Analysis and Prediction of Meso Shrinkage of Lightweight Aggregate Concrete Based on Digital Image Correlation

ABSTRACT Existing shrinkage models are based on macroscopic linear strain measurements and have not been extended to the mesoscale. In lightweight aggregate concrete (LWAC), nonuniform deformation between aggregate and mortar phases governs early-age cracking risk, making mesoscale prediction essential. To address this gap, this study uses three-dimensional digital image correlation to measure full-field deformation on cross sections of LWAC specimens with water-to-cement ratios of 0.35, 0.40, and 0.45 over 21 days. The mesoscale shrinkage is characterized by the minimum principal strain. Four classical models (B3, ACI209, GL2000, and CEB-FIP) are reformulated in terms of the minimum principal strain and compared with experimental data. Results show that the shrinkage of LWAC develops in distinct phases governed by moisture transport from prewetted lightweight aggregates, and the B3 model best fits the measured data among the original models. The GL2000 and CEB-FIP models are then revised by introducing time-function correction coefficients and a strength discount factor, respectively. Evaluated against the unmodified models using root mean squared error and a cumulative absolute-error loss function, the revised GL2000 model is more accurate for water-to-cement ratios of 0.35 and 0.40, whereas the revised CEB-FIP model performs better for 0.45.