A multi modal fusion coal gangue recognition method based on IBWO-CNN-LSTM.

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Accurate identification of coal and gangue is essential for efficient and safe mining operations.
• Current methods may lack the precision required for complex mining environments.

Purpose of the Study:

• To develop a novel coal-gangue recognition method using a multi-modal fusion model.
• To enhance the accuracy and robustness of coal-gangue identification in mining.

Main Methods:

• An improved beluga whale optimization (IBWO) algorithm was developed with mutation and memory library mechanisms.
• A convolutional neural network (CNN) and long short-term memory (LSTM) network model was employed for feature extraction and classification.
• Mel-Frequency Cepstral Coefficients (MFCC) were extracted from audio and vibration signals, and a multi-head attention mechanism was integrated into the CNN-LSTM model.

Main Results:

• The IBWO algorithm demonstrated superior performance compared to other optimization algorithms in benchmark tests.
• The proposed IBWO-CNN-LSTM model achieved a high accuracy rate of 95.238% in coal-gangue recognition.
• The multi-modal fusion strategy significantly improved the accuracy and robustness of the recognition system.

Conclusions:

• The developed IBWO-CNN-LSTM model offers an effective solution for automatic coal-gangue recognition.
• The integration of advanced AI algorithms and signal processing techniques enhances mining safety and efficiency.
• The multi-modal approach provides a robust framework for real-time identification in challenging mining conditions.