Maneuvering MEPDG Input Variables to Improve Level-3 and Level-2 Mastercurve Predictions to the Accuracy of Level-1 Input Hierarchy

摘要

In this study, mix and binder input variables were optimized to investigate the problems related to the accuracy of mastercurves developed using Mechanistic Empirical Pavement Design Guide (MEPDG). Dynamic modulus testing over a wide temperature and frequency range was performed on Superpave mixes typically used for structural and surface coarse pavement construction by New Mexico Department of Transportation (NMDOT). Dynamic modulus mastercurves were produced using the MEPDG software and using Microsoft Excel for the actual test data. Mastercurves developed using actual test results were compared with mastercurves produced using MEPDG software at level-1, level-2, and level-3 input hierarchies. Finally, mix and binder input variables were optimized to determine appropriate shift factors to improve accuracy of mastercurves developed at level-2 and level-3 input hierarchies. The results show that mastercurves developed using level-1 input hierarchy accurately represent the test results. However, all predicted mastercurves at level-2 and level-3 input hierarchies underpredict actual test results and level-3 prediction results showed better accuracy than level-2 outputs for all mixes in this study. The MEPDG software was also re-run to predict mix mastercurves using optimized (shifted) input values and the resulting mastercurves from level-3 and level-2 MEPDG input hierarchies were found to overlap with mastercurves produced using level-1 MEPDG input hierarchy. Optimized mix input values suggest that aggregate variables can be eliminated from the E~* model.