Font Size: a A A

Study On Electrical Resistance Tomography

Posted on:2008-12-23Degree:MasterType:Thesis
Country:ChinaCandidate:Y J ShiFull Text:PDF
GTID:2178360212495342Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
Two-phase flow widely exists in industry. The research on two-phase flow is developing rapidly and has been one of the frontier subjects concerned in the world. However, two-phase flow measurement is difficult owing to its complexity. In many cases, its measurement technique is the bottleneck of the development of research on two-phase flow.Electrical resistance tomography (ERT) is a kind of Process Tomography (PT) technology and has the advantages of low cost, simplicity, non-invasion, wide applicability, etc. ERT has become one of important techniques for two-phase flow measurement. Unfortunately, ERT technique for two-phase flow is still on the stage of laboratory research, though many research works have been reported. There are some problems which should be solved in ERT.The main works are listed as follows:(1)On the basis of flow regime of two-phase flow, direct problem of ERT (forward problem) is solved. Direct problem of ERT is to find the projection data which contains the flow pattern information. Finite-element method is applied in finite-element analysis of electrostatic field's distribution of imaging area is simulated.(2)Inverse problem of ERT, on the other hand, is imaging reconstruction derive from known projection data. There are two varieties imaging reconstruction's algorithm in the dissertation: Improving BP algorithm and WNN algorithm. In view of the BP algorithm insufficiency, Improving BP imaging reconstruction algorithm trains neural network applied Improving BP algorithm, the trained neural network is saved as known network and uses the known network to reconstruct image. WNN imaging reconstruction algorithm, the pretreating layer reduces the dimension of the input sample, the hidinglayer neural with the frame wavelet function, the output layer with the Sigmoid function, trains neural network applied algorithm, the trained neural network is saved as known network and uses the known network to reconstruct image.
Keywords/Search Tags:Two-phase Flow, Electrical Resistance Tomography, Finite Element Method, Neural Networks, Imaging Reconstruction
PDF Full Text Request
Related items