Artificial neural network-assisted thermogravimetric analysis of thermal degradation in combustion reactions: A study across diverse organic samples.

During gasification the kinetic and thermodynamic parameter depend on both the feedstock and the process conditions. As a result, one needs to enhance the understanding of how to model numerically these parameters using thermogravimetric analyzer. Consequently, there exists a pressing need to computationally devise gasification model that can efficiently account to thermodynamic and kinetic parameter from thermogravimetric data. In this study, we numerically model gasification process kinetic and thermodynamic parameters, which vary with feedstock and operational conditions. Our novel approach involves creating an ANN model in MATLAB using a carefully optimized 8-20-20-10-1 architecture. Based on thermogravimetric analyzer (TGA) data, this model uniquely predicts critical kinetic (activation energy, pre-exponential factor) and thermodynamic parameters (entropy, enthalpy, Gibbs free energy, ignition index, boiling temperature). Our ANN model, trained on over 80 diverse samples with the Levenberg-Marquardt algorithm, excels at prediction, with an MSE of 6.185e and an R value exceeding 0.9996, ensuring highly accurate estimates. Based on time, temperature, heating rate, and elemental composition, it accurately predicts thermal degradation. The model can predict TGA curves for many materials, demonstrating its versatility. For instance, it accurately estimates the activation energy for pure glycerol at 73.84 kJ/mol, crude glycerol at 67.55 kJ/mol, 12.12 kJ/mol for coal, and 111.3 kJ/mol for wood. These results, particularly for Kissinger-validated glycerol, demonstrate the model's versatility and efficacy in various gasification scenarios, making it a valuable tool for thermochemical conversion studies.