基于电导波动信号的EMD分解和RBF神经网络的气液两相流流型识别方法  

A Method for Identifying Gas-liquid Two Phase Flow Regimes Based on EMD Decomposition Using Conductance Fluctuation Signals and RBF Neural Network

作　　者：周云龙[1] 张学清[2] 张南[2] 张艳艳[3] 

机构地区：[1]东北电力大学能源与机械工程学院,吉林吉林132012 [2]东北电力大学自动化工程学院,吉林吉林132012 [3]莱芜钢铁集团有限公司自动化部,山东莱芜271104

出　　处：《化工自动化及仪表》2009年第1期52-56,共5页Control and Instruments in Chemical Industry

基　　金：吉林省科技发展计划项目(20040513)

摘　　要：针对电导波动信号具有非线性、非平稳的特征,提出一种应用经验模态分解(EMD)和RBF神经网络进行流型识别的新方法。首先对电导波动信号进行EMD分解,提取出IMF能量参数特征,然后将其输入到RBF神经网络,从而实现了对流型的识别。研究结果表明:该方法能够准确地识别出泡状流、弹状流、塞状流和混状流四种流型,且具有较好的识别效果。该方法为流型识别提开辟了新的方向。According to the non-linear and non-stationary characteristics of conductance fluctuation signals, a flow regime identification method based on empirical mode decomposition （EMD） and RBF neural network was presented. First of all,the conductance fluctuation signals were decomposed to gain energy parameter features of intrinsic mode functions （IMF） by means of empirical mode decomposition （EMD）. Secondly, energy parameter features of IMF could be served as input parameters of RBF neural networks to identify flow regimes. The results show that the method can identify bubble flow, slug flow, plug flow and mixed flow regimes successfully and take on better identification effect. The method develops a new direction for the flow regime identification

关 键 词：气液两相流 流型识别 电导探针 EMD分解 RBF神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] O359[自动化与计算机技术—计算机科学与技术]