Comparison between artificial neural network (multi-layer perceptron) and mathematical Peleg's model for moisture content estimation of dried potato cubes

In this study an artificial neural network, multi-layer perceptron (MLP), and Peleg's mathematical model were used to find the best model for the prediction of rehydration kinetic of air dried potato cubes. For rehydration, samples were immersed in water during different periods of time and temperatures (23 +/- 2 degrees C and 100 +/- 2 degrees C). Rehydration kinetic was monitored by measuring samples weights at regular intervals. In MLP neural network, water temperature, soaking time and potato varieties (Agria, Satina and Kenebek) were used as input parameters and the moisture content was used as output parameter. The results were compared with experimental data. Both Peleg's model and MLP had a proper correlation coefficient for each variety and water temperature but the correlation coefficient of the generalized Peleg's model was lower than of MLP.