Streamlining Use of Mechanistic–Empirical Pavement Design Guide

Abstract

The NCHRP Mechanistic–Empirical Pavement Design Guide (MEPDG) is a powerful tool for the design and analysis of highway pavements. Currently, the designer uses an iterative process to select design parameters and predict performance. If performance is unacceptable, the designer must change the design parameters until acceptable performance is achieved. A procedure outlined here used multiple adaptive regression splines (MARS) to develop predictive models obtained from a series of actual pavement design solutions from MEPDG software. Two model structures were developed: a series of models to predict individual pavement distress (rutting, fatigue cracking, and roughness) and a forward solution to predict pavement thickness from a desired level of distress. This model can be used to determine a starting thickness, which then can be used with the MEPDG software to obtain a refined design, potentially with fewer iterations to define an acceptable design thickness. These thickness prediction models can be developed for any subset of desired MEPDG solutions, such as typical designs within a state. The new DARWin-ME software can facilitate the development of these subsets through its sensitivity analysis procedure, which would allow the designer to develop a subset of solutions across a typical range of input values. Results could then be modeled with the MARS process to produce an efficient design solution for pavement thickness along with a quick performance prediction for specific pavement thickness. This procedure can significantly reduce the MEPDG iterations necessary to develop a viable design, to as few as two to three on either side of the predicted MARS model thickness. These results could be merged with a smart decision tree structure to execute the MEPDG procedure efficiently.