Neural networks for process control : application to the temperature control of batch chemical reactors
Abstract
The purpose of this chapter is to review the main applications of neural networks in chemical engineering and the fundamentals of feed forward multilayered neural networks. This chapter presents the various problems for controlling batch reactors, pilot-plant reactor, the dynamic simulation model, and the advanced control algorithm. It examines different questions encountered for the choice of the learning database and the neural network architecture. It discusses the reasons why the learning phase of a neural controller is not an easy task. This chapter tackles two specific methods for the design of the neural controller: learning from another command law and learning from a neural inverse model of the process. These two methods are illustrated by the dynamic simulation model of the reactor and/or by the pilot-plant reactor. This chapter also provides several experiments that evaluate the performances of these neural controllers and compare the performances when an advanced control algorithm is used.