Analysis of the drying kinetics of freeze-dried persimmon at different cabin pressures using artificial neural network method

Abstract

The main objective of this study is to freeze dry persimmon (Diospyros kaki) at three different cabin pressures (0.008 mbar, 0.010 mbar and 0.012 mbar) and product thicknesses (3 mm, 5 mm, and 7 mm) and, examine the drying kinetics, and assess the accuracy of artificial neural networks (ANN) in forecasting critical drying parameters, including Moisture Content (MC), Drying Rate (DR), and dimensionless Mass loss Ratio (MR). In this study, a feed forward ANN with a Multilayer Perceptron (MLP) architecture was designed to simulate and predict the freeze-drying behavior of persimmons. The ANN modeling, developed using MATLAB software while accounting for different product thicknesses and cabin pressures, demonstrated a test performance value of 0.99781 and an overall performance value of 0.99896. The drying time for persimmons ranged from 1080 minutes (3 mm, 0.008 mbar) to 2160 minutes (7 mm, 0.012 mbar). It was observed that reducing cabin pressure and product thickness resulted in decreased drying time. The highest drying rate (0.213%/min) was achieved with a 3 mm thick product at 0.008 mbar cabin pressure. Depending on the product thickness and cabin pressure, the Alibas model (3 mm, 0.008 mbar), the Improved Midilli-Kucuk model (3 mm, 0.010 mbar; 5 mm, 0.008 mbar; 5 mm, 0.012 mbar; and 7 mm, 0.010 mbar), and the Balbay & Sahin model (3 mm, 0.012 mbar; 5 mm, 0.010 mbar; 7 mm, 0.008 mbar; and 7 mm, 0.012 mbar) were found to be the most effective in describing the drying process of persimmons. These results suggest that ANNs are capable of effectively modeling the freeze-drying process of persimmons.