Investigation of concrete creep-shrinkage models using residual method

This study proposes a novel residual-based correction approach to improve the accuracy of concrete creep and shrinkage predictions. A comprehensive residual analysis was performed using a database of 2,254 creep and 5,399 shrinkage test results. The analysis revealed systematic biases in current code predictions: 55.5% of creep predictions showed overestimation, particularly at low stress levels (<20 MPa) and small theoretical thicknesses, while 66.2% of shrinkage predictions exhibited underestimation, especially for high-strength concrete and low water-cement ratios (<0.4). A power function-based residual model was developed, with strong internal parameter correlations (correlation coefficients of -0.998 and -0.96 for creep and shrinkage models respectively). Validation using independent experimental data demonstrated significant improvements in prediction accuracy, with average errors reduced by up to 40%. This study provides a practical methodology for enhancing existing prediction models through systematic residual analysis, offering a new perspective for addressing the uncertainties in concrete time-dependent behavior predictions.