Research On Improving Reconstruction Accuracy Of ERT Based On Combination Algorithm

Electrical Resistance Tomography(ERT)is one of the four basic forms of Electrical Resistance Tomography,and also an important branch of process Tomography.In this technology,the sensitive field is formed by exciting electrode array,and the projection data is obtained through the data acquisition system,and the distribution of the medium in the field is reduced by the appropriate image reconstruction algorithm.With the rapid development of science and technology,resistance tomography technology is constantly improving and improving.With its outstanding advantages of non-damage,real-time observation and online data acquisition,it is widely used in engineering science,seismic monitoring,pollution detection,medical diagnosis and other fields.In actual production and life,the successful application of ERT depends on the high performance of image reconstruction algorithm.However,the current traditional image reconstruction algorithms,whether iterative or non-iterative,have the advantages of small computation and fast convergence,but do not consider the "soft field" error,so the quality of the reconstructed image still has a large space for improvement.Aiming at the above problems,this paper proposes a combined algorithm that combines traditional image reconstruction algorithm with intelligent optimization algorithm to improve the reconstruction accuracy of ERT image.In this paper,the finite element model of the two-dimensional ERT sensor is established by using the finite element simulation software COMSOL.The basic principles of four classical image reconstruction algorithms,namely LBP,Landweber,Tikhonov regularization and CG conjugate algorithm,are analyzed in depth.The reconstruction results of the four algorithms and the evaluation error index are compared.The simulation results show that the reconstruction effect of Landweber is better,and it can be used as the input initial value of the intelligent optimization algorithm.In addition,this paper analyzes and explains two basic intelligent optimization algorithms,namely the standard particle swarm optimization algorithm and the crowd algorithm.On this basis,two improved algorithms are proposed,namely the double particle swarm optimization algorithm and the particle swarm and crowd collaborative optimization algorithm.In this paper,initial position values of the optimization algorithm are set as Landweber iteration results,and the fusion of traditional algorithm and intelligent optimization algorithm is realized.An optimization objective function considering the "soft error" was constructed,and the improved intelligent optimization algorithm was used to optimize the location and improve the reconstruction results.The normalized voltage deviation in the optimized objective function is predicted by machine learning.This paper introduces two machine learning models,namely least squares support vector machine and random forest.The principle of the model,the construction of data set,the learning and training of the model and the process of predictive regression are described.The prediction performance of the two models is compared.Four combined ERT image reconstruction algorithms are formed by combining the two optimization algorithms and the two prediction models.The reconstruction results based on simulation data and measured data show that,compared with Landweber,the reconstruction quality of the four groups of combined algorithms is significantly improved,and the combined algorithm of particle swarm optimization and crowd optimization under random forest prediction is the most obvious to improve the reconstruction effect.