Advances in Scientific Machine Learning for Coupled Fluid Flow and Transport

2026-06-17 · Source: Machine Learning · Field: Science & Research — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences, Engineering & Applied Sciences · Depth: Expert, quick

Summary

A recent chapter reviews advances in Scientific Machine Learning (SciML) for modeling coupled fluid flow and transport phenomena, governed by incompressible Navier-Stokes and scalar transport equations. Systems like turbidity currents and thermal convection are computationally expensive due to strong nonlinear coupling and multiscale behavior. The review surveys state-of-the-art SciML surrogate models, including linear reduced-order techniques (Singular Value Decomposition, Dynamic Mode Decomposition) and nonlinear neural network approaches (Physics-Informed Neural Networks, β-Variational Autoencoders). It details the authors' integration of these models with High Performance Computing strategies, such as Adaptive Mesh Refinement/Coarsening (AMR/C) and scientific floating-point data compression. New contributions include PINN-based surrogate modeling for turbidity currents and β-VAE-based disentangled nonlinear mode extraction for thermal flows, illustrated with benchmarks like lock-exchange flows. SciML is shown to enable fast, accurate approximations of complex coupled systems, substantially reducing computational cost.

Key takeaway

For research scientists developing models for coupled fluid flow and transport, this review highlights how Scientific Machine Learning (SciML) offers a path to significantly reduce computational expense. You should consider integrating surrogate models like PINNs or β-VAEs with HPC strategies such as Adaptive Mesh Refinement/Coarsening to achieve faster, accurate approximations. This approach can accelerate high-fidelity simulations for complex systems like turbidity currents or thermal convection, enabling more efficient research and development.

Key insights

SciML offers efficient surrogate models for complex, computationally expensive coupled fluid flow and transport systems.

Principles

• Strong nonlinear coupling characterizes fluid systems.
• Multiscale behavior drives high simulation costs.
• SciML enables efficient surrogate modeling.

Method

The chapter combines linear reduced-order techniques (SVD, DMD) and nonlinear neural networks (PINNs, β-VAEs) with HPC strategies like Adaptive Mesh Refinement/Coarsening and data compression for surrogate modeling.

In practice

• Model turbidity currents via PINNs.
• Extract disentangled modes using β-VAEs.
• Apply AMR/C for HPC efficiency.

Topics

• Scientific Machine Learning
• Fluid Dynamics
• Surrogate Models
• Physics-Informed Neural Networks
• Variational Autoencoders
• High Performance Computing

Best for: AI Scientist, Research Scientist

Related on AIssential

• FLAME 2023: Diving into the Future of Fluid Dynamics and Machine Learning — AI techniques are transforming simulation by integrating physics-based modeling with data-driven machine learning.
• Engineering Hybrid Physics-Informed Neural Networks for Next-Generation Electricity Systems: A State-of-the-Art Review — Physics-informed machine learning integrates domain physics into models, enhancing accuracy and efficiency for data-scarce electricity systems.
• A Validated LBM Dataset and Pipeline for Surrogate Modeling of Turbulent 3D Obstructed Channel Flows — A validated LBM dataset and pipeline enable standardized evaluation of neural operators for 3D turbulent flows.
• Physics-Informed Neural Network Modeling of Biodegradable Contaminant Transport through GCL/SL Composite Liners — Hard-constrained PINNs significantly improve accuracy and stability for modeling contaminant transport in composite liners.
• Neural Operator-Based Surrogate Model for CFD:Helical Coil Steam Generator in Small Modular Reactor — Neural operator-based surrogates, combined with ROMs and multi-scale techniques, enable real-time CFD for SMR digital twins.
• Droplet-LNO: Physics-Informed Laplace Neural Operators for Accurate Prediction of Droplet Spreading Dynamics on Complex Surfaces — PI-LNO uses Laplace transforms and physics-informed neural networks to model droplet spreading dynamics efficiently.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.