Using Bayesian networks to estimate bridge characteristics in early road designs

Vassilis Panopoulos, Apostolos Bougas, Borja Garcia De Soto, Bryan T. Adey

Research output: Contribution to journalArticlepeer-review

Abstract

When deciding where to build new roads, it would be useful to obtain quickly and reliably an idea of the necessary characteristics of any potential bridges, using limited information and without considerable effort, as there is a considerable amount of information on built bridges in a standardised form, and there are robust algorithms for analysing these data. This study presents a methodology for estimating the likely bridge characteristics using the information available in a bridge database and Bayesian networks. The methodology is demonstrated by estimating the bridge characteristics of 1793 bridge records using nine situational characteristics-for example, the cross-section of the bridge superstructure and number of bridge spans. It is concluded that the methodology is a useful tool when estimating the characteristics of new bridges using only situational information. Compared with naïve-search databases queries, the prediction capability of all networks developed using the proposed methodology showed an estimated accuracy above 86.5%, which is considerably higher than that found when the methodology was not used-that is, 66.5%. Additionally, it is shown that Bayesian networks based on expert experience can obtain results similar to, and in many cases even better than, those of Bayesian networks based solely on learning algorithms.

Original languageEnglish (US)
JournalInfrastructure Asset Management
DOIs
StateAccepted/In press - 2021

Keywords

  • artificial intelligence
  • bridges
  • infrastructure planning

ASJC Scopus subject areas

  • Building and Construction
  • Geography, Planning and Development
  • Safety Research
  • Transportation
  • Public Administration

Fingerprint

Dive into the research topics of 'Using Bayesian networks to estimate bridge characteristics in early road designs'. Together they form a unique fingerprint.

Cite this