MACHINE LEARNING BASED ESTIMATION OF DRYING CHARACTERISTICS OF APPLE SLICES

Abstract

Machine learning algorithms have been usually used in food drying. These models are also effectively used for nonlinear processes such as heat and mass transfer. Estimation of drying characteristics is also important for optimizing drying conditions. Estimating of moisture rate and drying rate ensures accurate and high quality drying of the product under air-convective drying conditions. In this study, drying rate (DR) and moisture ratio (MR) were estimated in air-convective conditions with the use of drying time, moisture content (d.b.), and effective moisture diffusivity as input. In addition, two different validation methodology was performed as k-fold cross validation and train test split. In the present study random forest-RF; multilayer perceptron-MLP; and k-nearest neighbor-kNN were performed to estimate of drying rate and moisture ratio. As a result, correlation coefficients were found above 0.8500 for moisture ratio and 0.8722 for drying rate. The findings show that algorithms could be successfully applied for the estimation of drying rate and moisture ratio.