SEARCH · Engineering Papers

Results for “physical feature”

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 559 records · Page 31

All sources NASA NTRS DOE OSTI

DOE OSTI2025

Physics-Informed Machine Learning Model for Ceramic Matrix Composite Creep

A physics-informed recurrent neural network (RNN) based surrogate model is developed to emulate the nonlinear, time-dependent constitutive behavior of ceramic matrix composites (CMCs) driven by matrix damage and constituent creep at the microscale. Physics-informed constraints are introduced into the surrogate model through regularization to ground the prediction in physics and improve its predictive capabilities. Training data is generated using the high-fidelity generalized method of cells (HFGMC) approach which calls appropriate creep and damage models for each of the constituents. This coupling permits simulating the nonlinear behavior of CMCs based on constituent response at the microscale along with microstructural features such as fiber and porosity volume fraction and fiber radius. The microscale repeating unit cell is loaded under creep fatigue conditions to replicate the material loading experienced in a turbine engine. Therefore, the RNN-based surrogate model is tasked with predicting, as a function of variable input stress sequence, temperature, and microstructural features, the resulting strain history response while satisfying physical constraints related to creep rate, isochoric inelastic deformation, and strain energy density. The trained surrogate model is shown to effectively match the strain history over quantified distributions of microstructural features and relevant loading regimes and temperatures. Neural network based surrogate models can offer efficient alternatives to running computationally intensive multiscale material models to simulate the nonlinear response of large structural models. Therefore, the presented work provides evidence towards the feasibility of developing, training, and running such models for CMCs with complex microstructures, nonlinear time-dependent material response, and under non-monotonic loading conditions.

ceramic matrix composites↗

DOE OSTI2025

Reduced Order Modeling conditioned on monitored features for response and error bounds estimation in engineered systems

Reduced Order Models (ROMs) form essential tools across engineering domains by virtue of their function as surrogates for computationally intensive digital twinning simulators. Although purely data-driven methods are available for ROM construction, schemes that allow to retain a portion of the physics tend to enhance the interpretability and generalization of ROMs. However, physics-based techniques can adversely scale when dealing with nonlinear systems that feature parametric dependencies. This study introduces a generative physics-based ROM that is suited for nonlinear systems with parametric dependencies and is additionally able to provide numerical error bounds associated with the respective estimates. A main contribution of this work is the conditioning of these parametric ROMs to features that can be derived from monitoring measurements, feasibly in an online fashion. This is contrary to most existing ROM schemes, which remain restricted to the prescription of the physics-based, and usually a priori unknown, system parameters. Our work utilizes conditional Variational Autoencoders to continuously map the required reduction bases to a feature vector extracted from limited output measurements, while additionally allowing for a probabilistic assessment of the ROM-estimated Quantities of Interest. An auxiliary task using a neural network-based parametrization of suitable probability distributions is introduced to re-establish the link with physical model parameters. We verify the proposed scheme on a series of simulated case studies incorporating effects of geometric and material nonlinearity under parametric dependencies related to system properties and input load characteristics.

Conditional VAEs↗

DOE OSTI2022

Low-Energy Physics in Neutrino LArTPCs

In this white paper, we outline some of the scientific opportunities and challenges related to detection and reconstruction of low-energy (less than 100 MeV) signatures in liquid argon time-projection chamber (LArTPC) detectors. Key takeaways are summarized as follows. 1) LArTPCs have unique sensitivity to a range of physics and astrophysics signatures via detection of event features at and below the few tens of MeV range. 2) Low-energy signatures are an integral part of GeV-scale accelerator neutrino interaction final states, and their reconstruction can enhance the oscillation physics sensitivities of LArTPC experiments. 3) BSM signals from accelerator and natural sources also generate diverse signatures in the low-energy range, and reconstruction of these signatures can increase the breadth of BSM scenarios accessible in LArTPC-based searches. 4) Neutrino interaction cross sections and other nuclear physics processes in argon relevant to sub-hundred-MeV LArTPC signatures are poorly understood. Improved theory and experimental measurements are needed. Pion decay-at-rest sources and charged particle and neutron test beams are ideal facilities for experimentally improving this understanding. 5) There are specific calibration needs in the low-energy range, as well as specific needs for control and understanding of radiological and cosmogenic backgrounds. 6) Novel ideas for future LArTPC technology that enhance low-energy capabilities should be explored. These include novel charge enhancement and readout systems, enhanced photon detection, low radioactivity argon, and xenon doping. 7) Low-energy signatures, whether steady-state or part of a supernova burst or larger GeV-scale event topology, have specific triggering, DAQ and reconstruction requirements that must be addressed outside the scope of conventional GeV-scale data collection and analysis pathways.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

