Research On Image Reconstruction Algorithm For Gas Solid Two Phase Flow Based On ECT

Two phase flow is widespread in key industrial fields such as electric power,environmental protection,chemical industry,metallurgy and nuclear energy,especially in metallurgical and electric power sectors.The parameter detection of two-phase flow is of great significance for the measurement and safety monitoring of production process.Electrical capacitance tomography(ECT)is a process tomography technology based on capacitive sensing mechanism.ECT technology uses the difference of relative permittivity between different substances to collect capacitance signals between electrode arrays through capacitive sensors and acquisition systems.Then,the distribution of dielectric constant in the field is calculated with image reconstruction algorithm,so as to obtain the distribution of multiphase flow medium in the measured section.It has the advantages of fast response,low cost,non-invasive,wide application range,safety and no radiation,especially suitable for measuring the mixed fluid composed of various insulating objects,in which the gas-solid twophase flow detection is a typical application.This thesis mainly focuses on the related problems of gas-solid two-phase flow imaging based on ECT technology.The main work of this thesis is as follows:First,the imaging characteristics,accuracy,speed and other attributes of different types of typical ECT reconstruction algorithms are analyzed.Firstly,the computational models of LBP,Landweber,Tikhonov,Newton Raphson and other traditional imaging algorithms are built respectively.Secondly,the influence of different parameters on imaging algorithms and the adaptability of different algorithms to different flow patterns are analyzed,and the advantages and disadvantages of these four typical electrical capacitance tomography image reconstruction algorithms are analyzed.Secondly,aiming at the problems that the existing common finite element commercial software are foreign software,complicated operation and high learning cost,the finite element metadata structure and finite element solver for ECT sensor modeling are designed and developed.According to the structural characteristics of ECT sensor,the finite element preprocessing module,solver model and post-processing module are designed and developed respectively,which are used to calculate the key information such as the measured capacitance and sensitivity matrix of the sensor under different conditions.The calculation results of the self-developed solver are compared with the commercial software COMSOL to verify the effectiveness of the solver.Thirdly,aiming at the defects of the existing imaging algorithms,an image reconstruction algorithm based on deep long and short-term memory network is proposed.During image reconstruction,this thesis not only considers the size of each capacitance value,but also incorporates the size order and time series of each capacitance value into the image reconstruction algorithm.Then,a double-layer long and short-term memory network(TLSTM)is built,and the construction and training method of GPU image reconstruction network is given to further improve the iterative training speed of T-LSTM.The simulation results show that the T-LSTM image reconstruction algorithm has significantly improved the imaging accuracy and speed.