Speaker
Description
A key task in the oil industry is the accurate characterization of pre-salt carbonate reservoir rocks, which display complex heterogeneity at multiple scales. These rocks’ intricate geological structure, shaped by a range of diagenetic processes—such as cementation, dissolution, and fracturing—significantly influences their petrophysical properties [1]. These characteristics demand for cutting-edge characterization methods.
Micro-computed tomography (micro-CT) is a non-destructive technique that creates three-dimensional volumes of objects. In the oil sector, micro-CT of rock samples is often used to evaluate properties like permeability, porosity and fluid connectivity. Furthermore, artificial intelligence (AI), particularly deep learning, has emerged as the state-of-the-art solution for computer vision tasks. Those models have been successfully applied to tasks including segmentation and classification, aiding specialists during the reservoir characterization[2, 3, 4, 5, 6].
Despite the recent progress, not many datasets of carbonate pre-salt rocks are publicly available, particularly those offering both high- and low-resolution images. This scarcity makes it harder to benchmark methods, validate models, and perform multi-scale analysis which are critical for understanding the hierarchical structure of carbonate reservoirs. The dataset titled "16 Brazilian Pre-Salt Carbonates: Multi-Resolution Micro-CT Images¹" [7, 8] is an open dataset of micro-CT images of high and low resolutions paired with their segmentations, as exemplified in Figure 1. Each sample’s porosity and permeability are also available.
The objective of this study is to establish benchmarks for image segmentation and the prediction of petrophysical properties using the recently released pre-salt carbonate dataset. We compare different deep learning architectures and work with either 2D slices or whole 3D volumes. In this work, we also analyse how image resolution affect the accuracy of the predictions. In the end, this benchmark could be used as an initial study of the dataset and verify how different methods and data resolutions affect the results of image-based characterization.
¹ - https://www.digitalrocksportal.org/projects/503
| References | [1] Huafeng Sun, Sandra Vega, and Guo Tao. “Analysis of heterogeneity and permeability anisotropy in carbonate rock samples using digital rock physics”. In: Journal of petroleum science and engineering 156 (2017), pp. 419–429. [2] Naif Alqahtani et al. “Machine learning for predicting properties of porous media from 2d X-ray images”. In: Journal of Petroleum Science and Engineering 184 (2020), p. 106514. [3] Carlos EM dos Anjos et al. “Deep learning for lithological classification of carbonate rock micro-CT images”. In: Compu- tational Geosciences (2021), pp. 1–13. [4] Ying Da Wanga et al. “Deep Neural Networks for Improving Physical Accuracy of 2D and 3D Multi-Mineral Segmentation of Rock micro-CT Images”. In: Applied Soft Computing 104 (2021), p. 107185. [5] Carlos Eduardo Menezes dos Anjos et al. “Permeability estimation on raw micro-CT of carbonate rock samples using deep learning”. In: Geoenergy Science and Engineering (2023), p. 211335. [6] Júlio de Castro Vargas Fernandes et al. “Absolute permeability estimation from microtomography rock images through deep learning super-resolution and adversarial fine tuning”. In: Scientific Reports 14.1 (2024), p. 16704. [7] Alyne Vidal et al. 16 Brazilian pre-salt carbonates: multi-resolution micro-CT images. 2024. DOI: 10.17612/xr50-s717. [8] Vidal, A.D., Neta, A.P.M., de Castro Vargas Fernandes, J. et al. Multi-resolution X-ray micro-computed tomography images of carbonate rocks from brazilian pre-salt. Sci Data 11, 1361 (2024). https://doi.org/10.1038/s41597-024-04198-9 |
|---|---|
| Country | Brazil |
| Student Awards | I would like to submit this presentation into both awards |
| Acceptance of the Terms & Conditions | Click here to agree |
