This review paper, published in the International Journal of Coal Science & Technology, provides a comprehensive roadmap of how we track, predict, and simulate the ground deformation caused by underground coal mining. It highlights a major technological shift: moving away from labor-intensive manual surveys toward an integrated "space-sky-surface" monitoring paradigm that combines satellites, drones, and ground sensors.
TL;DR
The paper synthesizes decades of research to show that while traditional ground surveys offer high precision, the future of mining safety lies in integrating remote sensing (InSAR, UAVs) with "black-box" AI models and mechanical simulations. The ultimate goal is to bridge the gap between empirical geometric predictions and the actual physical mechanics of rock strata.
What Problem Does This Paper Address?
Underground coal mining creates voids that cause the overlying rock layers to collapse, eventually leading to surface subsidence. This isn't just a "hole in the ground"—it involves complex 3D displacements, tilts, and strains that can destroy infrastructure, disrupt water tables, and ruin ecosystems.
Historically, researchers relied on "point-based" data (manually checking a few stakes in the ground). However, modern "green mining" requires "area-based" data—high-resolution, real-time maps of entire mining regions. The authors address the challenge of balancing the high accuracy of traditional methods with the massive data coverage provided by emerging remote sensing technologies.
Methods
The authors categorize the field into three pillars:
- Monitoring: They compare conventional geodetic methods (Leveling, GNSS) against remote sensing. Key technologies discussed include InSAR (satellite radar), UAV-based LiDAR/Photogrammetry, and Terrestrial Laser Scanning (TLS).
- Calculation (Prediction): The paper reviews the Probability Integral Method (PIM)—the gold standard in China—which uses influence functions to estimate subsidence. They also highlight the rise of Artificial Neural Networks (ANNs) and genetic algorithms to optimize these predictions.
- Simulation: This involves Numerical Modeling (using software like FLAC3D or UDEC to solve stress-strain equations) and Physical Modeling (building scaled-down versions of mines using sand and plaster to watch them collapse in a lab).
Datasets
As a review paper, it does not introduce a single new dataset. Instead, it references a vast library of historical data, specifically:
- The Manuals of Coal Mining (China): A massive compilation of geological and subsidence data from decades of Chinese mining operations.
- Satellite Data: Case studies utilizing Sentinel-1 (C-band), ALOS-PALSAR (L-band), and TerraSAR-X (X-band) imagery.
- Field Observations: Data from specific sites like the Daliuta Mine and Juye Coalfield are used to compare the accuracy of different monitoring methods.
Key Insights and Ideas
- The "Space-Sky-Surface" Integration: No single method is perfect. The authors argue for a "distributed global earth observation system" where satellites provide broad coverage, UAVs provide high-resolution "on-demand" snapshots, and ground GNSS stations provide real-time "truth" for calibration.
- The Mechanical Disconnect: A major "aha" moment is the critique of current prediction models. Most (like PIM) are purely geometric or statistical—they treat the ground like a mathematical surface rather than a complex mechanical structure. The authors call for a "unification of calculation methods and mechanical principles."
- InSAR’s "Goldilocks" Problem: While InSAR is revolutionary, it struggles with the "large gradient" (rapid, steep) sinking typical of active mining. It works best for monitoring the slow "residual" sinking that happens years after a mine closes.
- UAVs as the Middle Ground: Drones have emerged as the most practical tool for active mines, offering centimeter-level accuracy (20–150 mm) without the extreme labor of manual surveying or the decorrelation issues of satellites.
Key Results
The paper provides a valuable comparison of monitoring performance:
- Accuracy: Conventional leveling remains the king of precision (±2 mm), while UAVs and InSAR typically hover in the ±10 mm to ±150 mm range depending on vegetation and terrain.
- Evolutionary Timeline: The authors identify three "research climaxes": the early 20th century (Europe), the 1950s-60s (Poland/Germany), and the current era (China/USA), which is defined by the "Internet of Things" and Big Data.
- Calculation Applicability: They provide a lookup table (Table 5) matching specific mathematical models (Pearson, Negative Exponential, etc.) to coal seam dip angles, helping engineers choose the right tool for their specific geology.
Limitations and Future Work
- The "Black Box" Problem: While AI and Neural Networks are accurate, they lack physical transparency. Future work needs to make these models "physics-informed."
- Real-Time Gap: Most current methods are "post-event." There is an urgent need for real-time monitoring systems that can provide pre-warnings for sudden surface collapses.
- Complex Terrain: Current remote sensing still struggles with heavy vegetation and steep mountains. Improving "ground filtering" algorithms to see through trees is a high priority.
- Smart Mine Integration: Subsidence data should not exist in a vacuum; it needs to be integrated into 3D "Smart Mine" platforms that combine geological models with live production data.
Summary updated