A deep learning method based on multi-scale fusion for noise-resistant coal-gangue recognition

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Coal-gangue recognition is crucial for intelligent mining and coal quality.
• Existing methods struggle with dust and noise, limiting industrial use.
• Accurate recognition in harsh environments remains a challenge.

Purpose of the Study:

• To develop a robust coal-gangue recognition method for noisy industrial settings.
• To improve the accuracy and stability of coal-gangue identification systems.
• To enhance the intelligent realization of integrated working faces.

Main Methods:

• An end-to-end multi-scale feature fusion convolutional neural network (MCNN-BILSTM) was proposed.
• Vibration signals were analyzed using multi-scale learning and attention mechanisms.
• Traditional filtering methods were combined with deep learning.

Main Results:

• The MCNN-BILSTM method demonstrated strong adaptability and robustness.
• The approach showed significant noise resistance in complex environments.
• Experimental validation was performed on a coal-gangue impact hydraulic support platform.

Conclusions:

• The proposed MCNN-BILSTM method is suitable for complex practical industrial sites.
• The technique effectively overcomes limitations of existing coal-gangue recognition systems.
• This advancement contributes to safer and more efficient coal mining operations.