A Machine Learning Approach for Load Rating of Bridges
Volume Title: ICASGE2025
Paper ID : 1015-ICASGE-FULL (R1)
Authors
Alfred Benesch and Company
Abstract
Robust Bridge Management System (BMS) is required for resilient communities. A fundamental aspect of BMS is the process of load rating, which equips bridge engineers and administrators with essential information for making decisions about potential bridge rehabilitation strategies, if necessary. However, current load rating methodologies are often time-consuming or financially burdensome. Nonetheless, artificial intelligence (AI) techniques have the potential to offer efficient alternatives to these conventional approaches for load rating. Therefore, this study undertakes a comparison of various regression-based Machine Learning (ML) techniques to assist bridge engineers with making decisions related to load rating of bridges. The study establishes a comprehensive database named Bridge Net (BrNet), encompassing more than 260 bridges situated in the state of Michigan. These bridges are meticulously modeled using the widely adopted AASHTOWare Bridge Rating (BrR) software, encompassing variables such as condition ratings, bridge material, system type, beam strengths, number of beams, and other pertinent parameters. Through the application of ML techniques to the BrNet database, the study strives to contribute to the optimization of bridge management systems. This has the potential to lead to more informed decision-making and optimized allocation of resources in the realm of bridge maintenance and rehabilitation efforts.
Keywords
Subjects