In an effort to move toward pavement designs that employ mechanistic principles, the AASHTO Joint Task Force on Pavements initiated an effort in 1996 to develop an improved pavement design guide. The project called for the development of a design guide that employs existing state-of-the-practice mechanistic-based models and design procedures. The product of this initiative became available in 2004 in the form of software called the Mechanistic-Empirical Pavement Design Guide (MEPDG). The performance prediction models in the MEPDG were calibrated and validated using performance data measured from hundreds of pavement sections across the United States. However, these nationally calibrated performance models in the MEPDG do not necessarily reflect local materials, local construction practices, and local traffic characteristics. Therefore, in order to produce accurate pavement designs for the State of North Carolina, the MEPDG distress prediction models must be recalibrated using local materials, traffic, and environmental data. The North Carolina Department of Transportation (NCDOT) has decided to adopt the MEPDG for future pavement design work and has awarded a series of research projects to North Carolina State University. The primary objective of this study is to calibrate the MEPDG performance prediction models for local materials and conditions using the data and findings generated from this series of research projects.
The work presented in this report focuses on four major topics: (1) the development of a GIS-based methodology to enable the extraction of local subgrade soils data from a national soils database; (2) the rutting and fatigue cracking performance characterization of twelve asphalt mixtures commonly used in North Carolina; (3) the characterization of local North Carolina traffic; and (4) calibration of the flexible pavement distress prediction models in the MEPDG to reflect local materials and conditions.
The scope of this research is limited to rutting and fatigue cracking. The total number of sections available for this study is 46 sections: 22 long-term pavement performance (LTPP) sections (6 SPS and 16 GPS sites) and 24 non-LTPP sites. Because the LTPP sites have more complete distress and materials information available than the other sites, the research team used all the LTPP sites for calibration and used the 24 non-LTPP sites for validation. Some of the LTPP sections and many of the NCDOT sections were found to lack data, however, so MEPDG defaults and engineering judgment were used to populate the missing data.
For the subgrade soil characterization, a GIS-based methodology was developed so that NCDOT engineers can superimpose road sections of interest on the NCHRP 9-23A soil maps and find the corresponding alphanumeric soil unit code that is required to extract the related soil information. Material-specific hot mix asphalt (HMA) rutting and fatigue cracking model coefficients were developed for the twelve most commonly used HMA mixtures in North Carolina using data obtained from the triaxial repeated load permanent deformation (TRLPD) test and the direct tension cyclic test, respectively. These test data, in addition to the dynamic modulus data determined from the HWY-2003-09 project, Typical Dynamic Moduli for North Carolina Asphalt Concrete Mixes, have resulted in the North Carolina MEPDG materials database.
A North Carolina MEPDG traffic database has been established based on the research efforts from the HWY-2008-11 project, Development of Traffic Data Input Resources for the Mechanistic Empirical Pavement Design Process, and from this study. This database was developed based on a multidimensional clustering methodology and a pavement damage-based approach in order to characterize local traffic and to develop traffic catalogs for the traffic parameters required as inputs in the MEPDG.
The initial MEPDG verification runs reveal that, when the MEPDG national default calibration values are used, the rut depth and fatigue cracking predictions are significantly different from the measured values. Two approaches were used to calibrate the rutting and fatigue cracking models for local conditions and materials. The first approach uses the generalized reduced gradient (GRG) method, whereas the second approach uses the genetic algorithm (GA) optimization technique. The GA-based approach is found to result in statistically better total rut depth and alligator cracking predictions than the GRG method.
The local calibration of the MEPDG is found to reduce bias and standard error between the predicted and measured rut depth and fatigue cracking percentage values. However, the improvement is not enough to accept the null hypothesis that the measured values are equal to the predicted values at the 95% confidence interval. The calibration results demonstrate the importance of using material specific performance test results, having detailed and reliable distress data, and taking permanent deformation measurements from individual layers through forensic investigation. This study results in a set of local calibration factors for the permanent deformation and fatigue cracking performance prediction models in the MEPDG for the State of North Carolina and the North Carolina MEPDG User Reference Guide, along with a list of future research recommendations.
