Geospatial Analysis
A novel fine-scale identification method for coal mining subsidence basin based on TS-InSAR and subsidence curve characteristics
This paper proposes a novel fine-scale identification method for coal mining subsidence areas by integrating time-series InSAR technology and surface subsidence curve characteristics. The method constructs a time-series feature dataset from typical subsidence curves, trains a weighted Support Vector Machine (SVM) model, and then identifies large-scale coal mining subsidence basins. The approach improves boundary extraction accuracy by approximately 80% compared to existing methods, providing high-precision technical support for coal mine safety production and ecological restoration.
Executive Impact
This research delivers critical advancements in monitoring and managing the impact of coal mining, offering substantial benefits for operational efficiency, safety, and environmental stewardship.
80%
Boundary Extraction Accuracy Improvement
98.98%
Overall Identification Accuracy
97%
F1-Score
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Overview
Methodology Highlights
Key Results & Performance
Future Directions
Overview
This research introduces a novel fine-scale identification method for coal mining subsidence basins, combining time-series InSAR (TS-InSAR) technology with surface subsidence curve characteristics. It addresses limitations of existing models by integrating temporal evolution features and achieving high generalization capability. The methodology involves constructing a standardized time-series subsidence curve feature dataset from both field-measured and simulated data, followed by training a weighted Support Vector Machine (SVM) model for accurate pixel-wise identification. The model was validated across three diverse mining areas in China, demonstrating its rapid and accurate identification capabilities without requiring prior InSAR data from target regions. It significantly improves boundary extraction accuracy, providing critical support for mine safety, ecological restoration, and infrastructure protection.
Methodology Highlights
The core of the methodology lies in generalizing the morphological and magnitude characteristics of 'S-shaped' subsidence curves into numerical features, which are then learned by a Weighted SVM. Key steps include: 1. Data Collection and Standardization: Amalgamating field-measured and simulated time-series subsidence data, followed by temporal and spatial normalization to create a standardized dataset with 11 features (10 normalized time points + maximum subsidence). 2. Weibull Time-Function Fitting: Using a genetic algorithm to fit Weibull functions to subsidence data, enabling extraction of values at normalized time points. 3. Weighted SVM Training: Employing a Weighted SVM to address class imbalance (subsidence vs. non-deformation areas), enhancing recognition of the minority class. This model ensures transferability and adaptive capability by learning generalizable curve features rather than relying on specific observational data.
Key Results & Performance
The proposed model achieved an overall identification accuracy of 98.98% on the test set, with a recall rate of 94% and an F1-score of 97%. It demonstrated strong robustness and transferability across eastern, central, and western Chinese mining areas. Crucially, the model significantly improved the extraction accuracy of the 10 mm subsidence boundary by approximately 80% compared to existing methods. This high-precision delineation of subsidence impact zones is vital for building protection, mine safety, and environmental planning. The integration of simulated data into the training dataset was particularly effective, increasing overall accuracy from 96% to 98% and enhancing the model's feature learning capability.
Future Directions
Future research will focus on several key areas to further enhance the model: 1. Diversity of Observational Data: Incorporating multi-source heterogeneous datasets (UAV, point cloud) and specific geological/mining conditions (mountainous areas, fault zones, intermittent mining) to build a hierarchically categorized knowledge base of temporal curves. This aims to improve applicability under diverse conditions. 2. Dynamic Migration Mechanisms: Investigating the dynamic migration and evolution mechanisms of overburden strata using discrete element-finite difference coupling simulations and multi-source monitoring to reveal complete mechanical processes. 3. Spatial-Temporal Asynchrony: Deepening research into the spatial-temporal asynchrony of surface point migration during mining, enabling differential identification of mining stages and establishing a refined full-lifecycle identification model for subsidence basins, moving from 'extent' to 'states and processes'.
Enterprise Process Flow
SAR Images (N+1)
→
Phase Unwrapping
→
Orbital Refinement & Reflattening
→
GCP Selection
→
Removing Residual Topography Error
→
Calculating Linear Deformation Rate
→
Estimation of Deformation by SVD Method
→
Calculating Total Deformation
→
Construct Time-series Dataset
→
Train Weighted SVM Model
→
Predict Subsidence Basins
98%
Integrating simulated data into the training set increased overall accuracy from 96% to 98%, demonstrating enhanced model generalization capability by mitigating bias from limited field data.
| Feature | Proposed Method | Xu et al. (2024) |
|---|---|---|
| Spatial Clustering |
|
|
| Data Dependency |
|
|
| Empirical Thresholds |
|
|
| Boundary Accuracy (10mm contour) |
|
Application in Eastern Mining Area (Tangshan, Hebei)
In the urbanized Eastern Mining Area, the model successfully identified subsidence area E (Fig. 11), confirming the effectiveness in regions with dense infrastructure. Despite extensive urban built-up areas and high protection requirements, the method accurately reconstructed the subsidence basin, demonstrating its utility for mining-induced damage control.
Application in Central Mining Area (Hebei Province)
The central mining area, characterized by villages and farmland, presented challenges with weaker backscatter and lower coherence. Yet, the model effectively identified subsidence areas A and B (Fig. 12), including residual subsidence basins with relatively small deformation magnitudes, showcasing its robustness across varied geological conditions and protection targets.
Application in Western Mining Area (Inner Mongolia)
For the high-intensity mining in the Western Mining Area, which focuses on ecological restoration and soil/water conservation, the model successfully delineated subsidence areas C and D (Fig. 13). It handled significant surface settlement magnitudes and large-gradient subsidence, crucial for land reclamation and environmental restoration.
Advanced ROI Calculator
Quantify the potential efficiency gains and cost savings for your enterprise with our AI solution.
Your AI Implementation Roadmap
A clear, actionable path to integrate AI seamlessly into your operations.
Phase 1: Data Integration & Feature Engineering
Consolidate existing field measurements with TS-InSAR data from your mining operations. Our AI specialists will engineer a tailored feature set based on subsidence curve characteristics to prepare your unique dataset for model training.
Phase 2: Weighted SVM Model Training & Validation
Leverage our robust Weighted SVM framework, trained on your refined dataset. This phase focuses on adapting the model to your specific geological and operational conditions, ensuring high accuracy and minimizing bias for minority classes (subsidence areas).
Phase 3: Automated Subsidence Basin Identification Deployment
Deploy the trained model to automatically identify and delineate coal mining subsidence basins across your target regions. This enables rapid, fine-scale mapping of deformation areas without requiring continuous, manual InSAR analysis.
Phase 4: Advanced Boundary Refinement & Impact Assessment
Utilize the model's high-precision boundary extraction capabilities to refine 10mm subsidence contours, crucial for infrastructure protection and regulatory compliance. Integrate these insights into your safety production and ecological restoration planning.
Phase 5: Continuous Monitoring & Predictive Analytics Integration
Establish a continuous monitoring pipeline, feeding new InSAR data into the system for ongoing updates. Explore predictive analytics features to anticipate future subsidence patterns and proactively manage mining-induced risks.
Ready to Revolutionize Your Enterprise?
Schedule a personalized consultation to explore how our AI solutions can drive unparalleled efficiency and innovation for your business.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat