Integrated Experimental and Data-Driven Approach to Strength Prediction of Mortars Incorporating Thermally Activated Construction Waste Powder

Abstract This study investigated the feasibility of waste construction powder (WCP), a by-product generated from construction waste recycling plants, as a partial replacement for Portland cement through thermal activation. Mortar specimens were prepared using thermally activated WCP at 600 °C and 800 °C, replacing cement at rates of 6%, 13%, and 20%. The flowability, compressive strength, and flexural strength were evaluated. The results showed that mortar with WCP activated at 800 °C met the KS L 5201 industrial standard even at a 20% replacement ratio, achieving up to 21% and 6% increases in compressive strength and flexural strength, respectively. In addition, a machine learning (ML) model based on the light gradient boosting (LGB) algorithm was developed using 405 literature-derived data entries to predict the compressive strength of cementitious composites incorporating thermally activated WCP. The predictive performance was validated against experimental results, achieving strong correlation with the measured data ( R 2 = 0.813), and yielding an RMSE of 3.044 MPa, MAE of 2.507 MPa, and SMAPE of 5.675%. More than 90% of the predictions fell within a ± 10% error margin. These findings demonstrate the practical applicability of the proposed ML model for strength prediction of WCP-based cementitious materials, as well as the technical feasibility of thermally activated WCP as cement replacement.