Early warning model for roof disaster in coal mines based on multi-source data fusion

With the gradual depletion of shallow coal resources in China, deep mining has become an inevitable trend. The complex factors associated with deep mining have led to frequent occurrences of major mine disasters. Roof disasters, in particular, remain a critical challenge in coal mine safety due to their high frequency and severe¹ fatality rates. Taking the Xiayukou Coal Mine as a case study, this paper analyzes the variation patterns of monitoring indicators such as support resistance in the 22306 working face, bolt and cable stress, and roof separation in roadways. The study identifies the correlation between these indicators and disaster occurrences, establishing a scientifically sound early warning index system for roof disasters. To address the current limitations in roof disaster prediction, such as low reliability, poor accuracy, and insufficient intelligence, a multi-source data fusion-based roof disaster early warning model is proposed. By dynamically adjusting warning thresholds, this model effectively resolves the adaptability issues of traditional fixed-threshold warning systems in complex mining environments. The model was applied to real-time monitoring data for roof disaster risk assessment, and validation results demonstrate an early warning accuracy rate of 98.8%.