6,587 Hits in 4.2 sec

Deep Fluids: A Generative Network for Parameterized Fluid Simulations [article]

Byungsoo Kim, Vinicius C. Azevedo, Nils Thuerey, Theodore Kim, Markus Gross, Barbara Solenthaler
2019 arXiv   pre-print
A convolutional neural network is trained on a collection of discrete, parameterizable fluid simulation velocity fields.  ...  This paper presents a novel generative model to synthesize fluid simulations from a set of reduced parameters.  ...  Our Deep Fluids CNN is designed to generate velocity fields for parameterizable scenes.  ... 
arXiv:1806.02071v2 fatcat:gy64u5helrbjrlcwjyk3r44c3e

Deep learning observables in computational fluid dynamics [article]

Kjetil O. Lye, Siddhartha Mishra, Deep Ray
2019 arXiv   pre-print
Numerical experiments are presented to demonstrate low prediction errors for the trained network networks, even when the network has been trained with a few samples, at a computational cost which is several  ...  Under the assumption that the underlying neural networks generalize well, we prove that the deep learning MC and QMC algorithms are guaranteed to be faster than the baseline (quasi-) Monte Carlo methods  ...  A large proportion of computations for this paper were performed on the ETH compute cluster EULER.  ... 
arXiv:1903.03040v2 fatcat:rhhgoaxisffmvlwcxn7wdo5uny

An advanced hybrid deep adversarial autoencoder for parameterized nonlinear fluid flow modelling [article]

M. Cheng, F.Fang, C.C. Pain, I.M. Navon
2020 arXiv   pre-print
In this work, we propose a hybrid deep adversarial autoencoder (DAA) to integrate generative adversarial network (GAN) and variational autoencoder (VAE) for predicting parameterized nonlinear fluid flows  ...  In addition, the low-representation representations are applied into the adversarial network for model training and parameter optimization, which enables a fast computation process.  ...  Hybrid deep adversarial autoencoder for nonlinear fluid flow modelling In this paper, for nonlinear fluid flow modeling, a hybrid deep learning fluid model based on deep adversarial autoencoder (DAA) is  ... 
arXiv:2003.10547v1 fatcat:hbn22jd3gzdo7pngzjyozoayqu

Data recovery in computational fluid dynamics through deep image priors [article]

Marc T. Henry de Frahan, Ray W. Grout
2019 arXiv   pre-print
Focusing specifically on computational fluid dynamics simulations, this work proposes a method that uses a deep convolutional neural network to recover simulation data.  ...  Data recovery is performed for two canonical fluid flows: laminar flow around a cylinder and homogeneous isotropic turbulence.  ...  Experiments showed that using a fixed smoothly varying input z for the neural network imposes a smoothness prior, which is beneficial for data recovery for fluid flows.  ... 
arXiv:1901.11113v2 fatcat:wst72nf6m5bzdi32jseriili3a

A review on Deep Reinforcement Learning for Fluid Mechanics [article]

Paul Garnier and Jonathan Viquerat and Jean Rabault and Aurélien Larcher and Alexander Kuhnle and Elie Hachem
2021 arXiv   pre-print
Deep reinforcement learning (DRL) has recently been adopted in a wide range of physics and engineering domains for its ability to solve decision-making problems that were previously out of reach due to  ...  In this work, we conduct a detailed review of existing DRL applications to fluid mechanics problems.  ...  Deep Q-networks Instead of updating a table holding Q-values for each possible (s, a) pair, ANNs can be used to generate a map S + × A −→ R (called deep Q-networks (DQN)) that provides an estimate of the  ... 
arXiv:1908.04127v2 fatcat:5w72evcehrcvhmqv25zziwlfvq

Real-Time Simulation of Parameter-Dependent Fluid Flows through Deep Learning-Based Reduced Order Models

Stefania Fresca, Andrea Manzoni
2021 Fluids  
To do so, they rely on deep neural networks, after performing a former dimensionality reduction through POD, enhancing their training times substantially.  ...  However, they might require expensive hyper-reduction strategies for handling parameterized nonlinear terms, and enriched reduced spaces (or Petrov–Galerkin projections) if a mixed velocity–pressure formulation  ...  and on the use of deep learning (DL)-based ROMs for the sake of real-time simulation of fluid flows, thus relying on nonlinear reduction techniques.  ... 
doi:10.3390/fluids6070259 fatcat:pnfylgazmrfgdoht6rwikbg5h4

A review on deep reinforcement learning for fluid mechanics: an update [article]

Jonathan Viquerat, Philippe Meliga, Elie Hachem
2021 arXiv   pre-print
In the past couple of years, the interest of the fluid mechanics community for deep reinforcement learning (DRL) techniques has increased at fast pace, leading to a growing bibliography on the topic.  ...  Based on these comparisons, general conclusions are drawn regarding the current state-of-the-art in the domain, and perspectives for future improvements are sketched.  ...  Deep deterministic policy gradient (DDPG) Deep deterministic policy gradient (DDPG) can be thought as a DQN algorithm for continuous actions spaces, that combines the learning of a Q-network Q θ (s, a)  ... 
arXiv:2107.12206v2 fatcat:i3y6os4e6zh3vgslsotoo7tac4

Model identification of reduced order fluid dynamics systems using deep learning

Z. Wang, D. Xiao, F. Fang, R. Govindan, C. C. Pain, Y. Guo
2017 International Journal for Numerical Methods in Fluids  
Chen et al. proposed a NIROM based on black-box stencil interpolation method [38]. Xiao et al. also presented a parameterized NIROM for general time-dependent nonlinear PDEs [3].  ...  Audouze et al. proposed a NIROM for nonlinear parameterized time-dependent PDEs using RBF and POD [39, 40].  ...  FP7/20072013) under grant agreement NO. 603663 for the research project PEARL (Preparing for Extreme And Rare events in coastaL regions), the Newton funding: Smart technologies for optimal design, drilling  ... 
doi:10.1002/fld.4416 fatcat:yf2l5tbj2rbt3eld2dx65cmml4

