A probabilistic and adaptive approach to modeling performance of pavement infrastructure

Access full-text files




Li, Zheng

Journal Title

Journal ISSN

Volume Title



Accurate prediction of pavement performance is critical to pavement management agencies. Reliable and accurate predictions of pavement infrastructure performance can save significant amounts of money for pavement infrastructure management agencies through better planning, maintenance, and rehabilitation activities. Pavement infrastructure deterioration is a dynamic, complicated, and stochastic process with its outcome as the aggregated impact from various factors such as traffic loading, environmental condition, structural capacities, and some unobserved factors. However, existing performance prediction models are still constrained by inadequate consideration of the dynamic and stochastic characteristics of pavement infrastructure deterioration. The goal of this research is to develop a probabilistic and adaptive methodological framework that is capable of capturing the dynamic and stochastic nature of pavement deterioration processes. The ordered probit model and the sequential logit model as probabilistic models are proposed to directly predict the performance of pavements in terms of their condition states by relating the performance to the structural, traffic, and environmental variables. The proposed probabilistic models were pilot-tested with pavement performance data collected during the AASHO Road Test, yielding promising preliminary results. In addition, these models were further enhanced as mechanistic-empirical models by incorporating certain primary response variables of pavements as explanatory variables. The comparison results show that the proposed models yield better predictions than the previously developed models. Then, a structural state space model is proposed to characterize the dynamic nature of pavement deterioration. The structural model allows the prediction of pavement deterioration to be adaptively updated with new inspection data, taking advantage of a polynomial trend filter and the Kalman filter algorithm. The preliminary results from a simulation case study indicate that the adaptive algorithm is robust and responsive to structural deviations of the pavement deterioration process.