SEARCH · Search NASA

Results for “LES”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 37 records · Page 2

All sources NASA NTRS DOE OSTI

DOE OSTI2024

Evaluation of RANS vs. LES simulation of fluid flow through 3 × 3 rod bundle with a simple spacer grid as a precursor to coupled fluid–structure interaction simulations

The research literature on Computational Fluid Dynamics (CFD) of coolant flow through rod bundles with spacer-grids and mixing vanes is replete, ranging from high fidelity Large Eddy Simulation (LES)/Direct Numerical Simulation (DNS) simulations to Reynolds-Averaged Navier–Stokes (RANS) modeled studies. The mixing of flow between subchannels and the pressure drop through the bundle are fundamental quantities useful for comparing and evaluating CFD methods. Less commonly observed and compared are the forces exerted onto the structure by the fluid. The present study seeks to evaluate the use of RANS simulations for predicting the structural response to fluid flow. Wall resolved RANS simulations are benchmarked against LES simulations of fluid flow at a Reynolds number of 15,000 through a 3 × 3 fuel rod bundle with a simple spacer grid. Velocity line-plots are compared showing good agreement between RANS and LES results, ascertaining that the former is capable of capturing the essential time-averaged velocity profile. Additionally, the distribution of forces on the spacer grid and fuel rods are collected as a function of time and space. The RANS methods are evaluated using the frequency and magnitude of the fluctuating forces on various portions of the structure as compared to LES. In conclusion, the power spectral density evaluation of the models reveal underprediction of force amplitude on the rod walls by RANS and also discrepancy in the prediction of high frequency spectra, especially in the immediate vicinity of spacer-grid structure, which may be attributed to the lack of random turbulence fluctuation or insufficient modeling of small-scale eddies in RANS simulation.

FIV↗

DOE OSTI2026

Comparison of URANS and LES predictions for the open phase of the OECD NEA CSNI fluid structure interaction CFD benchmark

The OECD NEA CSNI WGAMA CFD Task Group ran a benchmark in 2020 and 2021 to assess the predictive capabilities of coupled fluid structure interaction (FSI) CFD analysis methods. This paper presents the predictions made for the open phase of the benchmark using URANS and LES turbulence modelling approaches, and a comparison of the results to the experimental data. The benchmark comprised a channel containing two inline cylinders in cross-flow. The cylinders were fixed at one end, free at the other, and had measured resonant frequencies and damping properties. The URANS modelling used ANSYS Fluent 2-way coupled to ANSYS Mechanical. The LES modelling used Nek5000, 1-way coupled to Diablo. Comparisons with cross-channel velocity profiles are presented, both for the mean flow and its RMS. Comparisons are also made to the frequency spectra for point measurements of fluid velocity and pressure, and for the accelerations of the free end of each cylinder. URANS predicts the average velocity profiles relatively well, and is able to predict the velocity and acceleration spectra at the shedding frequency. However, the frequency content at the 4th harmonic of the shedding frequency is low in the URANS flow fields, and so does not excite accelerations at the resonant frequency of the cylinders. LES makes better predictions of the average profiles, and the velocity spectra agree well at both the shedding frequency and at higher frequencies. In conclusion, the 1-way coupled LES results show good agreement for acceleration spectra.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2022

LES of HCCI combustion of iso-octane/air in a flat-piston rapid compression machine

