Machine learning for predicting gas-dynamic phenomena in coal mines

I.Kh. Utakaeva, S.M. Doguchaeva, R.M. Magomedov, S.V. Savina, T.L. Fomicheva
Financial University under the Government of the Russian Federation, Moscow, Russian Federation
Russian Mining Industry №3/ 2026 p. 105-113

Abstract: Gas-dynamic phenomena in coal mines remain the leading factor of man-made deaths of miners. Their share in fatalities at coal enterprises reaches 28%. Traditional forecasting methods (slump drillability index, initial gas emission rate of wells) are characterized by an accuracy of 65–75%, which is insufficient for effective prevention of disasters at deep horizons with gas pressure over 2 MPa. The aim of the study is to develop and verify a hybrid machine learning model based on the Stacking ensemble method with an XGBoost meta-learner for predicting risk classes and intensity of gas-dynamic phenomena. The tasks include: comparative analysis of five algorithms (SVM, Random Forest, LSTM, Bi-LSTM, StackingXGBoost), assessment of the significance of geomechanical and gas-dynamic features, validation on an independent test sample. The empirical base is 847 documented cases of gas-dynamic phenomena recorded at 12 coal mines during 2018–2024 at mining depths of 450–820 m. The dataset includes 12 predictor features: gas pressure, gas content of the seam, depth, coal strength coefficient, rock pressure, gas emission rate and others. Results: the Stacking-XGBoost hybrid model achieved the highest predictive ability among all tested algorithms – determination coefficient R2 = 0.971, root mean square error RMSE = 4.83, F1-score = 0.943 for risk level classification, area under the ROC curve AUC = 0.982. For comparison: the basic SVM model showed R2 = 0.883, RMSE = 9.73, F1 = 0.841. Feature significance analysis (SHAP analysis) established the dominant role of gas pressure (22.4%) and gas content of the seam (19.8%). The mean absolute percentage error of forecasting the methane concentration time series by the LSTM model was 4.2% with a horizon of 15–60 min. The practical significance of the results lies in creating an architecture of an early warning system for gas-dynamic phenomena with a response latency of less than 5 s, capable of integrating into existing mine gas aerology systems. Prospects for further research are related to the introduction of physics-informed neural networks (PINN) including filtration and geomechanics equations, as well as the development of a digital twin of the gas-dynamic state of the coal seam.

Keywords: machine learning, gas-dynamic phenomena, coal mine, Stacking ensemble method, XGBoost, coal and gas outburst prediction, LSTM

For citation: Utakaeva I.Kh., Doguchaeva S.M., Magomedov R.M., Savina S.V., Fomicheva T.L. Machine learning for predicting gas-dynamic phenomena in coal mines. Russian Mining Industry. 2026;(3):105–113. https://doi.org/10.30686/1609-9192-2026-3-105-113

Article info

Received: 09.02.2026

Revised: 24.03.2026

Accepted: 07.04.2026

Information about the authors

Irina Kh. Utakaeva – Cand. Sci. (Phys. & Math.), Associate Professor of the Department of Mathematics and Data Analysis, Financial University under the Government of the Russian Federation, Moscow, Russian Federation; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Svetlana M. Doguchaeva – Cand. Sci. (Phys. & Math.), Associate Professor of the Department of Mathematics and Data Analysis, Financial University under the Government of the Russian Federation, Moscow, Russian Federation; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Ramazan M. Magomedov – Cand. Sci. (Educ.), Associate Professor of the Department of Mathematics and Data Analysis, Financial University under the Government of the Russian Federation, Moscow, Russian Federation; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Svetlana V. Savina – Cand. Sci. (Phys. & Math.), Associate Professor of the Department of Mathematics and Data Analysis, Financial University under the Government of the Russian Federation, Moscow, Russian Federation; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Tatyana L. Fomicheva – Cand. Sci. (Econ.), Associate Professor of the Department of Mathematics and Data Analysis, Financial University under the Government of the Russian Federation, Moscow, Russian Federation; e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

References

1. Zhang C., Wang P., Wang E., Chen D., Li C. Characteristics of coal resources in China and statistical analysis and preventive measures for coal mine accidents. International Journal of Coal Science & Technology. 2023;10(1):22. https://doi.org/10.1007/s40789-023-00582-9

2. Zhang G., Wang E., Liu X., Li Z. Research on risk assessment of coal and gas outburst during continuous excavation cycle of coal mine with dynamic probabilistic inference. Process Safety and Environmental Protection. 2024;190(Part B):405–419. https://doi.org/10.1016/j.psep.2024.08.054

3. Cheng Y.P. Mine Gas Prevention. Xuzhou: China University of Mining and Technology Press; 2023. 412 p. (In Chinese).

4. Zhai C., Cong Y., Chen A., et al. Reflection and prospect on the prevention of gas outburst disasters in China’s coal mines. Journal of China University of Mining & Technology. 2023,52(6):1146–1161. (In Chinese).

