Point cloud–ANN integrated framework for rapid prediction of fresh properties of self-consolidating concrete

This study presents an integrated digital framework combining three-dimensional point cloud analysis and artificial intelligence (AI) modeling to quantitatively evaluate the fresh properties of self-consolidating concrete (SCC). Experimental tests, including slump flow, L-box, and U-box, revealed strong correlations among testing results, with slump flow showing a high relationship with T₅₀₀ (r = –0.81) and L-box ratio (r = 0.84). Although plastic viscosity exhibited weaker direct correlation, its association with T₅₀₀ remained significant (r = 0.76). Point cloud–based monitoring achieved high accuracy, with an average deviation of only 3.1% from direct measurements, demonstrating its capability for precise, non-contact evaluation of spreading kinetics. Based on these validated relationships, 1000 augmented datasets were generated to support AI training. The optimized artificial neural network (3 layers, 10 neurons per layer, and learning rate of 0.4) achieved strong multi-output predictive performance (R² = 0.79–0.88 and MAPE < 8%), effectively reproducing multiple experimental parameters. The proposed digital evaluation framework enables real-time prediction of fresh SCC behavior, offering a promising foundation for automated quality control and data-driven mix design in intelligent concrete construction.