Multifactorial analysis of AAR development: Integrating laboratory and field data with statistical and probabilistic modelling

Alkali-aggregate reaction (AAR) is among the most harmful durability issues affecting concrete infrastructure, with prevention being the most effective strategy. Widely used laboratory tests, such as the Accelerated Mortar Bar Test (AMBT) and Concrete Prism Test (CPT), assess aggregate reactivity and often present discrepancies with field performance that remain unquantified. This study introduces a probabilistic, risk-based framework to evaluate the reliability of these tests using a multifactorial analysis that integrates field and laboratory data with Bayesian inference and Beta distribution modelling. The likelihood of AAR occurrence is evaluated considering test outcomes, environmental exposure, and alkali loading. Results show AMBT outperforms in identifying non-reactive cases (41% vs. 61% posterior probability for mixes without supplementary cementitious materials [SCMs]; 16% vs. 30% with SCMs), while both tests perform similarly for reactive cases (i.e., 74% for mixes without SCMs and 50% with SCMs). Moreover, warm climates, high alkali content, and the absence of SCMs increase the risk of the tests’ misclassification, while cold environments with low alkali levels and SCMs enhance their reliability.