Abstract
With the continuous depletion of natural river sand resources and the escalating ecological degradation caused by excessive sand mining, manufactured sand has emerged as a sustainable and environmentally favorable alternative aggregate, playing an increasingly important role in the advancement of green construction materials. Nevertheless, the shrinkage behavior of manufactured sand concrete (MSC) exhibits significant deviations from that of conventional natural sand concrete due to differences in the material characteristics. Existing shrinkage prediction models-such as ACI 209, CEB-FIP 2010, B3, and GL 2000-fail to adequately incorporate the specific properties and substitution effects of manufactured sand, thereby limiting their predictive accuracy and applicability. To bridge this gap, the present study conducted a systematic evaluation of the four aforementioned classical shrinkage prediction models based on experimental data from MSC specimens incorporating varying replacement rates of manufactured sand. The findings revealed that models such as B3 and CEB-FIP 2010 neglected the influence of critical characteristics of manufactured sand-namely, particle morphology, gradation, and stone powder content-on the cementitious matrix and interfacial transition zone, which led to substantial prediction discrepancies. Accordingly, a nonlinear regression-based correction function was developed, introducing the manufactured sand content as a key influencing variable to recalibrate and enhance the ACI 209 and GL 2000 models for a more accurate application to MSC. The modified models exhibited markedly improved fitting performance and predictive robustness across the full range of manufactured sand replacement ratios (0-100%), thereby offering a more reliable framework for modeling the shrinkage development of MSC.