Pavement ME Transverse Cracking Model Calibration Using California’s Pavement Management System

This paper presents a new approach for calibration of mechanistic-empirical (ME) pavement simulation models. It uses network level performance data from the pavement management system (PMS), and was applied for the local calibration of the Pavement ME transverse cracking models in California, using 30,155 pavement sections with a combined length of approximately 4,400 lane-miles built on 446 lane replacement projects completed between 1947 and 2017. This represents two orders of magnitude more observations, sections, and length of pavement than are typically used when following the traditional ME calibration approach for all pavement types. The new approach does not require sampling and testing of materials from all sections in the network, but rather uses the median values from a representative sample of materials across the network (state-wide median). This approach is appropriate because pavement designers only know pavement material property specification limits in a design-bid-build delivery process. Variability of performance and reliability of design (probability that the design will meet or exceed the design life) are accounted for through separate consideration of within-project, between-contractor, and between-project variability. The calibration against PMS performance data reduced the significant prediction bias and standard error (13.3 percent and 23.03 percent cracked slabs, respectively) that would result from use of the nationally calibrated cracking models in California, with the bias and standard error reduced to 0.039 percent and 5.69 percent, respectively.