On the normalisation and mapping of influence lines

Bridges are vital components of transportation infrastructure, necessitating effective management of structural conditions to ensure safety and cost-effectiveness. Structural health monitoring (SHM) systems, utilising sensor data, offer a promising approach to tracking structural condition changes. By exploiting the increasing availability of structural information as a result of this activity, population-based SHM (PBSHM) aims to enhance approaches for practically managing structures by utilising the similarities between structures within a population to permit knowledge transfer between different population members.

Influence lines have proven insightful for bridge condition assessment but are typically limited to individual structures and often rely on pre-damage baselines, which are unavailable for structures not instrumented and tested before damage or deterioration occurs, a common scenario on older structures and even newer ones. To address these limitations, we propose a population-based bridge condition assessment methodology that utilises normalised influence lines, where the resulting non-dimensional functions permit comparisons across different bridges, alleviating the requirement for repeated load testing.

The originality of this approach lies in the fact that normalised influence lines are independent of specific material and geometry properties, which enables the creation of transformed influence lines (via simple parametric maps) for comparing varying sensor locations and structural responses across bridges. For SHM researchers and practitioners alike, the significance of this approach lies in its potential to address a gap in current influence line work by systematically facilitating transfer learning between structures, thus providing the potential to unearth patterns that individual analyses may miss.

Our method is validated using experimental case studies at both laboratory scale, with known damage states, and full scale, demonstrating its effectiveness in harmonising influence lines across diverse structures and test set-ups. This study demonstrates (i) an approach to normalising influence lines with straightforward interpretations; (ii) how these can map (and compare) influence lines between different set-ups, facilitating transfer learning; and (iii) the application of Bayesian model calibration to use influence lines for condition assessment between structures. By uncovering patterns in bridge behaviour across structures, this population-based approach advances current bridge condition assessment methods and offers a foundation for more robust, scalable infrastructure management techniques.