A Bayesian approach for improved pavement performance prediction

Abstract

We present a method for predicting future pavement distresses such as longitudinal cracking. These predicted distress values are used to plan road repairs. Large inherent variability in measured cracking and an extremely small number of observations are the nature of the pavement cracking data, which calls for a parametric Bayesian approach. We model theoretical pavement distress with a sigmoidal equation with coefficients based on prior engineering knowledge. We show that a Bayesian formulation akin to Kalman filtering gives sensible predictions and provides defendable uncertainty statements for predictions. The method is demonstrated on data collected by the Texas Transportation Institute at several sites in Texas. The predictions behave in a reasonable and statistically valid manner.

Keywords:

• pavement management information system
• Bayesian adjustment
• state-space models
• Kalman filtering
• Markov chain Monte Carlo

Acknowledgements

The authors wish to thank the Texas Department of Transportation (TxDOT) for financial support (Grant 0-4186) for this work. The contents of this paper reflect the views of the authors who are responsible for the opinions, findings, and conclusions presented herein. The contents do not necessarily reflect the official view or policies of the TxDOT. This paper does not constitute a standard, specification, or regulation.