Neural network modeling of mango (mangiféra indica l.) slice drying

Mango is a seasonal fruit subject to significant post-harvest losses due to its high water content. Drying is an effective method of reducing water content. It prolongs the shelf life of dried mangoes and can be carried out naturally or artificially. The main objective of the study is to develop a mathematical model capable of predicting the drying kinetics of mango slices in hot air, using artificial neural networks (ANN).To achieve this, machine learning algorithms were used to analyze the drying data and create a predictive model. The parameters studied include slice thickness, temperature, drying time, initial moisture content and mango Brix level. The optimal neural network identified is of 5-6-1 architecture. Results with it give high coefficients of determination (R²) and low root mean square errors (RMSE), indicating good agreement between predicted values and experimental data. The R² values for the training, test and validation sets are 0.9827, 0.9885 and 0.9836 respectively, with an RMSE of 0.004, demonstrating the effectiveness of the RNA model in predicting the drying process.