Coal and gas outburst prediction based on data augmentation and neuroevolution

Area of Science:

• Mining Engineering
• Artificial Intelligence
• Geological Hazards

Background:

• Coal and gas outbursts (CGO) pose significant risks in underground mining.
• Accurate and efficient CGO risk prediction is crucial for smart mine development.
• Existing methods struggle with imbalanced data and insufficient sample diversity.

Purpose of the Study:

• To develop an advanced method for predicting coal and gas outburst (CGO) risks.
• To enhance prediction accuracy by addressing data limitations in CGO prediction.
• To implement a neuroevolutionary approach for intelligent mine safety.

Main Methods:

• Proposed a CGO risk prediction method, ANEAT, integrating data augmentation and a neuroevolution algorithm.
• Utilized pointwise intensity transformation for data augmentation, addressing imbalanced and diverse samples.
• Employed feature importance sorting and Sparse PCA for dimensionality reduction, feeding into an evolutionary neural network.

Main Results:

• ANEAT demonstrated optimal CGO prediction effect with MAE of 0.0816, RMSE of 0.1322, and EVAR of 0.8972.
• Effectiveness validated through data augmentation analysis, deep learning comparisons, and swarm intelligence algorithm comparisons.
• Achieved high-precision mapping of feature parameters to outburst risk with a lightweight architecture.

Conclusions:

• ANEAT provides a highly accurate and efficient solution for coal and gas outburst (CGO) prediction.
• The method's lightweight architecture makes it suitable for practical application in smart mines.
• ANEAT effectively overcomes challenges of data imbalance and diversity in CGO risk assessment.