Enhancing and comparing shrinkage prediction models for High-Strength Concrete with and without admixtures

Abstract

Abstract This study aimed to improve and compare the parameterization of three prominent shrinkage prediction models—RILEM B4, MC 2010, and WITS—tailored specifically for High-Strength Concrete (HSC), both with and without the inclusion of admixtures. The dataset used for refining model parameters consisted of 220 experiments related to drying shrinkage and 342 experiments concerning autogenous shrinkage. Model performance evaluation involved various statistical metrics applied to the entire HSC dataset, subdatasets, and distinct time periods of shrinkage (0–99 days, 100–199 days, 200–499 days, and ≥500 days). The statistical indicators included Root Mean Square Error (RMSE), R-squared adjusted (R2 adj), Akaike’s Information Criterion (AIC), and the overall coefficient of variation (C.o.Vall). Modified models exhibited significantly improved predictions compared to the original models, with most predictions falling within ±20% of the measured shrinkages. For HSC drying shrinkage, the original model accuracy ranked as WITS, RILEM B4, and MC 2010. However, after parameter adjustments, WITS, MC 2010, and RILEM B4 were the best-performing models. Conversely, for HSC autogenous shrinkage predictions, the RILEM B4 model surpassed the MC 2010 model, demonstrating superior accuracy and reliability in forecasting this specific type of shrinkage behaviour within High-Strength Concrete.