Multi-gate mixture-of-different-experts model for the prediction of multiple properties in multi-phase rock
Abstract
The prediction of petrophysical properties in multi-phase rock is critical for various geoenergy applications. To this end, deep learning-based methods have recently emerged as a prominent research focus in rock physics. However, due to the inherent scarcity of structural data in multi-phase rocks, coupled with the limitations of convolutional neural network in capturing multi-scale dynamic phase interface and mitigating parameter interference in multi-task learning, predictive performance has not yet reached a satisfactory level. To address this shortcoming, a new multi-task learning framework based on a multi-gate mixture-of-different-experts model is proposed to predict multiple properties of multi-phase rocks. A semi-supervised fractal-informed generative adversarial network is employed to reconstruct multi-phase rocks images, while the finite element method is used to compute their transport properties. The gating network allocates bespoke expert subsets to each task, and an automatic weighted loss function dynamically balances the task-specific loss contributions, enhancing performance and generalization. The results show that the statistical average of the predicted relative permeability aligns with the Brooks-Corey equation, demonstrating the reliability of the proposed model in preserving fundamental physical principles. On the validation dataset, the model achieves high predictive accuracy across all target properties, including fractal dimension, porosity, saturation, permeability of gas phase, permeability of water phase, and effective permeability. Comparative evaluation on the test dataset demonstrates that the proposed model significantly outperforms other multi-task models. These findings confirm that the proposed framework can simultaneously and accurately predict multi-phase rock properties under limited data conditions, holding a promise for guiding assessments in hydrocarbon and geothermal exploration, CO₂ sequestration, nuclear waste disposal, and geological hazard mitigation, among others.
Document Type: Original article
Cited as: Tang, Z., Qiu, S., Cai, Z., Law, C.-L., Mao, X., Xu, P. Multi-gate mixture-of-different-experts model for the prediction of multiple properties in multi-phase rocks. Advances in Geo-Energy Research, 2026, 19(2): 182-196. https://doi.org/10.46690/ager.2026.02.06
DOI:
https://doi.org/10.46690/ager.2026.02.06Keywords:
Multi-phase rock, structural characteristics, permeability prediction, multi-task learning, generative adversarial networkReferences
Al Balushi, F., Taleghani, A. D. Digital rock analysis to estimate stress-sensitive rock permeabilities. Computers and Geotechnics, 2022, 151: 104960.
Alzahrani, M. K., Shapoval, A., Chen, Z., et al. Micrographnets: Automated characterization of the micro-scale wettability of porous media using graph neural networks. Capillarity, 2024, 12(3): 57-71.
Argilaga, A. FEM-GAN: A physics-supervised deep learning generative model for elastic porous materials. Materials, 2023a, 16(13): 4740.
Argilaga, A. Fractal informed generative adversarial networks (FI-GAN): Application to the generation of X-ray CT images of a self-similar partially saturated sand. Computers and Geotechnics, 2023b, 158: 105384.
Blunt, M. J., Bijeljic, B., Dong, H., et al. Pore-scale imaging and modelling. Advances in Water Resources, 2013, 51: 197-216.
Brown, G. O. Henry darcy and the making of a law. Water Resources Research, 2002, 38(7): 1106.
Cai, J., Jiao, X., Wang, H., et al. Multiphase fluid-rock interactions and flow behaviors in shale nanopores: A comprehensive review. Earth-Science Reviews, 2024, 257: 104884.
Cao, D., Ji, S., Cui, R., et al. Multi-task learning for digital rock segmentation and characteristic parameters computation. Journal of Petroleum Science and Engineering, 2022, 208: 109202.
Cao, L., Jiang, F., Chen, Z., et al. Data-driven interpretable machine learning for prediction of porosity and permeability of tight sandstone reservoir. Advances in GeoEnergy Research, 2025, 16(1): 21-35.
Chakraborty, A., Rabinovich, A., Moreno, Z. A data-driven physics-informed deep learning approach for estimating sub-core permeability from coreflooding saturation measurements. Advances in Water Resources, 2025, 198: 104919.