NASA NTRS2025

Large Impact Features on Ganymede and Callisto as Revealed By Geological Mapping and Morphometry

The icy Galilean satellites are host to a broad range of impact feature morphologies. Hypotheses seeking to explain the diversity of these impact features consider the effects of impact melt, the physical state of the subsurface at the time of impact, and the impactor characteristics. As part of a larger effort to assess the role of these factors in the formation and evolution of these impact features, we have performed topographic and geological mapping of 19 large impact features on Ganymede and Callisto. These are divided into two main morphological groups: craters (subdivided into pit, dome, and anomalous dome craters), and penepalimpsests/palimpsests. The transitions from pit, dome, to anomalous dome craters appear to be size-dependent up to diameters of ∼170 km. The morphologies of pit and dome craters appear to be independent of their age or geologic context. The impacts that formed them only affected a cold, rigid ice layer, with the development of pits and raised annuli on their floors possibly stemming from the evolution of a pocket of impact melt. The subdued rims and floors of anomalous dome craters indicate the increasing effect of a weak, warm ice layer on impact feature morphology with increasing size, but their prominent annuli and pits indicate that mobilization of impact melt is also a factor. The very low topographic relief of older penepalimpsests and palimpsests indicates that their impacts penetrated the ice shell to mobilize very large volumes of pre-existing liquid from a subsurface layer, with little contribution to the final feature morphology from impact melt. Penepalimpsests are distinguished from palimpsests by the higher frequency of concentric ridges within their interiors, indicating a generally more robust state of the subsurface that could better support the rotation and uplift of solid material during impact, even if a crater-like depression could not be supported. A few impact features seem to be transitional between anomalous dome craters and penepalimpsests, and the overlap of anomalous dome craters, penepalimpsests, and palimpsests in terms of diameter as well as age indicates that impactor size and subsurface properties over time are major factors in determining which of these morphologies emerges.

Oliver L White↗

DOE OSTI2025

Physics-informed latent neural operator for real-time predictions of time-dependent parametric PDEs

Deep operator network (DeepONet) has shown significant promise as surrogate models for systems governed by partial differential equations (PDEs), enabling accurate mappings between infinite-dimensional function spaces. However, when applied to systems with high-dimensional input-output mappings arising from large numbers of spatial and temporal collocation points, these models often require heavily overparameterized networks, leading to long training times. Latent DeepONet addresses some of these challenges by introducing a two-step approach: first learning a reduced latent space using a separate model, followed by operator learning within this latent space. While efficient, this method is inherently data-driven and lacks mechanisms for incorporating physical laws, limiting its robustness and generalizability in data-scarce settings. Here, in this work, we propose PI-Latent-NO, a physics-informed latent neural operator framework that integrates governing physics directly into the learning process. Our architecture features two coupled DeepONets trained end-to-end: a Latent-DeepONet that learns a low-dimensional representation of the solution, and a Reconstruction-DeepONet that maps this latent representation back to the physical space. By embedding PDE constraints into the training via automatic differentiation, our method eliminates the need for labeled training data and ensures physics-consistent predictions. The proposed framework is both memory and compute-efficient, exhibiting near-constant scaling with problem size and demonstrating significant speedups over traditional physics-informed operator models. We validate our approach on a range of parametric PDEs, showcasing its accuracy, scalability, and suitability for real-time prediction in complex physical systems.

Latent representations↗

DOE OSTI2021

Hybrid feature-driven learning system for abnormality detection and localization

A cyber-physical system may have a plurality of monitoring nodes each generating a series of current monitoring node values over time representing current operation of the system. A data-driven features extraction computer platform may receive the series of current monitoring node values and generate current data-driven feature vectors based on the series of current monitoring node values. A residual features extraction computer platform may receive the series of current monitoring node values, execute a system model and utilize a stochastic filter to determine current residual values, and generate current residual-driven feature vectors based on the current residual values. An abnormal detection platform may then receive the current data-driven and residual-driven feature vectors and compare the current data-driven and residual-driven feature vectors with at least one decision boundary associated with an abnormal detection model. An abnormal alert signal may then be transmitted when appropriate based on a result of said comparison.

