Artificial Neural Network Modelling of Steady State Chemical Engineering Systems

This paper presents the development artificial neural network (ANN) models for three steady state chemical engineering systems, which are 1) a crude oil distillation column for use in real time optimisation, 2) physical properties of palm oil components, and 3) pore size determination for membrane characterization. Although studies on ANN applications in chemical engineering in the literature are more concentrated on utilising dynamic models, there has been an increasing trend for diverse application of ANN to model steady state systems. For the crude oil distillation column standard radial basis function (RBF) gave sufficiently accurate predictions. For the physical properties of palm oil components, a multi layer perceptron (MLP) network model was able to give a much better prediction of the density of trilaurin than a thermodynamic correlation that is based on group contribution method. For pore size determination of an asymmetric membrane, stacked network gave slightly better prediction than the more commonly used single MLP network. On the whole, this study shows that there is high potential for various applications of ANN models in chemical engineering.