Initiative overview
AVAIL addresses the challenge of extracting reliable insight from large volumes of time-series data generated by bridge monitoring systems. As sensor networks expand, sensor anomalies make manual data review time-consuming and reduce the data’s usefulness for engineering analysis and long-term planning.
The initiative improves the quality and usability of structural health monitoring data by applying semi-supervised machine learning methods for anomaly detection and data cleaning. It isolates unreliable observations, producing cleaner datasets that support downstream structural integrity assessment, safety oversight and infrastructure lifespan extension.
Developed by the Data Science and Major Bridges and Projects teams at HICC, AVAIL is expected to evolve by expanding to additional sensor types, increasing automation and integrating with analytical dashboards to strengthen data-driven analysis and eventual infrastructure decision-making.
Results, outcomes and impacts
The AI system enabled faster and more systematic identification of anomalous sensor readings, reducing reliance on manual data screening. By automating anomaly detection, the project improved consistency and traceability in how atypical data points are flagged and reviewed, producing higher-quality datasets in significantly less time. Overall, these results demonstrate AVAIL’s effectiveness in isolating data quality issues within large-scale sensor datasets. The cleaned outputs provide higher-quality data, forming a stronger analytical foundation for future assessments of structural integrity and long-term infrastructure management.
Challenges and lessons learned
- Data labelling effort;
- sensor heterogeneity;
- and the importance of explainability and human oversight in safety-critical systems.
