northeast Petroleum University,Daqing 163318,China)

机构地区：[1]东北石油大学电气信息工程学院,黑龙江大庆163318

出　　处：《压力容器》2021年第2期14-21,共8页Pressure Vessel Technology

基　　金：黑龙江省自然科学基金项目“基于变分模态分解和流形学习的油气管道泄漏信号检测技术研究”(LH2020F005);东北石油大学国家基金培育基金项目“气体管道泄漏声波信号特征提取与状态识别”(2018GPYB-03)。

摘　　要：研究了气体管道各种运行状态下的声波信号特征参数,通过仿真分析验证可变模态分解(VMD)后前两个分量的中心频率IMF1和IMF2,以及经VMD-Wavelet处理后重构信号的云模型特征熵En和重心频率FC可以作为气体管道运行状态识别的特征参数;研究了反向传播神经网络,提出VMD-En-BP模型,通过测试分析发现,该模型能够准确识别气体管道的正常运行、敲击、渗漏、小泄漏和大泄漏等五种运行状态。In this paper,firstly,the characteristic parameters of acoustic signal under various operating states of gas pipeline were studied.Through simulation analysis,the center frequency IMF1 and IMF2 of the first two components after VMD decomposition were verified,and the cloud model characteristic entropy En and gravity center frequency FC of the reconstructed signal after VMD-Wavelet processing could be used as the characteristic parameters for gas pipeline operation state recognition.Then,the back propagation neural network was studied,and the VMD-EN-BP model was proposed.Through the test and analysis,it was found that the model could accurately identify five operating states of gas pipeline,including normal operation,knocking,seepage,small leakage and large leakage.

关 键 词：气体管道泄漏检测 可变模态分解 特征提取 运行状态识别 

分 类 号：TH49[机械工程—机械制造及自动化] TE973.6[石油与天然气工程—石油机械设备]