Estimating bridge characteristics with only situation characteristics using Bayesian networks

Abstract

When deciding where a new road or rail line should be built, it is useful to have some idea as to the characteristics of the bridges to be built. One way to do this is to have engineers develop preliminary designs for the bridges that would likely be included in each route. The developed preliminary design, then, stands as a basis for cost estimation and the detailed design of the structure. This would, of course, require a significant amount of time and effort. Another way to do this would be to exploit existing data on the types of bridges built in different situations using Bayesian networks. In this article, it is shown how this can be done using the information in existing databases to estimate bridge characteristics, knowing only the values of a number of situational characteristics, e.g., span to be covered, what the bridge carries, and what the bridge traverses.