Combustion and pyrolysis of dairy waste: A kinetic analysis and prediction of experimental data through Artificial Neural Network (ANN)

Abstract

The thermochemical conversion of biomass into energy is increasingly recognized as a sustainable alternative, yet analyzing biomass thermal decomposition is complex and resource intensive. In addition, kinetic modeling is a crucial step for process design and optimization of thermochemical degradation of biomass, where limited thermogravimetric (TG) data forms the basis of this analysis. Leveraging machine learning can expedite this process by extrapolating and interpolating experimental data, reducing time and costs. This study focuses on using Artificial Neural Network (ANN) models to predict the thermal degradation behavior of dairy dung during pyrolysis and combustion, validated by a Multistage Kinetic Model (MKM). Thermogravimetric analysis (TGA) data were collected at four heating rates (20, 40, 60, and 80 °C/min), revealing four stages in pyrolysis and three in combustion. A linearized MKM was applied to derive kinetic parameters (Ea, A, and n) from experimental data. The TGA data were then trained in ANN (backpropagation) taking heating rate and temperature as input variables and mass change as an output variable. The ANN accurately predicted data for 30 and 50 °C/min, subsequently applied in the MKM. Comparison of activation energies (Ea) values showed strong agreement between experimental and predicted values, indicated by a high regression coefficient (R²). This study demonstrates the utility of ANN in computing kinetic parameters for biomass thermal degradation, offering time savings and accurate prediction of non-experimental data.