AET-FRAP-A Periodic Reshape Transformer Framework for Rock Fracture Early Warning Using Acoustic Emission...

Area of Science:

• Geotechnical Engineering
• Signal Processing
• Machine Learning

Background:

• Accurate rock fracture identification is vital for deep subterranean engineering safety.
• Existing methods struggle with reliable warning indicators and effective warning levels for fractures.
• Acoustic emission (AE) signals contain valuable information for predicting rock failure.

Purpose of the Study:

• To develop a novel framework, AET-FRAP, for predicting rock fractures using AE feature parameters.
• To enhance the accuracy and reliability of fracture prediction in deep subterranean environments.
• To establish robust warning criteria that minimize false alarms.

Main Methods:

• Utilized Empirical Mode Decomposition (EMD) and Fast Fourier Transform (FFT) for signal processing and feature selection.
• Developed a multi-input time series forecasting framework (AET-FRAP) incorporating InceptionNeXt architecture.
• Implemented a secondary criterion using cosine similarity and kurtosis for robust change detection.

Main Results:

• AET-FRAP demonstrated superior accuracy compared to LSTM models, with R² approaching 1.
• Significant reductions in Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE) were observed.
• The framework accurately identified pre-fracture energy accumulation spikes and provided advanced warnings.

Conclusions:

• AET-FRAP offers a stable and significant advancement for rock fracture prediction in engineering applications.
• The developed collaborative thresholds effectively reduce noise-induced false alarms.
• The framework provides quantifiable trigger criteria for enhanced safety in subterranean operations.