The mechanistic empirical pavement design guide (MEPDG) is a state-of-the-art design approach that incorporates material properties, traffic, and climate to estimate the incremental damage using mechanical responses of the pavement. The cumulative damage is used to predict the field distress using empirical transfer functions. The Pavement-ME transfer functions have been nationally calibrated using long-term pavement performance (LTPP) pavement sections and other experimental test section data such as MnRoad. These nationally calibrated models may not represent the construction practices, materials, and climatic conditions of a particular state/region. Studies have calibrated the Pavement-ME transfer functions using the least squares method. Least squares is a widely used simplistic method based on the normal independent and identically distributed (NIID) assumption. Literature shows that these assumptions may not apply to non-normal distributions. This study introduces a new methodology for calibrating the bottom-up cracking, total rutting, and international roughness index (IRI) models in new flexible pavements and the transverse cracking and IRI models in new rigid pavements using Maximum Likelihood Estimation (MLE). The approach in this study includes MLE using synthetic and observed data, and the results are compared with those of the least squares approach. The MLE and least squares methods were also combined with resampling techniques to improve the robustness of calibration coefficients. The data are analyzed from the Michigan Department of Transportation's (MDOT) Pavement Management System (PMS) database to obtain the pavement sections and observed performance data for calibration.Despite several calibration efforts, limited research is available on the impact of calibration on pavement design. The calibrated models using the least squares method were then used for pavement design to estimate the calibration effects and compare them with AASHTO93 designs. Based on the newly calibrated coefficients, 44 new flexible and 44 rigid sections were designed. This study also identifies the controlling distresses for pavement design. It is often not viable to calibrate all coefficients at the same time. Therefore, it is crucial to identify the most sensitive transfer function coefficients. Moreover, the sensitivity also indicates the impact of each coefficient on the performance prediction. Typically, the sensitivity is obtained using a normalized sensitivity index (NSI). This study estimated the sensitivity of the Pavement-ME transfer function coefficients using scaled sensitivity coefficients (SSCs). The results show that MLE outperformed the least squares method for non-normally distributed data, such as transverse cracking and bottom-up cracking models for synthetic and observed data. Using the calibrated models for pavement design showed that, on average, the surface thicknesses using locally calibrated coefficients were thinner by 0.22 and 0.44 inches for flexible and rigid pavements, respectively. Critical design distresses for flexible pavements include bottom-up and thermal cracking. On the other hand, transverse cracking and IRI control the designs for rigid sections. The sensitivity of Pavement-ME model coefficients showed that SSCs provide a more reliable sensitivity on a range of independent variables rather than a point estimate, unlike NSI. Overall, this study helps improve the calibration process for local conditions.
Read
