MiningCase Study 02

Underground Mine Roof Collapse Prediction

This case study involves a deep underground coal mine operating at depths between 600m and 900m below surface.

Summary

At depth, coal mine roofs are subjected to a combination of vertical gravitational stress and horizontal tectonic stress. Where horizontal stress is high relative to vertical stress, roof layers can buckle and fail in a mode known as cutter failure — an angled shear fracture that propagates through the roof strata with little surface expression until collapse is imminent. This mechanism is particularly difficult to detect visually and is the dominant failure mode at this site. Previous monitoring relied on periodic extensometer readings (typically weekly) and visual inspection by a shift boss. Both were entirely inadequate for detecting the rapid progression from stable to critical conditions that characterises cutter failure.

Background & Context

Underground mining at extreme depths exposes workers to severe seismic hazard. Stress redistribution around excavations creates complex failure environments where conventional observation-based risk assessment is inadequate. The mine had experienced 14 rockfall incidents in the previous 12 months, two of which resulted in lost-time injuries. Technical Problem: At 2,800 m depth, the in-situ stress field exceeds 70 MPa. Seismic events triggered by distant stope blasts can destabilise hangingwall over areas of several thousand square metres within seconds. The challenge was to distinguish harmless routine seismicity from escalating damage sequences that precede major collapses.

Sensor Deployment

Microseismic Network
24 triaxial geophones forming a volumetric array with < 5 m source location accuracy.
Crown Pillar Extensometers
18 wire extensometers measuring hangingwall convergence to 0.01 mm resolution
Stress Cells
8 hollow inclusion stress cells monitoring principal stress changes.
In-stope Cameras
6 IP cameras providing visual confirmation of conditions in high-risk areas.
Key Outcomes & Results
Rockfall Incidents
Reduced by 79% in the 18 months post-deployment
LTI Frequency Rate
Dropped from 2.3 to 0.4 per million man-hours
Seismic Events Detected
>12,000 events classified — 94% classification accuracy verified
Re-entry Efficiency
Average re-entry delay reduced by 35% through evidence-based clearance
94 %
Classification accuracy
79 %
Rockfall reduction
24
Geophone array
2.8 km
Geophone array
Deployment Snapshot
Location
South Africa, Witwatersrand Basin
Mine Depth
2,800 m
Deployment Phase
18 months
Sensor Count
56 Instruments
Ready to Deploy?

Contact OctaSense for a pilot programme at your site.

Get in Touch
Chat with us