Data-driven fault diagnosis in structural health monitoring systems for civil infrastructure (feasibility study)

Problem statement
Structural health monitoring (SHM) is increasingly used to ensure functionality and serviceability of civil infrastructure. SHM for civil infrastructure comprises automated data acquisition, data analysis, data communication, and data storage. Despite the advancements in information and communication technologies, the quality of monitoring can be significantly affected by sensor faults. Most types of sensor faults that commonly occur in SHM cannot automatically be detected by SHM systems.

Project goal
This project aims at transferring a fault diagnosis concept into practice, which has been proposed and theoretically validated by the Chair of Computing in Engineering in previous research. By means of data-driven fault diagnosis, SHM systems will be enabled to automatically perform fault diagnosis tasks to detect and isolate different types of sensor faults. Unlike model-based approaches, the data-driven approach pursued in this project does not require physics-based computer models, which may be highly complex and computationally intensive. Instead, automated fault diagnosis is performed solely based on measurement data recorded from civil infrastructure.

Implementation
A system model for fault detection and isolation based on an artificial intelligence approach will be implemented. In addition, a decision logic will be designed to enable SHM systems automatically making decisions with respect to fault identification. Finally, the modules are integrated into an existing SHM system and the feasibility is validated through field tests. Implemented on a railway bridge, the approach will be extended to allow automated train detection by pattern recognition conducted on the sensor data recorded from the bridge. 

• Press release (in German)
• Newspaper article (in German)