Abbaszadeh, Masoud↗

DOE OSTI2023

Low-energy physics in neutrino LArTPCs

Here, in this paper, we review scientific opportunities and challenges related to detection and reconstruction of low-energy (less than 100 MeV) signatures in liquid argon time-projection chamber (LArTPC) neutrino detectors. LArTPC neutrino detectors designed for performing precise long-baseline oscillation measurements with GeV-scale accelerator neutrino beams also have unique sensitivity to a range of physics and astrophysics signatures via detection of event features at and below the few tens of MeV range. In addition, low-energy signatures are an integral part of GeV-scale accelerator neutrino interaction final-states, and their reconstruction can enhance the oscillation physics sensitivities of LArTPC experiments. New physics signals from accelerator and natural sources also generate diverse signatures in the low-energy range, and reconstruction of these signatures can increase the breadth of Beyond the Standard Model scenarios accessible in LArTPC-based searches. A variety of experimental and theory-related challenges remain to realizing this full range of potential benefits. Neutrino interaction cross-sections and other nuclear physics processes in argon relevant to sub-hundred-MeV LArTPC signatures are poorly understood, and improved theory and experimental measurements are needed; pion decay-at-rest sources and charged particle and neutron test beams are ideal facilities for improving this understanding. There are specific calibration needs in the low-energy range, as well as specific needs for control and understanding of radiological and cosmogenic backgrounds. Low-energy signatures, whether steady-state or part of a supernova burst or larger GeV-scale event topology, have specific triggering, DAQ and reconstruction requirements that must be addressed outside the scope of conventional GeV-scale data collection and analysis pathways. Novel concepts for future LArTPC technology that enhance low-energy capabilities should also be explored to help address these challenges.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2023

Project Update for “Designing Nuclear-data Measurements that Resolve Discrepancies in Existing Data” [Slides]

AIACHNE has made key progress this past year and will contribute to the larger scientific community. We recovered input data for the current 252 Cf(sf) PFNS evaluation that was previously lost. We render a standard to the best of our ability reproducible. We critically reviewed past data as input for ML & new standard evaluation that will impact PFNS of all major actinides. We developed a unique AI/ ML code that highlights which measurement features are related to bias and are working towards open-sourcing it for the community. Features that were identified as related to bias follow physics’ intuition and bring new understanding of exp effects and might help us for other reactions and isotopes. The results highlight that EXFOR is a goldmine of features that could help us understand experiment bias (if they are easy to parse).

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

DOE OSTI2025

Probing the shear-induced microstructure of a smectite clay aqueous suspensions using rheo-USANS and rheo-SIPLI measurements

Hypothesis: The static microstructure of aqueous sodium-montmorillonite (Na-Mt) suspensions at low ionic strengths (where Particle Size/Debye Length ≈1) exhibits both the particle–particle ordering as well as aggregation with repulsive ordered domains having characteristic optical birefringence and attractive aggregated entities larger than 20 µm resulting in ever-increasing yield stresses also known as physical aging-rejuvenation behavior. We hypothesize that the attractive particle–particle aggregation is the underlying cause behind the physical aging-rejuvenation behavior observed in Na-Mt suspensions with no contribution from structural dynamics driven by repulsive particle–particle ordering or jamming. Experiments: We investigate the shear-induced microstructure of aqueous Na-Mt suspensions in the sol and gel state using rheo-ultra-small angle neutron scattering (rheo-USANS) experiments at shear rates of 1, 50, 500, and 2000 s −1 . We also perform rheo-shear-induced polarization light imaging (rheo-SIPLI) experiments to relate ordering with shearing and aging. Findings: Shearing the suspensions at low to moderate shear rates induces particle–particle aggregation and shearing at high shear rates induces the breakage of particle–particle aggregation in the sol and gel states, suggesting the microstructural aggregation in the sol and gel state is shear sensitive and a full rejuvenation or breakage of particle–particle aggregation is only achieved at a minimum critical shear rate. The rheo-SIPLI experiments reveal that the sol and gel state exhibited strong Maltese cross patterns at a shear rate of 1000 s −1 , indicating particle–particle ordering. Post shearing, the gel exhibited temporal evolution of storage modulus without any noticeable influence on the appearance of the Maltese cross pattern indicating physical aging and particle ordering are distinct length scale phenomena in Na-Mt suspensions and the physical aging-rejuvenation behaviour is a feature of particle–particle aggregation as opposed to ordering.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2021

