Moisture content prediction of paddy drying in a fluidized-bed drier with a vortex flow generator using an artificial neural network

This research presents a numerical and experimental study to improve the drying process of paddy in a rectangular fluidized-bed dryer by applying the principle of vortex flow creation. Paddy drying process was compared in three different drying chamber configurations: a smooth surface chamber, a chamber with upstream pointing inclined baffles and chamber with downstream pointing inclined baffles. For each case study, two inlets hot-air temperatures (60 degrees C and 80 degrees C) and two air flow velocities (2.24 +/- 0.02 and 2.52 +/- 0.02 m/s, at about 1.6 and 1.8 times the minimum fluidized-bed velocity, respectively) within 5 h of drying time were investigated. The Rapid-Miner Studio 7 software was used to design an optimal multi-layered, feed-forward, artificial neural network (MLFF-ANN) model for predicting the moisture ratio of paddy during the drying process. The structure of the MLFF-ANN model with different numbers of hidden layers, neuron node numbers in the hidden layer, momentum coefficients and training epoch numbers were investigated. The results indicated that inclined baffles had significant effects on the flow behavior and the drying rate. A fluidized-bed with baffles reduced the drying time by about 7-18% compared with the smooth surface fluidized-bed at a moisture content of 13%(w.d.). The best performing MLFF-ANN model consisted of four layers, with the number of neuron nodes in each layer being 3, 2, 2 and 1, respectively, at a training epoch number of 1500 and a momentum coefficient of 0.4. The prediction results had a regression coefficient of determination (R-2) of 0.99556, a mean squared error (MSE) of 1.988 x 10(-4) and a mean absolute error (MAE) of 0.00127.