Hierarchical Early Warning Indicators Optimization and Correlation Analysis of Coal Spontaneous Combustion in Sealed Fire Areas: A Case Study

Monitoring coal spontaneous combustion is difficult in sealed fire zones. Air leakage and barometric-pressure-driven breathing can dilute or redistribute indicator gases without changing oxidation intensity. We developed a four-level early-warning framework for these disturbance-prone settings. The framework combines programmed-heating oxidation experiments, gray relational analysis, and field monitoring. For the tested coal, 30°C was defined as the baseline temperature (T0), and a marked rise in CO began at 70°C (T1). Under the present experimental and analytical conditions, C2H4 and C2H2 were first reported as detected at the 100°C (T2) and 200°C (T3) sampling points, respectively. These observations were treated as condition-specific experimental landmarks rather than universal gas-formation temperatures. Accordingly, the retained indicators were CO concentration, the reported first-detection points of C2H4 and C2H2, and the dimensionless CO-to-CO2 ratio RC,O.The gray relational grades of CO and RC,O with O2 concentration were 0.8013 and 0.7858, respectively. Their corresponding grades with pressure difference ΔH were 0.6968 and 0.6721. In the field case, RC,O first exceeded the dimensionless blue-warning threshold of 0.00169 on day 4, reaching 0.00241. On day 5, it remained above the threshold at 0.00271. These observations provide preliminary evidence that the framework can detect abnormal oxidation trends in the investigated sealed fire zone. By combining gas evolution with oxygen and pressure variables, the framework supports site-specific interpretation of abnormal oxidation trends under air leakage and barometric-pressure-driven breathing.