Bulletin of the American Physical Society
76th Annual Meeting of the Division of Fluid Dynamics
Sunday–Tuesday, November 19–21, 2023; Washington, DC
Session J43: Turbulence: Modeling II |
Hide Abstracts |
Chair: Dhawal Buaria, New York University Room: 207B |
Sunday, November 19, 2023 4:35PM - 4:48PM |
J43.00001: Multilevel Lagrangian model for passive scalar gradients in turbulence at high Reynolds and Schmidt numbers Soumak Bhattacharjee, Andrew D Bragg Zhang et al. (J. Fluid Mech., 964, A39, 2023) formulated a model for the Lagrangian evolution of passive scalar gradients in isotropic turbulence in which the diffusion term in the equation was closed using the recent deformation of Gaussian fields (RDGF) approach. The velocity gradient appears in the scalar gradient equation, and this was specified using the model of Johnson & Meneveau (Phys. Rev. Fluids, vol. 2 (7), 2017, 072601), which is based on a multi-level recent deformation of Gaussian fields (ML-RDGF) closure, enabling predictions at arbitrary Taylor Reynolds numbers Re_{λ}. The predictions from the model of Zhang et al. are in very good agreement with DNS data over the range Re_{λ} ≤ 250. However, the model blows up when Re_{λ} ≥ 500, and at lower Re_{λ} when the Schmidt number Sc is larger than one. To address these issues, we extend the Zhang et al. model to utilize the full ML-RDGF closure approach for the scalar gradient diffusion term, but with a recent deformation timescale based on the scalar gradient dynamics. This closure also enables the scalar gradient model to make robust predictions in the regime Sc>1, where there is a difference in the scale at which the largest velocity and scalar gradients exist. |
Sunday, November 19, 2023 4:48PM - 5:01PM |
J43.00002: Predicting scalar gradient dynamics in turbulent mixing using deep neural networks Dhawal Buaria, Katepalli R Sreenivasan A defining characteristic of turbulence in fluid flows is that it dramatically enhances the mixing and transport rates of scalars, such as heat or substance concentration. While turbulence stirs the scalars across a wide range of scales, the mixing efficiency is ultimately controlled by scalar gradients at the smallest diffusive scales, where the scalar fluctuations are dissipated. Consequently, understanding the dynamics of scalar gradients is important for both improving our fundamental understanding and in various modeling endeavours. However, measuring scalar gradients in experiments is extremely challenging, especially when scalar diffusivities are very low, or the Schmidt number, the ratio of kinematic viscosity to scalar diffusivity, is high. In contrast, direct numerical simulations (DNS) can provide any quantity by very design. However, due to prohibitive cost of resolving the smallest scales, they are restricted to low Reynolds numbers, particularly for mixing at high Schmidt numbers. In this work, we propose an alternative approach, whereby the power of deep learning is utilized to learn scalar gradient dynamics from available data at lower Reynolds and Schmidt numbers, and predict unseen dynamics at higher Reynolds and Schmidt numbers. To this end, we consider the evolution equation for scalar gradients, and model the unclosed diffusive Laplacian term as a function of velocity and scalar gradients using a physics informed vector-based neural network (VBNN), which imbues various physical constraints and symmetries. Training is performed using a massive DNS database. For validation, the trained model is run at both seen and higher unseen Reynolds and Schmidt numbers, demonstrating excellent prediction of various statistical properties, such as probability distributions of scalar gradients and joint structure of velocity and scalar gradients. |
Sunday, November 19, 2023 5:01PM - 5:14PM |
J43.00003: Assessment of RANS models for Rayleigh-Taylor mixing using the Macroscopic Forcing Method Dana Lynn Lavacot, Jessie Liu, Brandon E Morgan, Ali Mani The Macroscopic Forcing Method (MFM) is a numerical tool for assessing Reynolds-averaged Navier-Stokes (RANS) closure operators (Mani & Park, 2021). It specifically can be used to assess nonlocality by measuring moments of the eddy diffusivity kernel of a turbulent flow. Previous work showed that nonlocality should not be neglected in RANS models for Rayleigh-Taylor (RT) mixing using MFM measurements for mean scalar transport in both 2D RT (Lavacot et al, in review) and 3D RT (Lavacot et al, 73^{rd} APS DFD Meeting, 2022). In this talk, the k-L-F model is presented, which incorporates nonlocality of the eddy diffusivity through the addition of an equation for the turbulent scalar flux to the k-L model (Dimonte & Tipton, 2006) based on MFM measurements. By applying MFM to the RANS model itself, this model is evaluated along with the standard k-L model and the BHR-4 model (Braun & Gore, 2021), both of which are used for predicting RT mixing. Assessments are made by comparing the MFM-measured eddy diffusivity moments from the RANS simulations against those from high-fidelity simulations. |
Sunday, November 19, 2023 5:14PM - 5:27PM |
J43.00004: A Robust Turbulent Combustion Closure model via Deep Operator Network Arsalan Taassob, Anuj Kumar, Tarek Echekki, Rishikesh Ranade In this research, we introduced a novel technique called Deep Operator Network (DeepONet) to assess closure terms for turbulence and chemical source terms in the Sydney turbulent non-premixed flames. To achieve this, we utilized temperature, major species, and velocity point measurements to develop closures for momentum and thermo-chemical transport. The DeepONet method was trained on three different flame conditions: Sydney Flames 57, 59, and 80, and its performance was validated on an additional flame, Flame 103. |
