Experimental And Theoretical Study On Online Identification And Prediction Of Gas-liquid Two-phase Flow Patterns In Small Channels

Under the background of the rapid development of micro-processing technology and new material technology,micro-equipment is widely used in many fields,such as petroleum,chemical industry,medical treatment and so on because of its fine structure,sensitive reaction,low energy consumption and high conversion efficiency.The characteristics of two-phase flow in a small channel are different from those in a conventional channel.This is because the surface effect dominates in small channels.At the same time,the influence of wall wettability,pipe section shape,viscous force and roughness on gas-liquid two-phase flow becomes important.The operation safety of two-phase flow systems is greatly influenced by the flow characteristics and heat transfer characteristics,and the flow pattern affects the flow characteristics and heat transfer characteristics.Therefore,the identification of flow patterns is very important.The differential pressure fluctuation signal of gas-liquid two-phase flow is determined by the flow condition of the fluid,which contains abundant information and has a very close relationship with the flow pattern.Therefore,the prediction of gas-liquid two-phase flow pattern based on differential pressure fluctuation signal can realize on-line identification and prediction of flow pattern and provide guarantee for efficient and safe operation of industrial equipment.In this paper,the linear statistical analysis of the analog signal of photoelectric sensor and the pressure difference wave signal is carried out to realize the preliminary identification of the gas-liquid two-phase flow pattern of small channel.The chaotic analysis of pressure difference wave signal studies the fluctuation process of two-phase flow from the perspective of nonlinear analysis,which further improves the accuracy of flow pattern identification.The Volterra adaptive short-term prediction model and LSTM cyclic neural network prediction model based on pressure difference fluctuation signal were established to realize the online identification and prediction of gas-liquid two-phase flow pattern of small channel.First,this paper uses the self-built small channel gas-liquid two-phase flow test experimental platform,by adjusting the flow rate of different gas phase and liquid phase,using high-speed camera can take a clear high-speed camera diagram of the flow pattern inthe small channel gas-liquid two-phase flow under the three pipe diameters selected in this experiment,and carry on the preliminary identification of the flow pattern.The analog signal diagram of photoelectric sensor corresponding to the preliminary identification results of the flow pattern of gas-liquid two-phase flow in small channel obtained by high-speed camera is analyzed and identified.Because the analysis of the photoelectric sensor analog signal diagram and the differential pressure fluctuation signal diagram of the test experiment is only the flow pattern analysis identification of the signal fluctuation diagram from the linear angle,it does not involve the analysis of the nonlinear characteristics of the gas-liquid two-phase flow.Because the differential pressure fluctuation signal contains abundant flow information,it is easy to collect in the experiment process and will not have an effect on the two-phase flow state.Through the chaotic nonlinear analysis based on chaos,the differential pressure fluctuation signal is analyzed and the attractor diagram is drawn,and the flow pattern of the small channel gas-liquid two-phase flow is more accurately analyzed and identified.The prediction model of differential pressure fluctuation signal based on Volterra adaptive filter and the prediction model of differential pressure fluctuation signal based on LSTM cyclic neural network are established.Five typical high-speed images of bubble flow,plug flow,elastic flow,dispersion flow and wave flow are obtained by high-speed camera.The advantages of gas-liquid two-phase flow pattern identification based on chaotic attractor diagram are confirmed by comparing photoelectric analog signal with differential pressure fluctuation signal.Through the comparison and analysis of the prediction model of Volterra adaptive filter based on phase space reconstruction and the prediction model of LSTM cycle neural network,it is found that both prediction models can realize the prediction of pressure difference fluctuation signal,while the Volterra adaptive prediction model based on chaotic phase space reconstruction is more accurate to predict the gas-liquid two-phase flow in small channel.The prediction results of pressure differential fluctuation signals of three kinds of pipe diameters under Volterra adaptive prediction model show that the mean square error mm the prediction results of pressure differential fluctuation signals of gas-liquid two-phase flow type of pipe diameteris the smallest and the prediction effect is the most accurate.Of the five flow patterns measured in the experiment,the Volterra adaptive prediction model based on chaotic phase space reconstruction has the least mean square error in predicting pressure difference fluctuation signals of bubble flow patterns,and the prediction mean square error of pressure difference fluctuation signals based on LSTM cyclic neural network prediction model is relatively small,and the prediction results are better.