Abstract
Many state highway agencies are in the process of implementing the AASHTOWare Pavement ME Design (PMED) software for routine pavement design. However, a recurring implementation challenge has been the need to locally calibrate the software to reflect an agency’s design and construction practices, materials, and climate. This study introduced a framework to automate the calibration processes of the PMED performance models. This automated technique can search PMED output files and identify relevant damages/distresses for a project on a particular date. After obtaining this damage/distress information, the technique conducts model verification with the global calibration factors. Transfer function coefficients are then automatically derived following an optimization technique and numerical measures of goodness-of-fit. An equivalence statistical testing approach is conducted to ensure predicted performance results are in agreement with the measured data. The automated technique allows users to select one of three sampling approaches: split sampling, jackknifing, or bootstrapping. Based on the sampling approach chosen, the automated technique provides the calibration coefficients or suitable ranges for the coefficients and shows the results graphically. Model bias, standard error, sum squared error, and p-value from the paired t-test are also reported to assess efficacy of the calibration process.
Get full access to this article
View all access options for this article.