The histories of differential pressure fluctuations and their Fast Fourier Transform spectrum have close relation with the flow regimes.Unfortunately,each type of flow regime is very difficult or impossible to be dist...The histories of differential pressure fluctuations and their Fast Fourier Transform spectrum have close relation with the flow regimes.Unfortunately,each type of flow regime is very difficult or impossible to be distinguished from the other on the basis of the fluctuations or the spectrum.The present paper provides a feasible solution, which the gas-liquid two-phase flow regimes can be recognized automatically and objectively on the basis of the combination of the Counter Propagation Network (CPN) and the FFT spectrum of the differential pressure fluctuations. The CPN takes advantages of simpler algorithm and fast training processes.Furthermore,it does not require a great deal of samples.The recognition possibility is determined by the clustering results of the Kohonen layer in the CPN.With the presented test cases,the possibility can be higher than 90 percent for different liquid phase velocity.展开更多
传统的流型识别方法对流型的特征缺乏一个量化的评价标准,只能由识别者采用模糊的语言来描述特征,在很大程度上依赖于识别者的主观判断。为了克服传统流型识别方法的缺点,对水平管内气液两相流进行测量,得到了反映两相流波动特性的压差...传统的流型识别方法对流型的特征缺乏一个量化的评价标准,只能由识别者采用模糊的语言来描述特征,在很大程度上依赖于识别者的主观判断。为了克服传统流型识别方法的缺点,对水平管内气液两相流进行测量,得到了反映两相流波动特性的压差波动信号,采用 Hurst 分析描述了水平管内气液两相流不同流型的压差波动特征,发现压差波动信号中存在着不同程度的周期成分。通过对不同流型的压差波动信号的 Hurst 指数 H 进行计算,发现不同流型的 H 值有很大差别,可根据 Hurst 指数 H 值的大小来识别流型,从而为流型识别提供了一种有效方法。展开更多
文摘The histories of differential pressure fluctuations and their Fast Fourier Transform spectrum have close relation with the flow regimes.Unfortunately,each type of flow regime is very difficult or impossible to be distinguished from the other on the basis of the fluctuations or the spectrum.The present paper provides a feasible solution, which the gas-liquid two-phase flow regimes can be recognized automatically and objectively on the basis of the combination of the Counter Propagation Network (CPN) and the FFT spectrum of the differential pressure fluctuations. The CPN takes advantages of simpler algorithm and fast training processes.Furthermore,it does not require a great deal of samples.The recognition possibility is determined by the clustering results of the Kohonen layer in the CPN.With the presented test cases,the possibility can be higher than 90 percent for different liquid phase velocity.
文摘传统的流型识别方法对流型的特征缺乏一个量化的评价标准,只能由识别者采用模糊的语言来描述特征,在很大程度上依赖于识别者的主观判断。为了克服传统流型识别方法的缺点,对水平管内气液两相流进行测量,得到了反映两相流波动特性的压差波动信号,采用 Hurst 分析描述了水平管内气液两相流不同流型的压差波动特征,发现压差波动信号中存在着不同程度的周期成分。通过对不同流型的压差波动信号的 Hurst 指数 H 进行计算,发现不同流型的 H 值有很大差别,可根据 Hurst 指数 H 值的大小来识别流型,从而为流型识别提供了一种有效方法。
