Local Calibration of the Fatigue Prediction Models in the MEPDG for Pavement Rehabilitation in Oregon

摘要

The performance prediction models within the AASHTO Mechanistic-Empirical PavementDesign Guide (MEPDG) were calibrated primarily using design inputs and performance datalargely from the national Long-Term Pavement Performance (LTPP) program. Beforeimplementing the MEPDG at the state level, performance (distress) prediction models warrantdetailed validation and calibration because of potential differences between national and localconditions. The Oregon Department of Transportation (ODOT) is in the process of implementingthe new MEPDG for new pavement sections. However, the vast majority of pavement workconducted by ODOT involves rehabilitation of existing pavements. Hot mix asphalt concrete(AC) overlays are the preferred rehabilitation treatment for both flexible and rigid pavements inOregon. However, like new work sections, AC overlays are also susceptible to fatigue cracking(alligator cracking and longitudinal cracking), rutting, and thermal cracking. Additional work istherefore needed to calibrate the design process for rehabilitation of existing pavementstructures. A detailed comparison of predictive and measured distresses was made using recentlyMEPDG released software Darwin M-E (version 1.1). It was found that Darwin M-E predictivedistresses did not accurately reflect measured distresses, calling for a local calibration ofperformance prediction models is warranted. While the local calibration of rutting and thermalcracking prediction models is currently underway, alligator (bottom-up) cracking andlongitudinal (top-down) cracking models were calibrated. The Microsoft Excel Solver wasemployed to optimize the calibration coefficients by minimizing the sum of the squared errors(SSR) between the predictive and measured distresses. A comparison was made between theresults before and after the calibration to assess the improvement in accuracy of the distressprediction models provided by the local calibration. Both alligator cracking and longitudinalcracking models were improved by local calibration. However, there was a high degree ofvariability between the predicted and measured distresses, especially for longitudinal cracking,even after the calibration. It is recommended that additional sites, which would contain moredetailed inputs (mostly Level 1), be established to be included in the future calibration effortsand thus, improve the accuracy of the prediction models.