Performance evaluation of bamboo and coconut shell biochar as low-carbon cement replacements: experimental and machine learning approaches

The growing demand for sustainable construction materials has accelerated the use of biochar as a partial cement replacement to enhance concrete performance while reducing environmental impacts. This study experimentally evaluated the influence of bamboo biochar (BB) and coconut shell biochar (CC) at replacement levels of 2%, 4%, and 6% on the fresh, mechanical, durability, and microstructural properties of concrete. Workability, compressive strength, split tensile strength, flexural strength, rapid chloride permeability (RCPT), sorptivity, and microstructural characteristics were investigated using SEM and EDS analyses. A machine learning dataset comprising 360 samples, compiled from experimental results and published literature, was used to develop Decision Tree, Random Forest, and XGBoost models for compressive strength prediction. The results demonstrated that 4% biochar provided the most favorable overall balance between compressive strength, durability, workability, and microstructural performance. Among the evaluated machine learning algorithms, Random Forest achieved the highest predictive accuracy for the bamboo biochar dataset (R2 = 0.98), whereas the coconut shell biochar dataset attained a maximum R2 = 0.91. These findings demonstrate the potential of biochar-modified concrete as a sustainable construction material while highlighting the need for larger datasets, comprehensive biochar characterization, and life-cycle assessment to improve model generalizability and quantify long-term environmental benefits.