A Method to Extract Spatially Resolved Polycyclic Aromatic Hydrocarbon Emission from Spitzer Spectra: Application to M51

The mid-infrared spectrum contains rich diagnostics to probe the physical properties of galaxies, among which the pervasive emission features from polycyclic aromatic hydrocarbons (PAHs) offer promising means of estimating the star formation rate (SFR) relatively immune from dust extinction. This paper investigates the effectiveness of PAH emission as a SFR indicator on subkiloparsec scales by studying the Spitzer/IRS mapping-mode observations of the nearby grand-design spiral galaxy M51. We present a new approach of analyzing the spatial elements of the spectral data cube that simultaneously maximizes spatial resolution and spatial coverage, while yielding reliable measurements of the total, integrated 5–20 μm PAH emission. We devise a strategy of extracting robust PAH emission using spectra with only partial spectral coverage, complementing missing spectral regions with properly combined mid-infrared photometry. We find that in M51 the PAH emission correlates tightly with the extinction-corrected far-ultraviolet, near-ultraviolet, and Hα emission, from scales of ∼0.4 kpc close to the nucleus to 6 kpc out in the disk of the galaxy, indicating that PAH serves as an excellent tracer of SFR over a wide range of galactic environments. But regional differences exist. Close to the active nucleus of M51 the 6.2 μm feature is weaker, and the overall level of PAH emission is suppressed. The spiral arms and the central star-forming region of the galaxy emit stronger 7.7 and 8.6 μm PAH features than the inter-arm regions.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

DOE OSTI2023

Improving materials property predictions for graph neural networks with minimal feature engineering *

Graph neural networks (GNNs) have been employed in materials research to predict physical and functional properties, and have achieved superior performance in several application domains over prior machine learning approaches. Recent studies incorporate features of increasing complexity such as Gaussian radial functions, plane wave functions, and angular terms to augment the neural network models, with the expectation that these features are critical for achieving a high performance. Here, we propose a GNN that adopts edge convolution where hidden edge features evolve during training and extensive attention mechanisms, and operates on simple graphs with atoms as nodes and distances between them as edges. As a result, the same model can be used for very different tasks as no other domain-specific features are used. With a model that uses no feature engineering, we achieve performance comparable with state-of-the-art models with elaborate features for formation energy and band gap prediction with standard benchmarks; we achieve even better performance when the dataset size increases. Although some domain-specific datasets still require hand-crafted features to achieve state-of-the-art results, our selected architecture choices greatly reduce the need for elaborate feature engineering and still maintain predictive power in comparison.

42 ENGINEERING↗

DOE OSTI2026

An interactive machine learning platform for analyzing multi-particle coincidence data from cold target recoil ion momentum spectroscopy

We present SCULPT (Supervised Clustering and Uncovering Latent Patterns with Training), a comprehensive software platform for analyzing tabulated high-dimensional multi-particle coincidence data from Cold Target Recoil Ion Momentum Spectroscopy (COLTRIMS) experiments. The software addresses critical challenges in modern momentum spectroscopy by integrating advanced machine learning techniques with physics-informed analysis in an interactive web-based environment. SCULPT implements uniform manifold approximation and projection for non-linear dimensionality reduction to reveal correlations in high-dimensional data. We also discuss potential extensions to deep autoencoders for feature learning and genetic programming for automated discovery of physically meaningful observables. A novel adaptive confidence scoring system provides quantitative reliability assessments by evaluating user-selected clustering quality metrics with predefined weights that reflect each metric’s robustness. The platform features configurable molecular profiles for different experimental systems, interactive visualization with selection tools, and comprehensive data filtering capabilities. Utilizing a subset of SCULPT’s capabilities, we analyze photo-double-ionization data measured using the COLTRIMS method for three-body dissociation of the D 2 O molecule, revealing distinct fragmentation channels and their correlations with physics parameters. The software’s modular architecture and web-based implementation make it accessible to the broader atomic and molecular physics community, significantly reducing the time required for complex multi-dimensional analyses. This opens the door to finding and isolating rare events exhibiting non-linear correlations on the fly during experimental measurements, which can help steer exploration and improve the efficiency of experiments.