Homogeneous Charge Compression Ignition (HCCI) engines promise better efficiency and cleaner emissions than conventional piston engines, but can be challenging to control. Rapid compression machines (RCM) provide a simplified configuration for investigating HCCI combustion behavior, which is necessary for effective control of engine ignition timing and peak pressures. Here, in this study, we assess the utility of large eddy simulations (LES) for predicting HCCI combustion in a 3-D configuration. To this end, LES with finite-rate chemistry employing a 99-species iso-octane/air mechanism of two RCM operating conditions are performed. The RCM configuration under consideration was designed by Strozzi et al. with a flat piston to introduce large amounts of thermal stratification representative of realistic HCCI engine conditions, through the generation of corner vortices. It is shown that the simulation provides reasonable agreement with temperature fluctuations (7% difference), as well as ignition delay in the short ignition case (1 ms difference), while the long ignition case (35 ms difference) highlights more substantial deficiencies that are still within the expected uncertainty from the employed chemical mechanism. Flame propagation modes predicted by LES agree with experimental observations: spontaneous ignition is seen in the short ignition case, while deflagration is more prominent in the long ignition case. Analysis of global and local quantities classify the short ignition case in a mixed ignition regime, and the long ignition case in the mild ignition regime. These results demonstrate the utility of FRC-LES for investigations of multimode combustion regimes of HCCI combustion in a 3-D configuration.

33 ADVANCED PROPULSION SYSTEMS↗

DOE OSTI2025

Tropical Cyclone Super Resolution using conditional diffusion denoising probabilistic model from mesoscale simulation to LES

Accurate modeling of tropical cyclone wind fields is essential for the design, risk assessment, and operational planning of offshore energy infrastructure. While mesoscale simulations are widely used thanks to their computational efficiency, they lack the necessary resolution to capture key features such as wind shear and veer profiles as well as the distribution turbulent kinetic energy (TKE). High-fidelity large-eddy simulation (LES) models on the other hand, can resolve turbulent structures and provide a more accurate representation of the complex wind field, albeit at a higher computational cost. To address this modeling gap, we introduce a two-part generative framework to enhance the resolution and physics-capturing ability of mesoscale simulations. First, a reduced-order model based on Karhunen–Loève (KL) decomposition is used to extract dominant spatial modes from one-dimensional mean wind profiles. A multilayer perceptron (MLP) is trained to map mesoscale mode weights to their LES counterparts, enabling accurate reconstruction of vertical velocity profiles. Second, a conditional Diffusion Denoising Probabilistic Model (DDPM) is developed to super-resolve coarse and low-fidelity mesoscale velocity fields, recovering fine-scale turbulence structures and stress distributions. The framework is evaluated across different tropical cyclone intensity categories defined by the Saffir–Simpson scale and demonstrates strong performance in both interpolation and extrapolation tasks. The generated fields accurately reproduce spatial coherence, stress distributions, and spectral energy characteristics observed in LES data. By bridging the fidelity gap between mesoscale and LES outputs, this approach offers a scalable, data-driven solution for enhancing the representation of tropical cyclone wind fields, enabling more robust offshore energy infrastructure systems design in tropical-cyclone-prone areas.

17 WIND ENERGY↗

DOE OSTI2022

Effects of Horizontal Resolution, Domain Size, Boundary Conditions, and Surface Heterogeneity on Coarse LES of a Convective Boundary Layer

Atmospheric properties in a convective boundary layer vary over a wide range of spatial scales and are commonly studied using large-eddy simulations (LES) in various configurations. We examine how the boundary layer depth and distribution of variability across scales are affected by LES grid spacing, domain size, inhomogeneity of surface properties, and external forcing. Two different setups of the Weather Research and Forecasting (WRF) model are analyzed. A semi-idealized configuration uses a periodic domain, flat surface, prescribed homogeneous surface heat fluxes, and horizontally uniform profiles of large-scale advective tendencies. A nested LES setup employs a larger domain and realistic initial and boundary conditions, including an interactive land surface model with representative topography and vegetation and soil types. Subdomains of identical size are analyzed for all simulations. Characteristic structure sizes are quantified using the variability scales L 50 and L 95 , defined such that features smaller than that contain 50% and 95% of the total variance, respectively. Progressive increase in L 50 from vertical velocity to temperature and moisture structures is systematically reproduced in all simulation configurations. This dependence of L 50 on the considered variable complicates the development of scale-aware parameterizations for models with grid spacing in the “terra incognita”. In simulations using a larger domain with heterogeneous surface properties, the development of internal mesoscale patterns significantly affects variance distributions inside analyzed subdomains. Sizes of boundary layer structures also strongly depend on the LES grid spacing and, in case of heterogeneous surface and topography, on location of the subdomain inside a larger computational domain.

