PoreML: A Data-Driven Framework for Learning Multiphase Flow in Porous Media

summary

Video file (mp4)

The gist

Detailed Research Summary of PoreML: A Data-Driven Framework for Learning Multiphase Flow in Porous Media PoreML is an open-source, physics-grounded framework designed to unify data generation, model

In short

PoreML is an open-source framework that unifies data generation, model training, and evaluation for learning multiphase flow in porous media. It introduces a validated Lattice Boltzmann solver to create a massive dataset of 560 runs and 158,546 time steps across diverse geometries. The framework uses advanced metrics like the Euler characteristic to rigorously test machine learning models for physical accuracy.

Key concepts

Lattice Boltzmann Method (LBM)

This is a GPU-native solver used to simulate fluid flow within porous structures. It models the physics by distributing fluid properties across a discrete grid, ensuring that the resulting motion is consistent with fundamental fluid dynamics equations like Navier–Stokes. It allows for reproducible data generation.

PoreML Dataset

This is an extensive collection of time-resolved, 3D fluid distribution data for multiphase flow in porous media. It includes 560 simulation runs and over 158,000 stored time steps across different geometries and wetting conditions. This dataset serves as the foundation for training machine learning models.

Euler Characteristic ($\chi$)

This mathematical tool is used to quantify the topological fidelity of a fluid flow simulation. It measures the connectivity of the fluid phases by counting features like islands, tunnels, and enclosed bubbles. A high Euler characteristic indicates that a model accurately captures how fluids are connected in space.

Autoregressive Rollouts

This is a machine learning evaluation technique where models are tested by predicting future states sequentially over long sequences of time steps. This assesses the model's ability to maintain physical consistency and predict the long-term evolution of fluid interfaces rather than just single-step accuracy.

Terminology used across episodes

• Reinforcement Learning (RL) — explained in 25 episodes
• Sample Complexity — explained in 19 episodes
• Federated Learning — explained in 17 episodes
• Retrieval-Augmented Generation (RAG) — explained in 17 episodes
• Quantum Fisher Information (QFI) — explained in 16 episodes
• Flow Matching — explained in 15 episodes

This episode discusses

• PoreML: A Data-Driven Framework for Learning Multiphase Flow in Porous Media · Paper Radio

The paper

PoreML: A Data-Driven Framework for Learning Multiphase Flow in Porous Media · Read on arXiv

Chunyang Wang, Mingrui Zhang, Yuyan Zhang, Linqi Zhu, Xin Ju, Edo Sicco Boek, Martin J. Blunt, Gege Wen

Imperial College London University of Stanford University Queen Mary University of London EarthFlow AI, Inc.

Multiphase flow in porous microstructures is central to CO 2 storage, fuel-cell operation, and flip-chip packaging. Predicting these flows remains challenging because wettability and complex pore geometry govern the nonlinear evolution of fluid interfaces. Machine learning holds substantial promise for advancing the field, but progress is constrained by scarce time-resolved 3D datasets and a lack of a unified workflow for training and evaluating models. To fill this critical gap, we introduce PoreML, an open-source framework unifying data generation, model training, and evaluation grounded in pore-scale physics. The framework comprises three core components. (a) A modern GPU-native lattice Boltzmann solver, validated against analytical solutions and published experiments, enables reproducible data generation. (b) A 3.3 TB dataset contains 560 simulation runs and 158,546 stored time steps across four application-driven scenarios. These trajectories span synthetic structures and geometries derived from micro-CT scans of real materials, covering diverse wetting conditions and viscosity ratios. (c) A unified learning framework evaluates one-step prediction and autoregressive rollouts. Its domain-specific evaluation protocols assess predictive accuracy and physical consistency. We evaluate five models of diverse architecture under these protocols. Two complementary challenges assess transfer to larger domains and from synthetic to micro-CT-derived structures. PoreML provides a shared foundation for machine-learning research on multiphase flow in porous media, with the aim of empowering the community to develop reliable predictive models and advance the field.

Transcript

More episodes

• 2610.10273-A Closed-Loop Non-Asymptotic Convergence Analysis of PPO with Learned Critics and Clipping
• 2610.10245-MorphCL: Morphological Contrastive Learning for Inertial-based Human Activity Recognition
• 2610.09877-Layerwise Error Attribution for Fast and Robust Mixed-Precision Post-Training Quantization
• 2610.10274-Sparse Planning in Visual World Models via Cost Gradients
• 2610.09946-Learning Traffic Flow Dynamics with Stochastic Physics-Informed Neural Cellular Automata
• 2610.09969-TR-PTQ: High-Accuracy Integer-Only Transformer Post Training Quantization via Taylor Region Reformulation
• 2610.09985-Marrying Pricing and Advertising with LLMs
• 2610.10037-Matching of signal, noise and hardware timescales for filtering and forecasting of correlated noise signals
• 2610.10057-WxFM-XL: Adapting Univariate Foundation Models to Multi-Station Weather Forecasting
• 2610.10122-What Can a Gaussian Process Design Test

← Home