Predicting mechanical properties of biochar-based composites and biomass heating value using artificial neural network

Abstract Machine learning (ML) has been utilized to predict the properties and biofuel yields of biomass-to-bioenergy products. Various biomasses have distinct fuel characteristics and structures, highlighting the need to account for biomass type in model development. In addition to biofuel production, biochar—a by-product of thermochemical processes—serves as a filler to produce bio-composite material. Achieving optimal mechanical properties requires determining the appropriate combination of biochar loading and production parameters. This study employed a machine learning approach to predict both the heating value of biomass and the mechanical properties of bio-composite filament. Using an Artificial Neural Network, a coefficient of determination of 0.96 was achieved for both validation and test data in predicting biomass heating value. Input feature analysis revealed significant correlations: HHV exhibited a positive relationship with volatile matter and fixed carbon, and an inverse relationship with ash content. In terms of bio-composite filament, predictions for mechanical properties yielded R² values of 0.99 for validation data and 0.96 for test data, demonstrating the effectiveness of machine learning as a predictive tool for assessing mechanical performance.