Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
Predicting the heat of combustion of alcohols and carboxylic acids using particle swarm optimization
Submission Author:
Bruno Pasa , RS , Brazil
Co-Authors:
Bruno Pasa, Mikael Maraschin, Nina Salau
Presenter: Bruno Pasa
doi://10.26678/ABCM.ENCIT2022.CIT22-0618
Abstract
The aim of this work is the determination of the heat of combustion of alcohols and carboxylic acids by means of prediction from the number of atoms of molecules, using linear and exponential correlations, and subsequent analysis of the confidence interval of the estimated parameters, for comparison of the proposed models. Determining the value of thermodynamic properties is an essential step in any laboratory experiment or industrial process. The way that we are using to predict the properties can be used for basically any property, however, for this work, it was chosen to work with the heat of combustion due to the wide applications that this property has. The heat of combustion is the energy generated from the complete combustion with oxygen of a substance. Its application ranges from comparison of fuel yields, to expressing the amount of energy within a food. A dataset with 67 substances composed of 34 alcohols and 33 carboxylic acids was used to fit correlations based on the number of atoms. For the minimization of error an ordinary mean squared error was applied, and a particle swarm optimization method was used to infer the elliptic confidence interval. For each parameter obtained from the heat of combustion the elliptic confidence intervals were obtained using iterative method particle swarm optimization and the maximum and minimum values of each parameter for expected likelihood with 95% of confidence intervals. The proposed correlation had a satisfactory performance in the prediction of heat of combustion. This performance can be verified from the three statistical parameters used, the model achieved a coefficient of determination (R²) equal to 0.997, a mean square error (MSE) of 1.7313e+03 and root-mean-square error (RMSE) equal to 41.6083. This demonstrates that the proposed correlation can be used as a good tool for the prediction of the property.
Keywords
heat of combustion, correlation, Particle Swarm Optimization
