基于神经网络的水平管三相分层流相分率测量

Study to Determine Phase Fraction of Three Phase Stratified Flow in Horizontal Pipeline with Neural Network

将采用伽马射线测量的高能计数和低能计数作为输入参数,截面含水率和含气率作为输出参数,构建了预测水平管油气水三相分层流相分率的径向基函数神经网络.通过设计的相分率标定装置获得了神经网络的学习样本.在一内径为80mm的大型油气水三相流实验环道上进行了预测效果检验实验,结果表明,神经网络预测值与实测值非常吻合,含气率预测最大误差为3 6%,含水率最大误差为2 5%,有效地克服了传统双能伽马密度仪对流型敏感,不适于分离流动测量的问题.

A radial basis function (RBF) network was applied to determine phase fraction of oil-gas-water stratified three-phase flow in horizontal pipe. The numbers of counts from the gamma-ray densitometer were regarded as the input and the gas fraction and water fraction as the output. The neural network was trained according to the learning samples obtained from a specially designed device. To examine the prediction accuracy, experiments were conducted in a large oil-gas-water loop, the phase fractions were predicted with an error of 3.6%. The results show that the neural network technique is a powerful one to overcome the flow regime dependency problem of traditional gamma-ray densitometry.