54 ENVIRONMENTAL SCIENCES↗

NASA NTRS1990

Heart rate and pulmonary function while wearing the launch-entry crew escape suit (LES) during + Gx acceleration and simulated Shuttle launch

Space shuttle crewmembers have been equipped with a launch-entry crew escape system (LES) since the Challenger accident in 1986. Some crewmembers, wearing the new pressure suit, have reported breathing difficulties and increased effort to achieve the desired range of motion. This study was conducted to quantify the reported increased physical workloads and breathing difficulty associated with wearing the LES. Both veteran astronauts and centrifuge panel members were exposed to various + Gx profiles (including simulated shuttle launch) + Gx on the USAF School of Aerospace Medicine (USAFSAM) human-use centrifuge. Maximum heart rate data showed no increased workload associated with arm and head movement in the LES when compared to the flight suit/helmet ensemble (LEH). However, the LES did impose a significant increase in breathing difficulty beginning at +2.5 Gx which was demonstrated by a decrease in forced vital capacity and subjected questionnaries.

Krutz, Robert W., Jr.↗

NASA NTRS2011

Validating LES for Jet Aeroacoustics

Engineers charged with making jet aircraft quieter have long dreamed of being able to see exactly how turbulent eddies produce sound and this dream is now coming true with the advent of large eddy simulation (LES). Two obvious challenges remain: validating the LES codes at the resolution required to see the fluid-acoustic coupling, and the interpretation of the massive datasets that result in having dreams come true. This paper primarily addresses the former, the use of advanced experimental techniques such as particle image velocimetry (PIV) and Raman and Rayleigh scattering, to validate the computer codes and procedures used to create LES solutions. It also addresses the latter problem in discussing what are relevant measures critical for aeroacoustics that should be used in validating LES codes. These new diagnostic techniques deliver measurements and flow statistics of increasing sophistication and capability, but what of their accuracy? And what are the measures to be used in validation? This paper argues that the issue of accuracy be addressed by cross-facility and cross-disciplinary examination of modern datasets along with increased reporting of internal quality checks in PIV analysis. Further, it is argued that the appropriate validation metrics for aeroacoustic applications are increasingly complicated statistics that have been shown in aeroacoustic theory to be critical to flow-generated sound.

Bridges, James↗

NASA NTRS2011

Validating LES for Jet Aeroacoustics

Engineers charged with making jet aircraft quieter have long dreamed of being able to see exactly how turbulent eddies produce sound and this dream is now coming true with the advent of large eddy simulation (LES). Two obvious challenges remain: validating the LES codes at the resolution required to see the fluid-acoustic coupling, and the interpretation of the massive datasets that are produced. This paper addresses the former, the use of advanced experimental techniques such as particle image velocimetry (PIV) and Raman and Rayleigh scattering, to validate the computer codes and procedures used to create LES solutions. This paper argues that the issue of accuracy of the experimental measurements be addressed by cross-facility and cross-disciplinary examination of modern datasets along with increased reporting of internal quality checks in PIV analysis. Further, it argues that the appropriate validation metrics for aeroacoustic applications are increasingly complicated statistics that have been shown in aeroacoustic theory to be critical to flow-generated sound, such as two-point space-time velocity correlations. A brief review of data sources available is presented along with examples illustrating cross-facility and internal quality checks required of the data before it should be accepted for validation of LES.

