Evaluation of Predictive Models for Estimating Dynamic Modulus of Hot-Mix Asphalt in Oklahoma

Long-term performance of an asphalt pavement depends not only on the material properties but also on the stiffness achieved during compaction. Because the determination of stiffness during construction is not straightforward, a common approach uses predictive models to estimate the dynamic modulus of hot-mix asphalt (HMA) specimens. Four predictive models—the Witczak 1999, Witczak 2006, Hirsch, and Al-Khateeb models—were evaluated for their use in estimating the dynamic modulus of selected HMA mixtures that are commonly used in Oklahoma. Five mixes representing various aggregate sources, aggregate sizes, binder grades, and air void levels were tested in the laboratory, and the measured dynamic modulus of each mix was compared with the value predicted by each of the models. The performance of each predictive model was evaluated by three approaches: goodness-of-fit statistics, comparison of the measured and predicted values, and local bias statistics (slope, intercept, and average error). Analyses of the results showed that the predictive power of each model varied with the temperature and air void levels of a compacted specimen. Calibration factors were developed for each model to obtain an accurate estimate of dynamic modulus. The calibration factors are helpful for Level 2 and Level 3 designs of the Mechanistic–Empirical Pavement Design Guide.