Modeling of thermal cracking of LPG: Application of artificial neural network in prediction of the main product yields

A three layer perceptron neural network, with back propagation (BP) training algorithm, was developed for modeling of thermal cracking of LPG. The optimum structure of neural network was determined by a trial and error method and different structures were tried. The model investigates the influence of the coil outlet temperature, steam ratio (H2O/LPG), total mass feed rate and composition of feed such as C3H8, C2H6, iC4, and nC4 on the thermal cracking product yields. Good agreement was found between model results and industrial data. A comparison between the results of mathematical model and designed neural networks was also conducted and ANOVA calculation was carried out. Performance of the neural network model was better than mathematical model.