The network-level Pavement Management System (NPMS) relies upon performance models to forecast needs and develop multi-year programs. These models are dominantly empirical and often utilized without considering the error surrounding them. Some local models have been shown to provide an R2 as low as 14%. Without demonstrating reasonable accuracy, these models cannot be considered as a rational means of supporting decisions in the NPMS. Furthermore, engineers must be provided a method of quantifying the risk associated with performance models before using them to make decisions. In a survey sent to all U.S. and Canadian state/regional highway agencies, 95% ofthe responding agencies reported neglecting model error in the NPMS. While material variability and model uncertainties are addressed in project-level pavement design, these concepts are absent at the network level. Using the Nebraska Department of Roads (NDOR) as a model case, this research implements a reliability concept in the NPMS. The proposed concept utilizes a recent study which implements mechanistic-empirical (M-E) modeling in the NPMS. In a retroactive 5-year needs assessment of 86 sections, field data shows 23 of the 86 sections would have qualified for entry into the 5-year program. Estimating needs with M-E models would have resulted in 44 candidate sections compared to 85 estimated with linear models. Results show a reliability concept can improve the estimated needs. When using an 85%level of reliability, the estimation provided by the M-E models improves to 32 sections, a 70% increase in accuracy with respect to estimated needs.
Samenvatting