Image Reconstruction And Flow Regime Identification Of Electrical Resistance Tomography

Electrical Resistance Tomography is a tomography technology, which is based on the different mediums have different conductivities, detects the conductance between the electrodes through the detection circuit, determines the field of media distribution in known objects by reconstructing conductivity distribution according to a certain image algorithm. As an effective means to solve two-phase flow parameter measurement and flow regime identification, ERT has significantly characteristics such as low cost, simple structure, non-invasive, safe and a wide range of applications, and it has become one of main research fields about flow tomography technology. In this paper, image reconstruction and flow regimes identification are studied deeply. The main contents are as follows:The working principle and structure of ERT system are discussed in detail firstly. The forward problem in ERT system is analyzed through the mathematical model of the measured object field built by using the finite element method, the disciplines of distribution of sensitive field were found under the uniform medium, discrete media and different flow regimes that provide the predefined information for image reconstruction and improve the accuracy of image reconstruction.To improve the accuracy and speed of Image reconstruction, an image reconstruction algorithm based on polynomial accelerate is proposed, the mathematical model of the algorithm was given then, and finally the convergence of the algorithm was proved by using spectral analysis.Considered the“soft field”effect and pathological problems in ERT system, based on matrix singular value decomposition theory of the sensitivity field, a weighted conjugate gradient truncated SVD ERT image reconstruction algorithm is proposed, the mathematical model of the algorithm was given then, finally the convergence of the algorithm was implemented.To solve low quality and slow speed problems of image reconstruction, after analyzing the neural network training learning and optimization theory, a neural network based Hopfield optimization algorithm for image reconstruction is proposed; the image reconstruction algorithm based on Hopfield neural network is derived; and the results obtained by least squares are took as input of Hopfield neural network, and optimized by the training study.For the flow pattern identification problem in ERT system, a flow pattern identification method based on HMM model is proposed; the method adopted MMC algorithm for feature extraction, use HMM model to train and identification.