基于希尔伯特-黄变换和隐马尔可夫模型的气液两相流流型识别方法  

Applied study on flow regime identification of gas-liquid two-phase flow based on Hilbert-Huang transform and HMM

作　　者：周云龙[1] 张学清[2] 张松林[2] 

机构地区：[1]东北电力大学能源与机械工程学院,吉林132012 [2]东北电力大学自动化工程学院,吉林132012

出　　处：《仪器仪表学报》2009年第7期1512-1517,共6页Chinese Journal of Scientific Instrument

基　　金：吉林省科技发展计划基金(20040513)资助项目

摘　　要：为了研究垂直上升管中的气液两相流的流型,利用自制的多电导探针的测量系统采集了四种典型流型的电导波动信息。由于气液两相流电导波动信号的非平稳特征以及神经网络学习收敛慢等问题,提出了一种基于希尔伯特-黄变换(hilberthuang transform,HHT)和隐马尔可夫模型(hidden markov model,HMM)的两相流流型识别方法。该方法首先将信号经验模态分解(empirical mode decomposition,EMD)后的固有模态函数(IMFs)进行希尔伯特变换得到其幅值能量,并将其作为特征向量,输入到已经训练完毕的各状态HMM中,实现了对气液两相流的流型识别。实验结果表明:该方法能很好的识别垂直管内的4种流型,而且优于BP神经网络,从而为流型识别开辟了一条新的途径。In order to study the flow regimes of gas-liquid two-phase flow in a vertical upward pipe, the conductance fluctuation information of four typical flow regimes was collected by a measuring system with self-made multiple conductivity probes. Owing to the nonstationarity of the conductance fluctuation signals of gas-liquid two-phase flow and the problem of slow convergence of neural network learning, a kind of two-phase flow regime identification method based on Hilbert-Huang Transform （HHT） and Hidden Markov Model （HMM） is put forward. First of all, the original signals are decomposed into a finite number of Intrinsic Mode Functions （IMFs） by means of empirical mode decomposition （EMD）. Secondly, the IMFs are transformed to get the amplitude energy features by means of Hilbert Transform. In the end, the amplitude energy features are regarded as the input characteristic vectors and inputted to all state HMM that has been well trained, then the regime identification of the gasiquid two-phase flow could be performed. Experimental result shows that the method can identify the four regimes of the gas-liquid two-phase flow in upward pipe, and it is superior to BP neural network. Thereby, the method develops a new direction for the flow regime identification.

关 键 词：气液两相流 流型识别 希尔伯特-黄变换 经验模态分解 隐马尔可夫模型 

分 类 号：O359.1[理学—流体力学]