APPLICATION OF VARIOUS METHODS OF PROCESSING MATHEMATICAL MODELS IN THE SIMULATION OF DRYING APPLES

Abstract

In this article, the artificial neural network model was compared with traditional regression models for drying food materials. High-intensity ultrasound was applied to the processing of apple slices of various thicknesses with amplitudes set at 25%, 50%, 75% and 100% of the maximum. The four most commonly used regression models for drying were selected based on experimental data and their applicability was tested on various sets of experiments. To create a backpropagation neural network, input parameters were used: ultrasound amplitude, sample thickness, and drying temperature. Humidity was the output parameter. After training and testing the networks, a statistical analysis was carried out and the best network was selected. Neural networks showed excellent agreement with the experimental data, regardless of the input parameters obtained in the experiments. In contrast to regression models, which fit perfectly on only one set of experimental data and show inadequate fit for small changes in one or more input parameters.