Artificial neural networks↗

DOE OSTI2025

Distinct effect of Kondo physics on crystal field splitting in electron and spin spectroscopies

Magnetic anisotropy is a key feature of rare earth materials from permanent magnets to heavy fermions. Here we explore the complex interplay of Kondo physics and anisotropy, and their effect on different experimental probes of magnetic anisotropy in a minimal 𝐽=3/2 Anderson impurity model using the numerical renormalization group. While anisotropy suppresses Kondo screening, virtual valence fluctuations enhance the anisotropy. We find distinct renormalization of the magnetic anisotropy measured via dynamical spin response (inelastic neutron scattering) versus electronic excitations in the impurity spectral function (resonant inelastic x-ray and scanning tunneling spectroscopy). The two measurement types have different responses and dependencies on the temperature and Kondo scales.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

NASA NTRS2007

Heliophysics Science and the Moon: Potential Solar and Space Physics Science for Lunar Exploration

This report addresses both these features new science enabled by NASAs exploration initiative and enabling science that is critical to ensuring a safe return to the Moon and onward to Mars. The areas of interest are structured into four main themes: Theme 1: Heliophysics Science of the Moon Studies of the Moons unique magnetodynamic plasma environment. Theme 2: Space Weather, Safeguarding the Journey Studies aimed at developing a predictive capability for space weather hazards. Theme 3: The Moon as a Historical Record Studies of the variation of the lunar regolith to uncover the history of the Sun, solar system, local interstellar medium, galaxy, and universe. Theme 4: The Moon as a Heliophysics Science Platform Using the unique environment of the lunar surface as a platform to provide observations beneficial to advancing heliophysics science.

Lunar exploration↗

DOE OSTI2026

Feedback, physics, and forecasts: The emerging paradigm of machine learning-driven battery research

Machine learning (ML) is reshaping how we understand, predict, and optimize electrochemical systems. In batteries, ML accelerates discovery across chemistry, design, and operation by transforming massive experimental and simulated datasets into predictive, interpretable models. This review consolidates a decade of progress in ML-driven battery innovation, from early-cycle feature extraction to operando image analysis and physics-informed modeling. We categorize approaches by data domain and physical fidelity, emphasizing interpretable ML for diagnostics, reinforcement learning for control, and multi-objective optimization for lifetime extension strategies. Additionally, we demonstrate how integrated models accelerate discovery, reduce testing time, and guide sustainable design. Economic analyses furthermore illustrate how these advances can lower cost per cycle and improve circularity. Together, these developments chart a path toward self-optimizing, sustainable battery technologies.

artificial intelligence↗

DOE OSTI2026

Virtual refrigerant charge sensor for variable-speed heat pumps based on feature selection

The refrigerant charge level in heat pump systems significantly impacts their energy efficiency. Virtual refrigerant charge (VRC) sensing technology has been comprehensively investigated and well-established due to its lower cost compared to physical sensors. However, the previous VRC research often relied on expert judgment and physical reasoning for their variable selection, which can potentially select redundant (or highly correlated) or insignificant features, and it is also primarily focused on single-speed systems. To address these challenges, this study proposes a VRC algorithm for variable-speed heat pumps that selects features through a rigorous feature selection method in combination with physical insights. We also propose a piecewise linear model structure segmented by subcooling temperature to accurately predict charge levels, particularly when subcooling temperatures are substantially low. The proposed algorithm was evaluated using experimental data of a residential R410A heat pump, and the performance was compared with two baseline VRC algorithms. The results are: (1) The proposed algorithm outperforms for the case with subcooling temperature less than 1 °C. (2) The proposed algorithm achieves a tested mean absolute percentage error (MAPE) of 4.23%, and improves the overall accuracy for cooling conditions by approximately 60%, compared with the two baseline algorithms. (3) The proposed algorithm uses two fewer features and improves the accuracy for undercharge cooling conditions by 68.0%, compared with baseline algorithm 2. These improvements enhance prediction accuracy and prevent overfitting, providing a more reliable refrigerant charge level prediction and helping improve the heat pump energy efficiency.

