Adaptive neural model predictive control of chemical process: an empirical study

Gives an empirical study in controlling a typical chemical process: distillation column, where the equations governing the system are unknown. The neural networks are online trained to model the process at various operating points and then employed as nonlinear predictors for use in model predictive control. Explicit control laws are derived by using Clark´s GPC performance index and linearization technique. The experimental results show that the proposed neural control strategies have good practical potential for processes control