Study Of Image Reconstruction Algorithms For Solid-Liquid Two-Phase Flow In Pipelines Based On Electrical Capacitance Tomography

Electrical Capacitance Tomography serves the purpose of monitoring process parameters by visualizing the media distribution inside a pipeline or vessel.Traditional reconstruction algorithms are ineffective and low in accuracy,and cannot provide a reliable basis for the analysis and control of industrial processes.In this paper,an ECT finite element model and data sets are established and image reconstruction algorithms are investigated for solid-liquid twophase flow in pipelines.The main work and results of this paper are as follows in addressing the problems in reconstructing images by Electrical Capacitance Tomography.1.The image reconstruction algorithm based on compressed sensing and image segmentation is proposed.For the problems of serious artifacts and blurred boundaries in ECT reconstruction images,CS-CNN reconstruction framework is proposed.A compressed sensing model of ECT is established,and a convolutional neural network based on upsampling segmentation method is designed to determine the boundary of medium distribution.The effectiveness and accuracy of the GPSR-CNN and IHT-CNN algorithms proposed in this paper are verified through ECT pipeline two-phase flow simulation experiments.2.The image reconstruction algorithm based on improved residual network is proposed.Aiming at the characteristics of "soft field",a deep residual network is designed to establish the nonlinear mapping relationship between the capacitance vector and the two-phase flow image.The RIR-RepVGG residual structure is proposed,and the connection method of long and short hops is used to improve the performance of the residual network.On the basis of it,a nonlinear mapping network is designed.Through ECT simulation experiments and comparison with Landweber,GPSR and ART algorithm,the results showed that the reconstruction result of the proposed algorithm is the closest to the real two-phase flow distribution,and the stability of the algorithm is guaranteed.3.The ECT experimental platform is developed and static experiments are completed.According to the needs of research and experiments,a 12-electrodes ECT system is designed,which consists of an ECT hardware system and a host computer experimental platform.The effectiveness of the image reconstruction algorithm proposed in this paper is verified through static experiments.In this paper,based on the basic principles of ECT,compressed sensing theory and deep learning methods,the image reconstruction accuracy is improved of ECT by establishing finite element models,self-built data sets,and improved algorithm models.In general,the reconstruction algorithm proposed in this paper has strong accuracy and reliability,and has certain research and application value.