Coupled Environmental Factors Affecting Coal Low-Temperature Oxidation: A Machine Learning Prediction Model

Addressing the issues of unclear multi-parameter coupling mechanisms and insufficient prediction accuracy in coal spontaneous combustion, long-flamed coal from Underground Mine No.3 was investigated. Programmed temperature experiments, gray correlation analysis, and machine learning methods were combined to study the dynamic coupling mechanisms of ambient temperature, oxygen (O2) volume fraction, and ventilation flow rate on low- temperature oxidation of coal. A risk prediction model was established. The study found that the oxidation reaction rate is jointly impacted by the O2 volume fraction and ventilation flow rate below 110°C coal temperature. The risk evolution above 130°C is driven by the ambient temperature-ventilation flow rate synergy. The coupling impacts between ventilation flow rate and O2 volume fraction are amplified by the elevated ambient temperature. A critical temperature-based risk grading model was developed. The XGBoost and RF achieve a better fit to the training data, whereas the SVM and LR offer greater generalization capability and stability. Finally, a “step-by-step blocking” prevention strategy is proposed, providing theoretical support and intelligent early-warning means for precise spontaneous combustion control.