Study Of ECT Image Reconstruction Based On Elman Neural Network

The main principle of Electrical Capacitance Tomography(ECT)is that the data acquired by the data acquisition system is transformed into an image and then presented on the computer screen,i.e.,the dielectric constant of the two-phase or multiphase flow medium inside the pipe under test is acquired by a capacitive sensor unit and then calculated using an appropriate image reconstruction algorithm to derive The corresponding two-phase or multiphase flow distribution image is then calculated using appropriate image reconstruction algorithms.At present,ECT systems have the advantages of simple structure,low cost and non-invasive,and are widely used in practical industrial fields such as petrochemical and natural gas pipeline transportation.However,there are still some problems in ECT technology.In this paper,the reconstruction algorithm of ECT technology is investigated,and the main work is as follows.First,this paper analyzes the technical characteristics and composition principles of the capacitance laminar imaging system in depth,analyzes the working principle of the ECT system theoretically,models and analyzes the system using ANSYS finite element simulation software,simulates the two-phase flow of natural gas and water,and writes programs using ANSYS and Matlab software respectively to obtain the sensor model given the relatively optimal structural parameters,calculates The simulated experimental data are used to obtain the corresponding capacitance values of the flow pattern distribution as the experimental data in the subsequent image reconstruction process.Secondly,for the problems of low reconstruction accuracy and slow reconstruction speed of classical image reconstruction algorithms Tikhonov and Landweber,a new image reconstruction algorithm based on improved Elman neural network is proposed,and Adam’s algorithm is introduced for calculating the error value in the backpropagation process,which effectively improves the problem that Elman neural network is easy to fall into the local optimal solution,and The number of neural units in the hidden layer of Elman network is determined by using the method of variable structure dynamics to increase the performance of the network and improve the convergence rate of the network.Through the results,it is shown that the algorithm is simple,fast and of high imaging quality.Finally,an Elman-Xg Boost classifier based on the concept of integrated learning is proposed for the problem of limited data as samples of the simulated flow pattern to obtain capacitance values.The Ada Boost algorithm and Xg Boost algorithm in the integrated learning method are not ideal in the classification accuracy of limited capacitance data samples,so a fusion of Elman-Xg Boost and Elman-Ada Boost classifier is constructed,and the optimal Elman-Xg Boost classifier is selected according to the classification results,and the capacitance sample data classified by the classifier The capacitance samples are used as the input of the Elman neural network for the image reconstruction process,and the results show that the reconstruction results are more satisfactory for the sample dataset processed by the classifier.