Study On Contactless Impedance Tomography Technique And Image Reconstruction Algorithm

Gas-liquid two-phase flow exists widely in daily life and industrial production.Due to its complexity,it is difficult to monitor the characteristics and measure the parameters of gas-liquid two-phase flow.At present,Capacitively Coupled Electrical Impedance Tomography(CCEIT)is a new technique to dynamically monitor and measure gas-liquid two-phase flow.It has the advantages of non-contact measurement and using the real part information and the imaginary part information of gas-liquid two-phase flow simultaneously to implement the parameter measurement and image reconstruction.However,the research works on CCEIT technique are still not sufficient.On the one hand,its hardware system still needs to be improved.The current CCEIT technique does not consider the shielding design of the sensor.On the other hand,the existing image reconstruction algorithms of gas-liquid two-phase flow are insufficient,so it is urgent to put forward new image reconstruction algorithm that can fully exploit the potential of CCEIT technique.This dissertation aims at the insufficiencies of CCEIT technique and carries out related research,including:1)study of three different sensor configurations,and the improvement of the hardware sy stem of CCEIT technique.2)introducing the unsupervised clustering algorithm into the image reconstruction of gas-liquid two-phase flow,and proposing two new image reconstruction methods to implement image reconstruction of gas-liquid two-phase flow.The main work and innovations of this dissertation are as follows:1.Three different sensor configurations are compared and analyzed.By introducing external shield and radial screens into CCEIT sensor design,the circuit models,boundary conditions and sensitivity field distributions under three sensor configurations are studied.Based on the research results,the hardware system of CCEIT technnique is improved and a 12-electrode contactless impedance tomography(CIT)system prototype is developed.2.Based on linear back projection(LBP)algorithm and Gaussian mixture model(GMM)clustering method,an image reconstruction algorithm for CIT technique,i.e.,LBP+GMM algorithm,is proposed.By the 12-electrode CIT system prototype,the image reconstruction experiment is carried out.Experimental results show that the proposed LBP+GMM algorithm is effective.Compared with other conventional image reconstruction algorithms,LBP+GMM algorithm can obtain higher quality reconstructed images with less prior knowledge and manual intervention.The relative image error is no more than 7.9%.3.Based on LBP algorithm and density based spatial clustering of applications with noise(DBSCAN)algorithm,this dissertation proposes a new image reconstruction algorithm,i.e.,LBP+DBSCAN algorithm.The image reconstruction experiment is carried out by the 12-electrode CIT system prototype.Experimental results show that the proposed LBP+DBSCAN algorithm is feasible.Compared with other image reconstruction algorithms,LBP+DBSCAN algorithm requires less prior knowledge and pre-set empirical parameters,and avoids the influence of manual intervention on image reconstruction.Its image reconstruction quality is higher,and the relative image error is less than 6.1%.4.To make full use of the impedance information of gas-liquid two-phase flow,this dissertation proposes an image fusion strategy based on the LBP+DBSCAN image reconstruction algorithm.Using the mean square error(MSE)image quality index,the image of the real part obtained by LBP+DBSCAN algorithm and the image of the imaginary part obtained by LBP+DBSCAN algorithm are fused by weighted image fusion.Image fusion experiments show that the proposed image fusion strategy is feasible,the quality of fusion image has been further improved.Compared with other image reconstruction algorithms,the relative image error of fusion image is less than 4.1%.