Study on Printability, Mechanical Properties, and Strength Prediction of 3D-Printed Modified Recycled Micro-Powder Concrete

The influence of recycled powder obtained by combined treatment of acidification and carbon activation on the printing feasibility and the mechanical properties of the material is investigated, and a predictive model for the development of mechanical properties has been developed by combining practical experiments with theoretical modeling. When the flow of the jet is between 160 mm and 170 mm, the extrusion performance of the material is optimal, allowing continuous printing of 20 layers with a stable forming height of up to 200 mm. Mechanical tests indicate a significant increase in interlayer adhesion strength, with marked anisotropy of compressive strength in different directions: Mechanical tests reveal significant anisotropy in compressive strength: the highest strength is observed in the T1 direction, followed by the T2 direction, while the T3 direction shows relatively lower strength. In a subsequent modeling analysis, three machine learning algorithms—XGBoost, Random Forest, and LightGBM—were compared in terms of compressive strength prediction. After training and validation, the XGBoost model performs the best, with MAE below 1.5 MPa and R2 not less than 0.91; Random Forest and LightGBM are second. Among them, LightGBM has a large prediction error in the Z direction, with a MAE of 2.30 MPa. The XGBoost model proved to be the most efficient. Thus, this study provides an effective tool for the design and regulation of 3D Printed Concrete (3DPC) properties while exploring viable paths for the recovery of construction waste in the field of advanced manufacturing, thus fostering the joint development of resource circulation and digital construction.