Prediction method of surface subsidence induced by block caving method based on UAV oblique photogrammetry

Area of Science:

• Geosciences
• Mining Engineering
• Geotechnical Engineering

Background:

• Unmanned Aerial Vehicle (UAV) oblique photogrammetry is widely used in mining for 3D reconstruction and change detection.
• Current UAV applications in mining lack predictive capabilities for surface subsidence.
• Surface subsidence poses significant hazards in mining operations, particularly with block caving methods.

Purpose of the Study:

• To develop a predictive model for surface subsidence using UAV oblique photogrammetry data.
• To forecast maximum surface subsidence height and affected area under block caving.
• To enhance the application of UAV techniques in mining for hazard mitigation.

Main Methods:

• Leveraged 3D real scene model data from UAV oblique photogrammetry.
• Developed a two-stage prediction model merging the probabilistic integral method with recurrent neural network (PIMF-RNN).
• Validated the model using surface data from Pulang copper mine (2018-2020).

Main Results:

• Achieved 91.47% prediction accuracy for maximum surface subsidence height.
• Achieved 87.52% prediction accuracy for surface subsidence area.
• Demonstrated the model's effectiveness in mitigating the impact of internal and external factors on subsidence.

Conclusions:

• The PIMF-RNN model effectively predicts surface subsidence height and area in mining.
• UAV oblique photogrammetry, combined with advanced modeling, offers expanded applications in mining.
• This research establishes a cost-effective and efficient procedure for predicting mine surface subsidence and mitigating associated hazards.