Bridges

Bridge maintenance relies on accurate and timely information to support inspection and repair strategies. Traditional methods, while effective, often struggle to process the large volumes of unstructured data found in maintenance logs and inspection reports. Large Language Models, a subset of artificial intelligence, provide an approach to analysing and extracting insights from textual data at scale. This paper examines a proof of concept conducted by Transport for New South Wales (TfNSW) to assess the application of LLMs in bridge maintenance.

The study explores how LLMs can assist in summarising reports, identifying trends in maintenance records, and supporting decision-making. Preliminary findings suggest that LLMs have the potential to improve data accessibility and enhance predictive maintenance practices. However, challenges such as data consistency, model transparency, and computational requirements remain. The paper discusses these factors and considers future applications for LLMs in infrastructure asset management.