Study on infrared drying kinetics of quince slices and modelling of drying process using genetic algorithm-artificial neural networks (GA-ANNs)

In this research infrared radiation was used for drying of quince slices. For this reason, the influence of drying temperature of 50, 60, 70 and 80 °C acquired by 51, 73, 98 and 12 W infrared lamp was investigated. Drying results showed that the drying rate increased with increasing temperature. The drying decreased up to 60% when temperature was increased from 50 to 80 °C. Affected by the lamp power from 51 to 125 W, the moisture content diminished from 453% (d.b.) to 16% (d.b.). Modeling of drying process using genetic algorithm-artificial neural networks (GA-ANNs) with 3 inputs (drying time, drying temperature and the slice enter temperature) and one output (the amount of moisture ratio (MR)) was done. The modeling results demonstrated a network with 7 neurons in hidden layer and tangent hyperbolic transport function could precisely predict the moisture content of slices during drying (R2 = 0.9997 and RMSE = 0.0044). This precision for optimized GA-ANNs was even higher than that of Midilli model -the best empirical model- (R2 = 0.9987-0.9994 and RMSE = 0.0068-0.0098) at all the temperatures tested. The results obtained from the sensitivity analysis by the optimized neural networks revealed that the center temperature of slices was the most pronounced factor (0.0044) to control the MR. Increase in temperature resulted in an increase in the effective diffusivity coefficient, so that this coefficient reached to 26.1×10-9 m2/s at 80 °C from the initial value of 10.8×10-9 m2/s at 50 °C. The activation energy (Ea) calculated for the quince slices were 28.68 kJ/mol.