Surrogate modeling for self-consolidating concrete characteristics estimation for efficient prestressed bridge construction

This paper investigates the effects of material constituents on fresh and hardened properties of Self- Consolidating Concrete (SCC) mixture necessary for efficient prestressed bridge girder fabrication using a surrogate modeling technique. Response surface methodology (RSM)-based surrogate models consisting of input parameters such as density of coarse and fine aggregate were created based upon the past laboratory testing results for different SCC mixture trials. These models were used to estimate various SCC material characteristics, including slump flow, J-ring flow, passing ability, filling capacity, Visual Stability Index (VSI), T₅₀(concrete spread time to reach the 50.8 cm [20 in] mark), column segregation, 16-hour compressive strength, and 28-days compressive strength, while examining the correlation between the input parameters on each material characteristic. To observe the effect of core input parameters in an efficient manner, 2D contour plot and 3D surface plot for material characteristics were also created. Then, statistical analyses with the testing results were performed to determine the accuracy of the surrogate models in terms of coefficient of regression (R²). Most of the R²values are higher than 90%, indicating a higher degree of correlation among the testing and surrogate data. Average predicted-to-measure ratios of the surrogate models were almost equal to or slightly greater than 1.00, showing good agreement with the testing results, and specifically the surrogate and testing values for J-ring flow and 28-days compressive strength were nearly identical. Key findings indicate that the coarse aggregate content significantly affected the characteristics of the SCC mixtures.