A Novel Neural Network and Its Apptication on Two-Phase Flow Electrical Capacitance Tomography

摘要

High quality reconstruction of capacitance tomography data is important to get the quantitative information from the cross sectional images of multiphase pipe flow. In this paper, a neural network method for two-phase electrical capacitance tomography (ECT) is studied. A novel neural network named as multi-avtivation function neural network (MAFNN), in which both orthogonal scaling functions and the corresponding mother wavelets are combined as the nonlinear activation function, is proposed. A simple and effective learning algorithm is presented, which is realized by using only backpropagation learning iterations and recursive least square (RLS) algorithm. The MAFNN is applied to reconstruct the cross sectional images. Experimental results verify that the proposed neural network for two-phase capacitance tomography is effective and accurate.