Research on rock burst prediction based on an integrated model

Area of Science:

• Mining Engineering
• Geotechnical Engineering
• Artificial Intelligence

Background:

• Rockbursts pose a significant safety hazard in coal mining due to complex nonlinear dynamics.
• Existing prediction methods often lack accuracy in capturing multi-factor coupling and spatial correlations.

Purpose of the Study:

• To develop and validate a novel rockburst risk prediction method for coal mining.
• To improve the accuracy and reliability of rockburst forecasting using advanced computational techniques.

Main Methods:

• A hybrid deep learning model integrating Convolutional Neural Networks (CNN) for local feature extraction and a modified Long Short-Term Memory (MoLSTM) network for temporal modeling.
• Incorporation of an attention mechanism to enhance feature recognition and a Sparrow Search Algorithm (SSA) for hyperparameter optimization.
• Utilizing working face advancement distance as the primary predictor, moving beyond traditional time-axis analysis to capture spatial correlations.

Main Results:

• The proposed SSA-CNN-MoLSTM-Attention model achieved a prediction accuracy of 93.62% and an F1-score of 93.54% on microseismic monitoring data.
• The model demonstrated superior performance compared to traditional methods, with lower Mean Absolute Error (MAE) and Root Mean Square Error (RMSE).

Conclusions:

• The developed model offers a more effective approach to rockburst risk assessment and disaster prevention in mining operations.
• The integration of spatial correlation analysis and advanced AI techniques provides valuable insights for enhancing mine safety.