Review of Network-Level Pavement Performance Modeling Approaches throughout the US and Critical Assessment of Deterministic Methods

Abstract

Abstract This paper presents a detailed review of pavement performance modeling methods found in literature and reported to be used by transportation agencies in the US. It was found that deterministic models are the dominant type for network-level pavement management applications in the US, and the majority of those models are equations found by sorting the pavement condition data into families and performing regression. Four different approaches to performance modeling were identified as being used in network-level pavement management models: a single family model fit to all data, models fit to individual pavement segments, and two approaches that combined family and segment model information. A detailed evaluation of these approaches using synthetically generated data and data from a highway agency showed that those that combined family and segment information consistently performed better in terms of minimizing prediction error. Overall, this paper shows that, while a significant amount of research on performance models exists in literature, there is a lack of research on methods to better leverage family and segment information in network-level pavement performance modeling.