Summary
This case study focuses on monitoring the structural health of an urban metro tunnel subjected to complex loads such as ground pressure, vibration, and environmental changes. Traditional inspection methods were insufficient to detect slow-developing defects like cracks and joint movements. OctaSense implemented an AI-driven monitoring system using advanced sensors and anomaly detection models to continuously track tunnel behavior, enabling early identification of structural issues, reducing maintenance costs, and ensuring uninterrupted metro operations.
Background & Context
The metro network serves 1.2 million daily commuters. Its tunnels were constructed using three different methods — cut-and-cover reinforced concrete, NATM sprayed concrete, and cast-iron segment bored tunnels — each with distinct structural behaviours and deterioration modes. The authority's infrastructure renewal budget requires careful prioritisation across a complex asset portfolio.
Monitoring Design:
The SHM system design was segmented by tunnel type:\n\n• Cast-Iron Segments (28 km): Focus on bolt hole cracking, segment joint opening, and corrosion-driven section loss. Eddy-current sensors and crack gauges at 120 m intervals.
• NATM Sprayed Concrete (9 km): Convergence monitoring using automated laser scanning and vibrating-wire strain gauges embedded in the lining.
• Cut-and-Cover RC (5 km): Surface crack monitors, rebar corrosion probes, and load cells at prop locations.All data streams feed into a unified OctaSense dashboard with GIS-linked asset condition maps
Sensor Deployment
Key Outcomes & Results
Structural Events Detected
14 genuine anomalies captured in Year 1 — all confirmed by inspection
Maintenance Savings
Condition-based prioritisation delivered 28% reduction in routine inspection costs
Alert Accuracy
96% precision on anomaly classification — significantly reducing engineer call-outs
Coverage
42 km of tunnel monitored continuously — first time in the authority's history
Passenger Safety
Zero service disruptions due to undetected structural deterioration
96 %
Alert precision
42 km
Tunnel network
2,672
Sensor channels
28 %
Inspection savings
Deployment Snapshot
Tunnel Segments
68
Network Length
42 km
Sensor Count
2,672 channels
Monitoring Mode
Permanent / real-time