Neural Network Approaches for Prediction of Drying Kinetics During Drying of Sweet Potato

Abstract

Drying kinetic of sweet potato was investigated considering different drying conditions. The drying experiments were performed at five levels of drying air temperature of 50-90^oC, together with five levels of air flow velocities of 1.5-5.5 m/s, and also three levels of thickness of 0.5-1.2 cm. A predictive model using artificial neural network was proposed in order to obtain on-line predictions of moisture kinetics during drying of Sweet potato. A three-layer network with tangent sigmoid transfer function in hidden layer and linear transfer functions in the output was used. A feedforward networks with two hidden neurons was used. The best fitting with the training data set was obtained with eight neurons in first hidden layer and 4 neurons in second hidden layer, which made possible to predict moisture kinetics (moisture content, drying rate and moisture ratio) with accuracy, at least as good as experimental error, over the whole experimental range. On validation data set, simulation and experimental kinetics test were in good agreement. Comparing the R² (coefficient of determination), MRE and STD_R  using the developed ANN model it was concluded that the neural network could be used for on-line state estimation of drying characteristics and control of drying processes.