Selection of Design Parameters for Construction of a Surrogate Model for the MEPDG

The American Association of State Highway and Transportation Officials’ (AASHTO) Mechanistic-Empirical Pavement Design Guide (MEPDG) is the most widely utilized M-E design procedure in the U.S. The MEPDG incorporates extensive climatic and empirical performance data. Inclusion of such extensive data hinders the computational efficiency of design with the MEPDG and likely includes design parameters of little significance to pavement performance. Construction of a surrogate model that emulates the behavior of the MEPDG can be utilized to reduce computational expense and is valuable in any situation where large numbers of evaluations are required. Development of a surrogate improves efficiency, but requires selection of a reduced quantity of design parameters compared to the MEPDG model that it is meant to emulate. This paper includes investigation of the required quantity of input parameters necessary to accurately imitate the MEPDG design procedure. This investigation will provide a tool for sensitivity analysis for the MEPDG. The objective of this paper is to develop a variable selection process for sensitivity analysis and construction of a surrogate model for the Level 1 MEPDG design procedure utilizing the DARWin-ME design software. Investigation of the importance of input variables is investigated for 3 distinct climate regions to examine the impact of climatic data on the relative significance of input variables. Sensitivity analysis will provide insight to the most significant design information for Level 1 analyses. Further, construction and verification of a surrogate model for a Level 1 MEPDG analysis incorporating the information obtained from the selection process is presented.