Predicción de la energía de activación para los residuos de la empresa “Real Flowers” empleando redes neuronales artificiales

Abstract

The objective of this study was to predict the thermogravimetric behaviour of the activation energy of rose stem residues using artificial neural networks. The data of the thermogravimetric analysis were obtained at two heating ramps of 5ºC / min and 15ºC / min in temperature ranges from 25 to 900ºC with a nitrogen flow of 20 ml/min. The calculation of the activation energy was carried out based on the proposed kinetic models, where the distributed activation energy model had a correlation factor for the 5 ºC / min ramp of 0.967012 and an R2 of 93.5113%, for the 15 ºC / min ramp, the distributed activation energy model had a coefficient of 0.955083 and an R2 of 91.2184%. The artificial neural network was designed with 3 input neurons temperature (K), time (s) and weight (mg), the data from the thermogravimetric analysis, in the hidden layer it has 200 neurons and an output neuron that is the activation energy (KJ / mol), the training algorithm was the Bayesian regularization algorithm with a correlation coefficient of 1 and a mean square error of 2.73E-3. The results were evaluated through the comparative statistical analysis of two samples, real activation energy and that predicted by the network, where the P-value is greater than 0.05, and there was no statistically significant difference between the means of the two variables. It is recommended to use the prediction model in projects for the use of agricultural residues for the generation of energy due to the reduction of analysis times in the laboratory.