Bulletin of the American Physical Society
77th Annual Meeting of the Division of Fluid Dynamics
Sunday–Tuesday, November 24–26, 2024; Salt Lake City, Utah
Session ZC15: Low-Order Modeling and Machine Learning in Fluid Dynamics: Turbulence Modeling II |
Hide Abstracts |
Chair: Cristian Ricardo Constante Amores, University of Illinois, Urbana Champaign Room: 155 E |
Tuesday, November 26, 2024 12:50PM - 1:03PM |
ZC15.00001: Tailwind turbulence: a possible bound on energy available from turbulence for transit Gregory P Bewley, Scott Bollt We investigate the unconstrained minimum energy performance of vehicles moving through turbulence that interact with their environment through thrust, weight and drag forces, such as rotorcraft or submersibles. Using a turbulence model, we find that for such vehicles there exists a speed that minimizes energy requirements, which depends on the turbulence intensity level. For intermediate turbulence intensities the energy required for transit can be substantially smaller than what is required in quiescent fluid. We describe a simple, analytically tractable, picture for how a flight trajectory could preferentially put vehicles in tailwinds rather than headwinds, predicated on the organization of turbulence around vortices. Supported by computer optimizations of flight trajectories in a model of turbulence, we find a bound on energy reduction by turbulence, which contains no free parameters and which we expect is most useful for low to intermediate turbulence levels. Furthermore, we predict the existence of an optimum level of turbulence for which the energy use is minimized. Thus, favorable trajectories are available to maneuverable vehicles if they have sufficient knowledge of the flow and computational resources for path planning. This work strengthens previous findings that turbulence can always reduce energy use, even for power-limited vehicles. |
Tuesday, November 26, 2024 1:03PM - 1:16PM |
ZC15.00002: Generative AI for Synthesizing Spatio-temporal Wall Pressure Fluctuations in Turbulent Boundary Layers Xiantao Fan, Meet Hemant Parikh, Yi Liu, Xinyang Liu, Meng Wang, Jian-Xun Wang Wall pressure fluctuations in turbulent boundary layers are key sources of flow-induced noise, vibration, and hydroacoustic effects in underwater vehicles. Accurately predicting frequency/wavenumber spectra of these fluctuations is critical. Existing models based on Lighthill's wave equation or Kraichnan's Poisson equation often depend on empirical data and lack spatiotemporal features. Simulating these fluctuations requires high-resolution, eddy-resolving simulations such as DNS or WMLES, which is costly and impractical for high Reynolds numbers. We propose a generative learning framework using a probabilistic latent diffusion model and conditional neural field to synthesize spatiotemporal wall pressure fluctuations across different adverse pressure gradients in turbulent boundary layers. This framework generates extended trajectories of instantaneous pressure fluctuation fields in an autoregressive manner given sparse sensor measurements and far-wall velocity. By comparing root-mean-square values, frequency spectra, and wavenumber-frequency spectra of generated spatiotemporal instantaneous pressure fluctuations with DNS reference, we demonstrate the merit and effectiveness of our proposed method. |
Tuesday, November 26, 2024 1:16PM - 1:29PM |
ZC15.00003: A DNS database of wall-pressure fluctuations beneath turbulent boundary layers with pressure gradients Yi Liu, Xiantao Fan, Meng Wang, Jian-Xun Wang Spatiotemporal characteristics of wall-pressure fluctuations induced by turbulent boundary layers with varying pressure gradients are examined using incompressible direct numerical simulations. A series of pressure gradient distributions are generated by applying suction and blowing of different magnitudes at the upper computational boundary, resulting in three separated and three attached flat-plate boundary layers. The root-mean-square values of wall-pressure fluctuations for the separated boundary layers exhibit two peaks: one before the separation in the adverse pressure gradient (APG) region and the other after reattachment in the favorable pressure gradient (FPG) region, consistent with findings from previous numerical studies. In contrast, for the attached boundary layers, only a single peak exists in the APG region, while the fluctuation levels in the FPG region are lower than those in a zero-pressure-gradient boundary layer. Additionally, correlation and spectral analyses are conducted to reveal the substantial impact of pressure gradients on the structure of wall-pressure fluctuations. The database is used to develop a machine learning-based surrogate model for efficient prediction. |
Tuesday, November 26, 2024 1:29PM - 1:42PM |
ZC15.00004: Generative Latent Diffusion Model for Stochastic Inflow Turbulence Synthesis Xinyang Liu, Meet H Parikh, Pan Du, Xiantao Fan, Jian-Xun Wang Accurate inflow conditions are essential for eddy-resolving simulations. Traditional approaches like the recycling method are effective but demand extensive computational resources due to their high-fidelity nature. Synthetic approaches, though computationally efficient, often fail to capture the intricacies of real turbulence, requiring larger computational domains for accurate physics representation. Recently, deep learning methods have emerged as competitive alternatives. However, deterministic autoregressive models like Long Short-Term Memory (LSTM) and Transformers struggle with long-term predictions and the inherent stochasticity of turbulence. Generative models like Generative Adversarial Networks (GAN) are promising but are difficult to train and lack the ability to generalize across diverse flow conditions and meshes. To address these limitations, we propose a novel generative learning approach that leverages the strengths of both conditional neural fields (CNF) and latent diffusion models (LDM). By encoding spatiotemporal features into a hidden space using CNF and generating new samples with LDM, we develop a robust, mesh-independent inlet turbulence generator that generalizes over a wide range of Reynolds numbers. This method offers an efficient and accurate solution for synthesizing inflow turbulence in eddy-resolving simulations. |
