Research On Image Reconstruction Algorithm Of Electrical Capacitance Tomography

As a new measurement technology with great development potential for parameter visualization,Electrical Capacitance Tomography(ECT)has gradually become a research hotspot in industrial production and scientific fields because of its advantages of wide application range,non-invasiveness,simple structure,no radiation,low cost and fast imaging speed.ECT technology is an effective means of real-time parameter detection,so it has broad application prospect s in the field of two-phase flow and multiphase flow.The speed and accuracy of image reconstruction is one of the key factors that determine the performance of ECT system.Based on the 12-electrode ECT imaging system,a series of researches on image reconstruction algorithms are carried out in this paper,aiming at improving the accuracy of image reconstruction.The main work and related achievements of this paper are summarized as follows:(1)Aiming at the problem that the solution method of linear forward problem will lead to reconstruction error in the process of ECT imaging,a solution method of nonlinear ECT forward problem based on Extreme Learning Machine(ELM)is proposed in this paper.The input of the network is the permittivity distribution,and the output is the measured capacitance value.The algorithm is combined with the Landweber algorithm to realize image reconstruction.The samples of the ELM network are randomly generated in the distribution position and size of the objects,making them more representative.Simulation and static experiments show that,compared with the Landweber iterative algorithm with adaptive step size,t he new algorithm has obvious advantages in imaging quality and convergence speed,which verifies the effectiveness of the new algorithm.(2)This paper attempts to apply the Robust Regularized Extreme Learning Machine(RELM)algorithm based on Iterative Reweighted Least Squares(IRLS)to ECT imaging.In the RELM-IRLS network,the role of the Bisquare loss function is to enhance the robustness,the role of the l2-norm regularization term is to avoid overfitting,and the IRLS algorithm is used to optimize the objective function.Five typical samples of random flow pattern distribution in oil/gas two-phase flow are constructed by matlab finite element simulation to complete the training of the network.Simulation and static experiments show that the new algorit hm has quite good imaging effect.This algorithm is superior to the classical ELM network in generalization performance,and has better imaging quality than the Landweber iterative algorithm.(3)A novel ECT image reconstruction algorithm based on an Adaptive Support Driven Bayesian Reweighted(ASDBR)algorithm was proposed.The great advantage of this algorithm is that it can accurately extract the main features of the flow pattern and remove redundant information.This algorithm transforms the original problem into a series of subproblems with iteratively reweighted weights,and solves these subproblems by the Iterative Shrinkage-Thresholding Algorithm(ISTA).Comparisons are made among the ASDBR algorithm,the Landweber iterative algorithm,the Sparse Bayesian Learning(SBL)algorithm,and Lasso.Both simulation and experiment results show that the proposed new method con siderably enhances the quality of the reconstructed image.(4)An novel ECT image reconstruction algorithm based on an Efficient Sparse Bayesian Learning(ESBL)algorithm is presented.This algorithm takes the Gaussian-scale mixture model as the prior distribution of the parameters to increase the flexibility of the model.Then,a surrogate function is used to replace the Gaussian likelihood function to reduce the computational complexity of the algorithm.In order to verify the effectiveness of this algorithm,the Laplace distribution and the Student’s T distribution are used as the prior distribution of the parameters to achieve two specific implementations of this algorithm,and simulation and experiments are carried out.Compared with the SBL algorithm,the Laplace Prior based Bayesian Compress Sensing(LPBCS)algorithm and the Landweber algorithm,the presented EBSL algorithm with the Laplace prior distribution has better image quality and excellent real-time performance.