CHARACTERIZATION OF THE BEHAVIOR AND PRODUCT DISTRIBUTION IN FLUID CATALYTIC CRACKING USING NEURAL NETWORKS

This paper demonstrates how neural networks can be used as an operational support tool for industrial fluid catalytic cracking (FCC) units by providing an insight into fluid bed behaviour and predicting a complete picture of product distribution. The fluidised bed behaviour is represented by a hydraulic model expressed in terms of the dense phase average bed density in both reactor and regenerator, vapour entrainment in the regenerator dilute phase, dilute phase bed density in the regenerator and dilute phase hold-up. A separate reaction model is used to predict the product distribution in terms of kerosene, naphtha, gasoline, residue, coke, gas and liquid hydrocarbons. The advantages of using neural networks to develop an optimal operating strategy are discussed.