ARTIFICIAL NEURAL NETWORK MODELING FOR DRYING KINETICS OF PADDY USING A CABINET TRAY DRYER

Abstract

The study of drying kinetics and characteristics of agricultural products is essential for drying time estimation, designing dryers, and optimizing the drying process. Moisture diffusivity under different drying conditions is crucial to process and equipment design. The drying kinetics of paddy using a cabinet tray dryer was modeled using an Artificial Neural Network (ANN) technique. For predicting moisture ratio and drying rate, the Levenberg-Marquardt (LM) training algorithm with TANSIGMOID and TANSIGMOID hidden layer activation function provided superior results. A comparative evaluation of the predicting abilities of ANN and 12 different mathematical drying models was also carried out. The Midilli model was adequate for fitting the experimental data with an R2 comparable to that of the ANN. However, the RMSE observed for ANN (0.0360) was significantly lower than that of the Midilli model (0.1673 to 0.712). Effective moisture diffusivity increased with an increase in temperature from 15.05 10-9 m2 /s to 28.5 10-9 m2 /s. The activation energy for drying paddy grains varied between 6.8 kJ/mol to 7.3 kJ/mol, which showed a moderate energy requirement for moisture diffusion.