Bibliographic citations
Soto, A., (2024). Predicción de la temperatura inicial de congelación de raíces y tubérculos andinos utilizando inteligencia artificial [Universidad Nacional Agraria La Molina]. https://hdl.handle.net/20.500.12996/6535
Soto, A., Predicción de la temperatura inicial de congelación de raíces y tubérculos andinos utilizando inteligencia artificial []. PE: Universidad Nacional Agraria La Molina; 2024. https://hdl.handle.net/20.500.12996/6535
@misc{renati/245349,
title = "Predicción de la temperatura inicial de congelación de raíces y tubérculos andinos utilizando inteligencia artificial",
author = "Soto Torres, Andres Isidro",
publisher = "Universidad Nacional Agraria La Molina",
year = "2024"
}
The objective of this research was to predict the initial freezing temperature (Tf) of Andean roots and tubers (mashua, maca, yacon, arracacha, olluco) using Artificial Intelligence techniques. Tf emerges as a crucial physical property, influencing thermophysical properties such as specific heat, thermal conductivity, and freezing time. Artificial neural networks (ANNs) were employed with varying layer configurations (3, 5, 10) and neurons (5, 10, 20), based on composition data from the USDA (2023). The database comprised 31 datasets and 6000 generated sets within established composition limits to analyze the effect of increased training data. Composition data served as inputs, and freezing temperatures were determined by the Boonsupthip and Heldman (2007) model, used as output variables for ANN training. Eight input variables were evaluated (water, proteins, fats, monosaccharides, disaccharides, fiber, minerals, and ascorbic acid), along with five (with reduced molecular weight). Results were validated by comparing them with theoretical freezing temperatures and analyzing the proximal and mineral composition of roots and tubers. The most accurate configuration utilized 31 datasets, 5 input variables, 3 hidden layers, and 5 neurons, achieving an R2 of 0.99 and an RMSE of 0.02 °C. Using this configuration, freezing temperatures of roots and tubers were predicted and validated by comparing them with experimental measurements of arracacha, maca, olluco, yacon, and mashua, obtaining values of -1.11, -2.03, -0.7, -1.37, and -0.86 °C, respectively. The comparison revealed an average absolute difference of 0.33 ± 0.17 °C and a percentage error of 31.82%, confirming the accuracy of the neural network. This study represents a significant advancement in understanding and applying these methodologies in the food industry.
This item is licensed under a Creative Commons License