Comparison of mathematical models and artificial neural networks for prediction of drying kinetics of mushroom in microwave vacuum dryer

Drying characteristics of button mushroom slices were determined using microwave vacuum drier at various powers (130, 260, 380, 450 W) and absolute pressures (200, 400, 600, 800 mbar). To select a suitable mathematical model, 6 thin-layer drying models were fitted to the experimental data. The fitting rates of models were assessed based on three parameters; highest R2, lowest chi square () and root mean square error (RMSE). In addition, using the experimental data, an ANN trained by standard back-propagation algorithm, was developed in order to predict moisture ratio (MR) and drying rate (DR) values based on the three input variables (drying time, absolute pressure, microwave power). Different activation functions and several rules were used to assess percentage error between the desired and the predicted values. According to our findings, Midilli et al. model showed a reasonable fitting with experimental data. While, the ANN model showed its high capability to predict the MR and DR quite well with determination coefficients (R2) of 0.9991, 0.9995 and 0.9996 for training, validation and testing, respectively. Furthermore, their predictions Mean Square Error were 0.00086, 0.00042 and 0.00052, respectively