Physics-informed neural networks for the Reynolds-Averaged Navier–Stokes modeling of Rayleigh–Taylor turbulent mixing

Xiao, Meng-Juan; Yu, Teng-Chao; Zhang, You-Sheng; Yong, Heng

doi:10.1016/j.compfluid.2023.106025

Cited by 9 publications

(2 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Increasing errors associated with higher Reynolds numbers were also reported by Sun et al [8] and Harmening et al [12]. Consequently, numerous studies were conducted incorporating measured or simulated training data to support training of the PINN [14][15][16][17][18][24][25][26], while investigations focusing unsupervised physics-informed DL of high Reynolds number flows without training data remain sparse [13].…”

Section: Discussionmentioning

confidence: 99%

Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data

Harmening,

Peitzmann,

el Moctar

2023

Preprint

View full text Add to dashboard Cite

Unsupervised physics-informed deep learning can be used to solve computational physics problems by training neural networks to satisfy the underlying equations and boundary conditions without labeled data. Parameters such as network architecture and training method determine the training success. However, the best choice is unknown a priori as it is case specific. Here, we investigated network shapes, sizes, and types for unsupervised physics-informed deep learning of the two-dimensional Reynolds-averaged flow around cylinders. We trained mixed-variable networks and compared them to traditional models. Several network architectures with different shape factors and sizes were evaluated. The models were trained to solve the Reynolds-averaged Navier-Stokes equations incorporating Prandtl's mixing length turbulence model. No training data were deployed to train the models. The superiority of the mixed-variable approach was confirmed for the investigated high Reynolds number flow. The mixed-variable models were sensitive to the network shape. For the two cylinders, differently deep networks showed superior performance. The best fitting models were able to capture important flow phenomena such as stagnation regions, boundary layers, flow separation, and recirculation. We also encountered difficulties when predicting high Reynolds number flows without training data

show abstract

Section: Discussionmentioning

confidence: 99%

Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data

Harmening,

Peitzmann,

el Moctar

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…The Reynolds mean (RANS) is the use of space-time averaging to deal with turbulence. It can reduce the calculation time cost while obtaining better results, and for most engineering projects, the details of turbulence fluctuations are often negligible . RANS is more widely used in the range of indirect numerical simulation methods.…”

Section: Experiments and Methodsmentioning

confidence: 99%

CFD Simulation Investigation on the Flow Field Distribution Pattern under Different Operating Environment Pressures in a Jet Impingement-Negative Pressure Reactor

Dong,

Qiu,

Chai

2023

ACS Omega

View full text Add to dashboard Cite

A jet impingement-negative pressure deamination reactor (JI-NPDR) has excellent ammonia removal efficiency. The CFD numerical simulation method based on the Euler−Euler model and the realizable k−ε turbulence model was used to investigate the effect of different negative pressures at the reactor top outlet on the distribution pattern of pressure, velocity, turbulent kinetic energy, and vortex. The results indicate that the overall water flow distribution in the reactor increases in axial cohesion with the increase in negative pressure. The scattered small eddies gradually connect to a large eddy current as a whole, and the small eddies generated in the jet area also become regular and flat with the increase of negative pressures. These findings can provide detailed information for the study of flow patterns in a jet impingement-negative pressure reactor.

show abstract

An improved physics-informed neural network with adaptive weighting and mixed differentiation for solving the incompressible Navier–Stokes equations

Wang,

Xiao,

Feng

et al. 2024

Nonlinear Dyn

View full text Add to dashboard Cite

Physics-informed neural networks for the Reynolds-Averaged Navier–Stokes modeling of Rayleigh–Taylor turbulent mixing

Cited by 9 publications

References 38 publications

Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data

Effect of Network Architecture on Physics-Informed Deep Learning of the Reynolds-Averaged Turbulent Flow Field around Cylinders without Training Data

CFD Simulation Investigation on the Flow Field Distribution Pattern under Different Operating Environment Pressures in a Jet Impingement-Negative Pressure Reactor

An improved physics-informed neural network with adaptive weighting and mixed differentiation for solving the incompressible Navier–Stokes equations

Contact Info

Product

Resources

About