Physics-informed neural networks for solving Reynolds-averaged Navier–Stokes equations

Physics of Fluids, Journal Year: 2022, Volume and Issue: 34(7)

Published: June 17, 2022

Physics-informed neural networks (PINNs) are successful machine-learning methods for the solution and identification of partial differential equations (PDEs). We employ PINNs solving Reynolds-averaged Navier$\unicode{x2013}$Stokes (RANS) incompressible turbulent flows without any specific model or assumption turbulence, by taking only data on domain boundaries. first show applicability laminar Falkner$\unicode{x2013}$Skan boundary layer. then apply simulation four turbulent-flow cases, i.e., zero-pressure-gradient layer, adverse-pressure-gradient over a NACA4412 airfoil periodic hill. Our results excellent with strong pressure gradients, where predictions less than 1% error can be obtained. For flows, we also obtain very good accuracy even Reynolds-stress components.

Language: Английский

---

A Review of Physics-Informed Machine Learning in Fluid Mechanics

Energies, Journal Year: 2023, Volume and Issue: 16(5), P. 2343 - 2343

Published: Feb. 28, 2023

Physics-informed machine-learning (PIML) enables the integration of domain knowledge with machine learning (ML) algorithms, which results in higher data efficiency and more stable predictions. This provides opportunities for augmenting—and even replacing—high-fidelity numerical simulations complex turbulent flows, are often expensive due to requirement high temporal spatial resolution. In this review, we (i) provide an introduction historical perspective ML methods, particular neural networks (NN), (ii) examine existing PIML applications fluid mechanics problems, especially Reynolds number (iii) demonstrate utility techniques through a case study, (iv) discuss challenges developing mechanics.

Language: Английский

---

A physics-informed convolutional neural network for the simulation and prediction of two-phase Darcy flows in heterogeneous porous media

Journal of Computational Physics, Journal Year: 2023, Volume and Issue: 477, P. 111919 - 111919

Published: Jan. 18, 2023

Language: Английский

---

A comprehensive review of advances in physics-informed neural networks and their applications in complex fluid dynamics

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(10)

Published: Oct. 1, 2024

Physics-informed neural networks (PINNs) represent an emerging computational paradigm that incorporates observed data patterns and the fundamental physical laws of a given problem domain. This approach provides significant advantages in addressing diverse difficulties field complex fluid dynamics. We thoroughly investigated design model architecture, optimization convergence rate, development modules for PINNs. However, efficiently accurately utilizing PINNs to resolve dynamics problems remain enormous barrier. For instance, rapidly deriving surrogate models turbulence from known characterizing flow details multiphase fields present substantial difficulties. Additionally, prediction parameters multi-physics coupled models, achieving balance across all scales multiscale modeling, developing standardized test sets encompassing dynamic are urgent technical breakthroughs needed. paper discusses latest advancements their potential applications dynamics, including turbulence, flows, multi-field flows. Furthermore, we analyze challenges face these outline future trends growth. Our objective is enhance integration deep learning facilitating resolution more realistic problems.

Language: Английский

---

Analysis of MHD Flow With Convective Boundary Conditions Over a Permeable Stretching Surface Using a Physics‐Informed Neural Network

Heat Transfer, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 3, 2025

ABSTRACT In this study, we examine the impact of heat and mass transfer magnetohydrodynamic (MHD) flow through a stretching permeable surface while considering chemical reaction convective boundary conditions. A physics‐informed neural network (PINN) approach is employed to obtain precise solutions, representing key novelty work. The governing partial differential equations were transformed into nonlinear ordinary by applying similarity transformations. These are integrated PINN's loss function enforce initial conditions, enabling model learn effectively during training. We analyze various parameters related velocity, thermal, concentration distributions present results graphically. findings indicate that injecting fluid leads reduction in velocity gradient as moves away from surface, whereas suction has opposite effect, increasing gradient. parameter significantly reduces layer thickness, an effect further enhanced magnetic parameter. thermal layers primarily affected Schmidt Prandtl numbers. Additionally, slows near sheet, increases temperature at plate's surface. Our proposed method shows significant agreement with previous studies, validating its effectiveness solving complex MHD problems. provide deeper insights dynamics flows have implications for applications involving transfer, such reactors, cooling systems, material processing, environmental management.

