Multi-Classification of Complex Microseismic Waveforms Using Convolutional Neural Network: A Case Study in Tunnel...

Area of Science:

• Geophysics
• Signal Processing
• Machine Learning

Background:

• Microseismic data analysis requires accurate multi-classification due to complex waveforms.
• Existing methods face challenges in distinguishing microseismic signals from noise.
• Precise classification is crucial for signal processing and rock mass stability assessments.

Purpose of the Study:

• To propose a novel microseismic multi-classification (MMC) model.
• To enhance the accuracy and robustness of microseismic data classification.
• To demonstrate the model's effectiveness in various geological conditions.

Main Methods:

• Utilizing short-time Fourier transform (STFT) for feature extraction.
• Employing convolutional neural networks (CNN) for data classification.
• Inputting real and imaginary parts of STFT coefficients into the CNN model.

Main Results:

• The proposed MMC model achieves optimal performance in Precision, Recall, and F1-score.
• The model exhibits low sensitivity to noise, validated using semi-synthetic data.
• Accurate detection of microseismic signals with M ≥ 0.2 is achieved across diverse geological settings.

Conclusions:

• The MMC model offers superior performance for microseismic data multi-classification.
• The method demonstrates robustness against noise and generalizability across geological conditions.
• Potential applications include exploration seismology and earthquake studies.