Chen, H., Xue, L., Liu, L., et al. Physics-informed graph neural network for predicting fluid flow in porous media. Petroleum Science, 2025, 22(10): 4240-4253.
Du, H., Zhao, Z., Cheng, H., et al. Modeling density-driven flow in porous media by physics-informed neural networks for CO2 sequestration. Computers and Geotechnics, 2023, 159: 105433.
Du, Q., Zhang, Z., Liu, M., et al. Prediction and optimization of flow and heat transfer performance for helium turbine disc cavity with hybrid deep neural network. International Communications in Heat and Mass Transfer, 2025, 168: 109497.
Geng, S., Zhai, S., Li, C., et al. Swin transformer based transfer learning model for predicting porous media permeability from 2D images. Computers and Geotechnics, 2024, 168: 106177.
Ghedira, A., Lataoui, Z., Benselama, A. M., et al. Numerical simulation of incompressible two-phase flows with phase change process in porous media. Results in Engineering, 2025, 25: 103706.
Han, R., Wang, Z., Guo, Y., et al. Multi-label prediction method for lithology, lithofacies and fluid classes based on data augmentation by cascade forest. Advances in Geo-Energy Research, 2023, 9(1): 25-37.
Hashemi, Z., Gholampour, M., Wu, M. C., et al. A physicsinformed neural networks modeling with coupled fluid flow and heat transfer – revisit of natural convection in cavity. International Communications in Heat and Mass Transfer, 2024, 157: 107827.
He, K., Zhang, X., Ren, S., et al. Deep residual learning for image recognition. Paper Presented at the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27-30 June, 2016.
Hou, Z., Cao, D. Estimating elastic parameters from digital rock images based on multi-task learning with multigate mixture-of-experts. Journal of Petroleum Science and Engineering, 2022, 213: 110310.
Hu, C., Sun, B. Multitask learning for petrophysical attributeprediction using convolutional neural network and imbalance dataset. Paper SEG 2020-W13-03 Presented at the SEG International Exposition and Annual Meeting, Virtual, 11-16 October, 2020.
James, G., Witten, D., Hastie, T., et al. An Introduction to Statistical Learning: With Applications in R. Springer Nature, New York, USA, 2013.
Kang, Q., Li, K., Fu, J., et al. Hybrid LBM and machine learning algorithms for permeability prediction of porous media: A comparative study. Computers and Geotechnics, 2024, 168: 106163.
Kato, M., Takahashi, M., Kawasaki, S., et al. Segmentation of multi-phase X-ray computed tomography images. Environmental Geotechnics, 2015, 2(2): 104-117.
Kendall, A., Gal, Y., Cipolla, R. Multi-task learning using uncertainty to weigh losses for scene geometry and semantics. Paper Presented at the IEEE Conference on Computer Vision and Pattern Recognition, Salt Lake City, UT, USA, 18-23 June, 2018.
Lee, J., Kwon, M., Hong, Y. Predicting porosity and water saturation from well-log data using probabilistic multitask neural network with normalizing flows. Paper OTC 31085 Presented at the Offshore Technology Conference, Virtual and Houston, Texas, 16-19 August, 2021.
Liu, L., Dai, S., Ning, F., et al. Fractal characteristics of unsaturated sands-implications to relative permeability in hydrate-bearing sediments. Journal of Natural Gas Science and Engineering, 2019a, 66: 11-17.
Liu, P., Zhao, J., Li, Z., et al. Numerical simulation of multiphase multi-physics flow in underground reservoirs: Frontiers and challenges. Capillarity, 2024, 12(3): 72-79.
Liu, S., Johns, E., Davison, A. J. End-to-end multi-task learning with attention. Paper Presented at the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Long Beach, CA, USA, 15-20 June, 2019b.
Ma, J. Im2mesh (2D image to triangular meshes). MATLAB Central File Exchange, 2019.
