A Data-Driven Approach for Automated Operational Safety Evaluation of the National Inventory of Reinforced Concrete Slab Bridges

The uncertainty and subjectivity in current bridge load rating and posting practices, especially for structures with limited or missing design plans or documentations calls for alternative data- driven operational safety evaluation solutions. This study proposes a method for automated assessment of bridge load postings that leverages the National Bridge Inventory data by using emerging machine learning techniques. The proposed method involves the extraction of patterns between general geometrical, functional and environmental bridge descriptors with the posting status in the national inventory of reinforced concrete slab bridges. Decision Tree and Random Forest classification algorithms were selected as the primary tools and were carefully trained, tested and tuned on a dataset of over 40,000 bridges considering several resampling and cost- sensitive classification scenarios to deal with the inherent class imbalance. Upon the selection and tuning of final models, validations were performed on an independent unseen dataset of over 5,000 bridges and generalizability of models was confirmed. Finally, the application of the same models on the population of bridges without plans demonstrated that 17% of currently unposted bridges and 8.4% of currently posted bridges need further evaluations thus highlighting the relative inefficiency of human judgment in detecting hidden deterioration processes. With the promising results obtained in this investigation, the proposed approach is expected to be extended on other classes of structures and other safety and functionality measures providing a step towards data-driven operational safety evaluation.

• 
• Supplemental Notes:
  • This paper was sponsored by TRB committee AHD35 Standing Committee on Bridge Management. Alternate title: Data-Driven Approach for Automated Operational Safety Evaluation of the National Inventory of Reinforced Concrete Slab Bridges
• Corporate Authors:

  Transportation Research Board

  500 Fifth Street, NW
  Washington, DC  United States  20001
• Authors:
  • Alipour, Mohamad
  • Gheitasi, Amir
  • Harris, Devin K
  • Ozbulut, Osman E
  • Barnes, Laura E
• Conference:
  • Transportation Research Board 95th Annual Meeting
  • Location: Washington DC, United States
  • Date: 2016-1-10 to 2016-1-14
• Date: 2016

Language

• English

Media Info

• Media Type: Digital/other
• Features: Figures; References; Tables;
• Pagination: 15p
• Monograph Title: TRB 95th Annual Meeting Compendium of Papers

Subject/Index Terms

• TRT Terms: Automation; Bridge decks; Condition surveys; Data analysis; Decision trees; Evaluation; Load factor; Reinforced concrete bridges; Safety
• Identifier Terms: National Bridge Inventory
• Uncontrolled Terms: Random forest algorithm
• Subject Areas: Bridges and other structures; Data and Information Technology; Design; Highways; Safety and Human Factors;

Filing Info

• Accession Number: 01593579
• Record Type: Publication
• Report/Paper Numbers: 16-6168
• Files: PRP, TRIS, TRB, ATRI
• Created Date: Mar 15 2016 10:10AM