Bridges, James↗

NASA NTRS2013

Influence of Computational Drop Representation in LES of a Droplet-Laden Mixing Layer

Multiphase turbulent flows are encountered in many practical applications including turbine engines or natural phenomena involving particle dispersion. Numerical computations of multiphase turbulent flows are important because they provide a cheaper alternative to performing experiments during an engine design process or because they can provide predictions of pollutant dispersion, etc. Two-phase flows contain millions and sometimes billions of particles. For flows with volumetrically dilute particle loading, the most accurate method of numerically simulating the flow is based on direct numerical simulation (DNS) of the governing equations in which all scales of the flow including the small scales that are responsible for the overwhelming amount of dissipation are resolved. DNS, however, requires high computational cost and cannot be used in engineering design applications where iterations among several design conditions are necessary. Because of high computational cost, numerical simulations of such flows cannot track all these drops. The objective of this work is to quantify the influence of the number of computational drops and grid spacing on the accuracy of predicted flow statistics, and to possibly identify the minimum number, or, if not possible, the optimal number of computational drops that provide minimal error in flow prediction. For this purpose, several Large Eddy Simulation (LES) of a mixing layer with evaporating drops have been performed by using coarse, medium, and fine grid spacings and computational drops, rather than physical drops. To define computational drops, an integer NR is introduced that represents the ratio of the number of existing physical drops to the desired number of computational drops; for example, if NR=8, this means that a computational drop represents 8 physical drops in the flow field. The desired number of computational drops is determined by the available computational resources; the larger NR is, the less computationally intensive is the simulation. A set of first order and second order flow statistics, and of drop statistics are extracted from LES predictions and are compared to results obtained by filtering a DNS database. First order statistics such as Favre averaged stream-wise velocity, Favre averaged vapor mass fraction, and the drop stream-wise velocity, are predicted accurately independent of the number of computational drops and grid spacing. Second order flow statistics depend both on the number of computational drops and on grid spacing. The scalar variance and turbulent vapor flux are predicted accurately by the fine mesh LES only when NR is less than 32, and by the coarse mesh LES reasonably accurately for all NR values. This is attributed to the fact that when the grid spacing is coarsened, the number of drops in a computational cell must not be significantly lower than that in the DNS.

Bellan, Josette↗

NASA NTRS2015

Investigation of Unsteady Tip Clearance Flow in a Low-Speed One and Half Stage Axial Compressor with LES And PIV

The primary focus of this paper is to investigate how a rotor's unsteady tip clearance flow structure changes in a low speed one and half stage axial compressor when the rotor tip gap size is increased from 0.5 mm (0.49% of rotor tip blade chord, 2% of blade span) to 2.4 mm (2.34% chord, 4% span) at the design condition are investigated. The changes in unsteady tip clearance flow with the 0.62 % tip gap as the flow rate is reduced to near stall condition are also investigated. A Large Eddy Simulation (LES) is applied to calculate the unsteady flow field at these three flow conditions. Detailed Stereoscopic PIV (SPIV) measurements of the current flow fields were also performed at the Johns Hopkins University in a refractive index-matched test facility which renders the compressor blades and casing optically transparent. With this setup, the unsteady velocity field in the entire flow domain, including the flow inside the tip gap, can be measured. Unsteady tip clearance flow fields from LES are compared with the PIV measurements and both LES and PIV results are used to study changes in tip clearance flow structures. The current study shows that the tip clearance vortex is not a single structure as traditionally perceived. The tip clearance vortex is formed by multiple interlaced vorticities. Therefore, the tip clearance vortex is inherently unsteady. The multiple interlaced vortices never roll up to form a single structure. When phased-averaged, the tip clearance vortex appears as a single structure. When flow rate is reduced with the same tip gap, the tip clearance vortex rolls further upstream and the tip clearance vortex moves further radially inward and away from the suction side of the blade. When the tip gap size is increased at the design flow condition, the overall tip clearance vortex becomes stronger and it stays closer to the blade suction side and the vortex core extends all the way to the exit of the blade passage. Measured and calculated unsteady flow fields inside the tip gap agree fairly well. Instantaneous velocity vectors inside the tip gap from both the PIV and LES do show flow separation and reattachment at the entrance of tip gap as some earlier studies suggested. This area at the entrance of tip gap flow (the pressure side of the blade) is confined very close to the rotor tip section. With a small tip gap (0.5mm), the gap flow looks like a simple two-dimensional channel flow with larger velocity near the casing for both flow rates. A small area with a sharp velocity gradient is observed just above the rotor tip. This strong shear layer is turned radially inward when it collides with the incoming flow and forms the core structure of the tip clearance vortex. When tip gap size is increased to 2.4 mm at the design operation, the radial profile of the tip gap flow changes drastically. With the large tip gap, the gap flow looks like a two-dimensional channel flow only near the casing. Near the rotor top section, a bigger region with very large shear and reversed flow is observed.

