Work

Optimizing rubberized Concrete: Machine learning insights into effects of particle size on the compressive strength

Year: 2026

Type: article

Abstract: End-of-life tires (ELTs) are a persistent solid waste stream worldwide, posing significant challenges to sustainable waste management. Incorporating crumb rubber (CR) derived from ELTs into concrete offers a promising recycling pathway; however, the complex interactions among CR particle size, aggregate replacement ratio, and water-to-cementitious (w/cm) ratio make compressive strength prediction and mix optimization challenging using conventional experimental approaches alone. This study develops a Categorical Boosting (CatBoost) based machine learning (ML) framework trained on a comprehensive dataset of 909 data points compiled from 58 peer-reviewed studies. The CatBoost model achieved the highest predictive accuracy among eight evaluated models, with a test R 2 of 92.2% and RMSE of 4.7 MPa. Shapley Additive exPlanations (SHAP) analysis revealed that the aggregate replacement ratio was the primary driver of strength degradation, exerting a substantially greater influence than rubber particle size. Partial Dependence Plot (PDP) and Individual Conditional Expectation (ICE) analysis establish practical design thresholds: rubber mass fraction below 0.10 and w/cm below 0.40 to 0.45 are identified as the primary boundaries for maintaining structural performance. Independent experimental validation yielded prediction errors of 1.1% to 6.8%, with all predicted values falling within 95% bootstrap confidence intervals. An inverse design framework identified three Pareto-optimal mix scenarios that simultaneously divert ELTs waste from landfills and reduce natural aggregate consumption, with carbon emission assessments demonstrating CO 2 reductions of 0.55% to 3.02% relative to a CR-free baseline. The framework provides a quantitative pathway for incorporating recycled rubber into structural concrete, supporting the transition toward cleaner construction materials

Source: Cleaner Materials

DOI: 10.1016/j.clema.2026.100402

Authors: Sung Eun Cho, Sung Eun Cho, Sung-Won Cho, Sung-Won Cho, Alexander S. Brand

Institution: Virginia Tech

Language: English

Topic: Innovative concrete reinforcement materials

Subfield: Civil and Structural Engineering

Field: Engineering

Domain: Physical Sciences

SDGs: Responsible consumption and production, Industry, innovation and infrastructure, Sustainable cities and communities

Open Access status: gold

Estimated APC paid: $1,710