Dynamic Modulus Prediction Models for Illinois’ Full-Depth Asphalt Pavement M-E Design Considering Modern Materials

In the state of Illinois, USA, the last major modification of the full-depth asphalt pavement mechanistic-empirical (M-E) design procedures and policies was based on research conducted in 2008. Significant changes and improvements in hot-mix asphalt (HMA) and other material technologies have taken place since that date, including the widespread use of recycled materials and modified binders. The aim of this study was to update the state’s algorithm for modulus prediction of binder courses, considering such materials’ advances. Four mix designs with different recycled material contents were selected, and seven different PG-grade binders (including neat and modified) were used in each mix for dynamic modulus (|E*|) measurements. The ground-truth laboratory results were compared to (i) the current Illinois |E*| algorithm, (ii) traditional Witczak and Hirsch |E*| models, (iii) a recently developed Bayesian Neural Network (BNN) |E*| prediction model from the Illinois Center for Transportation, and (iv) a non-destructive |E*| measurement technique by means of the Ultrasonic Pulse Velocity (UPV). Results showed (i) the current Illinois algorithm underpredicts modulus, (ii) the ICT’s BNN model outperformed Witczak and Hirsch models, raising as a potential replacement for the current algorithm, and (iii) the use of UPV as a modulus surrogate has potential due to its practicality and low cost, but it still relies on assumptions that can compromise accuracy.