Sunday, November 19, 2023 5:27PM - 5:40PM |
J43.00005: Compressed representations and reduced order modeling of the turbulent flow response to roughness Miles J Chan, Ugo Piomelli, Beverley J McKeon All surfaces are hydrodynamically rough at sufficiently large Reynolds numbers. While the drag penalty of different engineering-relevant surfaces (sandgrain, biofouled, painted, etc.) can be characterized by various statistical parameters of the surfaces, finding a universal relationship between the flow response and arbitrary surfaces remains an open area of research. |
Sunday, November 19, 2023 5:40PM - 5:53PM |
J43.00006: Physics-Aware Spatio-Temporal Dynamics and Test-Time Refinement for Turbulent Flow Reconstruction. Shengyu Chen, Peyman Givi, Can Zheng, Xiaowei Jia A new physics-guided neural network is developed for reconstructing the full-scale DNS field from low-resolution LES data in turbulent flows. The method utilizes the transport equations that underlie the flow dynamics to design the spatio-temporal model architecture. A degradation-based refinement method is also developed to enforce physical constraints, and to reduce accumulated reconstruction errors over long periods. The model is shown to reconstruct the long-term continuous spatial and temporal dynamics of the flows in an accurate manner. Data from two incompressible turbulent flow configurations are used to evaluate the performance of the model and to compare it with previous super-resolution models. Detailed qualitative and quantitative comparative assessments demonstrate the effectiveness of each of the components of the new model. |
Sunday, November 19, 2023 5:53PM - 6:06PM |
J43.00007: Towards a physical interpretation of machine-learned turbulence models Jiaqi Li, Yuanwei Bin, George P Huang, Xiang Yang This study aims at obtaining a physical understanding of the existing machine-learning turbulence models. We apply three established methods, i.e., tensor-basis neural networks (TBNN), physics-informed machine learning (PIML), and field inversion & machine learning (FIML), to the one-equation Spalart-Allmaras model, the two-equation Wilcox k-omega model, and the seven-equation full Reynolds stress model. The machine learning corrections are trained against plane channel flow and temporally-evolving mixing layer flow. The goal is to assess if the ML methods can preserve the law of the wall. Our results show that FIML preserves the law of the wall for the one- and two-equation models and improve the predictions of the seven-equation model in the context of channel flow---although the improvement offered by FIML is not entirely physical. TBNN and PIML, on the other hand, do not preserve the law of the wall, which proves to be a consequence of the choice of inputs. |
Sunday, November 19, 2023 6:06PM - 6:19PM |
J43.00008: Wavelet-based predictions of bursting events in 2D Kolmogorov flow Anagha Madhusudanan, Rich R Kerswell Prediction of intermittent high-energy events (bursting events) is complicated by the fact that the characterization of such events using conventional Fourier-based methods involves many different frequencies. In this study, we therefore explore the use of wavelet-based techniques to predict such intermittent events. Two wavelet-based methods are compared: (1) a purely data-driven method using a wavelet-based Proper Orthogonal Decomposition (WPOD) and (2) a method that, along with data, uses the Navier-Stokes equations in the form of a wavelet-based resolvent analysis. The flow considered is 2D Kolmogorov flow, i.e. the Navier-Stokes equations forced by a sinusoidal body forcing over a 2-torus, which exhibits intermittent bursts of energy that are localised in time. Firstly, we find that the WPOD method is able to predict an oncoming bursting event. Secondly, the use of a wavelet-based resolvent analysis can give an improvement in prediction times albeit with an increase in the number of false-positives. |
Sunday, November 19, 2023 6:19PM - 6:32PM |
J43.00009: Symmetries Based Invariant Modeling for Second Moment Turbulence Modeling Felician C Putz, Nils Benedikt, Dario S Klingenberg, Martin Oberlack Attempts to develop semi-empirical model turbulence date back to the late 19th century; however, their limitations often stem from missing symmetries, which are axiomatic invariance properties in physics. The key basis for turbulence description, die Navier-Stokes equations, admit the Galilean group extended by scaling. Over the decades modelers have implicitly included increasingly more symmetries into turbulence model. Classical two-equation models eventually included all of them, though have too many and unphysical symmetries, being invariant under constant rotations. Recently, new symmetries of the infinite set of multi-point moment equations (MPME) have been discovered, dubbed statistical symmetries, as they only occur in the statistical descriptions of turbulence such as the MPME and have been linked with intermittency and non-Gaussianity. In PRL 2022 (Oberlack et al.) it was shown that these symmetries are essential for high-moment scaling laws. All turbulence models covered so far are not invariant under these statistical symmetries. We show very how a second moment turbulence model can be developed from all symmetries known so far, and further we present a recently developed model. |
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. |
© 2024 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