Sound Identification Method for Gas and Coal Dust Explosions Based on MLP

Area of Science:

• Mine Safety Engineering
• Acoustic Signal Processing
• Artificial Intelligence

Background:

• Current gas and coal dust explosion alarm technologies in mines suffer from limitations in accuracy and rely on single monitoring methods.
• There is a critical need for improved identification systems to prevent catastrophic mine accidents.

Purpose of the Study:

• To develop and validate a novel sound identification method for detecting gas and coal dust explosions in coal mines.
• To enhance the accuracy and reliability of real-time safety monitoring and alarm systems.

Main Methods:

• Analysis of various sound features including short-time energy, zero crossing rate, spectral features, MFCC, GFCC, and short-time Fourier coefficients.
• Utilized the Relief algorithm for optimal feature extraction and a Multi-Layer Perceptron (MLP) for sound recognition model development.
• Selected the top 35-dimensional feature values as the optimal feature vector for characterizing sound signals.

Main Results:

• The developed MLP-based sound recognition model achieved high computational efficiency, meeting real-time monitoring requirements.
• The algorithm demonstrated superior accuracy in distinguishing between gas explosions, coal dust explosions, and other underground sounds.
• Recognition experiments yielded an average recognition rate of 95%, recall rate of 95%, and accuracy rate of 95.8%.

Conclusions:

• The proposed sound identification method significantly improves the accuracy of gas and coal dust explosion detection in coal mines.
• The algorithm's high performance and efficiency make it suitable for real-time safety monitoring and alarm systems.
• This approach offers a robust solution for enhancing mine safety and preventing explosions.