Flow Regime Identification Based Multi-Sensor Information Fusion For Gas-liquid Two-phase Flow

Multiphase flow widely exists in nature and industrial production.The flow process of multiphase flow fluctuates frequently,the flow structure is complex and changeable,and multiphase flow has the characteristics of nonlinearity,interphase deformation and frequent energy exchange.As one of the important process parameters characterizing the structure and state of multiphase flow,the accurate identification of flow regime is of great significance.To overcome the influence of flow state fluctuation and complex flow structure,the comprehensive analysis and processing of the multiphase flow process,which includes the local fluctuation characteristic and the global structure relationship,can be used to obtain more abundant flow process state information and realize accurate identification of multiphase flow regime.In the research,the flow process of gas-liquid two-phase flow is taken as the object,and the fluctuation signal of the flow process is obtained by using two non-disturbed sensors with different sensitivity principles.The methods of multi-domain feature preprocessing,comprehensive dimensionality reduction,identification are used to realize the flow regime identification scheme,and the scheme can realize the multi-domain characterization of local fluctuation characteristic and the characterization of the flow regime global structural relationship.The main research contents are as follows:(1)For the problem of low utilization of gas-liquid two-phase flow detection signal fluctuation information and incomplete characterization of local fluctuation characteristic in the flow process,a flow regime identification scheme based on multidomain fluctuation characteristic analysis is proposed.The conductance rings sensor is used to obtain the water holdup data.By comparing the analysis results of five different domain combination methods,the empirical wavelet transform(EWT)method is determined and improved to obtain the fluctuation information of water holdup signal in multi-scale domain.Combined with the fluctuation information of water holdup in time domain and frequency domain,the multi-domain joint characterization and quantification of the local fluctuation characteristic of different flow states are realized.And support vector machine(SVM)is used to identify the flow regime.The analysis results of experimental data show that the average identification rate of the proposed scheme for four typical gas-water two-phase flow regimes is 84%.(2)To further improve the accuracy of flow regime identification and comprehensively consider the global structure relationship of the flow state,a flow regime identification scheme with the flow regime global structure characterization by feature vectors dimensionality reduction is proposed.The scheme realizes the dimensionality reduction of multi-domain high-dimensional feature vectors and the characterization of the flow regime global structure relationship,and further improves the speed and accuracy of flow regime identification.The analysis results of experimental data show that the average identification rate of the proposed scheme for four typical gas-water two-phase flow regimes is improved to 96%.(3)For the influence of gas-liquid two-phase flow on the regime identification when the liquid phase is the mixture of oil and water,by using an ultrasonic Doppler sensor,and the probability density function(PDF)distribution method is used for the measured ultrasonic echo signal to obtain the ultrasonic echo energy distribution characteristic of the gas liquid interface.The short-time Fourier transform(STFT)method is used for the Doppler frequency shift information to obtain the flow velocity frequency shift fluctuation characteristics of discrete phases such as bubbles,air plugs,and oil droplets.Combined with the multi-domain features extracted by the EWT method from the water holdup,in which the water holdup is detected by conductance rings sensor.The flow regime global structure is characterized by multi-domain high-dimensional feature vectors dimensionality reduction processing,and SVM is used to realize the flow regime identification of oil-gas-water three-phase flow.The analysis results of the experimental data show that the average flow regime identification rate of the five gasliquid flow regimes of oil-gas-water three-phase mixed flow is 95%.