3-COMPONENT TOMOGRAPHIC FLOW IMAGING USING ARTIFICIAL NEURAL-NETWORK RECONSTRUCTION

An introduction to electrical capacitance tomography and a brief backg round of multi-modal tomography is given. The motivation for a neuroco mputing solution to the inverse problem of image reconstruction is dis cussed together with a brief overview of previous work in this field. The techniques developed form the basis for training a single artifici al neural network to perform three-component flow imaging using simula ted data. A;dedicated thresholding function is demonstrated to accommo date three distinct and stable regions for gas, oil and water componen t estimation. Preliminary results indicate the feasibility of this neu ral solution for three-component imaging. Noise performance of the thr ee-component reconstructor is also analysed, and the image reconstruct ion for test patterns and noise performance of this network are illust rated. (C) 1997 Elsevier Science Ltd. All rights reserved.