compressor↗

NASA NTRS2015

Investigation of Unsteady Tip Clearance Flow in a Low-Speed One and Half Stage Axial Compressor with LES And PIV

unsteady↗

DOE OSTI2024

Three-step LES-C models for flows at high Reynolds numbers

Here, we investigate the need for the second correction step in the recently proposed LES-C models for fluid flows at high Reynolds numbers. These models use a predictor-corrector idea to enhance the efficiency of the existing Large Eddy Simulation models. Different three-step(one defect step, two corrections) LES-C models, based on the Leray-α, ADM and NS-ω LES models, are tested in three different situations. The new Leray-α-C2 model (C2 stands for two correction steps) is applied to the Navier–Stokes equations; the ADC2 is applied to the MagnetoHydroDynamic flow; and the NS-ω-C2 is used in the fluid-fluid interaction problem. We evaluate the effectiveness of the second correction step in all these settings, using qualitative and quantitative numerical tests.

97 MATHEMATICS AND COMPUTING↗

DOE OSTI2024

LES and RANS Spray Combustion Analysis of OME 3-5 and n-Dodecane

Clean-burning oxygenated and synthetic fuels derived from renewable power, so-called e-fuels, are a promising pathway to decarbonize compression–ignition engines. Polyoxymethylene dimethyl ethers (PODEs or OMEs) are one candidate of such fuels with good prospects. Their lack of carbon-to-carbon bonds and high concentration of chemically bound oxygen effectively negate the emergence of polycyclic aromatic hydrocarbons (PAHs) and even their precursors like acetylene (C 2 H 2 ), enabling soot-free combustion without the soot-NO x trade-off common for diesel engines. The differences in the spray combustion process for OMEs and diesel-like reference fuels like n-dodecane and their potential implications on engine applications include discrepancies in the observed ignition delay, the stabilized flame lift-off location, and significant deviations in high-temperature flame morphology. For CFD simulations, the accurate modeling and prediction of these differences between OMEs and n-dodecane proved challenging. This study investigates the spray combustion process of an OME 3 − 5 mixture and n-dodecane with advanced optical diagnostics, Reynolds-Averaged Navier–Stokes (RANS), and Large-Eddy Simulations (LESs) within a constant-volume vessel. Cool-flame and high-temperature combustion were measured simultaneously via high-speed (50 kHz) imaging with formaldehyde (CH 2 O) planar laser-induced fluorescence (PLIF) representing the former and line-of-sight OH* chemiluminescence the latter. Both RANS and LES simulations accurately describe the cool-flame development process with the formation of CH 2 O. However, CH 2 O consumption and the onset of high-temperature reactions, signaled by the rise of OH* levels, show significant deviations between RANS, LES, and experiments as well as between n-dodecane and OME. A focus is set on the quality of the simulated results compared to the experimentally observed spatial distribution of OH*, especially in OME fuel-rich regions. The influence of the turbulence modeling is investigated for the two distinct ambient temperatures of 900 K and 1200 K within the Engine Combustion Network Spray A setup. The capabilities and limitations of the RANS simulations are demonstrated with the initial cool-flame propagation and periodic oscillations of CH 2 O formation/consumption during the quasi-steady combustion period captured by the LES.

