Research On ECT Image Reconstruction Base On RBF Neural Networks

Electrical Capacitance Tomography (ECT) technique is a kind of Process Tomography (PT) technique based on electrical capacitance sensors. ECT technique has the advantage of being non-intrusive, simple in structure, fast in response, low in cost, wide in application and good in security (non- radiation). ECT has gained considerable progress in recent years. As a new technology of two-phase flow parameters detection, ECT has great development potential in the industrial application.Image reconstruction algorithm is an important factor to improve the image reconstruction quality in ECT system research. It reconstructions the medium distribution image of the measured area by limited observation data, that is to gain the pixel-gray-scale value of image area. This is a non-linearity, ill-posed inverse problems.In this article, the principle of RBF neural network was introduced. The image reconstruction of 16-electrode ECT system based on RBF neural networks was investigated. RBF neural network is a kind of local approximation neural networks. In theory, it can approximate any continuous function if there is enough neuron. The image reconstruction algorithm based on RBF neural networks uses RBF networks to build the mapping relationship between the capacitance value and the image gray-scale value. In this article, the number of hidden nodes of the RBF neural networks was determined using maximal matrix element method;the centers and the widths of RBF functions were determined using the nearest neighbor-clustering algorithm. Thus the RBF neural networks converting the electrode capacitance measurements to the image gray-scale value was established.The RBF neural networks for image reconstruction were trained in MATLAB environment. The training samples were obtained by using finite element method. Simulation experiment results indicate that the image reconstruction algorithm based on RBF networks can provide images superior to those obtained with the linear back-projection algorithm, with a similar reconstruction time.In order to control ECT system running and view the medium distribution in the pipeline realtime, the software for the 16-electrode ECT system was developed in DELPHI. The functions of human-computer interaction, image reconstructing, communication with lower-computer, and the like, were implemented using DELPHI.