摘要
对垂直上升气液两相流测得的电导波动信号,在频域中采用语音信号处理中的线性预测方法提取了4个特征量,在时域中用时间序列统计分析方法提取了6个特征量。将这10个反映气液两相流流动特性的特征量作为径向基神经网络软测量模型的输入量,在水相流量1～10 m^3/h及气相流量1～130 m^3/h的范围内,较好地实现了气液两相流持水率预测,为两相流相含率测量提供了一种新的软测量途径。
From conductive wave signals measured in a gas/liquid two-phase flow rising vertically, 4 characteristic indexes are extracted in frequency domain by a linear prediction method of speech signal processing, while 6 statistics are derived in time domain as other characteristic indexes. Application of these 10 characteristic indexes which describe features of the two-phase flow as the inputs of a RBF neural network results in an excellent prediction of water holdup of the two-phase flow within flow rate ranges of water from 1 to 10 m^3/h and of gas from 1 to 130 m^3/ h. This study provides a soft way to measure the phase volume fraction of two-phase flow.
出处
《测井技术》
CAS
CSCD
2007年第5期425-429,共5页
Well Logging Technology
基金
国家自然科学基金项目(No.50674070
60374041)
关键词
生产测井
气液两相流
持水率
电导传感器
特征提取
软测量
production logging, gas/liquid two-phase flow, water holdup, conductance sensor, characteristic index extraction, soft measurement
