Optimal Design of Experiments Based on Artificial Neural Network Classifiers for Fast Kinetic Model Recognition

Developing mathematical models for the description of reaction kinetics is fundamental for process design, control and optimisation. The problem of model discrimination among a set of candidate models is not trivial, and recently a new and complementary approach based on artificial neural networks (ANNs) for kinetic model recognition was proposed. This paper extends the ANNs-based model identification approach by defining an optimal design of experiment procedure, whose performance is assessed through a simulated case study. The proposed design of experiments method allows to reduce the number of experiments to be conducted while increasing the ability of the artificial neural network in recognising the proper kinetic model structure.