Pretreatment optimization and machine learning-based prediction of methane production from fruit wastes

The increasing need for sustainable waste management and renewable energy has enhanced interest in anaerobic digestion as an alternative valorization pathway for organic residues/wastes. Although fruit wastes offer considerable biogas potential, their composition often limits efficient conversion, making pretreatment essential for the improvement of methane yields. This study investigates how different fruit wastes and applied pretreatment methods affect methane production by combining laboratory data assessment with computational modeling. Six typical fruit wastes (banana, pear, kiwi, orange, tangerine, and apple) underwent mechanical, thermal, alkaline, ultrasonic, and combined pretreatments to assess their effects on biogas production. The mechanical–thermal pretreatment showed the strongest enhancement, yielding a maximum of 344 mL CH 4 /g VS for apple waste, whereas alkaline pretreatment inhibited methanogenic activity. The application of predictive modeling using Response Surface Methodology (RSM) and Artificial Neural Networks (ANNs) yielded high accuracy, with ANNs better capturing the nonlinear effects of the main applied pretreatment variables. Overall, the findings demonstrate that combining optimized pretreatment strategies with data-driven modeling can significantly improve methane production/recovery from fruit-derived residues/wastes. • Fruit wastes responded differently to pretreatments, precluding a uniform approach • Combined pretreatments rarely gave additional benefits compared to individual ones • Alkaline pretreatment consistently inhibited methanogenic activity • RSM and ANN models predicted methane yield based on fruit waste and pretreatment • RSM and ANN predicted methane yield with errors below 5%