5. Anani A., Adewuyi S.O., Risso N., Nyaaba W. Advancements in machine learning techniques for coal and gas outburst prediction in underground mines. International Journal of Coal Geology. 2024;285:104471. https://doi.org/10.1016/j.coal.2024.104471

6. Wang L., Jia B., Su G. Prediction of coal and gas outbursts based on physics informed neural networks and traditional machine learning models. Scientific Reports. 2025;15:29984. https://doi.org/10.1038/s41598-025-02320-4

7. Xie Q., Yan J., Dai Z., Du W., Wu X. Study on coal and gas outburst prediction technology based on multi-model fusion. Frontiers in Big Data. 2025;8:1623883. https://doi.org/10.3389/fdata.2025.1623883

8. Paty S., Biswas A., Djebali S., Guerard G., Kamilya S. IoT-enabled methane monitoring and LSTM-based forecasting system for enhanced safety in underground coal mining. ACM Transactions on Internet of Things. 2025;6(1):4. https://doi.org/10.1145/3703460

9. Panina O.V., Zavalko N.A., Eremin S.G., Kharchenko K.V., Zudenkova S.A. Development of neural networks to analyze vibration signals of mining equipment and prevent emergency situations. Russian Mining Industry. 2025;(2):97–104. (In Russ.) https://doi.org/10.30686/1609-9192-2025-2-97-104

10. Krasyukova N.L., Panina O.V., Eremin S.G., Zubenko A.V., Laffakh A.M. Smart prediction of ground displacement using parallel neural network models and high-precision geodetic measurements. Russian Mining Industry. 2025;(2):106–112. (In Russ.) https://doi.org/10.30686/1609-9192-2025-2-106-112

11. Panina O.V., Belyaev A.M., Zavalko N.A., Eremin S.G., Sagina O.A. Application of deep machine learning methods for structural analysis of ore bodies and prediction of optimal mining zones. Russian Mining Industry. 2025;(1):177–183. (In Russ.) https://doi.org/10.30686/1609-9192-2025-1-177-183

12. Kadyrova G.M., Krasyukova N.L., Rozhdestvenskaya I.A., Tokmurzin T.M., Voronova E.I. Adaptive optimization of traffic flows in underground mine workings based on artificial intelligence methods. Russian Mining Industry. 2025;(1):137–146. (In Russ.) https://doi.org/10.30686/1609-9192-2025-1-137-146

13. Muzalyov S.V., Abdikeev N.M., Obolenskaya L.V. Improving the system of criteria for evaluating the effectiveness of Russia's industrial policy under sanctions pressure. Finance: Theory and Practice. 2025;29(4):6–18. https://doi.org/10.26794/2587-5671-2025-29-4-6-18

14. Kharchenko K.V. Improving the cluster landscape of the region as a factor of ensuring economic security. Economics, Taxes & Law. 2025;18(2):98–106. (In Russ.) https://doi.org/10.26794/1999-849X-2025-18-2-98-106

15. Lin H., Li W., Li S., Wang L., Ge J., Tian Y., Zhou J. Coal mine gas emission prediction based on multifactor time series method. Reliability Engineering & System Safety. 2024;252:110443. https://doi.org/10.1016/j.ress.2024.110443

16. Tutak M., Krenicky T., Pirník R., Brodny J., Grebski W.W. Predicting methane concentrations in underground coal mining using a multi-layer perceptron neural network based on mine gas monitoring data. Sustainability. 2024;16(19):8388. https://doi.org/10.3390/su16198388

17. Liu X., Du Z., Wang E., Jiang A., Tian H., Wang K. et al. Intelligent recognition of coal and gas outburst precursor and potential hazards using YOLOv8 neural network. Geomechanics and Geophysics for Geo-Energy and Geo-Resources. 2024;10:117. https://doi.org/10.1007/s40948-024-00837-y

18. Wang N. Research on coal and gas outburst risk prediction based on improved search algorithm optimized deep learning network. Scientific Reports. 2025;15:40976. https://doi.org/10.1038/s41598-025-24744-8

19. Shenderov V.I. The use of big data and neural networks for monitoring and forecasting environmental changes in Russia. Voprosy Ecologii. 2024;37(1):98–127. (In Russ.) https://doi.org/10.25726/u2912-9651-3244-z