Processing anomalies in potential prospecting methods using probabilistic machine learning
Keywords:
machine learning, statistics, interpretation, algorithm, gravity and magnetic potential fieldsAbstract
Machine learning is currently a disruptive tool for digital signal processing and decision making. In particular, probabilistic machine learning makes it possible to approximate the probability density function under which the recorded signals are distributed. Using probabilistic machine learning, we propose to assist interpreters of potential field prospecting methods. By generating alternatives that are statistically consistent with the working data, the interpreter can visualize realistic variations of the original data that allow them to expand their insights. To do this, the interpreter enters the anomaly of interest into the system. The system deduces from the data provided an approximation to its probability density function. Then, the system allows the user to select a region of the input data to generate distributed realizations under the same probability density of the observed data. We evaluate the novelty of the data generated with respect to the original data in the selected region, allowing the interpreter to weigh the proposals obtained. To infer the probability density function, we use a method of adding and removing random noise on the supplied grid. In generating data, we use a Monte Carlo method based on a Markov chain known as Langevin dynamics. Some of the challenges of the proposal are the training of a probabilistic machine learning method with a single database and the limitation in hardware and computing time involved in using the method on a personal computer. We present an experience on synthetic data and magnetic total intensity scalar anomaly field data. The results show that the proposal can assist the interpreter in the spatial delineation of anomalous bodies and in the inversion of parameters, such as the magnetization direction.
Downloads
References
Bhattacharyya, B. K. (1964). Magnetic anomalies due to prism-shaped bodies with arbitrary polarization. Geophysics, 29(4), 517-531.
Blakely, R. J. (1996). Potential theory in gravity and magnetic applications. Cambridge University Press.
Burkov, A. (2019). The hundred-page machine learning book. Andriy Burkov.
Chollet, F. (2021). Deep learning with Python. Simon and Schuster.
de Lima, R. P., Suriamin, F., K. J. Marfurt y M. J. Pranter. (2019). Convolutional neural networks as aid in core lithofacies classification. Interpretation, 7, SF27-SF40. http://dx.doi.org/10.1190/int-2018-0245.1
Dutra, A.C. y Marangoni, Y. R. (2009). Gravity and magnetic 3-D inversion of Morro do Engenho complex, central Brazil. Journal of South American Earth Sciences, 28(2), 193-203.
Dutra, A. C., Marangoni, Y. y Trindade, R. I. F. (2014). Aeromagnetic and physical-chemical properties of some complexes from Goiás Alkaline Province. Brazilian Journal of Geology, 44, 361–373, https://doi.org/10.5327/Z2317-4889201400030003
Elad, M., Kawar, B. y Vaksman, G. (2023). Image denoising: The deep learning revolution and beyond - a survey paper. SIAM Journal on Imaging Sciences, 16, 1594-1654. http://dx.doi.org/10.1137/23M1545859
Fedi, M., Florio, G. y Rapolla, A. (1994). A method to estimate the total magnetization direction from a distortion analysis of magnetic anomalies. Geophysical Prospecting, 42(3), 261-274.
Flusser, V. (2023). El universo de las imágenes técnicas: elogio de la superficialidad. Caja negra Editora.
Gonzalez, R.C. y Woods, R. E. (2018). Digital image processing. (4ª ed.). Pearson.
Goodfellow, I., Bengio, Y. y Courville, A. (2016). Deep learning. MIT press.
Hu Y., Xiaolong W, Wu X., Sun J., Huang Y. y Chen J. (2024). Three-dimensional cooperative inversion of airborne magnetic and gravity gradient data using deep-learning techniques. Geophysics, 89(1), WB67-WB79. http://dx.doi.org/10.1190/geo2023-0225.1
Journel, A. G. (1989). Fundamentals of geostatistics in five lessons. (Vol. 8). American Geophysical Union.
