Machine Learning Applications for Roadway Pavement Deterioration Modeling

Abstract

Roadway and highway agencies across the globe spend a sizable fraction of their annual budget for the upkeep and maintenance of roadways. Different road segments deteriorate at different rates owing to variable traffic flow along the segments. In previous works, various forms of mathematical formulations were provided for roadway maintenance and pavement deterioration modeling. Numerical solutions algorithms using linear programming, dynamic programming, and genetic algorithms were proposed. The solution algorithms, however, did not benefit from the prescriptive and predictive capabilities of machine learning (ML) algorithms (e.g., random forest classifier, support vector machine, and artificial neural networks). Furthermore, previous methods treated transition probabilities of condition states of a pavement in future years to be static. In this paper, a variable transition probability is introduced based on the deterioration rate of a pavement over time. A modified capacitated arc routing formulation is developed for a highway infrastructure management information system. Prescriptive and predictive analytics are performed using ML to analyze the road network in simulation studies and from Montgomery County, Maryland, USA. The pavement condition index (PCI) for the road network is predicted using ML algorithms. The results show a good promise for PCI prediction based on variable deterioration rate and for obtaining condition states in future years subject to varying transition probabilities.

Received: 2 November 2023 | Revised: 27 November 2023 | Accepted: 5 December 2023

Conflicts of Interest

Manoj K. Jha is an editorial board member for Journal of Computational and Cognitive Engineering and was not involved in the editorial review or the decision to publish this article. The author declares that he has no conflicts of interest to this work.