Alternatives for calibration of dynamic modulus prediction models of asphalt concrete containing RAP

Dynamic modulus is a measure of stiffness of an asphalt concrete (AC) mix when subjected to cyclic sinusoidal compressive stresses. In the Mechanistic-Empirical Pavement Design Guide (MEPDG) and its software Pavement ME, dynamic modulus (E∗) is an essential parameter for the prediction of asphalt pavement distresses such as rutting and fatigue cracking. Several empirical models have been developed to estimate the E∗ from AC mix properties to compliment laboratory measured E∗ values. Two models developed under NCHRP 1-37A and NCHRP 1-40D projects have been integrated into the MEPDG program. The models estimate E∗ values when Level 2 and Level 3 inputs for AC mixes are used. This paper presents the evaluation of uncalibrated E∗ values obtained from NCHRP 1-37A and NCHRP 1-40D and compares the results to two calibrated techniques; an exponential fit of uncalibrated model outputs, and updated model coefficients using nonlinear multiple regression. In total, 51 specimens from 17 types of AC mixes containing 0-50% RAP were prepared and tested in the laboratory. E∗ was determined in the laboratory and compared to estimated values based on calibrated and uncalibrated models. The results showed that uncalibrated NCHRP 1-37A produced lower error in predicting E∗ than uncalibrated NCHRP 1-40D. Both calibration techniques enhanced the accuracy of the models, however nonlinear multiple regression showed the best potential for predicting E∗. Calibrated models showed improvement in prediction of E∗ for all RAP mixes for Level 3 inputs, although high RAP mixes showed the least improvement among the other RAP mixes.