Применение методов машинного обучения для автоматизированной классификации магнитных аномалий Северного Кавказа по данным аэрогеофизической съемки

V.S. Tynchenko; M.V. Zhuravlev; N.A. Yagudaeva; V.V. Toroptsev; A.G. Grigoryan

doi:10.46698/VNC.2026.48.42.001

V.S. Tynchenko Bauman Moscow State Technical University, 5, 2nd Baumanskaya Str., Moscow, 105005, Russian Federation
M.V. Zhuravlev Russian State Agrarian University – Moscow Timiryazev Agricultural Academy, 49 Timiryazevskaya Str., Moscow, 127434, Russian Federation
N.A. Yagudaeva Russian State Agrarian University – Moscow Timiryazev Agricultural Academy, 49 Timiryazevskaya Str., Moscow, 127434, Russian Federation
V.V. Toroptsev Russian State Agrarian University – Moscow Timiryazev Agricultural Academy, 49 Timiryazevskaya Str., Moscow, 127434, Russian Federation
A.G. Grigoryan Admiral F.F. Ushakov State Maritime University, 93 Lenin Ave., Novorossiysk 353918, Russian Federation

DOI: https://doi.org/10.46698/VNC.2026.48.42.001

Keywords: magnetic anomalies, airborne geophysical survey, machine learning, classification, neural networks, geophysics, North Caucasus, magnetic field, Random Forest, geological interpretation

Abstract

Relevance. The improvement of accuracy of reservoir property prediction accuracy from well logging data is an important task in modern petroleum geophysics. Conventional interpretation methods often fail to provide reliable predictions of reservoir properties in geologically heterogeneous formations. Aim. To assess the effectiveness of neural network algorithms for predicting porosity and permeability of reservoir rocks using well logging data. Methods. Gamma-ray, neutron, acoustic, density, and electrical logging data, together with laboratory core analysis results, were used. Data preprocessing included filtering, normalization, and correlation analysis. A multilayer perceptron (MLP) neural network was applied for reservoir property prediction. The obtained results were compared with conventional interpretation approaches and other machine learning algorithms. Results. The neural network model demonstrated higher prediction accuracy compared with traditional methods. The coefficient of determination reached 0.89, while the root mean square error of porosity prediction was approximately 2.1 %. Acoustic, neutron, and density logging data were identified as the most informative predictors. The results confirm the effectiveness of artificial intelligence technologies for digital interpretation of geological and geophysical information and prediction of reservoir properties in oil and gas-bearing formations.

References

Кузьмина А.О., Ильина М.Ю., Лукичев К.Е., Преснякова Н.Г., Татарников А.А. ATTCNN картирование порфировых Cu-Mo месторождений Удоканского типа // Горная промышленность. – 2026. –№ 1. – С. 135–142. DOI: 10.30686/1609-9192-2026-1-135-142.

Шедько Ю.Н., Харченко К.В., Зуденкова С.А., Галкин А.И., Бабаян Л.К. Deep Q-learning в диспетчеризации карьерных самосвалов БелАЗ на месторождениях Курской магнитной аномалии // Горная промышленность. – 2026. – № 1. – С. 105–113. DOI: 10.30686/1609-9192-2026-1-105-113.

Chen H., Wu G., Chen H., Xu C., Wu C. Modeling of high-precision sea surface geomagnetic field in the Northern South China Sea Based on PSO-BP neural network // Journal of Marine Science and Engineering. – 2026. – Vol. 14. No. 1. – Art. No. 108. DOI: 10.3390/jmse14010108.

Fan L., Lin G., Zheng Q., Ni F. Condition monitoring of levitation systems for maglev trains using Bayesian physics-informed neural networks // Measurement Science and Technology. – 2026. – Vol. 37. No. 2. – Art. No. 026109. DOI: 10.1088/1361-6501/ae26a7.

Ghamizi S., Kanli G., Deng Y., Palissot V., Perquin M., Keunen O. Foundation models for brain imaging: A systematic review // NeuroImage. – 2026. – Vol. 333. – Art. No. 121877. DOI: 10.1016/j.neuroimage.2026.121877.

Ghosh J., Thoram S., Sun J., Sager W.W. Enhancing marine magnetic anomaly interpretation with anisotropic diffusion and deep transfer learning // Scientific Reports. – 2026. – Vol. 16. No. 1. – Art. No. 1185. DOI: 10.1038/s41598-025-30926-1.

