欧州実験生物学ジャーナル オープンアクセス

抽象的な

Mathematical and neural networks modeling of thin-layer drying of peach (Prunus persica) slices and their comparison

Majid Yazdani, Ali Mohammad Borghaee, Shahin Rafiee, Saeid Minaei and Babak Beheshti

Fast ripening and decay after harvesting are the factors limiting peach storage, so drying peach slices is a solution for this problem. To introduce an accurate model to simulate the drying curves under different conditions, peach slices were dried to equilibrium moisture in thin-layer dryer set using different air temperatures (40, 50, 60, 70 and 80°C) and velocities (1, 1.5 and 2 m/s). In this research, 13 mathematical equations have been used to fit peach slices drying curve. Results indicated that Midilli et al. model with RMSE between 0.01273 to 0.00463 and r2 0.9996 to 0.9982 was the best mathematical model that satisfactorily represented the experimental values. Artificial neural network is a well-known tool for solving complex, non-linear biological systems. The Multi-layer perceptron network was used for modeling drying kinetics. With regard to the results, the network with LOGSIG-TANSIGPURELIN activation function and 3-6-4-1 topology showed the best performance, in which RMSE was 0.00196 and r2 was 0.99996. At last, by comparing final unique neural network model with Midilli et al. model for the whole range of experiments, it was clear that neural network model is more accurate than Midilli model in each experiment conditions.