The dynamic modulus (E*) of asphalt mixtures is essential for Mechanistic-Empirical (ME) pavement designs but is seldom measured by the departments of transportation (DOTs) because of the high cost and complexity of traditional E* tests. This research introduces an Artificial Intelligence (AI)-powered IDEAL-E* test that integrates IDEAL cracking test at various temperatures, finite element analysis, and machine learning. First, the Wiczak E* model was used to generate the E* dataset using a full factorial combination of variables (asphalt binder, aggregate gradations, binder content by volumes, and air voids) and ranges of those variables. That resulted in a total of 11,220 mixtures. The AI model for force-displacement and E* was then trained on 8,976 of those mixtures, and the remaining 2,244 mixtures were used to test the AI model's accuracy. Additionally, the AI model was calibrated and verified with 16 mixtures. The comparison between AMPT-measured and AI-predicted E* values is highly promising with a R² value of 0.97. This innovative approach addresses the limitations of existing models that struggle with evolving asphalt compositions. By simplifying E* data generation, it facilitates its use in AASHTOWare Pavement ME design software and is aligned with current practices with IDEAL cracking tests in DOTs’ Quality Assurance laboratories.
The final report is available here.