Two Phase Flow Pattern Recognition Based On Ultrasonic And Electrical Multi-sensor Information Fusion

Two phase flow is widely encountered and of high importance in the manufacturing process and related scientific research,which is complex and involves many parameters.The accurate identification of the two-phase flow pattern helps to reveal flow mechanism and improve the accuracy of parameter measurement.The traditional way of studying the flow patterns uses a single type of sensors,to discover and identify the flow patterns,which lacks a comprehensive description of the flow process.While more comprehensive information can be obtained when different principle sensors are used.Horizontal gas-water two-phase flow and oil-water two-phase flow are studied with the Electrical resistance tomography(ERT)system measuring phase distribution information,and the continuous wave ultrasonic Doppler(CWUD)system measuring the flow rate information.Extract features from this two different sensors,separately,and fuse the proposed features to improve the accuracy of flow pattern recognition.The main work of the research includes:(1)Study the static connection and dynamic connection of ERT compression data.The experimental data is compressed and treated as a 16-dimension time-series corresponding to the average response of the fraction distribution in the field of each exciting electrode,which can be studied by graph-based techniques.Three static connection indicators are obtained from the ERT compression data--correlation,coherence and phase lag index.Different gas-water two-phase flows are distinguished by the highly correlated distribution region in the static adjacency matrix.In the dynamic connectivity,plug and slug appears periodically,which reveals the dynamic characteristics of the flow pattern.(2)Study the modular energy distribution of ERT compression data.Based on the analysis of the graph variable signal,combined with the horizontal two-phase fluid distribution characteristics,the electrodes of the ERT system are divided into four modules and the modular Dirichlet energy is calculated.The energy ratio between modules is extracted to characterize the difference of phase fraction distribution in different flow patterns.(3)Identify the flow patterns with the fusion of acoustic and electrical.Make time-frequency analysis of the received signals from the CWUD sensor,and characterize flow patterns by the bandwidth of the energy distribution.In order to further realize the accurate identification of flow patterns,an acoustic-electric multi-sensor information fusion method is proposed.The three-dimensional feature space is formed by the energy ratios between modules and bandwidth,which can effectively distinguish various flow patterns of the gas-water two-phase flow(and oil-water two-phase flow).The above three features are input into the support vector machine,and a high recognition rate is obtained.