Paper contribution

Predicting fluid and mechanical response in 3D discrete fractures using graph neural network

Botong Du; Zhao Feng; Fengshou Zhang; Bicheng Yan

Proceedings
EUROCK 2026
Pages
1323-1328
Submitted
28 January 2026
Accepted
15 June 2026
Published
12 August 2026

Summary

Abstract

Accurate prediction of fluid flow and mechanical deformation in fractured rock formation is critical for deep subsurface projects. Although high-fidelity numerical simulations can resolve these coupled processes, their computational cost limits rapid screening and optimization of injection strategies. This study presents a graph neural network (GNN) based surrogate for predicting the hydro-mechanical response in 3D discrete fracture networks (DFNs). The DFN is represented as a fracture-level graph, in which fractures are modelled as nodes with features encoding fracture geometry and spatial context, and fracture intersections are represented as edges with features encoding network connectivity and topology. The training data are generated using a distinct element method simulator 3DEC, which solves coupled hydro-mechanical processes during hydraulic fracturing. Separate graph convolutional networks (GCNs) are trained to predict the node-wise average fracture aperture and pore pressure under different injection conditions. Results show that the surrogate achieves high predictive accuracy for both fracture aperture and pore pressure, with test R² values of 0.9937 and 0.9894, respectively. For aperture prediction, the mean absolute error (MAE) is 0.01 mm. The model captures the response of both the primary hydraulic fracture and natural fractures. The trained model delivers a prediction for a single scenario in approximately 6.43 ms on a CPU, compared to the 11.5 h required by the simulator. The proposed workflow enables rapid estimation of the extent of activated fractures and permeability changes under varying operating conditions, providing a practical basis for injection parameter screening, optimization, and uncertainty analysis in fractured reservoirs.

Contributors

Authors

  1. Botong Du Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092, China; Key Laboratory of Geotechnical & Underground Engineering of Ministry of Education, Tongji University, Shanghai 200092, China; Physical Science and Engineering Division, King Abdullah University of Science and Technology (KAUST), Thuwal, 23955-6900, Saudi Arabia ORCID 0009-0005-9227-7502
  2. Zhao Feng Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092, China; Key Laboratory of Geotechnical & Underground Engineering of Ministry of Education, Tongji University, Shanghai 200092, China
  3. Fengshou Zhang Department of Geotechnical Engineering, College of Civil Engineering, Tongji University, Shanghai 200092, China; Key Laboratory of Geotechnical & Underground Engineering of Ministry of Education, Tongji University, Shanghai 200092, China ORCID 0000-0002-4998-6259
  4. Bicheng Yan Physical Science and Engineering Division, King Abdullah University of Science and Technology (KAUST), Thuwal, 23955-6900, Saudi Arabia

Index terms

Keywords

Full text

Attachments

Works cited

References

  1. Bronstein MM, Bruna J, LeCun Y, Szlam A, Vandergheynst P (2017) Geometric deep learning: Going beyond Euclidean data. IEEE Signal Processing Magazine 34(4): 18–42. DOI:10.1109/MSP.2017.2693418.
  2. Detournay E (2016) Mechanics of hydraulic fractures. Annual Review of Fluid Mechanics 48: 311–339. DOI:10.1146/annurev-fluid-010814-014736.
  3. Dontsov EV, Zhang F (2018) Calibration of tensile strength to model fracture toughness with distinct element method. International Journal of Solids and Structures 144: 180–191. DOI:10.1016/j.ijsolstr.2018.05.001.
  4. Du B, Zhang F, Zhang C (2025) Distinct element modeling of hydraulic fracture propagation with discrete fracture network at Gonghe enhanced geothermal system site, northwest China. Journal of Rock Mechanics and Geotechnical Engineering 17(6): 3435–3448. DOI:10.1016/j.jrmge.2024.04.028.
  5. Itasca Consulting Group, Inc. (2020) 3DEC: Three-dimensional distinct element code, version 7.0. Itasca Consulting Group, Minneapolis, MN.
  6. Karniadakis GE, Kevrekidis IG, Lu L, Perdikaris P, Wang S, Yang L (2021) Physics-informed machine learning. Nature Reviews Physics 3(6): 422–440. DOI:10.1038/s42254-021-00314-5.
  7. Lei Q, Latham J-P, Tsang C-F (2017) The use of discrete fracture networks for modelling coupled geomechanical and hydrological behaviour of fractured rocks. Computers and Geotechnics 85: 151–176. DOI:10.1016/j.compgeo.2016.12.024.
  8. McClure MW, Horne RN (2014) An investigation of stimulation mechanisms in Enhanced Geothermal Systems. International Journal of Rock Mechanics and Mining Sciences 72: 242–260. DOI:10.1016/j.ijrmms.2014.07.011.
  9. Pfaff T, Fortunato M, Sanchez-Gonzalez A, Battaglia PW (2021) Learning mesh-based simulation with graph networks. International Conference on Learning Representations, Virtual Event, 3–7 May 2021. DOI:10.48550/arXiv.2010.03409.
  10. Riahi A, Pettitt W, Damjanac B, Varun, Blanksma D (2019) Numerical modeling of discrete fractures in a field-scale FORGE EGS reservoir. Rock Mechanics and Rock Engineering 52(12): 5245–5258. DOI:10.1007/s00603-019-01894-6.
  11. Tang M, Liu Y, Durlofsky LJ (2020) A deep-learning-based surrogate model for data assimilation in dynamic subsurface flow problems. Journal of Computational Physics 413: 109456. DOI:10.1016/j.jcp.2020.109456.
  12. Zhang F, Yin Z, Chen Z, Maxwell S, Zhang L, Wu Y (2020) Fault reactivation and induced seismicity during multistage hydraulic fracturing: Microseismic analysis and geomechanical modeling. SPE Journal 25(2): 692–711. DOI:10.2118/199883-PA.