An internal model control strategy using artificial neural networks for a class of nonlinear systems

The use of an artificial neural network (ANN) in model based control: the internal model control (IMC), both as process model and as a controller is considered in this paper. The neural network is trained with observed input-output data from the system to represent its inverse dynamics. The resulting inverse model neural network can then be used as a controller, typically in a feedforward fashion. The proposed procedure is presented to design a control law for a class of nonlinear systems with separable nonlinearity. An IMC with a neural network controller, in which the linear part of the plant and its inverse are replaced by neural networks, cancels the effects of the nonlinear dynamics and measured disturbances, with satisfying performance. The linear conjecture is so verified for the considered nonlinear system class. Simulation results, for different slopes k of the nonlinearity, show control performance and give limitations of proposed strategy application, beyond which, the neural controller yields unstable behaviour.