Kinetic study of rigid polyurethane foams thermal decomposition by artificial neural network

TL;DR

This study develops a neural network approach to improve the accuracy of kinetic modeling in the thermal decomposition of rigid polyurethane foams, outperforming traditional models significantly.

Contribution

A novel neural network method integrating kinetic models as activation functions for better fit of thermal decomposition data is introduced.

Findings

Neural network residual errors are 102 times lower than traditional kinetic models.

R3 model best describes the decomposition process for all temperatures.

Method enables more precise determination of kinetic parameters.

Abstract

Kinetic models of solid thermal decomposition are traditionally used for individual fit of isothermal decomposition experimental data. However, this methodology can provide unacceptable errors in some cases. To solve this problem, a neural network (MLP) was developed and adopted in this work. The implemented algorithm uses the rate constants as predetermined weights between the input and intermediate layer and kinetic models as activation functions of neurons in the hidden layer. The contribution of each model in the overall fit of the experimental data is calculated as the weights between the intermediate and output layer. In this way, the phenomenon is better described as a sum of kinetic processes. Two rigid polyurethane foam samples: loaded with Al2O3 and no inorganic filler were used in this work. The R3 model described the thermal decomposition kinetic process for all temperatures for both foams with smaller residual error. However, the network residual errors are on average 102 times lower compared to this individual kinetic model. This improved methodology allows detailed study of physical processes and therefore a more accurate determination of kinetic parameters such as the activation energy and frequency factor.