Modeling and Optimization of Mass Transfer during Osmosis Dehydration of Carrot Slices by Neural Networks and Genetic Algorithms

Osmotic dehydration, as a minimal processing method, has found increasingly wide prospects during the past few decades. This process involves mass transfer which is commonly modeled by applications of different procedures, mostly based on Fick's law. In this research, we approach the modeling process by first obtaining experimental measurement of carrot's solid gain and water loss under different conditions of solution concentrations (20, 40 and 60% w/w), temperatures (40, 60 and 80 degrees C) as well as time intervals (1-6h). Then two paradigms of artificial neural networks (ANN), feed forward neural networks (FFNN) and radial basis function neural networks (RBFNN) are applied and compared for modeling this process. Additionally, genetic algorithm is used to determine optimal conditions for osmotic dehydration.