Evaluation of Measured and Estimated Dynamic Moduli for Selected Asphalt Mixes

An accurate measurement or estimation of dynamic modulus (|E*|) of a hot mix asphalt (HMA) mix is important to understand stress–strain behavior of flexible pavements under loading and unloading conditions. The Mechanistic Empirical Pavement Design Guide (MEPDG) (Transportation Research Board, Washington, D.C., 2004) recommends that |E*| be used in all three levels of design (i.e., Level 1, Level 2, and Level 3). For Level 1, |E*| is measured in the laboratory, while it is estimated using the Witczak models for Level 2 and Level 3 designs. The measurement of |E*| in the laboratory is not always feasible because it requires costly equipment and skilled personnel. Consequently, use of empirical models seems to be a reasonable approach to estimate |E*|. Several researchers have reported that the accuracy of the Witczak models varies with local materials and volumetric properties. The present study was undertaken to compare the measured and the estimated |E*| for some commonly used mixes in Oklahoma. Specifically, |E*| of five different HMA mixes, comprised of aggregates from several sources and sizes, binder grades, and air voids, were measured in the laboratory. The Witczak 1999 model (Andrei , 1999) was used to estimate |E*| for each of these mixes. A comparison was made between the measured and the estimated |E*| at four different levels of air voids, namely, 6%, 8%, 10%, and 12%. It was observed that the Witczak 1999 model overestimates |E*| at all four levels of air voids. To address these overestimates, the Witczak 1999 model was calibrated. The calibrated model was similar in form to the Witczak 1999 model but having different numerical coefficients. Verification of this model was done using a mix that was not used in the calibration process. Furthermore, two full depth field cores were obtained to further verify the accuracy of the calibrated model. Three different criteria, namely, goodness-of-fit statistics, matching the measured and the estimated |E*|, and average relative error (%), revealed that the calibrated model exhibits much better performance compared to the Witczak 1999 model. It is expected that the calibrated model would be useful in estimating |E*| for the Level 2 and Level 3 designs for the implementation of the MEPDG in Oklahoma.