Perspectives on predicting and controlling turbulent flows through deep learning

Physics of Fluids, Год журнала: 2024, Номер 36(3)

Опубликована: Март 1, 2024

The current revolution in the field of machine learning is leading to many interesting developments a wide range areas, including fluid mechanics. Fluid mechanics, and more concretely turbulence, an ubiquitous problem science engineering. Being able understand predict evolution turbulent flows can have critical impact on our possibilities tackle sustainability problems (including climate emergency) industrial applications. Here, we review recent emerging context predictions, simulations, control flows, focusing wall-bounded turbulence. When it comes flow control, refer active manipulation improve efficiency processes such as reduced drag vehicles, increased mixing processes, enhanced heat transfer exchangers, pollution reduction urban environments. A number important areas are benefiting from ML, identify synergies with existing pillars scientific discovery, i.e., theory, experiments, simulations. Finally, I would like encourage balanced approach community order harness all positive potential these novel methods.

Язык: Английский

---

3D elastic wave propagation with a Factorized Fourier Neural Operator (F-FNO)

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2024, Номер 420, С. 116718 - 116718

Опубликована: Янв. 9, 2024

Язык: Английский

---

Latent Neural PDE Solver: a reduced-order modelling framework for partial differential equations

Journal of Computational Physics, Год журнала: 2025, Номер unknown, С. 113705 - 113705

Опубликована: Янв. 1, 2025

Язык: Английский

---

Accelerating phase field simulations through a hybrid adaptive Fourier neural operator with U-net backbone

npj Computational Materials, Год журнала: 2025, Номер 11(1)

Опубликована: Янв. 13, 2025

Abstract Prolonged contact between a corrosive liquid and metal alloys can cause progressive dealloying. For one such process as liquid-metal dealloying (LMD), phase field models have been developed to understand the mechanisms leading complex morphologies. However, LMD governing equations in these often involve coupled non-linear partial differential (PDE), which are challenging solve numerically. In particular, numerical stiffness PDEs requires an extremely refined time step size (on order of 10 −12 s or smaller). This computational bottleneck is especially problematic when running simulation until late horizon required. motivates development surrogate capable leaping forward time, by skipping several consecutive steps at-once. this paper, we propose U-shaped adaptive Fourier neural operator (U-AFNO), machine learning (ML) based model inspired recent advances learning. U-AFNO employs U-Nets for extracting reconstructing local features within physical fields, passes latent space through vision transformer (ViT) implemented (AFNO). We use U-AFNOs learn dynamics mapping at current into later step. also identify global quantities interest (QoI) describing corrosion (e.g., deformation interface, lost metal, etc.) show that our proposed able accurately predict dynamics, spite chaotic nature LMD. Most notably, reproduces key microstructure statistics QoIs with level accuracy on par high-fidelity solver, while achieving significant 11, 200 × speed-up high-resolution grid comparing expense per Finally, investigate opportunity using hybrid simulations, alternate leaps stepping. demonstrate advantageous some design choices, fully auto-regressive settings consistently outperforms schemes.

Язык: Английский

---

ViTs as backbones: Leveraging vision transformers for feature extraction

Information Fusion, Год журнала: 2025, Номер unknown, С. 102951 - 102951

Опубликована: Янв. 1, 2025

Язык: Английский

---

Multi-scale time-stepping of Partial Differential Equations with transformers

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2024, Номер 426, С. 116983 - 116983

Опубликована: Апрель 13, 2024

Developing fast surrogates for Partial Differential Equations (PDEs) will accelerate design and optimization in almost all scientific engineering applications. Neural networks have been receiving ever-increasing attention demonstrated remarkable success computational modeling of PDEs, however; their prediction accuracy is not at the level full deployment. In this work, we utilize transformer architecture, backbone numerous state-of-the-art AI models, to learn dynamics physical systems as mixing spatial patterns learned by a convolutional autoencoder. Moreover, incorporate idea multi-scale hierarchical time-stepping increase speed decrease accumulated error over time. Our model achieves similar or better results predicting time-evolution Navier–Stokes equations compared powerful Fourier Operator (FNO) two transformer-based neural operators OFormer Galerkin Transformer. The code data are available on https://github.com/BaratiLab/MST_PDE.

Язык: Английский

---

Physics informed token transformer for solving partial differential equations

Machine Learning Science and Technology, Год журнала: 2024, Номер 5(1), С. 015032 - 015032

Опубликована: Фев. 9, 2024

Abstract Solving partial differential equations (PDEs) is the core of many fields science and engineering. While classical approaches are often prohibitively slow, machine learning models fail to incorporate complete system information. Over past few years, transformers have had a significant impact on field Artificial Intelligence seen increased usage in PDE applications. However, despite their success, currently lack integration with physics reasoning. This study aims address this issue by introducing Physics Informed Token Transformer (PITT). The purpose PITT knowledge embedding PDEs into process. uses an equation tokenization method learn analytically-driven numerical update operator. By tokenizing derivatives, transformer become aware underlying behind physical processes. To demonstrate this, tested challenging 1D 2D operator tasks. results show that outperforms popular neural has ability extract physically relevant information from governing equations.

Язык: Английский

---

Transformers as neural operators for solutions of differential equations with finite regularity

Computer Methods in Applied Mechanics and Engineering, Год журнала: 2024, Номер 434, С. 117560 - 117560

Опубликована: Ноя. 28, 2024

Язык: Английский

---

Superstep wavefield propagation

Wave Motion, Год журнала: 2025, Номер unknown, С. 103489 - 103489

Опубликована: Янв. 1, 2025

Язык: Английский

---

A Hybrid CNN-Transformer Surrogate Model for the Multi-Objective Robust Optimization of Geological Carbon Sequestration

Advances in Water Resources, Год журнала: 2025, Номер unknown, С. 104897 - 104897

Опубликована: Янв. 1, 2025

Язык: Английский

---