Dynamic modelling and control of heaters using neural network computation techniques

Abstract

Classical approaches to modelling of nonlinear process systems such as the Volterra series method and the Hammerstein model have proven to be cumbersome due to the large number of parameters to be used in the models. In contrast, artificial neural networks offer a promising alternative for modelling nonlinear systems. Although typical neural networks also contain numerous number of parameters which are characterised by the neuron connections, the internal structure of neural networks provide a convenient method to organise and to determine the values of the connections. The internal structure of a neural network is considered to include the choice of the number of neuron layers in the network as well as the number of neurons in each layer, and also the form of the internal neuron activation functions.