Prediction of coal and gas outbursts based on physics informed neural networks and traditional machine learning models

Coal and gas outbursts pose significant risks to underground mining operations, and accurate and reliable prediction is crucial for improving mine safety. Traditional machine learning models struggle to balance prediction accuracy and interpretability, particularly in cases of limited data or complex geological conditions. To address this challenge, this study proposes a prediction model based on Physics-Informed Neural Networks (PINN), which integrates physical monotonicity constraints with data-driven learning to ensure that the predictions align with physical laws. Using actual data from a coal mine, this study compares the performance of the PINN model with traditional machine learning models, including Random Forest (RF), Support Vector Machine (SVM), and Backpropagation Neural Network (BPNN). The results show that the PINN model achieves a coefficient of determination (R2) of 0.966 and a root mean square error (RMSE) of 6.452, outperforming the traditional models in both prediction accuracy and generalization ability. Furthermore, interpretability is significantly enhanced by incorporating known physical behaviors and monotonicity constraints. The proposed PINN-based prediction framework provides a more reliable and theoretically grounded approach to assessing coal and gas outburst risks. Integrating it into mining safety management systems can significantly improve early warning mechanisms and risk mitigation strategies.