Dynamic neural networks for input-output linearisation

A dynamic recurrent neural network is proposed to model, linearise and control a process plant. In this approach the plant is first modelled using artificial neural networks (ANNs) based on input-output data and then the dynamic neural network model acting as a plant emulator is feedback linearised. The approach outlined in this work uses a novel ANN based structure to feedback linearise the process plant. Once the plant emulator is linearised, standard linear control strategy is used to control the plant emulator. Simulation results using actual industry data reveal the accuracy of the modelling, linearisation and control achieved.