GeONet: Geometry-Encoded Operator Network for Predicting Flow in Heterogeneous Porous Media
Abstract
Rapid prediction of pressure-driven flow in microfluidic porous-media analogs is important for experimental design, tracer-test interpretation, and real-time digital-twin development for subsurface energy and critical mineral science applications. High-fidelity numerical simulators resolve geometry-controlled flow in heterogeneous pore networks, but their computational cost limits repeated evaluation across large geometry ensembles. Here, we introduce GeONet, a Geometry-Encoded Operator Network that predicts steady-state pressure fields by jointly representing material properties, spatial coordinates, and explicit solid–void geometry. GeONet extends the conventional Branch–Trunk architecture with a dedicated Leaf Network that encodes the binary domain mask through multiscale features. The Branch Network represents porosity and permeability, the Trunk Network combines coordinates with local geometric descriptors and random Fourier features, and a gated DeepMixer integrates the three latent representations before decoding pressure. GeONet was trained and evaluated using 5,000 PFLOTRAN simulations of heterogeneous Reduced Complexity Platform domains, comprising nearly 30 TB of data. Direct comparisons with pore-scale Stokes solutions across three geometries yielded Pearson correlations of 0.987–0.991 and R² values of 0.974–0.982, supporting PFLOTRAN solutions as reference data. Under matched data partitions and two-stage hyperparameter optimization, GeONet substantially outperformed a boundary-condition-aware DeepONet baseline. Across three held-out test domains, GeONet achieved NRMSE of 0.0381–0.0500, MAPE of 1.76–2.59%, and R² of 0.9673–0.9822, compared with 0.1483–0.2040, 7.81–10.45%, and 0.5658–0.7126 for DeepONet, respectively. GeONet generalized to three additional unseen geometries, achieving NRMSE of 0.0448–0.0676 and R² of 0.9511–0.9770. These results demonstrate that explicit geometry encoding improves accuracy, pressure-range preservation, and cross-geometry generalization of neural operators for flow in heterogeneous porous domains.
Related
- A Geometry-Informed Neural Operator surrogate of LISFLOOD-FP for rapid pluvial-fluvial flood forecasting: the case of Forlì, Italy
- GenONet: A Generative operator Network for High-Resolution Precipitation Nowcasting
- Predicting hourly thermal demand of buildings for digital twin-based energy network modeling using a steady-state model capturing thermal inertia
- Predicting nitrogen dioxide column density from the International Wealth Index with a convolutional neural network in Africa
- Utilising nuclear encoded plastid DNA to identify donors of grass-to-grass lateral gene transfer
BibTeX
@misc{mamud2026geonet,
title = {{GeONet: Geometry-Encoded Operator Network for Predicting Flow in Heterogeneous Porous Media}},
author = {Md Lal Mamud and Maruti Mudunuru and Yucheng Fu and Aramy Troung and Satish Karra and Ivani Patel and Arunima Bhattacharjee and Tamas Varga and Pratanu Roy and Michael Penwarden and Rajesh K. Singh and Arjun Chakrawal and Kalyana B. Nakshatrala and Christof Meile and Erin Rooney and Glenn Hammond},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x57j84},
url = {https://eartharxiv.org/repository/view/15512/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - GeONet: Geometry-Encoded Operator Network for Predicting Flow in Heterogeneous Porous Media
AU - Md Lal Mamud
AU - Maruti Mudunuru
AU - Yucheng Fu
AU - Aramy Troung
AU - Satish Karra
AU - Ivani Patel
AU - Arunima Bhattacharjee
AU - Tamas Varga
AU - Pratanu Roy
AU - Michael Penwarden
AU - Rajesh K. Singh
AU - Arjun Chakrawal
AU - Kalyana B. Nakshatrala
AU - Christof Meile
AU - Erin Rooney
AU - Glenn Hammond
PY - 2026
DA - 2026/10/15
PB - EarthArXiv
DO - 10.31223/x57j84
UR - https://eartharxiv.org/repository/view/15512/
AB - Rapid prediction of pressure-driven flow in microfluidic porous-media analogs is important for experimental design, tracer-test interpretation, and real-time digital-twin development for subsurface energy and critical mineral science applications. High-fidelity numerical simulators resolve geometry-controlled flow in heterogeneous pore networks, but their computational cost limits repeated evaluation across large geometry ensembles. Here, we introduce GeONet, a Geometry-Encoded Operator Network that predicts steady-state pressure fields by jointly representing material properties, spatial coordinates, and explicit solid–void geometry. GeONet extends the conventional Branch–Trunk architecture with a dedicated Leaf Network that encodes the binary domain mask through multiscale features. The Branch Network represents porosity and permeability, the Trunk Network combines coordinates with local geometric descriptors and random Fourier features, and a gated DeepMixer integrates the three latent representations before decoding pressure. GeONet was trained and evaluated using 5,000 PFLOTRAN simulations of heterogeneous Reduced Complexity Platform domains, comprising nearly 30 TB of data. Direct comparisons with pore-scale Stokes solutions across three geometries yielded Pearson correlations of 0.987–0.991 and R² values of 0.974–0.982, supporting PFLOTRAN solutions as reference data. Under matched data partitions and two-stage hyperparameter optimization, GeONet substantially outperformed a boundary-condition-aware DeepONet baseline. Across three held-out test domains, GeONet achieved NRMSE of 0.0381–0.0500, MAPE of 1.76–2.59%, and R² of 0.9673–0.9822, compared with 0.1483–0.2040, 7.81–10.45%, and 0.5658–0.7126 for DeepONet, respectively. GeONet generalized to three additional unseen geometries, achieving NRMSE of 0.0448–0.0676 and R² of 0.9511–0.9770. These results demonstrate that explicit geometry encoding improves accuracy, pressure-range preservation, and cross-geometry generalization of neural operators for flow in heterogeneous porous domains.
ER -