Ma, J., Zhao, Z., Yi, X., et al. Modeling task relationships in multi-task learning with multi-gate mixture-of-experts. Paper Presented at the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, London, United Kingdom, 19-23 August, 2018.
Meng, Y., Jiang, J., Wu, J., et al. Transformer-based deep learning models for predicting permeability of porous media. Advances in Water Resources, 2023, 179: 104520.
Misra, I., Shrivastava, A., Gupta, A., et al. Cross-stitch networks for multi-task learning. Paper Presented at the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 27-30 June, 2016.
Montero, F. A. C., Coelho, B. Z., Smyrniou, E., et al. Schemagan: A conditional generative adversarial network for geotechnical subsurface schematisation. Computers and Geotechnics, 2025, 183: 107177.
Ng, C. W. W., Zhou, C., Ni, J. Advanced Unsaturated Soil Mechanics: Theory and Applications. London, UK, CRC Press, 2024.
Otsu, N. A threshold selection method from gray-level histograms. Automatica, 1975, 11: 285-296.
Rene, N. N., Liang, W., Chen, Y., et al. Modeling of multiphase fluids flow in anisotropic rock mass during CO2 sequestration in fractured reservoirs. Geoenergy Science and Engineering, 2025, 251: 213904.
Ruder, S., Bingel, J., Augenstein, I., et al. Latent multi-task architecture learning. Paper Presented at the Thirty-Third AAAI Conference on Artificial Intelligence, Honolulu, HI, USA, 27-31 January, 2019.
Russ, J. C. The Image Processing Handbook. Boca Raton, USA, CRC Press, 2006.
Shazeer, N., Mirhoseini, A., Maziarz, K., et al. Outrageously large neural networks: The sparsely-gated mixture-ofexperts layer. ArXiv Preprint ArXiv: 1701.06538, 2017.
Siddiqa, S., Chang, K., Naqvi, S. B., et al. AI-assisted proton exchange membrane (PEM) fuel cell performance prediction using CFD and data-driven surrogate models. International Communications in Heat and Mass Transfer, 2024, 156: 107616.
Temam, R. Navier-Stokes Equations: Theory and Numerical Analysis. Providence, USA, American Mathematical Society, 2024.
Wang, T., Wang, Y. D., Sun, C., et al. Surface wetting characterization in pore-scale multiphase flow simulations: A ketton carbonate case study. Geoenergy Science and Engineering, 2024, 240: 212933.
Wu, H., Fang, W., Kang, Q., et al. Predicting effective diffusivity of porous media from images by deep learning. Scientific Reports, 2019, 9(1): 20387.
Xu, P., Qiu, S., Yu, B., et al. Prediction of relative permeability in unsaturated porous media with a fractal approach. International Journal of Heat and Mass Transfer, 2013, 64: 829-837.
Yang, M., Wang, D., Dong, Z., et al. Insight into the permeability effect on forced convective heat transfer characteristics in porous media based on the pore-scale numerical study. International Communications in Heat and Mass Transfer, 2025, 165: 109004.
Yin, P., Song, H., Ma, H., et al. The modification of the Kozeny-Carman equation through the lattice Boltzmann simulation and experimental verification. Journal of Hydrology, 2022, 609: 127738.
Yu, Y., Wei, W., Cui, W., et al. Multi-phase segmentation methods for micro-tomographic images based on deep learning. Geoenergy Science and Engineering, 2025, 252: 213962.
Zarin, T., Eshkaftaki, H. A., Sharifi, A. Machine learningbased prediction of oil-water relative permeability using core flooding and CT-scan data. Journal of Molecular Liquids, 2025, 27:105735.
Zhang, T., Bian, N., Liu, Q., et al. 3D super-resolution reconstruction of porous media based on GANs and CBAMs. Stochastic Environmental Research and Risk Assessment, 2024, 38(4): 1475-1504.
Zhao, J., Wu, J., Wang, H., et al. Single phase flow simulation in porous media by physical-informed Unet network based on lattice Boltzmann method. Journal of Hydrology, 2024, 639: 131501.
Downloads
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Author(s)
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.