CFD↗

NASA NTRS2017

Large-Eddy Simulation (LES) of a Compressible Mixing Layer and the Significance of Inflow Turbulence

In the context of Large Eddy Simulations (LES), the effects of inflow turbulence are investigated through the Synthetic Eddy Method (SEM). The growth rate of a turbulent compressible mixing layer corresponding to operating conditions of GeobelDutton Case 2 is investigated herein. The effects of spanwise width on the growth rate of the mixing layer is investigated such that spanwise width independence is reached. The error in neglecting inflow turbulence effects is quantified by comparing two methodologies: (1) Hybrid-RANS-LES methodology and (2) SEM-LES methodology. Best practices learned from Case 2 are developed herein and then applied to a higher convective mach number corresponding to Case 4 experiments of GeobelDutton.

les↗

NASA NTRSDate not supplied

LES-Based Analysis of Film Cooling in a High-Pressure Turbine Rotor Including Effects of Purge and Tip Flow

Accurate modeling of coolant airflows, which form protective films over turbine blades, is essential for designing fuel-efficient and environmentally sustainable gas turbine engines. Excessive coolant reduces thermal efficiency, while insufficient coolant leads to blade overheating and causes damage. Therefore, precise prediction of flow field interactions with cooling air is critical for optimizing turbine performance. This study numerically investigates the cooling effectiveness of purge and film cooling flows within a high-pressure turbine (HPT) rotor using Large Eddy Simulation (LES). The study utilizes NASA Glenn Research Center’s Glenn-HT solver. The simulation models ethe conditions of the Penn State University START rotating rig. A high-fidelity structured mesh comprising up to 800 million cells is employed to resolve high-Reynolds number flow (Re ≈ 350,000) and to capture intricate secondary flow structures, including tip leakage and purge-induced vortices. Film cooling effectiveness computations are highly sensitive to boundary conditions at the cooling holes and to grid resolution. Even with well-resolved grids and included plena, strong mixing challenges traditional eddy viscosity models. To address this, a simplified configuration is simulated: a truncated row of shaped holes on the suction side near the leading edge and a row on the pressure side, both fed from internal plena while the purge slot and tip clearance are also modeled. Two isothermal LES cases are conducted at two distinct wall temperatures, which yield the adiabatic wall temperature and the heat transfer coefficient. The definition and means of computation of the effectiveness is discussed in this paper. The simulations reveal detailed three-dimensional unsteady flow features, including coherent vortical structures and secondary flows originating from the purge cavity. Film cooling effectiveness and Nusselt number distributions are presented for both the blade surface and tip, highlighting regions of elevated heat transfer and complex thermal behavior. These findings underscore the importance of high-resolution LES and realistic boundary conditions in capturing the dynamics of purge and film cooling, offering valuable insights for improving turbine blade design and thermal management strategies.

Secondary Flows↗

DOE OSTI2023

Machine learning from RANS and LES to inform coarse grid simulations

Nuclear system thermal hydraulic analysis has historically relied on computationally inexpensive 1D codes. However, such tools are unable to capture multiscale multidimensional effects in large nuclear reactor enclosures. On the other hand, simulations with higher fidelity can be too expensive for such purposes. One of the ways to reduce computational cost is to perform simulations on a coarse grid, which, unfortunately, introduces large discretization errors. In this paper, two high-to-low data-driven approaches are investigated: (1) a coarse grid turbulence model to predict eddy viscosity and (2) correction of errors in coarse grid velocity fields. The approaches aim to reduce grid- and turbulence model-induced errors in coarse grid Reynolds-averaged Navier–Stokes (RANS) simulations. Two sources of high-fidelity data, RANS and large eddy simulations (LES), are explored. To extract the eddy viscosity from the LES data, an inverse optimization problem is solved. However, the LES eddy viscosity is shown to be comparable to the RANS eddy viscosity in terms of error reduction. Therefore, the directly available RANS eddy viscosity was used to develop a coarse grid data-driven turbulence model. Additionally, error correction in velocity is used to reduce the remaining uncertainties and bring the results closer to reality. In conclusion, the performance of the frameworks is demonstrated for a scaled upper plenum of a gas-cooled reactor facility.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗

