Modeling and Optimization of the Osmotic Dehydration of Cantaloupe
Journal
Journal of Food Chemistry & Nanotechnology
ISSN
2471-4291
Date Issued
2024-10-09
Author(s)
Hamid Bakhshabadi
Mohammad Ganje
Masoumeh Moghimi
Alireza Ghodsvali
Toktam Mohammadi-Moghaddam
Elham Taghavi
Afsaneh Morshedi
Abstract
In current research, the optimization of osmotic dehydration of cantaloupe pieces aimed to maximize water loss (WL) and minimize moisture reabsorption using artificial neural network (ANN). The effects of three parameters were studied: osmotic solution temperature (40-60°C), immersion time (40-240 min), and solution concentration (40-60°Brix), employing central composite design (CCD).
Various parameters including WL, solid gain (SG), reduction in WL to SG ratio, and reduction in sample weight were analyzed. The results indicated that the optimal conditions for osmotic dehydration were achieved with a solution temperature of 60°C, immersion time of 85.71 min, and solution concentration of 40% sucrose (sugar). Under these conditions, the following parameters were observed: WL of 3.79%, SG of 43.74%, WL to SG ratio of 14.48, and sample weight reduction of 47.71%.
Furthermore, results from the ANN revealed that a network structure with one hidden layer comprising 5 nodes (3-5-4 network with 3 inputs, 5 nodes in the hidden layer, and 4 outputs) provided the most accurate predictions. This network achieved correlation coefficients (R2) of 0.999 and root mean squared error (RMSE) of 0.000039, demonstrating high reliability and precision in predicting the selected responses.
Various parameters including WL, solid gain (SG), reduction in WL to SG ratio, and reduction in sample weight were analyzed. The results indicated that the optimal conditions for osmotic dehydration were achieved with a solution temperature of 60°C, immersion time of 85.71 min, and solution concentration of 40% sucrose (sugar). Under these conditions, the following parameters were observed: WL of 3.79%, SG of 43.74%, WL to SG ratio of 14.48, and sample weight reduction of 47.71%.
Furthermore, results from the ANN revealed that a network structure with one hidden layer comprising 5 nodes (3-5-4 network with 3 inputs, 5 nodes in the hidden layer, and 4 outputs) provided the most accurate predictions. This network achieved correlation coefficients (R2) of 0.999 and root mean squared error (RMSE) of 0.000039, demonstrating high reliability and precision in predicting the selected responses.