Temperature Control of Reactors Using Neural Networks

Abstract

The automation and control of industrial batch and semibatch processes, like chemical reactors and high pressure extraction vessels, are complex tasks that do not allow the use of standard project procedures. This work presents a predictive controller based on a feedforward neural network and on a quadratic performance criteria. This controller was able to compensate for non-linearity and for steady-state absence, usual characteristics of batch processes. Experiments were executed in a pilot unit where a multivariable temperature control of a jacketed semibatch reactor was implemented. The results showed that the proposed controller is able to maintain the reactor in the desired conditions during the whole operation cycle.