Gas-liquid Two-phase Flow Pattern Identification Based On ESMD And Support Vector Machine

Gas-liquid two-phase flow widely exists in industries,agriculture and daily life.Many parameters can be used to describe the behaviors of gas-liquid two-phase flow,and measurement of the flow pattern,which is of great scientific and engineering importance for operation monitoring,process controlling and safety assurance of gas-liquid two-phase system.Due to the complexity of two-phase flow,flow pattern identification is a quite difficult problem around the world,which has not been well solved.Hilbert-Huang Transform(HHT)is a nonlinear signal processing method.It can obtain feature parameters which can describe the gas-liquid two-phase flow more effective.Therefore the accuracy flow regime identification can be improved.The iteration number of traditional HHT method is difficult to determine.Traditional HHT method also has other disadvantages like rough trend function and mode mixing.Extreme-point symmetric mode decomposition(ESMD)is proposed by scholar to improve traditional HHT method.ESMD is first applied to flow regime identification.ESMD calculate the optimal iteration number from variance ratio.Then decompose the differential pressure fluctuation signals measured in venturi tube to IMFs.According to the denoising threshold,the denoising signal is obtained by denoising of IMFs.The mean values,skewness,kurtosis and the root amplitude of the denoising signal are obtained as the time-domain features.Meanwhile,ESMD and EMD are used to decompose the denoising pressure signals of gas-liquid two-phase flow into various characteristic scales respectively.DI and HT are used to extract the energy of IMFs as the frequency-domain features.Based on the time-domain features and the frequency-domain features obtained,and combining with the support vector machine technique,the flow pattern identification of gas-liquid two-phase flow is studied.Identification results show that the ESMD can improve the effect of flow regime identification,and the accuracies of three kinds of flow pattern identification are above 90%.