Solution of non-linear forward problems in electrical capacitance tomography using neural networks

A feedforward neural network (NN) is used to solve non-linear forward problems in electrical capacitance tomography (ECT). Data front ECT measurements is used to train and test an NN solution. The NN approach of solving forward problems is based on training a network with pre-known capacitance data corresponding to different permittivity distributions. The network is then used to predict forward problem solutions for various permittivity distributions. The output is also compared to results from linearization methods through iterative image reconstruction.