Design of the Neuro-Adaptive Control of Nonlinear Slow Processes

Abstract

A new control design approach based on the Internal Model Control (IMC) structure is presented. This approach consists of three steps. The first step of it is related to the tuning of the Proportional plus Derivative (PD) controller in an internal loop of the controlled process. We considered some typical cases of linear plant models and presented the rules for tuning the parameters of the PD controller in the internal loop. Such internal loop is further on treated as an equivalent plant for which one has to decide whether to use the linear or nonlinear model and its corresponding implementation. As a second step, the choice and implementation of the internal loop model is considered. A zero-steady state error in cases of the piecewise constant changes of the reference and disturbance at output of the plant is achieved thanks to the embedding integral action into the IMC structure. The tuning of parameters of the embedded integral control law is the third step of the whole design approach. In the case of strong and variable process nonlinearity, despite the improving the overall dynamic characteristics of the obtained internal loop closed with the PD controller, in this paper we use the Fast Clustered Radial Basis Function Network (FCRBFN) equipped by the Stochastic Gradient Descent (SGD) learning algorithm to implement internal model of the equivalent plant. Adaptive and robustness characteristics of the proposed control design approach are provided by on line learning of the varying process dynamics model and the used IMC structure. We illustrate the performance of the proposed control design approach by presenting the simulation results for the control of a double tank system.