Predicting workability and mechanical properties of bentonite plastic concrete using hybrid ensemble learning.

Heavy metal contamination in wastewater poses severe environmental challenges, highlighting the urgent need for efficient and cost-effective solutions. While bentonite incorporation in concrete mixtures has shown promise in adsorbing heavy metals, its experimental validation-through Bentonite Plastic Concrete (BPC)-is hindered by high costs, labor-intensive procedures, and the need for specialized equipment. This study overcomes these barriers by introducing hybrid ensemble learning models, optimized with Forensic-Based Investigation Optimization (FBIO), to predict BPC's workability and mechanical properties, including slump (S), tensile strength (TS), and elastic modulus (E). Using input parameters such as gravel, bentonite, silty clay, curing time, sand, cement, and water, models including Random Forest (RF), Adaptive Boosting (ADB), Extreme Gradient Boosting (XGB), and Gradient Boosting Regression Tree (GBRT) were developed. Notably, GBRT-FBIO achieved the highest accuracy for E predictions, while XGB-FBIO excelled for TS and S. Shapley Additive Explanation (SHAP) analysis identified water as the most critical factor influencing slump (+ 0.11) predictions while curing time emerged as the key determinant for TS (+ 0.18) and E (+ 0.12) predictions. Additionally, a user-friendly online tool was developed to enable the real-time application of these models, reducing reliance on costly experimental methods. This work addresses key challenges in experimental BPC testing, offering a transformative computational approach for advancing civil engineering materials research.