Abnormal Pressure Event Recognition and Dynamic Prediction Method for Fully Mechanized Mining Working Face Based on...

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Geotechnical Engineering

Background:

• Intelligent longwall mining requires accurate strata pressure prediction for safety and efficiency.
• Existing machine learning models struggle with the temporal dependencies and periodic nature of strata behavior.

Purpose of the Study:

• To develop a novel framework for identifying and predicting abnormal strata pressure in intelligent longwall mining faces.
• To improve the accuracy and reliability of strata pressure prediction by capturing complex temporal characteristics.

Main Methods:

• Proposed a Gated Recurrent Unit (GRU) integrated with an attention mechanism (AM) for strata pressure prediction.
• The GRU-AM model was designed to capture both short-term fluctuations and long-term cyclic patterns in support resistance.
• Evaluated model performance against conventional LSTM and CNN models using metrics like RMSE, MAE, MAPE, and Pearson correlation coefficient (R).

Main Results:

• The GRU-AM model demonstrated high prediction accuracy in both single-support and multi-support scenarios.
• Achieved an accuracy of 0.9741 for abnormal pressure identification at a 1-minute step length and 0.9195 at a 10-minute step length.
• Outperformed LSTM and CNN models across various evaluation metrics and prediction tasks.

Conclusions:

• The GRU-AM framework offers an effective and scalable solution for real-time abnormal strata pressure recognition and early warning in intelligent coal mining.
• The model's robustness, computational efficiency, and practical reliability were confirmed through field applications.
• This approach significantly enhances safety and production efficiency in intelligent mining environments.