مرکز منطقه ای اطلاع رساني علوم و فناوري -

چكيده لاتين :

In this research, Artificial Neural Networks (ANNs) used for prediction of Equilibrium Moisture Content (EMC) of three varieties of paddy (Sadri, Tarom and Khazar) as a new method. Feed forward back propagation and cascade forward back propagation networks with Levenberg-Marquardt and Bayesian regularization training algorithms used for training of input patterns. Optimized trained network has the ability ofEMC prediction to test patterns at thermal boundary of 20-40°C and relative humidity boundary of 13.5-87°% with R2 = 0.9929 and mean absolute error 0.0229. Comparison between optimized ANN result and empirical model of Henderson showed that artificial neural network not only can simultaneously predict the EMC of samples of all varieties but also has better coefficient of determination and less mean absolute error.