Language: Английский

---

Numerical prediction of cavitation erosion risk in an axisymmetric nozzle using a multi-scale approach

Physics of Fluids, Journal Year: 2022, Volume and Issue: 34(6)

Published: June 1, 2022

In the present study, a two-way coupling Eulerian–Lagrangian approach is developed to assess cavitation erosion risk in an axisymmetric nozzle. Macroscopic structures are simulated using large eddy simulation along with volume of fluid method. The compressible Rayleigh–Plesset equation and bubble motion introduced resolve microscopic dynamics. calculated results agree favorably experimental data can capture more flow details, which associated potential risk. Based on information multi-scale cavitating flow, new asymmetric collapse model proposed calculate impact pressure, then used quantitatively show that, compared traditional Euler method, location value maximum predicted by this method closer measurement. advantages newly further elaborated systematically. study found that high environmental pressure triggered shedding clouds cause near-wall bubbles shrink even collapse, releasing impulsive directly damages material surface. This phenomenon considered be actual process. Finally, analyzing relationship between reveals mainly due oscillation generated near attached cavity closure line or surrounding clouds.

Language: Английский

---

Physics-informed neural networks for heat transfer prediction in two-phase flows

International Journal of Heat and Mass Transfer, Journal Year: 2023, Volume and Issue: 221, P. 125089 - 125089

Published: Dec. 21, 2023

This paper presents data-driven simulations of two-phase fluid processes with heat transfer. A Physics-Informed Neural Network (PINN) was applied to capture the behaviour phase interfaces in flows and model hydrodynamics transfer flow configurations representative established numerical test cases. The developed PINN approach trained on simulation data derived from physically based Computational Fluid Dynamics (CFD) interface capturing. present study considers fundamental problems, including tracking rise a single gas bubble denser exploring wake rising close heated wall. Tracking fluids disparate properties performed, revealing maximum error only 5.2% at edge 2.8% position centre mass. Inferred (hidden variable) are studied addition purely extrapolative inverse isothermal case. When no velocity supplied, field predictions remained accurate. Rise an inferred unseen found produce mean-squared 0.28 mass 1.25%. For case hot wall, temperature domain using specified boundary conditions 6.8%, while analysis reveals positional 3.6%. These results demonstrate that is agnostic geometry when studying combined effects convection buoyancy for first time. work serves as starting point multiphase cases involving over range geometries. Eventually, will be used such provide solutions forward, inverse, Each which represent dramatic saving computational cost compared traditional CFD.

Language: Английский

---

Theoretical prediction model of transient performance for a mixed flow pump under fast start-up conditions

Physics of Fluids, Journal Year: 2023, Volume and Issue: 35(2)

Published: Feb. 1, 2023

There always appear unsteady characteristics during start-up periods of pumps, which can lead to instability the entire system. However, lack a method for quickly and accurately predicting pump performance makes it difficult analyze overall system period. To this end, theoretical model predict transient under fast conditions is established in present study. The prediction steady built based on loss modeling first. Then, balance between head pipeline considered determine performance. A time stepping algorithm proposed solve periods. corresponding are applied mixed flow with various acceleration time. predicted evolution shows good agreement experimental measurements, average relative errors within 10% both conditions. In addition, oscillating curves impact head. mechanism results relation peak rotation speed revealed.

Language: Английский

---

The Application of Physics-Informed Machine Learning in Multiphysics Modeling in Chemical Engineering

Industrial & Engineering Chemistry Research, Journal Year: 2023, Volume and Issue: 62(44), P. 18178 - 18204

Published: Oct. 26, 2023

Physics-Informed Machine Learning (PIML) is an emerging computing paradigm that offers a new approach to tackle multiphysics modeling problems prevalent in the field of chemical engineering. These often involve complex transport processes, nonlinear reaction kinetics, and coupling. This Review provides detailed account main contributions PIML with specific emphasis on momentum transfer, heat mass reactions. The progress method development (e.g., algorithm architecture), software libraries, applications coupling surrogate modeling) are detailed. On this basis, future challenges highlight importance developing more practical solutions strategies for PIML, including turbulence models, domain decomposition, training acceleration, modeling, hybrid geometry module creation.

Language: Английский

---

Physics-informed neural networks for transonic flow around a cylinder with high Reynolds number

Physics of Fluids, Journal Year: 2024, Volume and Issue: 36(3)

Published: March 1, 2024

The physics-informed neural network (PINN) method is extended to learn and predict compressible steady-state aerodynamic flows with a high Reynolds number. To better the thin boundary layer, sampling distance function hard condition are explicitly introduced into input output layers of deep network, respectively. A gradient weight factor considered in loss implement PINN methods based on averaged Navier–Stokes (RANS) Euler equations, respectively, denoted as PINN–RANS PINN–Euler. Taking transonic flow around cylinder an example, these first verified for ability complex then applied global part physical data. When predicting velocity data local key regions, can always accurately field including layer wake, while PINN–Euler inviscid region. subsonic under different freestream Mach numbers (Ma∞= 0.3–0.7), fields predicted by both avoid inconsistency real phenomena pure data-driven method. insufficient shock identification capabilities. Since does not need second derivative, training time only 1/3 times that at same point network.

Language: Английский

---