A Direct Adaptive Controller for a Class of Nonlinear Processes Using Neural Networks

This paper proposes a direct adaptive control strategy for a class of nonlinear systems for which the dynamics is incompletely known and time varying. The nonlinear controller design is based on the input-output linearizing technique. The only information required about the process is the measurements of the state variables and its relative degree. Unknown controller functions are approximated using neural networks. The form of the controller and the adaptation laws for the neural controller are derived from a Lyapunov analysis of stability. Under certain conditions the state vector remains bounded and the plant output tracks with user specified dynamics the output of a linear reference model. The technique is applied to a nonlinear and time varying biotechnological process model. Computer simulations are included to demonstrate the performances of this controller by comparison to an exactly linearizing non-adaptive controller.