Local High Reliability Calibration of Faulting Model Using Pavement Management Data

Conventional approaches to the local calibration of mechanistic-empirical (ME) pavement performance prediction models require the use of field data with low variability. However, these strict data requirements often lead to datasets with a limited number of observations. Local calibrations using such datasets permit the elimination of biases in predictions with 50% reliability, but might not provide an accurate evaluation of predictions of higher design reliability, which can exceed 90%. In this paper, we propose a novel approach to faulting model evaluation for concrete pavements that specifically focuses on high levels of reliability. To address the issue of limited dataset size, our approach leverages pavement management system (PMS) data, which are collected regularly and in large quantities at a local level. We also account for the presence of censored data from out-of-service or modified pavement sections. This permits the local calibration of performance prediction models, with a particular emphasis on the accurate prediction of pavement distress with high reliability levels, which is critical for the design of high-volume roads. To validate our methodology, we applied it to evaluate, modify, and calibrate the Pavement ME faulting model using Pennsylvania PMS data. The proposed methodology can be applied for the local calibration of other Pavement ME models.