Tuesday, November 26, 2024 1:42PM - 1:55PM |
ZC15.00005: Neural ODEs for RANS Verification Mustafa Aljabery, Cesar A Leos, Arvind T Mohan, Daniel M. Israel Calibration of model coefficients is critical to ensuring the accuracy RANS model simulations of turbulent flow. Typically, these models are calibrated using data that is taken in a state which is presumed to be a self-similar state. However, the available data is typically not from a self-similar regime, as we have previously shown using reduced-order models. These models capture the essential behavior of the RANS model as a dynamical system. Recent developments in Neural Ordinary Differential Equations (Neural ODEs) allow for the dynamical system to be rewritten and parameterized by neural networks to represent model coefficients. The model coefficients can be learned better to calibrate these quantities against experimental or high-fidelity simulation data. This approach allows the calibration to consider the entire trajectory of the data, not just the self-similar fixed point. In addition to coefficient calibration, the method can also be used for model validation, by comparing the trajectories of the experimental data and the model over a range of flow regimes. |
Tuesday, November 26, 2024 1:55PM - 2:08PM |
ZC15.00006: An Exact Formalism for Passive Scalar Mixing in Turbulent Flows Andre N Souza, Glenn R Flierl The parameterization of fluxes in turbulent flows, particularly for systems such as the atmosphere and ocean, remains a complex challenge due to the vast range of dynamical scales involved. As a first step, we explore new methodologies for representing the unresolved dynamics in passive scalar mixing. We begin by introducing a formalism that expresses the mean flux as a functional of mean gradients, highlighting its non-local dependence in both space and time. Next, we analyze a class of stochastic advection problems, deriving analytic (and exact) expressions for turbulent diffusivity as a function of flow statistics, thereby relating mixing to Koopman modes. We then apply the formalism to two-dimensional turbulent simulations and provide insights into when local eddy diffusivity is a sufficient approximation and when non-local effects must be considered. |
Tuesday, November 26, 2024 2:08PM - 2:21PM |
ZC15.00007: Field-inversion machine learning for unsteady aerodynamics Karim S Ahmed, Sudeep Menon, Anupam Sharma, Paul Allen Durbin Field inversion machine learning (FIML) has been successfully used to improve turbulence and transition models for steady problems. In FIML, a scalar field, β(x), which augments turbulence |
Tuesday, November 26, 2024 2:21PM - 2:34PM |
ZC15.00008: An immersed boundary (IB) resolvent framework for wall-bounded flows with spatial disturbances Zoey Flynn, Jane Bae, Andres Goza Turbulent wall-bounded flows often prove challenging to model, due to the complex flow structures that arise along the surface. This is especially true for fluid problems involving non-smooth surfaces, such as riblets, or compliant materials which further complicate modeling as a result of fluid-structure interactions. Resolvent analysis is useful for modeling turbulent wall-bounded flows, due to its ability as an operator-based method to generate response-forcing mode pairs. However, in its typical form resolvent analysis is limited to simple wall boundary conditions. Therefore, we have developed an immersed boundary (IB)-based resolvent analysis framework, capable of handling various surface conditions. The proposed method is fully derived in a resolvent setting and retains the property of streamwise-spanwise-homogeneity obeyed by wall-bounded flows, keeping the overall size of the resolvent system small. At the same time, the IB treatment enables significant generality in the boundary conditions applied at the wall surface. We present this IB resolvent method, including its application on a meaningful set of turbulent wall-bounded flow problems. |
Tuesday, November 26, 2024 2:34PM - 2:47PM |
ZC15.00009: Labelling forced two-dimensional turbulence with vortex crystals Andrew Cleary, Jacob Page Recent methods for finding unstable periodic orbits (UPOs) in turbulence have dramatically expanded the number of known solutions in two-dimensional, turbulent Kolmogorov flow (Page et al, Proc. Nat. Acad. Sci., 121 (23), 2024). Despite this, it is still unknown which of these remain dynamically relevant as Re → ∞, while the self-sustaining processes encapsulated in this set of solutions have not been identified. In this talk, we perform an arclength continuation of a library of O(150) UPOs upwards from Re = 100 to beyond Re ≈ 1000, with several apparently connecting to solutions of the Euler equation. Motivated by this connection, we compute exact coherent solutions for a system of point vortices that mimic the Navier-Stokes UPOs. This is achieved via gradient-based optimisation of a scalar loss function which seeks to both (1) match the positions of the point vortices to the turbulent vortex cores and (2) insist that the point vortex evolution is itself time-periodic. We categorise the Kolmogorov UPOs into classes based on the corresponding point vortex solution. We also modify our approach to find stationary vortex crystals which can describe the large-scale vortices in decaying two-dimensional turbulence. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2025 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700