Evaluating mechanical and environmental impacts of sustainable natural fiber reinforced recycled aggregate concrete incorporating supervised machine learning methods

Construction industry is increasingly focusing on sustainable decarbonized concrete using alternative materials. One promising innovation is natural fiber recycled aggregate concrete (NFRAC), which combines recycled aggregates with natural fibers to enhance concrete performance and reduce the carbon footprint from quarrying and fiber production. This study analyzes 534 data points from existing literature to predict the compressive strength of NFRAC made with fibers from jute, sisal, kenaf, ramie, coir, and bamboo, across varying water-cement ratios and curing ages. Five machine learning models: eXtreme Gradient Boosting (XGB), Random Forest (RF), Light Gradient Boosting Machine (LGBM), Multilayer Perceptron (MLP), and Categorical Boosting (CAT), were employed to predict compressive strength, with hyperparameters optimized using Particle Swarm Optimization (PSO). SHapley Additive exPlanations (SHAP) and partial dependency plots (PDP) assessed the impact of key factors, showing that the water-binder ratio significantly affects compressive strength. The XGB model achieved the best results with an RMSE of 4.2 MPa and R² of 0.94. Life cycle analysis indicated that using 25% recycled concrete aggregate (RCA) and natural fibers reduces embodied CO₂ emissions by 2.7%; with 50% RCA replacement, the reduction reaches 5.4%. A cost-benefit analysis revealed that NFRAC offers significant economic advantages over traditional concrete, particularly at higher replacement rates. The study validated the machine learning models and established a user-friendly web interface for predicting NFRAC compressive strength. This integration of advanced machine learning, carbon analysis, and economic evaluation highlights the potential for adopting natural fibers and recycled aggregates in sustainable concrete construction, effectively mitigating environmental impact.