An artificial neural network (ANN) model for determining the steady-state behaviour of an industrial Fluid Catalytic Cracking (FCC) unit is presented in this paper. Industrial data from a Creek petroleum refinery were used to develop, train and check the model. FCC is one of the most important oil refinery processes. Due to its complexity the modelling of the FCC poses a great challenge. The proposed model is capable of predicting the volume percent of conversion based on six input variables. This work is focused on determining the optimum architecture of the ANN, in order to gain good generalization properties. The results show that the ANN is able to accurately predict the measured data. The prediction errors in both training and validation data sets are almost the same, indicating the capabilities of the model to accurately generalize when presented with unseen data. The neural model developed is also compared to an existing non-linear statistical model. The comparison shows that the neural model is superior to the statistical model.