Deep learning of mixing by two 'atoms' of stratified turbulence

Hesam Salehipour, W. R. Peltier
2019 Journal of Fluid Mechanics  
In this paper, we describe the application of deep learning methods to the discovery of a generic parameterization of diapycnal mixing using the available DNS dataset.  ...  Recently, an unprecedented volume of data has been generated through direct numerical simulation (DNS) of these flows.  ...  A convolutional neural network (CNN) is a special type of neural network architecture that relies on the convolution operator in lieu of general matrix multiplication in at least one layer of its configuration  ... 
doi:10.1017/jfm.2018.980 fatcat:h2wimzplnzfspndffhh4vqnzee

A deep learning enabler for nonintrusive reduced order modeling of fluid flows

S. Pawar, S. M. Rahman, H. Vaddireddy, O. San, A. Rasheed, P. Vedula
2019 Physics of Fluids  
In this paper, we introduce a modular deep neural network (DNN) framework for data-driven reduced order modeling of dynamical systems relevant to fluid flows.  ...  Our a posteriori analysis shows that the proposed data-driven approach is remarkably accurate, and can be used as a robust predictive tool for non-intrusive model order reduction of complex fluid flows  ...  (Dated: 12 July 2019) In this paper, we introduce a modular deep neural network (DNN) framework for data-driven reduced order modeling of dynamical systems relevant to fluid flows.  ... 
doi:10.1063/1.5113494 fatcat:egbzpmlcoreuzlaxwqpmwu4ify

Deep convolutional recurrent autoencoders for learning low-dimensional feature dynamics of fluid systems [article]

Francisco J. Gonzalez, Maciej Balajewicz
2018 arXiv   pre-print
Our approach constructs a modular model consisting of a deep convolutional autoencoder and a modified LSTM network.  ...  In this work we propose a deep learning-based strategy for nonlinear model reduction that is inspired by projection-based model reduction where the idea is to identify some optimal low-dimensional representation  ...  Blue Waters is a joint effort of the University of Illinois at Urbana-Champaign and its National Center for Supercomputing Applications.  ... 
arXiv:1808.01346v2 fatcat:g46ttwgcdbfj5prd2l7f6lbzma

Learning Incompressible Fluid Dynamics from Scratch – Towards Fast, Differentiable Fluid Models that Generalize [article]

Nils Wandel, Michael Weinmann, Reinhard Klein
2021 arXiv   pre-print
Recent deep learning based approaches promise vast speed-ups but do not generalize to new fluid domains, require fluid simulation data for training, or rely on complex pipelines that outsource major parts  ...  In this work, we propose a novel physics-constrained training approach that generalizes to new fluid domains, requires no fluid simulation data, and allows convolutional neural networks to map a fluid  ...  Recent advances in deep learning aim for fast and accurate fluid simulations but rely on vast datasets and / or do not generalize to new fluid domains.  ... 
arXiv:2006.08762v3 fatcat:6vfzottjgzdodjzx6l74mjcd5i

Sub-grid scale model classification and blending through deep learning

Romit Maulik, Omer San, Jamey D. Jacob, Christopher Crick
2019 Journal of Fluid Mechanics  
In this article we detail the use of machine learning for spatio-temporally dynamic turbulence model classification and hybridization for large eddy simulations (LES) of turbulence.  ...  Our test case for the demonstration of this concept is given by Kraichnan turbulence, which exhibits a strong interplay of enstrophy and energy cascades in the wavenumber domain.  ...  We note that these assessments are all carried out for a Reynolds number value that is equal to that used for DNS data generation and subsequent network optimization and while the network is tasked with  ... 
doi:10.1017/jfm.2019.254 fatcat:6i25ui5gtjgilphyevz7rr7rv4

A Point-Cloud Deep Learning Framework for Prediction of Fluid Flow Fields on Irregular Geometries [article]

Ali Kashefi, Davis Rempe, Leonidas J. Guibas
2020 arXiv   pre-print
We present a novel deep learning framework for flow field predictions in irregular domains when the solution is a function of the geometry of either the domain or objects inside the domain.  ...  Grid vertices in a computational fluid dynamics (CFD) domain are viewed as point clouds and used as inputs to a neural network based on the PointNet architecture, which learns an end-to-end mapping between  ...  ACKNOWLEDGMENTS The authors would like to thank the Vannevar Bush Faculty Fellowship (VBFF) from the Department of Defense (DoD) and gifts from Amazon AWS and Google for providing the funding support for  ... 
arXiv:2010.09469v1 fatcat:nedgkez7hvcw3f7tmtug4zcmiq

Hidden Fluid Mechanics: A Navier-Stokes Informed Deep Learning Framework for Assimilating Flow Visualization Data [article]

Maziar Raissi, Alireza Yazdani, George Em Karniadakis
2018 arXiv   pre-print
We present hidden fluid mechanics (HFM), a physics informed deep learning framework capable of encoding an important class of physical laws governing fluid motions, namely the Navier-Stokes equations.  ...  Consequently, the predictions made by HFM are among those cases where a pure machine learning strategy or a mere scientific computing approach simply cannot reproduce.  ...  Generation of data was performed on XSEDE resources supported by award No. TG-DMS140007.  ... 
arXiv:1808.04327v1 fatcat:43lfrnzcprhijjnoqig5axi3w4
« Previous Showing results 1 — 15 out of 6,587 results