Comparison of Markovian-Based Bridge Deterioration Model Approaches

Abstract

Bridge management systems are a critical component in the toolbox of those who are responsible for maintaining a population of bridges. Deterioration models are generally incorporated in bridge management systems, but minimal consideration is paid to how those models work and how the assumptions inherent to the model might influence the prediction. This paper identifies, synthesizes, and assesses typical bridge deterioration model approaches from the stochastic family of models. Each model considered is applied to two data sets for bridges in Texas and compared. A novel modeling approach that considers all models together is described. The novel approach demonstrates the value of considering multiple models when attempting to predict a future condition or behavior. It was found that a simple multiple model approach inherently and transparently reduces the uncertainty of a single model approach.