Simulation and model predictive control of the fluid catalytic cracking unit using artificial neural networks

An Artificial Neural Network (ANN) model has been developed for an industrial Fluid Catalytic Cracking Unit (FCCU). ANN design and training are presented. Successful training procedure is proved when the prediction capability of the network is investigated on the testing set of data. The trained ANN model has been subsequently used to implement FCCU control using the Model Predictive Control (MPC) algorithm. Main process variables have been controlled in the presence of typical disturbances. Setpoint tracking and disturbance rejection show good control performance and, associated to important decrease of computation time, reveal incentives of the ANN based MPC approach for industrial implementation.