RAM-based neural networks for image reconstruction in process tomography

The paper describes preliminary investigations into the application of RAM-based neural networks to image reconstruction for tomographic systems. Amenability to hardware implementation and the trivial mathematics involved in recall suggest that the RAM-based approach may allow for high speed reconstruction of images at a fraction of the cost of traditional reconstruction methods. Simulated data for a 12 electrode capacitance tomography system, with a 2-phase flow regime, have been generated using finite element modelling. Through extensive software simulations, image flows have been reconstructed from capacitance measurements. Results for two flow regimes (stratified and bubble) are presented. Careful selection of the training patterns and network parameters reveals that high fidelity images can be reconstructed