Probabilistic characterization of the viscoelastoplastic behavior of asphalt-aggregate mixtures -

Abstract

The objective of this work is to provide accurate and realistic characterization of different types of asphalt concrete mixtures using advanced material modeling within a probabilistic framework. The methodology adopted builds on and enhances a viscoelastoplastic continuum damage (VEPCD) material model by utilizing a suite of associated experimental testing protocols and incorporating the uncertainties associated with the different material properties. The developed framework is then applied to assess the behavior of different types of unconventional asphalt concrete mixtures such as Warm Mix Asphalt (WMA), Fiber-Reinforced Asphalt Concrete (FRAC), and those containing Reclaimed Asphalt Pavement materials (RAP). The modeled uncertainties address the variabilities and errors associated with the linear viscoelastic (LVE) functions achieved from the complex modulus test and damage characteristic curves obtained from constant crosshead rate testing. A probablistic scheme using First Order approximations and Monte Carlo simulations is developed to characterize the inherent uncertainty of each of the LVE functions (dynamic modulus, relaxation modulus E(t), and creep compliance D(t)) over the time domain of their mastercurves. The results show that the quantified uncertainties are significant especially at high temperatures and-or slow loading rates. Based on the results of several investigated mixtures, the inherent uncertainty of LVE properties of asphalt concrete becomes higher for mixtures with a larger nominal maximum aggregate size, mixtures with modified binders, and-or mixtures with WMA additives. At small reduced times, the uncertainty in dynamic modulus, E(t), and D(t) are similar in magnitude; however, differences become significant at large reduced times implying that modeling the uncertainty of either of these functions is not enough to represent that of the other ones. The sources of uncertainty in these functions are categorized and their influence are tested where fitting techniques yield uncertainty u