Machine learning applications in identifying outliers within asphalt mixture fracture test data*

This study investigates the application of machine learning models to analyze the post-peak section of load-displacement curves in asphalt mixture fracture tests, aiming to enhance the detection of abnormal curve shapes before the computation of performance indices related to crack propagation resistance. Utilizing four distinct regression models – exponential decay, logistic decay, regression splines, and Gaussian process models – this research fits these models to the post-peak section of Load-CMOD curves. The parameters derived from these models are then utilized in outlier detection analysis. The results demonstrate that while the exponential decay model offers a more general representation of the post-peak curve, it significantly outperforms the other models in outlier detection accuracy, as measured by a robust estimator of the Mahalanobis distance. This finding highlights the potential of machine learning to identify atypical fracture behaviors in asphalt mixtures, advancing materials engineering and quality control in pavement construction.