Infrared drying kinetics study of lime slices using hybrid GMDH-neural networks

Modeling of drying kinetics is one of the most appropriate methods to control the time or any condition related to the drying process. In this research, the drying process of Persian lime slices with 5 and 10 mm thicknesses was conducted at four temperature levels of 100, 125, 150 and 175 °C in an experimental infrared dryer. For modeling of the drying kinetics of lime slices, 7 well-known thin-layer models were used. In addition, for this modeling, a hybrid GMDH-neural network with four layers (one input, two hidden and one output layer) was applied. The results demonstrated that the optimized GMDH model was completely efficient to predict the moisture content of lime slices during infrared drying process (R2 = 0.9909). This efficiency was almost similar to the observed efficiency for the Page model as the best mathematical model used (R2 = 0.9793-0.9950). It was found that the sensitivity to the drying time was more than other inputs, so that sensitivity of this parameter was near 45%. Increase in temperature resulted in an increase in the effective diffusivity coefficient (Deff), so that this coefficient reached to 2.76×10-9 at 175 °C from the initial value of 9.90×10-10 at 100 °C. The activation energy (Ea) calculated for the lime slices was 87.61 kJ/mol.