Modelling of dried apple rehydration indices using ANN

The purpose of the research was to study the effect of different drying and rehydration conditions on the rehydration indices of apple and to model the rehydration indices of apple using artificial neural networks. The research involved the examination of the rehydration process of 10 mm apple cubes, which were dried in natural convection (drying air velocity), forced convection and fluidization at the following drying temperatures 50, 60 and 70°C. The process of rehydration was conducted in distilled water at the following temperatures 20, 45, 70 and 95°C. Five rehydration indices were used to express rehydration. Artificial neural networks (MLP 3-5-1 and MLP 3-4-1) were used to make the rehydration indices dependent on both drying and rehydration parameters: drying temperatures, v and following temperatures. Five statistical tools, i.e. correlation coefficient, mean bias error, root mean square error, reduced chi-square, and t-statistic method (t-stat), were applied to determine the fit. To identify critical parameters and their impact on the ANN outputs, a sensitivity analysis (backward stepwise method) was performed.