Kinetic analysis and machine learning insights in the production of biochar from Artocarpus heterophyllus (jackfruit) through pyrolysis

Abstract

According to International Energy Agency (IEA) Task 40, biomass contributes approximately 10 % of global energy production. This includes waste from agriculture and forestry, generating around 140 billion tons of biomass each year—posing a major challenge for efficient management and disposal. The Food and Agriculture Organization (FAO) reports that global jackfruit production reached 3.7 million tons between 2015 and 2017, while 2.96 million tons of bioenergy feedstock were produced in 2018. Utilizing jackfruit waste as a renewable bioenergy source not only adds economic value to agricultural residues but also helps reduce overall waste generation. The bark of the jackfruit tree (Artocarpus heterophyllus (AHB)) possesses considerable economic importance and exhibits an enormous distribution throughout several regions in Asia. This study involves the production of biochar from AHB biomass through fast pyrolysis at temperatures between 400 and 600 °C. The biochar produced has a carbon content of 66.69 wt% and a calorific value of 27.15 MJ/kg, respectively, which have similar properties to coal. The kinetic analysis of biomass employed three distinct models (OFW, KAS, and TANG) to determine the activation energy. The current study employed machine learning (ML) models to forecast the mass loss of biomass during pyrolysis, which is challenging because of the intricate characteristics of biomass and the extensive range of operating circumstances. Temperature and heating rate were used as input data, while mass loss was the desired output, to train a variety of machine learning models, including ensemble learning, support vector regression, Gaussian process regression, and neural network models. Among these models, the Gaussian process regression model showed superior performance compared to others, achieving a perfect R2 of 1 and minimal errors on both the validation and test sets, making it the best model to predict mass loss of biomass. © 2025 Elsevier Ltd