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

Abstract

This study investigates Natural Fiber Recycled Aggregate Concrete (NFRAC), using fibers including jute, sisal, kenaf, ramie, coir, and bamboo with recycled aggregates. A total of 534 data points analyzed using five machine learning (ML) models: eXtreme Gradient Boosting (XGB), Random Forest (RF), Light Gradient Boosting Machine (LGBM), Multilayer Perceptron (MLP), and Categorical Boosting (CAT); optimized with Particle Swarm Optimization (PSO) to predict compressive strength of NFRAC. The water-binder ratio was identified as a key factor using SHapley Additive exPlanations (SHAP) and Partial Dependency Plots (PDP). XGB achieved the best performance [Root Mean Square Error (RMSE) 4.2 MPa, Coefficient of Determination (R²) 0.94] among five ML models. Life cycle analysis showed NFRAC reduces embodied Carbon Dioxide (eCO 2 ) by 2.7 % with 25 % Recycled Concrete Aggregate (RCA) and 5.4 % with 50 % RCA. Cost-benefit analysis confirmed economic advantages over traditional concrete. A user-friendly web interface for predicting NFRAC strength was developed.