Convective drying of garlic (Allium sativum L.): Artificial neural networks approach for modeling the drying process

Abstract

In this study, artificial neural networks (ANNs) was utilized for modeling and the prediction of moisture content (MC) of garlic during drying. The application of a multi-layer perceptron (MLP) neural network entitled feed forward back propagation (FFBP) was used. The important parameters such as air drying temperature (50, 60 and 70°C), slice thickness (2, 3 and 4 mm) and time (min) were considered as the input parameters, and moisture content as the output for the artificial neural network. Experimental data obtained from a thin-layer drying process were used testing the network. The optimal topology was 3-25-5-1 with LM algorithm and TANSIG threshold function for layers. With this optimized network, R2 and mean relative error were 0.9923 and 9.67 %, respectively. The MC (or MR) of garlic could be predicted by ANN method, with less mean relative error (MRE) and more determination coefficient compared to the mathematical model (Weibull model).

Keywords

• artificial neural networks
• back propagation
• convective drying
• garlic
• moisture content