A proposal of a fuzzy-neuro predictive control, tuned by genetic algorithms, with application to the start-up control of a distillation column

This paper presents the development of a fuzzy predictive controller that can be applied to the control of multi-stage processes. Recurrent neural networks are used to identify each stage of the controlled process, and then provide predictions about the stage behaviour. These predictions are used by a set of fuzzy controllers, in order to accomplish several control procedures needed to start-up a distillation column from chemical industry. Genetic algorithms (GA´s) are used to automatically tune the controllers.