Estimación del tiempo de secado de fresas (Fragaria sp) del cantón Guano utilizando redes neuronales artificiales.

To find the total drying time of food, long and traditional methods still persist, where the use of dehydrators implies the application of complex procedures in the laboratory, they are also expensive because the direct and indirect raw material is required. This research compares the time taken by the traditional methodology to calculate the total drying time of strawberries (fragaria sp.) from the Valparaíso parish with the technique that uses Artificial Neural Networks (ANN). It began by drying 300 g/day of strawberries in the traditional way in the Unach laboratory of the Faculty of Engineering using a tray dryer, a drying time of 360 min was calculated, then five quantitative characteristics necessary to form the ANN were determined, which were: relative humidity, globe temperature, dry bulb temperature, wet bulb temperature and the initial weight of the fruit, considering the sigmoid function for the activation of the backpropagation algorithm (RP) using Matlab. As a result, four layers were determined: a) the first one has 16 static receptor neurons; b) the second one has 10 neurons; c) the third one has 1 dynamic neuron and d) the fourth one is the static output neuron, with calculation errors of less than 3.47 %. To determine the best model, the MSE was calculated, which measures the amount of error between two sets of data, comparing the experimental value with the predicted one and it was found that both in training and in validation and testing the values are close to zero, which indicates that it is an optimal model. This experiment was developed under the following initial conditions: altitude of 2754 meters above sea level corresponding to the city of Riobamba, with strawberries with a sweetness of 12 ° Brix and the dehydrator at a constant temperature of 65 ° C. For comparison purposes, the same quantitative characteristics defined above have been entered into Matlab, but instead of entering the 11 values obtained in the experimental part, only the first one was entered, which corresponds to the mass of the fruit after 30 min in the dehydrator under the same conditions, in order to determine the error in the calculation if only 1 drying data were available. It was determined that the average error when all 11 variables are available is 0.76 %, while when one variable is available it rises to 1.58 %, for both cases the error is acceptable, indicating that if the first drying variable is known, the value could be determined after 360 min of drying, which makes this method even more efficient.

• Main Author: Duque Enriquez, Samantha Daniela (author)
• Format: bachelorThesis
• Language: spa
• Published: 2024
• Subjects: Tiempo de secado; RNA; Fresa (fragaria sp); Retro propagación
• Online Access: http://dspace.unach.edu.ec/handle/51000/13847