Experimental evaluation and modeling of paddy rice drying in convective dryer by artificial neural network

Harvested rice has a high moisture content and it should be reduced to avoid corruption. In this study, the hot air drying was used to evaluate the kinetics of drying. This research was carried out under three variables including hot air velocity ( 0.65 , 0.8 m/s), hot air temperature ( 50 , 65 0C), and final moisture content (11, 13%). Modeling of rice drying was done by a multilayer perceptron artificial neural network. In order to evaluate the performance of training algorithms and transfer functions in predicting the drying behavior of paddy rice, three algorithms including Levenberg Marquardt, Resilient Bach Propagation and Scale Conjugate Gradient and two transfer functions including logsig and tansig were used. The results showed that the maximum drying time was approximately 8 hours at temperature of 500C, hot air velocity of 0.65 m/s and final moisture content of 11% and the minimum drying time was 1.21 hours at temperature of 650C, hot air velocity of 8 m/s and final moisture content of 13%. In general, the drying time decreased with increasing the temperature, hot air velocity and final moisture content. Also, the results of modeling showed that the levenberg Marquardt training algorithm had the best performance compared to the other algorithms. In general the topology of 3-11-1 with levenberg Marquardt training algorithm and logsig transfer function had the lowest mean square error and the highest correlation coefficient.