Stability evaluation of open-pit mine slope based on Bayesian optimization 1D-CNN.

Area of Science:

• Mining Engineering
• Geotechnical Engineering
• Artificial Intelligence

Background:

• Mechanized open-pit coal mining requires accurate slope stability assessment.
• Traditional methods for slope stability analysis have limitations.
• Developing advanced predictive models is crucial for mine safety.

Purpose of the Study:

• To develop a novel model for predicting slope stability in open-pit mines.
• To overcome limitations of existing mechanical, numerical, and experimental techniques.
• To identify key factors influencing slope stability and create a comprehensive dataset.

Main Methods:

• Developed a Bayesian-optimized one-dimensional convolutional neural network (B-1D MCNN) model.
• Utilized Bayesian optimization for hyperparameter tuning and incorporated enhanced convolutional layers.
• Employed Adam optimizer with dropout for improved feature extraction in one-dimensional convolutional neural networks (1D-CNN).

Main Results:

• The B-1D MCNN model accurately depicts nonlinear correlations between influencing factors and slope stability.
• B-1D MCNN demonstrated significant performance enhancements: 10.96-27.85% in Accuracy, 8.98-25.05% in Precision, and 10.26-28.55% in F1-Score compared to other models.
• Model performance improved with increased training dataset length, showing a generalization power of 87.5%.

Conclusions:

• The B-1D MCNN model offers a superior approach for predicting open-pit mine slope stability.
• The enhanced model provides more accurate and reliable assessments than conventional methods.
• The findings highlight the potential of AI-driven models for improving safety and efficiency in mining operations.