<![CDATA[This study investigated the development of sustainable Controlled Low Strength Material (CLSM) using industrial by-products pond ash, fly ash, and red mud as alternatives to conventional concrete constituents. This research employs a dual methodology: comprehensive experimental testing aligned with ASTM standards and the implementation of advanced machine learning (ML) techniques to predict the unconfined compressive strength (UCS) of CLSM mixes. Experimental datasets, generated through the variation of key material and mix design parameters, were utilized to train ensemble-based supervised ML models, including ADAboost, XGBoost, gradient boosting machine (GBM), and random forest (RF). A comparative performance evaluation was conducted, and the XGBoost model emerged as the most accurate predictor, achieving R² values of 0.969 for training and 0.933 for testing, surpassing GBM, ADAboost, and RF across multiple performance indicators. The optimal model was subsequently embedded into a graphical user interface (GUI) for UCS prediction. A sensitivity analysis based on the XGBoost model revealed that cement, water, and curing age were the most influential parameters affecting UCS, with cement exhibiting the highest impact value of 0.86 and a relative contribution of 19%. These findings emphasize the significance of these variables in strength development and mix optimization. The integration of experimental validation with predictive modeling not only advances the understanding of CLSM behavior but also underscores the utility of ML in the formulation of sustainable construction materials. This research supports the beneficial reuse of industrial waste, aligns with environmental sustainability goals, and provides an efficient and reliable tool for CLSM mix design.]]>
Copyrights © 2025
