Validation of Various Australian Concrete Code (AS3600) Shrinkage Prediction Models: A Novel Approach

Shrinkage strain is a requirement of the concrete design process that has proven difficult to model. The consequences of unacceptable shrinkage magnitudes may lead to negative effects on the structural performance of concrete, including durability implications or even loss of serviceability. Although laboratory tests may be undertaken to quantify the shrinkage, these are time consuming. For this reason, empirical code-type prediction models are used to predict shrinkage strain. This paper considers the accuracy of four versions of the Australian Standard Concrete Code (AS3600) shrinkage prediction models (AS3600: 1994, 2001, 2009 and 2018). In addition, five new models, using a novel approach based on combinations of the differing components of the similar 2009 and 2018 models, were proposed and evaluated. All nine models were evaluated, when compared with the actual shrinkage strains measured, over a period of approximately six months, on a range of concretes under laboratory-controlled conditions, for six mixes. The six mixes comprised three aggregate types (quartzite, granite and andesite) and two water cement ratios (0.56 and 0.4). All the models generally over-predicted shrinkage strain with age. Furthermore, the superseded AS3600 (1994), which considered the least number of factors, was the most accurate model, yielding an overall coefficient of variation (ωall) of 88.9%. The currently applied AS3600 (2018) model yielded a ωall of 168.6%. Two of the newly proposed models were more accurate that the AS3600 (2018) Model, with a ωall of 156%.