Kinetic modeling of artemisinin supercritical extraction using artificial neural networks

Abstract

Artemisinin is the major compound synthesized from Artemísia annua L. It is of great interest for the pharmaceutical, cosmetic and/or food industries. Several research groups have been studying mathematical models to describe the process's kinetic behavior. Thus, this work aimed to develop an artificial neural network to model the kinetics of supercritical extraction of artemisinin. Eight experiments at different operational conditions were used as dataset. Two strategies were used to train the network. The first used as input variables: pressure, temperature, solvent flow, and extracted mass at times t and t-1. The output variable was the extracted mass at the time t+1. The second strategy used ‘time’ as input - to substitute the variables mass of extract at times t and t-1 used in the first strategy, and the output variable was the mass of extract at time t. The best network was one structure from the second strategy, with 7 neurons in the 1^st intermediate layer and 1 neuron in the 2^nd intermediate layer (structure 4-7-1-1), which was able to predict and describe the kinetic profile accurately when compared to the experimental data.