Prediction Model for the Indirect Tensile Strength of Engineered Emulsion-Stabilized Cold Recycled Materials for Efficient Quality Assurance Practices

Abstract The growing use of cold recycling for flexible pavement rehabilitation offers a compelling approach to advancing sustainable infrastructure developments. Despite its many advantages, one of the challenges hindering broader implementation of cold recycling is the lack of real-time methods to quickly assess as-constructed material properties, which are essential for evaluating construction quality and to ensure compliance with the original material design. Integrating fast and efficient methods for determining as-constructed material properties can greatly enhance quality assurance and design processes on a wider scale. This study leveraged advanced statistical modeling techniques to develop a robust predictive model for estimating the indirect tensile strength of engineered emulsion-stabilized cold-recycled materials. A comprehensive database of laboratory-produced mixtures which incorporates material attributes such as varied engineered emulsion contents, the presence of chemical additives, gradation characteristics, and density, replicating as-constructed conditions, was utilized to train and validate predictive models, using the support vector regression (SVR) and self-validating ensemble modeling (SVEM) techniques. Among the advanced statistical modeling techniques explored, SVR consistently demonstrated better predictive accuracy, based on a range of statistical error metrics assessed. Model sensitivity analyses identified coarseness and mixture density as the most influential variables for the prediction of the indirect tensile strength. The resulting model was incorporated into a user-friendly Microsoft Excel–based prediction tool, developed to aid engineers and practitioners in quality assurance by enabling rapid estimation of as-constructed mechanical properties, thereby minimizing dependence on extended laboratory testing and support pavement design processes.