Improving Prediction Accuracy in Mechanistic-Empirical Pavement Design Guide.

Model calibration plays a fundamental role in the implementation of the Mechanistic-Empirical Pavement Design Guide. While it is seldom possible to predict pavement distress behavior exactly, statistical methods provide maximum agreement between measured and predicted distress through calibration. The Long Term Pavement Performance (LTPP) database afforded the default calibration effort in the Design Guide with data spanning the climatic conditions of several states and provinces, encompassing all traffic levels, materials, and thicknesses. State Highway Agencies vying for calibration data must rely upon the local Pavement Management System (PMS) when LTPP data availability proves scarce. As part of the local calibration effort for the state of Nebraska, this paper explores a concept for maximizing its accuracy potential. The primary objective for this study investigates the use of "focus calibrations" as a vehicle to improve prediction accuracy in the Design Guide. "Focus calibration" follows the concept of narrowing a dataset to reflect similar properties found at the project-level rather than representing the entire network. The narrowed or "focused" dataset is then used to calibrate the pavement distress model. Under this study, Nebraska Department of Roads (NDOR) PMS data was used to calibrate 2 Design Guide smoothness models using annual daily truck traffic (ADTT) and surface layer thickness as focus parameters. Results showed that focus calibrations reduced default model prediction error by nearly twice that of network calibration. This study offers a window into the accuracy that can be achieved with local focus calibrations of Design Guide prediction models.