Kingma, D. P. y Ba, J. (2014). Adam: A method for stochastic optimization (v. 1). https://doi.org/10.48550/arXiv.1412.6980
Li, P., Liu, M., Alfarraj, M., Tahmasebi, P. y Grana, D. (2024). Probabilistic physics-informed neural network for seismic petrophysical inversion. Geophysics, 8(2), M17-M32. https://doi.org/10.1190/geo2023-0214.1
Liu, Q. y Ma, J. (2024). Generative interpolation via a diffusion probabilistic model. Geophysics, 89(1), V65-V85. http://dx.doi.org/10.1190/geo2023-0182.1
McAliley, W. A. y Li, Y. (2024). Stochastic inversion of geophysical data by a conditional variational autoencoder. Geophysics, 89(1), WA219–WA232. https://doi.org/10.1190/geo2023-0147.1
Meju, M. A. (1994). Geophysical data analysis: Understanding inverse problem. Theory and Practice. Society of Exploration Geophysicists.
Mosegaard, K. y Tarantola, A. (1991). Monte Carlo analysis of geophysical inverse problems. SEG Technical Program Expanded Abstracts, 940-940. https://doi.org/10.1190/1.1888770
Murphy, K. P. (2023). Probabilistic machine learning: Advanced topics. MIT press.
Reis, A. L., Oliveira Jr, V. C. y Barbosa, V. C. (2020). Generalized positivity constraint on magnetic equivalent layers. Geophysics, 85(6), J99-J110. https://doi.org/10.1190/geo2019-0706.1
Ronneberger, O., Fischer, P. y Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention, 234–241. https://doi.org/10.48550/arXiv.1505.04597
Saharia, C., Ho, J., Chan, W., Salimans, T., Fleet, D. J. y Norouzi, M. (2023). Image super–resolution via iterative refinement. IEEE Transactions on Pattern Analysis and Machine Intelligence, 45(4), 4713–4726. https://doi.org/10.1109/TPAMI.2022.3204461
Song, Y. y Ermon, S. (2019). Generative modeling by estimating gradients of the data distribution. Advances in Neural Information Processing Systems, 32. https://doi.org/10.48550/arXiv.1907.05600
Song, Y. y Ermon, S. (2020). Improved techniques for training score-based generative models. Advances in Neural Information Processing Systems, 33, 12438–12448.
Spector, A. y Grant, F. (1970). Statistical models for interpreting aeromagnetic data. Geophysics (1970), 35(2), 293–302.
https://doi.org/10.1190/1.1440092
Zhang, H., Ravat, D., Marangoni, Y. R., Chen, G. y Hu, X. (2018). Improved total magnetization direction determination by correlation of the normalized source strength derivative and the reduced-to-pole fields. Geophysics, 83(6), J75–J85. https://doi.org/10.1190/geo2017-0178.1
Zhou, H., Guo, R., Li, M., Yang, F., Xu, S. y Abubakar, A. (2024). Feature-based magnetotelluric inversion by variational autoencoder using a subdomain encoding scheme. Geophysics, 89(1), WA67–WA83. https://doi.org/10.1190/geo2022-0774.1
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.
Starting in 2022 (Vol. 43 number 2) articles will be published in the journal under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International license (CC BY-NC-SA 4.0)
According to these terms, the material can be shared (copied and redistributed in any medium or format) and adapted (remixed, transformed and created from the material another work), provided that a) the authorship and the original source of its publication (journal and URL of the work), b) is not used for commercial purposes and c) the same license terms are maintained.
Prior to this date the articles were published in the journal under a Creative Commons Attribution license (CC BY)
In both cases, the acceptance of the originals by the journal implies the non-exclusive assignment of the economic rights of the authors in favor of the editor, who allows reuse, after editing (postprint), under the license that corresponds according to the edition.
Such assignment means, on the one hand, that after its publication (postprint) in the GEOACTA Magazine of the Association of Geophysicists and Geodesists, the authors can publish their work in any language, medium and format (in such cases, it is requested that it be recorded that the material was originally published in this journal); on the other, the authorization of the authors for the work to be harvested by SEDICI, the institutional repository of the National University of La Plata, and to be disseminated in the databases that the editorial team considers appropriate to increase visibility. of the publication and its authors.
Likewise, the journal encourages the authors so that after their publication in the Journal of the Association of Geophysicists and Geodesists, they deposit their productions in other institutional and thematic repositories, under the principle that offering society scientific production and Unrestricted academic scholarship contributes to a greater exchange of global knowledge.