DOE OSTI2023

Dynamic Subgrid Turbulence Modeling for Shallow Cumulus Convection Simulations beyond LES Resolutions

A scale-dependent dynamic Smagorinsky model is implemented in the Met Office/NERC Cloud (MONC) model using two averaging flavors, along Lagrangian pathlines and local moving averages. The dynamic approaches were compared against the conventional Smagorinsky–Lilly scheme in simulating the diurnal cycle of shallow cumulus convection. The simulations spanned from the LES to the near-gray-zone and gray-zone resolutions and revealed the adaptability of the dynamic model across the scales and different stability regimes. The dynamic model can produce a scale- and stability-dependent profile of the subfilter turbulence length scale across the chosen resolution range. At gray-zone resolutions the adaptive length scales can better represent the early precloud boundary layer leading to temperature and moisture profiles closer to the LES compared to the standard Smagorinsky. As a result, the initialization and general representation of the cloud field in the dynamic model is in good agreement with the LES. In contrast, the standard Smagorinsky produces a less well-mixed boundary layer, which fails to ventilate moisture from the boundary layer, resulting in the delayed spinup of the cloud layer. Moreover, strong downgradient diffusion controls the turbulent transport of scalars in the cloud layer. However, the dynamic approaches rely on the resolved field to account for nonlocal transports, leading to overenergetic structures when the boundary layer is fully developed and the Lagrangian model is used. Introducing the local averaging version of the model or adopting a new Lagrangian time scale provides stronger dissipation without significantly affecting model behavior.

54 ENVIRONMENTAL SCIENCES↗

DOE OSTI2024

An Investigation of LES Wall Modeling for Rayleigh–Bénard Convection via Interpretable and Physics-Aware Feedforward Neural Networks with DNS

Abstract The traditional approach of using the Monin–Obukhov similarity theory (MOST) to model near-surface processes in large-eddy simulations (LESs) can lead to significant errors in natural convection. In this study, we propose an alternative approach based on feedforward neural networks (FNNs) trained on output from direct numerical simulation (DNS). To evaluate the performance, we conduct both a priori and a posteriori tests. In the a priori (offline) tests, we compare the statistics of the surface shear stress and heat flux, computed from filtered DNS input variables, to the stress and flux obtained from the filtered DNS. Additionally, we investigate the importance of various input features using the Shapley additive explanations value and the conditional average of the filter grid cells. In the a posteriori (online) tests, we implement the trained models in the System for Atmospheric Modeling (SAM) LES and compare the LES-generated surface shear stress and heat flux with those in the DNS. Our findings reveal that vertical velocity, a traditionally overlooked flow quantity, is one of the most important input features for determining the wall fluxes. Increasing the number of input features improves the a priori test results but does not always improve the model performance in the a posteriori tests because of the differences in input variables between the LES and DNS. Last, we show that physics-aware FNN models trained with logarithmic and scaled parameters can well extrapolate to more intense convection scenarios than in the training dataset, whereas those trained with primitive flow quantities cannot. Significance Statement The traditional near-surface turbulence model, based on a shear-dominated boundary layer flow, does not represent near-surface turbulence in natural convection. Using a feedforward neural network (FNN), we can construct a more accurate model that better represents the near-surface turbulence in various flows and reveals previously overlooked controlling factors and process interactions. Our study shows that the FNN-generated models outperform the traditional model and highlight the importance of the near-surface vertical velocity. Furthermore, the physics-aware FNN models exhibit the potential to extrapolate to convective flows of various intensities beyond the range of the training dataset, suggesting their broader applicability for more accurate modeling of near-surface turbulence.

54 ENVIRONMENTAL SCIENCES↗