Research on coal and gas outburst security situations based on expert knowledge and graph convolutional models

Coal and gas outbursts are typical gas dynamic disasters in coal mines, and their risk assessment is highly challenging. This study proposes a security situation assessment model based on expert knowledge and a graph convolutional network, establishing a theoretical foundation for the application of deep learning in situation assessment and providing decision support for the scientific management of gas dynamic disasters. This study achieves the extraction of gas outburst security situation elements and the evaluation of gas outburst safety situation based on deep learning. It explores the impact of key situation elements on long-term situation forecasting. Theoretical risk analysis of situation elements is conducted through expert knowledge to construct risk features. The graph convolutional network captures the coupling relationships between situation elements and aggregates risk features from multi-source heterogeneous data. Finally, this paper presents a case study demonstrating that the proposed model can accurately identify the risk of coal and gas outburst during the mining process. The application of the model achieves a 24-day security situation assessment of coal and gas outburst. Additionally, by analyzing the distribution of attention coefficients in deep learning, sensitive factors under risk conditions are identified. The historical trend of the situation and the trend of risk changes in sensitive factors provide theoretical support for long-term security situation forecasting.