Accurate Identification of Pavement Materials Susceptible to Moisture Damange with Advanced Test Methods and Machine Learning

Veeraragavan, Ram Kumar

Etd

Accurate Identification of Pavement Materials Susceptible to Moisture Damange with Advanced Test Methods and Machine Learning

Public

Moisture induced damage in Hot Mix Asphalt (HMA) mixture is a prevalent problem all over the world. It is one of the leading causes of premature failures in asphalt pavements and a significant concern to the paving industry. It is, therefore, necessary to identify mixes that are susceptible to moisture damage during the mix design process. Extensive research has been carried out by several researchers over the years to develop a reliable and practical laboratory test procedure that can simulate field moisture damage conditions and that can make predictions that are likely to correlate to field performance. However, it is inferred from literature that no single laboratory test method can accurately predict the moisture induced damage performance HMA mixtures.\n\nThe objectives of the present study are to: Develop a framework that considers different test methods to predict the moisture induced damage of Hot Mix Asphalt (HMA); Develop a suitable machine learning method to achieve significantly high accuracy in predicting the moisture damage potential of Hot Mix Asphalt (HMA); and develop a tool (App) for use by practicing engineers to identify HMA mixes that are likely to be susceptible to moisture induced damage.\n\nA total of 35 in-plant produced asphalt mixtures with known field performance were sampled, and compacted in the laboratory, and the compacted samples were subjected to mechanical tests before and after moisture conditioning with the Moisture Induced Stress Tester (MiST). In addition, the effluent from the MiST was checked for Dissolved Organic Carbon (DOC) content and gradation of dislodged aggregates. Fourier-Transform Infrared Spectroscopy (FTIR) analysis of the asphalt extracted from HMA samples was performed to observe changes in the functional groups before and after the MiST test.\n\nStatistical analysis showed that seismic modulus and indirect tensile strength were effective in distinguishing poor-performing mixes from the well-performing mixes. Principal component analysis was conducted on the test data, and a reduced set of dimensions that were capable of explaining significant variance in the data was identified. The significant test properties were used to develop machine-learning models with two supervised classification approaches. The k-nearest neighbor model was found to be very accurate in differentiating the mixes. The use of MiST conditioning, specified physical tests, and machine learning methods are recommended for the identification of moisture-susceptible hot mix asphalt.\n\nContribution of this Work\nThe major contribution of this work is the creation of a framework or a system that combines appropriate test methods and suitable machine learning models to achieve high accuracy (84%) in predicting the moisture damage potential of Hot Mix Asphalt (HMA). A secondary contribution is that this study, for the first time, combines the principles of Artificial Intelligence (AI), in the form of Machine Learning (ML), with the field of pavement performance, specifically for the evaluation of mixes that are subjected to moisture damage. Finally, the work provides users with a highly accurate ML model as well as an app, which can be used and further improved.

Creator