Computing interface curvature from volume fractions: A machine learning approach

Patel, H.V.; Panda, A.; Kuipers, J.A.M.; Peters, E.A.J.F.

doi:10.1016/j.compfluid.2019.104263

Cited by 38 publications

(35 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The following study investigates this neural network-based curvature prediction approach for an algebraic VOF method. Following the work of Qi et al [13] and Patel et al [10], who used geometric VOF methods, the approach is adapted and extended for the underlying computational fluid dynamics (CFD) framework to increase its accuracy and robustness. Stencils of 7×7 volume fraction values are used here.…”

Section: Deep Learning Framework For Curvature Predictionmentioning

confidence: 99%

“…The data generation process is similar to the one proposed by Patel et al [10]. However, it is extended to include a broader range of possible volume fraction stencil configurations and to account for not perfectly sharp interfaces, as described in detail in Section 3.1.…”

Section: Hidden Layersmentioning

confidence: 99%

“…The data set should cover a wide range of different interface configurations to represent as many situations of a 'real' CFD simulation as possible. The algorithm used in this study to generate such a data set is loosely based on the work of Patel et al [10] and is outlined in the following.…”

Section: Data Generationmentioning

confidence: 99%

“…The neural network has to be trained on a data set before being used within a flow computation. Patel et al [10] further developed and investigated this method and evaluated its performance within a flow solver.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

A Deep Learning Approach for Curvature Prediction in Algebraic Volume of Fluid Methods

Kraus¹,

Köhler²,

Friedrich³

et al. 2021

9th Edition of the International Conference on Computational Methods for Coupled Problems in Science and Engineering

View full text Add to dashboard Cite

The accurate prediction of the curvature of fluid-fluid interfaces is crucial for appropriately modeling the surface forces when computing two-phase flows with immiscible fluids. The volume of fluid (VOF) method is often used for these computations to specify the different fluids and the interface by the so-called volume fraction field. In this study, a deep artificial neural network is trained to predict the interface curvature from the volume fraction values. This approach is investigated within an algebraic VOF framework. A rudimentary interface resharpening algorithm is introduced for the input stencils to enhance the accuracy and robustness when the interface can not be captured entirely sharp. The performance of different neural network architectures is evaluated by generic test data and the computation of two oscillating droplet flow configurations.

show abstract

Section: Deep Learning Framework For Curvature Predictionmentioning

confidence: 99%

Section: Hidden Layersmentioning

confidence: 99%

Section: Data Generationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

A Deep Learning Approach for Curvature Prediction in Algebraic Volume of Fluid Methods

Kraus¹,

Köhler²,

Friedrich³

et al. 2021

9th Edition of the International Conference on Computational Methods for Coupled Problems in Science and Engineering

View full text Add to dashboard Cite

show abstract

“…Upon proper training, they are then plugged in the solver for evaluation of down-stream tasks. Examples include training of explicit subgrid scale models in large eddy simulations [2], interface reconstruction in multiphase flows [3,4], and cell-face reconstruction in shock-capturing schemes [5,6]. However, the successful development of powerful general-purpose automatic-differentiation frameworks, such as Tensorflow [7], Pytorch [8], and JAX [9] has enabled online training of ML models.…”

Section: Introductionmentioning

confidence: 99%

A fully-differentiable compressible high-order computational fluid dynamics solver

Bezgin¹,

Buhendwa²,

Adams³

2021

Preprint

View full text Add to dashboard Cite

Fluid flows are omnipresent in nature and engineering disciplines. The reliable computation of fluids has been a long-lasting challenge due to nonlinear interactions over multiple spatio-temporal scales. The compressible Navier-Stokes equations govern compressible flows and allow for complex phenomena like turbulence and shocks. Despite tremendous progress in hardware and software, capturing the smallest length-scales in fluid flows still introduces prohibitive computational cost for real-life applications. We are currently witnessing a paradigm shift towards machine learning supported design of numerical schemes as a means to tackle aforementioned problem. While prior work has explored differentiable algorithms for one-or two-dimensional incompressible fluid flows, we present a fully-differentiable three-dimensional framework for the computation of compressible fluid flows using high-order state-of-the-art numerical methods. Firstly, we demonstrate the efficiency of our solver by computing classical two-and three-dimensional test cases, including strong shocks and transition to turbulence. Secondly, and more importantly, our framework allows for end-to-end optimization to improve existing numerical schemes inside computational fluid dynamics algorithms. In particular, we are using neural networks to substitute a conventional numerical flux function. * D. A.B. and A.B.B. contributed equally to this work.

show abstract

Error-Correcting Neural Networks for Two-Dimensional Curvature Computation in the Level-set Method

Larios-Cárdenas

Gibou

2022

J Sci Comput

View full text Add to dashboard Cite

Computing interface curvature from volume fractions: A machine learning approach

Cited by 38 publications

References 26 publications

A Deep Learning Approach for Curvature Prediction in Algebraic Volume of Fluid Methods

A Deep Learning Approach for Curvature Prediction in Algebraic Volume of Fluid Methods

A fully-differentiable compressible high-order computational fluid dynamics solver

Error-Correcting Neural Networks for Two-Dimensional Curvature Computation in the Level-set Method

Contact Info

Product

Resources

About