Liang, Chenjiyu↗

DOE OSTI2025

Neutron Identification Capabilities in MicroBooNE Through the Application of Machine Learning with Blips

Neutrinos (ν) are subatomic particles first observed in 1956 by LANL physicists Clyde Cowan and Frederick Reines but first theorized by Wolfgang Pauli in 1930. Neutrinos are the least massive known particle and are classified as leptons with 3 flavors corresponding to their leptonic counterparts (electron, muon, and tau). We know that they are abundant, 65 billion neutrinos travel through your fingertip every second, and elusive, a single neutrino could fly through a lightyear of lead without interacting at all. Though, there is much still unknown and a better characterization of these “ghostly” particles can give us clues as to the matter/anti-matter asymmetry in the early universe and possible glimpses into new physics. To measure a particle that is extremely light and rarely interacting, physicists have developed an extremely sensitive detector known as a Liquid Argon Time Projection Chamber (LArTPC). The fiducial volume (or TPC) is bombarded with neutrinos, some of which interact with argon (Ar) atoms to produce particles that in turn excite and ionize the Ar. The products of these are free electrons which then drift through the TPC’s applied magnetic field towards a multi-plane wire readout system. The electrons’ charge is collected at this anode and the light from the initial interactions is collected by photomultiplier tubes (PMTs). In conjunction, these mechanisms allow LArTPCs to achieve millimeter spatial resolution and sub-MeV energy thresholds. The detector of interest in this study is the MicroBooNE Experiment at Fermilab. MicroBooNE is an above ground LArTPC with dimensions of approximately 10m × 2.5m × 2.3m, about the size of a school bus. Its purpose is to study neutrinos, so to improve rates of measured ν interactions, the detector is squarely in the path of the Booster Neutrino Beam (BNB) at Fermilab. A major challenge in neutrino studies is energy reconstruction, much of the neutrino’s original energy is lost in interactions that the detector is not sensitive to, often due to low-energy products. The initial goal of this analysis was to better identify neutrons, the main source of poor energy reconstruction in neutrino events. Because neutrons are neutral particles, like neutrinos, we can only directly measure the products of their interactions in LArTPCs. Most of these products are low-energy signals and while each one contributes a negligible amount of energy, collectively these signals make up most of the lost energy in each neutrino event. We define these signals as blips; point-like, isolated depositions of charge in the detector. Blips have MeV-scale energies and are the size of a single hit (charge deposition) or a cluster of a few hits on at least two wire planes. Blips are the principal detector features used to study low-energy physics; thus, they are the key to unlocking information not only about neutrons but gamma photons, supernova and solar neutrinos as well as helping us better identify certain particles. Therefore, this analysis strives to use blips for improved neutron identification (ID) and characterization.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

DOE OSTI2025

An intercomparison of wall fluxes in a turbulent thermal convection chamber: Direct numerical simulations and wall-modeled large-eddy simulations enhanced by machine learning

Thermal convection in a closed chamber is driven by a warm bottom, a cold top, and side walls at various temperatures. Although wall fluxes are the source of convection energy, accurately modeling these fluxes (i.e., the wall model) is challenging. In large-eddy simulations (LESs), many wall models are traditionally derived from the canonical boundary layer, which may be unsuitable for thermal convection bounded by both horizontal and vertical walls. This study conducts a model intercomparison of dry convection in a cubic-meter chamber using three direct numerical simulations (DNSs) and four LESs with different wall models. The LESs employ traditional wall models, a new wall model employing physics-aware neural networks, and a refined grid near the walls. The experiment involves four cases with varying sidewall temperatures. Our results show that LESs capture the main flow features and the trends of mean fluxes. The physics-aware neural networks and refined wall grids can improve the temporally averaged local fluxes when the large-scale circulation has a preferred direction. Even without the local improvement of wall fluxes, the LES flow quantities (temperature and velocities) can still largely match those in DNSs, provided the mean flux largely matches the DNSs. Additionally, DNSs reveal that a variation in corner treatments has minimal impacts on the flow quantities away from corners. Finally, LESs underestimate the mean fluxes of the entire wall due to their inability to resolve corner regions, but their mean flux away from the corner can better match DNS.

54 ENVIRONMENTAL SCIENCES↗