Hui R., Liang X., Zuo C., Wang Z. 2D CNN based multifeature fusion detection method for the magnetic anomaly generated by submarine wake // Journal of Ocean Engineering and Science. – 2025. – Vol. 10. No. 6. – pp. 1139–1154. DOI: 10.1016/j.joes.2023.11.001.

Li Q., Li Z., Wang Z., Lang Y., Liu H., Zhao W. et al. Optical fiber sensor based transportable active seismic survey system solution to OSI // Pure and Applied Geophysics. – 2025. – Vol. 182. No. 12. – pp. 5303–5313. DOI: 10.1007/s00024-024-03539-4.

Ma L., Wang J. High-precision delineation of curie point depth under remanent magnetization using physics-informed neural networks // Geophysical Research Letters. – 2026. – Vol. 53. No. 7. – Art. No. e2025GL120275. DOI: 10.1029/2025GL120275.

Rahmaniar W., Suzuki K. Multi-AD: cross-domain unsupervised anomaly detection for medical and industrial applications // Pattern Recognition. – 2026. – Vol. 172. – Art. No. 112486. DOI: 10.1016/j.patcog.2025.112486.

Shen J., Hou Z., Wei J., Zhao X., Wang J. SDCNet: Stable downward continuation of magnetic anomalies by depth information fusion // IEEE Geoscience and Remote Sensing Letters. – 2026. – Vol. 23. – Art. No. 7501505. DOI: 10.1109/LGRS.2026.3657949.

Teknik V. Physics-informed neural networks for super-resolution of geospatial data: a case study from the aeromagnetic data of Central Iranian Plateau // Geophysical Prospecting. – 2026. – Vol. 74. No. 4. – Art. No. e70175. DOI: 10.1111/1365-2478.70175.

Tyagi C., Ansari M.A., Usmani H., Yadav R.K. Recent trends of CNN in medical imaging: a comprehensive analysis // In: Proceedings of the 2nd International Conference on Cognitive Computing in Engineering, Communications, Sciences and Biomedical Health Informatics (IC3ECSBHI 2026). Greater Noida, India, 2026. – pp. 1625–1630. DOI: 10.1109/ IC3ECSBHI67834.2026.11469172.

Wang Y.-L., Li C.-S., Zhang H., Wang X.-F. Research on a rapid prediction method for magnetic anomaly responses in the penetration process based on an improved proper orthogonal decomposition (POD)-radial basis function (RBF) neural network // Measurement. – 2025. – Vol. 256. – Art. No. 118266. DOI: 10.1016/j.measurement.2025.118266.

Wang H., Jiang C., Ren Y., Ding L. Research on magnetic anomaly signal detection based on CNN-LSTM // Measurement Science and Technology. – 2026a. – Vol. 37. No. 12. – Art. No. 126111. DOI: 10.1088/1361-6501/ae5334.

Wang H., Tang Z., Qi X., Zhang D., Zhao C. A Magnetic anomaly inversion method integrating convolutional block attention module and physical consistency constraints // Journal of Visualized Experiments. – 2026b. – No. 229. – Art. No. e69539. DOI: 10.3791/69539.

Ye J., Mansour A., Huang F. Enhancing real-time heading estimation for pedestrian navigation via deep learning and smartphone embedded sensors // Scientific Reports. – 2025. – Vol. 15. No. 1. – Art. No. 31672. DOI: 10.1038/s41598-025-13390-9.

Yurtsever M.M.E., Eken S., Arslan B., Atay Y., Sagıroglu S., Celtikci E. et al. Evaluating augmented deep networks for MRI-based brain anomaly classification // Cluster Computing. – 2026. – Vol. 29. No. 3. – Art. No. 148. DOI: 10.1007/s10586-025-05893-x.

Zhou X., Wang H., Wang Z., Li X., Long Z. Anomaly Detection Method for HighSpeed Maglev Train Suspension System Based on Isolation Forest and Improved LSTM // IEEE Transactions on Transportation Electrification. – 2026. – Vol. 12. No. 2. – pp. 3907–3919. DOI: 10.1109/TTE.2025.3645233.

Application of machine learning methods for automated classification of magnetic anomalies of the North Caucasus based on airborne geophysical survey data

Abstract

References