Inherent Variability in the Parameters Describing the Linear Viscoelastic Response of Asphalt Concrete

Mechanistic methods are introduced to provide a realistic prediction of the performance of asphalt pavements. Despite the advantage of these methods in providing accurate and advanced material characterization, they should take into account the uncertainties associated with material properties and models. Such material properties include the linear viscoelastic response functions that characterizes the behavior of asphalt concrete such as relaxation modulus E(t). Typically, complex modulus E* data is obtained by lab testing and then used to determine E(t) through mathematical interconversion methods. The objective of this study is to propagate the inherent variability in E* data for various reduced times into E(t). This was achieved by analyzing the data of E* testing and interconverted E(t) from three different mixes that include at least eight replicates with a probabilistic framework that allows for a preliminary quantification of uncertainty. Monte Carlo simulations were used to propagate the uncertainties of the Prony series coefficients to determine the mean, coefficient of variation (COV), and probability distribution in E(t) as a function of reduced time.