Research on a dynamic early warning model for gas outbursts using adaptive fractal dimension characterization

To address the issues of single warning indicators, fixed thresholds, and insufficient adaptability in coal and gas outburst early warning models, this study proposes a dynamic early warning model for gas outbursts based on adaptive fractal dimension characterization. By analyzing the nonlinear characteristics of gas concentration data, an adaptive window fractal analysis method is introduced. Combined with box-counting dimension and variation of box dimension metrics, a cross-scale dynamic warning model for disaster prevention is established. The implementation involves three key phases: First, wavelet denoising and interpolation methods are employed for raw data preprocessing, followed by validation of fractal characteristics. Second, an adaptive window cross-scale fractal dimension method is proposed to calculate the box-counting dimension of gas concentration, enabling effective capture of multi-scale complex features. Finally, dynamic threshold partitioning is achieved through membership functions and the 3σ principle, establishing a graded classification standard for the mine gas disaster (M_GD) index. Validated through engineering applications at Shoushan #1 Coal Mine in Henan Province, the results demonstrate that the adaptive window fractal dimension curve exhibits significantly enhanced fluctuation characteristics compared to fixed window methods, with local feature detection capability improved and warning accuracy reaching 86.9%. The research reveals that this model effectively resolves the limitations of traditional methods in capturing local features and dependency on subjective thresholds through multi-indicator fusion and threshold optimization, providing both theoretical foundation and practical tool for coal mine gas outburst early warning.