Void-fraction Measurement Based On XGBoost Of Electrical Capacitance Tomography Pipeline And Three Dimensional Image Reconstruction

Electrical capacitance tomography(ECT)is a kind of visualization technology which transforms the collected datas into images and presents them on the computer screen,that is,the capacitance datas of two-phase flow medium in the measured area of pipeline or device are obtained by capacitance sensor,and the corresponding algorithms are used to calculate the flow medium distribution images.ECT technology has the characteristics of simple design structure,low development cost,fast response speed,non-invasive,wide application range and others.It has been widely used in chemical industry,aerospace,nuclear industry and other fields for many years.As one of the hot research directions of process tomography(PT),it has very good research significance and practical value.In order to improve the accuracy of the capacitance measurement of the detection circuit,based on the working principle of ECT system,24 electrodes were used to establish the image reconstruction model of ECT system.Due to the non-uniformity of the potential field,the finite element method is used to mesh the model,the vertical single electrode excitation mode is used to apply the excitation voltage to each electrode plate of the capacitance sensor.Then,the capacitance measurement datas of block flow,hierarchical flow and ring flow are obtained by the post-processing module and calculate their sensitivities in the sensitive field.The ‘soft field'effect of electrical capacitance tomography system leads to the low accuracy of image reconstruction.Because the distribution of void-fraction is not related to the distribution of dielectric constant,the measurement results are not affected by the ‘soft field',XGBoost algorithm is used to measure the void-fraction distribution in two-phase flow.The relationship between capacitance datas and flow pattern identification is classification,the relationship between capacitance datas and viod-fraction is regression,firstly,the flow pattern recognition is realized by the classification model of capacitance datas,record the capacitance datas and the corresponding flow pattern,then,according to different flow patterns,the regression models of each flow pattern are established to make void-fraction prediction for them,record the datas of void-fraction.Due to the number of independent capacitance measurements obtained from the simulation model is smaller,the accuracy of image reconstruction is difficult to meet the needs of industrial applications.In order to solve this problem,dual tree complex wavelet transform(DTCWT)is used to extract the void-fraction distribution characteristics datas,the unit void-fraction datas and the extracted void-fraction distribution characteristics datas are used as the input of DTCWT void-fraction BP neural network algorithm to reproduce the distribution state of medium in the pipeline,finally,complete the three-dimensional image reconstruction.