Unified Approach for Uncertainty Analysis Using the AASHTO Mechanistic-Empirical Pavement Design Guide

Abstract

This paper addresses the problem of uncertainty management for the Mechanistic-Empirical Pavement Design Guide (MEPDG). Specifically, development of an all-inclusive approach to uncertainty management in which model uncertainty and input parameter variability are taken into consideration is presented in this paper. A Gaussian process surrogate model is chosen in this paper as a potential model for the MEPDG design software, and model verification metrics are included to assess the ability of the surrogate model to predict MEPDG output. The construction of a surrogate model that emulates the MEPDG will reduce the computational expense for uncertainty analysis but will introduce another source of uncertainty. Uncertainty propagation of these major sources of uncertainty is performed for a case study. The relative effects of these three major sources of uncertainty are investigated, and the influences of these sources of uncertainty are discussed utilizing the results from the uncertainty propagation. Reliability analysis incorporates the proposed methodology to investigate the effect of these sources of uncertainty for a typical flexible pavement design.