The stratigraphic breakdown of the section according to well logging data is the basis of all types of geological work. Stratigraphic correlation of layers is a routine task, when specialist has good knowledge of the geological structure of the region and the features of formations. Automatic correlation algorithm is proposed to facilitate the work of specialists. For automatically correlation of formations, it is necessary to select several reference wells that are evenly distributed over the area. As a rule, 20 % of the total number of wells under consideration is used. The correlation of layers is carried out by a geologist in reference wells, the selected algorithm is used between reference wells. The algorithm consists of several stages. In the first stage, a three-dimensional geological surface is constructed based on the data from reference wells. This provides the initial approximation of the stratigraphic boundary. Next, in the vicinity of the obtained initial approximation, the position of the stratigraphic boundary is refined. For this purpose the similarity in the behavior of well logging curves is analyzed. The article considers two approaches for assessing similarity: the first is a classical method using a target function dependent on the correlation coefficient, and the second is correlation using a neural network. As part of the study, experiments were conducted to evaluate the quality and efficiency of the proposed methods. Models using autoencoders performed relatively well when analyzing test data. These approaches can greatly facilitate the work of geologists in process of stratigraphic correlation.
References
1. Hinton G.E., Salakhutdinov R.R., Reducing the dimensionality of data with neural networks, Science, 2006, V. 313, no. 5786, pp. 504–507,
DOI: http://doi.org/10.1126/science.1127647
2. Ma H., Wei Y., Cui X., Image denoising with convolutional autoencoders, Proceedings of IEEE International Conference on Systems, Man and Cybernetics (SMC), 2018, pp. 3076–3081.
3. Qian S., Zhang S., Jia R., A new method for environmental sound classification based on convolutional autoencoder, Proceedings of 14th IEEE Conference on Industrial Electronics and Applications (ICIEA), 2019, pp. 2644–2649.
4. Bromley J., Bentz J.W., Bottou L. et al., Signature verification using a «siamese» time delay neural network, International Journal of Pattern Recognition and Artificial Intelligence, 1993, V. 7(04), pp. 669–688, DOI: https://doi.org/10.1142/s0218001493000339
5. Taigman Y., Yang M., Ranzato M., Wolf L., DeepFace: Closing the gap to human-level performance in face verification, Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2014, DOI: https://doi.org/10.1109/cvpr.2014.220
6. Mueller J., Thyagarajan А., Siamese recurrent architectures for learning sentence similarity, Proceedings of the AAAI Conference on Artificial Intelligence, 2016,
V. 30 (1), DOI: https://doi.org/10.1609/aaai.v30i1.10350
7. Koch G., Zemel R., Salakhutdinov R., Siamese neural networks for one-shot image recognition, ICML deep learning workshop, 2015, V. 2,
