Research On Image Reconstruction And Flow Pattern Identification For Electrical Capacitance Tomography

Electrical capacitance tomography(ECT)is a visualization measurement technique.As one of the most mature electrical tomography techniques,it is widely used for multiphase flow measurement in petroleum,chemical,pharmaceutical,electric power,metallurgy and other industrial processes because of its simple structure,low cost,fast response,non-invasive and good safety.Meanwhile,the flow pattern and reconstructed image are significant parameters reflecting the multiphase flow state in multiphase flow measurement.Therefore,it is important to use electrical capacitance tomography technology for accurately identifying flow patterns and imaging the inside of pipes with high precision and speed.This thesis focuses on the image reconstruction and flow pattern identification of ECT technology,and its main research content can be divided into two directions:1.The influence of capacitance data on the reconstruction quality of ECT images is studied:The differences in the spatial distribution of each electrode pair in the ECT sensor cause the acquired projection data to have a decisive influence on the good or bad ECT imaging results.The use of all acquired capacitance data often does not get ideal reconstruction results,and the presence of redundant data can seriously affect the quality of image reconstruction.In order to improve the imaging quality,this thesis first conducted a test study to examine the effect of valid data on the imaging quality.It is shown that the use of valid data can greatly improve the image reconstruction.Later,the effect of valid data on imaging quality was studied in detail using the Landweber algorithm as an example.Firstly,the causes of poor ECT imaging quality are analyzed from the perspective of negative sensitivity field.Secondly,a data screening method is proposed to reduce the influence of negative sensitivity field on imaging quality.The capacitance data are screened based on the capacitance value and the set threshold.A small volume of capacitance data is used for imaging.Finally,the new method is evaluated by correlation coefficient and computation time.The results show that the new method can not only effectively suppress the semi-convergence of the algorithm,but also improve the image reconstruction quality and speed.2.The ECT flow pattern identification algorithm based on SGAN network is studied:In order to address the problems that the flow pattern recognition algorithm based on ECT technology has low recognition accuracy for complex flow patterns and requires a large amount of label data for learning.A flow pattern identification method based on SGAN(semi-supervised generative adversarial network)and electrical capacitance tomography is proposed.Firstly,the principles of the ECT technique and general GAN are briefly described,and the model parameters,loss function,and training process of the SGAN are explained in detail.Secondly,a data sample set of 11,400 random flow patterns is constructed by co-simulations of COMSOL and MATLAB,and then the SGAN,BP,and SVM network models are trained and validated by the training set in this data set.Finally,static experiments were conducted on the self-developed ECT system,and the identification results of different algorithms were compared and analyzed by modifying the label sample size of the training set.The experimental results show that SGAN maintains a higher average identification accuracy under the training condition where the number of label samples of SGAN is ten times smaller than the other two algorithms.