Modelling of corn kernel pre-treatment, drying and processing for ethanol production using artificial neural networks

Abstract

Two artificial neural network (ANN) models were developed to predict the main quality parameters of the corn used for ethanol production. Five Croatian corn hybrids were evaluated in this study (introducing the hybrid type as the first input categorical variable for ANN modelling), grown during three vegetation periods (the second categorical variable), under two levels of agrotechnology (the third categorical variable), dried at four temperatures (the fourth input variable), using two different heating and pressure pre-treatments of corn kernels (the fifth variable for ANN calculation) in order to improve the properties of corn for ethanol production. The first model (ANN1) was used to predict the hectolitre weight, 1000-kernels weight, the gelatinisation rate, and the contents of: glucose, reducing sugars and ethanol during the drying process, according to the type of the corn hybrid and the drying temperature. The ANN2 model was developed to predict the corn weight and moisture during the process, based on the input parameters. The artificial neural network models gave a good fit to experimental data and were able to predict the output variables successfully, showing a reasonably good predictive capability (overall r2 for the corn kernel weight and moisture was 0.989, while r2 for other outputs was 0.856). On the basis of a developed ANN models, multi-objective optimization was performed showing the possible practical use in the corn kernel drying process.