Developing Bridge Deterioration Model Using Artificial Neural Network and Markov Chain
No Thumbnail Available
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Most transportation agencies in the U.S. are facing the challenge of fixing the aging transportation infrastructures with insufficient budget. Pavements and bridges are the two major components of transportation infrastructures. Bridges in very poor condition could become unsafe for the traveling public to drive across. Deteriorating bridge condition coupled with ever increasing costs to maintain, repair, and rehabilitate bridges means difficult budget allocation decisions must be made to keep all bridges in safe operating condition and extending the service life of existing bridges. The existing and projected condition of a bridge is therefore an important input for the decision-making process.
Many transportation agencies utilize Bridge Management System (BMS) to help with managing thousands or, sometimes, tens of thousands of bridges. BMS enable agencies to make critical rehabilitation and reconstruction decisions based on systematically collected bridge condition data and projected deterioration trends.
This study focuses on developing bridge condition deterioration models to help provide a more accurate prediction of future bridge conditions. Historical bridge condition data for bridges under the jurisdiction of the Ohio Department of Transportation from 1992 to 2019 were obtained from the National Bridge Inventory (NBI) database. These data include ratings for bridge deck, superstructure, and substructure of each bridge, as well as various characteristics of that bridge, such as age of bridge (years in service), bridge materials, structure type, length, width, maintenance done, etc.
Two condition prediction models, one based on the Artificial Neural Network (ANN) method, and the other based on the Markov Transitional Probability method, were developed. The results show that the ANN model can produce significantly better results than the Markov model in predicting future bridge condition ratings.