Calibration of Fatigue Cracking and Rutting Prediction Models in Pennsylvania Using Laboratory Test Data for Asphalt Concrete Pavement in AASHTOWare Pavement ME Design

The Pennsylvania Department of Transportation (PennDOT) is transitioning from empirical design procedures to the American Association of State Highway and Transportation Officials (AASHTO) Mechanistic-Empirical Pavement Design Guide (MEPDG) procedure and accompanying software AASHTOWare Pavement ME Design (Pavement ME) into its routine pavement design practice. Some aspects of the PennDOT pavement design practices were modified to implement the Pavement ME as a PennDOT standard. In addition, the implementation of the Pavement ME in Pennsylvania required local calibration to develop unbiased distress and smoothness prediction models suited to Pennsylvania’s local conditions. This paper discusses the local calibration efforts undertaken to accurately predict the performance of fatigue cracking and total pavement rutting of asphalt concrete pavements in Pennsylvania using laboratory data from intermediate temperature indirect tensile strength and repeated load permanent deformation tests. As part of this study, the fatigue cracking and total pavement rutting global models were calibrated to reflect Pennsylvania-specific traffic, materials and soils, and environmental conditions, and PennDOT’s commonly adopted pavement construction practices. The local calibration of the Pavement ME global performance prediction models was conducted using nonlinear model optimization techniques. The new local calibration coefficients were developed for each performance prediction model using Pennsylvania-specific design inputs and performance data. Overall, the local calibration process helped remove the bias (consistent over- or under-prediction), reduce the prediction error, and improve the accuracy of performance models. The improvement in these models was notable, thus indicating that the Pennsylvania-calibrated models would provide much more accurate, reliable, and cost-effective designs than the global performance prediction models.