Hierarchical Markov Chain Monte Carlo and pavement roughness model

Date
2010
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Traditionally, pavement roughness has been modeled to mimic heterogeneity across pavement sections. Modeling heterogeneity is challenging and can generate models that are unable to reflect true pavement conditions. Heterogeneity is fundamental to modeling pavement roughness and describes how road conditions change continuously with corresponding time change. However, road conditions are unpredictable and this feature raises inherent challenges when modeling heterogeneity across pavement sections. This thesis seeks to model the roughness of road pavements in Kansas using hierarchical Markov Chain Monte Carlo (MCMC) simulation. The aim is to investigate how efficient this technique is at estimating and predicting pavement roughness without neglecting inherent heterogeneity across pavement sections. Hierarchical MCMC models use Bayesian approach in their estimation process which allows them to account for heterogeneity in pavement roughness. Models easily lend themselves to validation and can be examined to see if they reflect roughness conditions on a specified length of roadway or a network of roads. Using individual lengths of pavements and a nineteen year time span, a hierarchical MCMC model is used to predict the IRI value for the twentieth year. Estimated IRI values are then compared with original IRI values to see how well they correlate and if they reflect prevailing road conditions. Once proven to be successful, this technique can be incorporated into pavement management systems and used as a basis for making sound decisions about the level of roughness on a given road network.
Description
Keywords
Citation