Quantification of the inherent uncertainty in the relaxation modulus and creep compliance of asphalt mixes

Abstract

Advanced material characterization of asphalt concrete is essential for realistic and accurate performance prediction of flexible pavements. However, such characterization requires rigorous testing regimes that involve mechanical testing of a large number of laboratory samples at various conditions and set-ups. Advanced measurement instrumentation in addition to meticulous and accurate data analysis and analytical representation are also of high importance. Such steps as well as the heterogeneous nature of asphalt concrete (AC) constitute major factors of inherent variability. Thus, it is imperative to model and quantify the variability of the needed asphalt material’s properties, mainly the linear viscoelastic response functions such as: relaxation modulus, $E(t)$ , and creep compliance, $D(t)$ . The objective of this paper is to characterize the inherent uncertainty of both $E(t)$ and $D(t)$ over the time domain of their master curves. This is achieved through a probabilistic framework using Monte Carlo simulations and First Order approximations, utilizing $E^{*}$ data for six AC mixes with at least eight replicates per mix. The study shows that the inherent variability, presented by the coefficient of variation (COV), in $E(t)$ and $D(t)$ is low at small reduced times, and increases with the increase in reduced time. At small reduced times, the COV in $E(t)$ and $D(t)$ are similar in magnitude; however, differences become significant at large reduced times. Additionally, the probability distributions and COVs of $E(t)$ and $D(t)$ are mix dependent. Finally, a case study is considered in which the inherent uncertainty in $D(t)$ is forward propagated to assess the effect of variability on the predicted number of cycles to fatigue failure of an asphalt mix.