Study On Identification Of Gas-liquid Two-phase Flow Patterns Based On The Complex Network Theory In The Double-contact Absorber

Gas-liquid two-phase flow in the double-contact absorber is widely used in energy engineering field with the advantages of large liquid holdup,large contact area,high gas-liquid two-phase flow rate and strong mixing.Flow patterns have a significant impact not only on the gas-liquid two-phase flow characteristics,but also on heat and mass transfer characteristics.Therefore,study on identification of gas-liquid two-phase flow patterns has important academic value and is significant for the practical project.Firstly,flow simulation experiment is carried out on the double-contact gas-liquid two-phase flow tower experiment platform,and the differential pressure signals of gas-liquid two-phase flow in 80 groups of flow conditions are collected by differential pressure transmitter.The differential pressure time series of five typical flow regimes are introduced and analyzed,which contain drum bubbly flow,panting flow,liquid-column flow,disorder flow and whirlpool flow.In the 80 groups of flow conditions,the pressure differential time series are pre-processed and feature extracted,and the normalized flow pattern pressure differential time series characteristic vectors Tp’ are obtained,which can be used to build a complex network of fluid.Secondly,the image signals of gas-liquid two-phase flow are collected by using high speed camera on the double-contact absorption tower.In the 80 groups of flow conditions,the obtained image signals are pre-processed,gray processed and drawn the gray histogram.The statistical feature gray mean value of gray histogram is selected to construct the image gray mean time series.The image gray mean time series are feature extracted,and the normalized flow pattern image gray mean time series characteristic vectors Ti’ are obtained,which can be used to build a complex network of fluid.Finally,the flow pattern complex networks are constructed based on the pressure signals,in which the flow pressure difference time series of 80 sets of flow conditions are taken as network nodes,and the characteristics correlation strength of pressure time series are taken as edges.The flow pattern complex networks are constructed based on the image signals,in which the flow image gray mean time series of 80 sets of flow conditions are taken as network nodes,and the characteristics correlation strength of image gray mean time series are taken as edges.Based on the AP clustering algorithm,the community structure of two kinds of complex networks are analyzed,and three distinct community structures are obtained.It is found that the corresponding relationship between the structure and flow pattern.The results obtained from the two methods are analyzed and compared,and the results are basically consistent.Therefore,the flow pattern recognition is basically achieved for drum bubbly flow,panting flow,liquid-column flow,disorder flow and whirlpool flow.