Optimizing rock characterization using digital rock physics: Artificial Intelligence applications to reproduce 3D natural structures and patterns in rocks

• Applicant: Professor Dr. Erik H. Saenger
• Subject Area: Geophysics
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 547420511

he use of cuttings to estimate rock properties often present issues of representativity and inaccuracy, as most of these rock fragments are transported from the drill head to the surface by the mud flow and are partially altered thereby. To improve the estimation accuracy of rock properties, we propose combining high-resolution X-ray computed tomography (micro-CT), petrographic, and SEM images with artificial intelligence. As micro-CT images display a 3D volume of the rock samples compared to petrographic and SEM images, we intend to start our study using micro-CT images and Artificial Intelligence (AI) to obtain the best digital replicas of rock samples (Generation of Intelligent Rock Replicas). However, micro-CT images are not always accessible due to high costs, extensive preparation, and post-processing. Whereas petrographic and SEM images are more accessible, the idea is to reconstruct 3D images from the 2D petrographic and SEM images. Therefore, we will combine the 3D micro-CT scans and AI techniques to obtain replicas that can numerically reproduce the original rock sample's properties. Concurrently, we will use the micro-CT images to test the reconstructed AI model (generated 3D volume based on petrographic and SEM 2D images). This workflow will contribute to predicting the rock properties from cuttings images more accurately using either micro-CT, petrographic, and /or SEM images. This project aims to develop and optimize new techniques to reproduce 3D volumes from 2D petrographic and SEM cuttings images. This newly developed workflow will dramatically expand the scope of digital cuttings by reducing acquisition costs, increasing calculation speed, and simplifying the overall workflow of deriving 3D models at the pore scale. In addition, we intend to develop new and improved techniques to reproduce 3D micro-CT images from new artificial 3D images. This will significantly impact the estimation of rock properties from small samples, given the possibility of obtaining accurate and more abundant information from available and economic samples. Especially, an uncertainty estimation of the estimated properties can be developed. In addition, all this will contribute to optimizing the rock characterization using digital rock physics (DRP). The present proposal intends to join UNAM (Universidad Autonoma de Mexico) and HBO (Bochum University of Applied Sciences) efforts to achieve its goals. The UNAM part will be focused on developing and implementing codes to replicate rock images (micro-CT, petrographic and SEM) and the HBO will be focused on developing and implementing new geological-driven segmentation techniques in combination with numerical physical property simulations to assist the generation of intelligent rock replicas.

• DFG Programme: Research Grants
• International Connection: Mexico
• Partner Organisation: Universidad Nacional Autónoma de México (UNAM)
• Co-Investigator: Dr. Martin Balcewicz
• Cooperation Partners: Dante Arteaga; Juan Eduardo Linares Perez; Dr. Debora Sandra Vega Ruiz