The moisture content of papaya fruit during drying in a cabinet dryer was estimated using both
mathematical and neural network models. The effect of air temperatures (40, 50 and 60 °C) and fruit
slice thickness (3, 5, 7 mm) on moisture ratio were investigated. A three-layer perceptron neural
network with different training algorithms was designed and used. To obtain better estimation of fruit
moisture content, topologies were established based on different threshold functions. The results
showed that the multi-layer perceptron network with 3-9-1 topology, the Levenberg-Marquardt training
algorithm and the logarithmic sigmoid threshold function provided the least errors. Furthermore, eight
well-known mathematical models were also used to simulate the drying process. Among the
mathematical drying models, the Two-term model was found to be more suitable for predicting the
drying of papaya fruit slices, with a coefficient of determination (R2
) of 0.9974 and a root-mean-square
error (RMSE) of 0.0123. However, these values were lower than those obtained for the designed neural
network model (R2
= 0.9994; RMSE = 0.0070). Therefore, estimation of the moisture content of papaya
fruit could be better modelled by a neural network than by the mathematical models.