应用神经网络法智能识别油气水多相流流型.pdfVIP

中国【程熟物理萼会 热杌气动热力学学术会议 992034 应用神经网络法智能识别油气水多相流流型 昊浩江周芳德何安定 (西安交通大学多相流国家重点实验室，710049，西安) 摘要本文采用分形理论和神经网络技术来实现智能识别油气水多相流流型．本文利t}}j 快速响应压阻式压差传感器测得油气水多相流各种流型F的瞬态压羊信号，对信号进行 出各种流型下的信号的特征向量，然后将已知流型的压差信号特征向量输入神经网络进 行训练，将某种未知流型的压差信号特征向量送入经过训练后的神经网络．这时网络将 能自动输出流型的类型，从而实现对流型的智能识别．结果证明．该识别方法十分有效 主题词：分形关联维数径向基函数神经网络 多相流流型 Flow ofOil--Gas--Water Identifying Intelligent Regime Multiphase FlowwithNeuralNetwork Wu Zhou He Haojiang FangdeAnding xijdH JiaotongUniversi以710049,Xran,Shannxi Abstract：nleflow ofOil—Gas—Waterthree flowis identified regime phase intelligentlyby meansoffractaI andneuralnetwork inthis InstantaneousdifferentiaI theory technologypaper flow alemeasured ofOil-Gas-Waterflowunderdifferent pressuresignals multiphase regimes with are witha differentialtransducerfast piezo—resistancepressure response．111esignals denoisedwavelet andthecharacteristicvectorsofvariousflow are using theory regimes of extractedfromthedenoisedsignalsmeans algorithm by reconstructingpseudo-phasespace known arefedintoaradiusbasis offractaI characteristicvectorsof flow theory．The regimes for vectorofsomekindofunknOWll fonctionlleufa】network