arXiv subjects

Guowei He

Publications and source records attributed to Guowei He.

35 records · Page 2Linked to original sources

arXiv2023

Combining direct and indirect sparse data for learning generalizable turbulence models

Learning turbulence models from observation data is of significant interest in discovering a unified model for a broad range of practical flow applications. Either the direct observation of Reynolds stress or the indirect observation of velocity has been used to improve the predictive capacity of turbulence models. In this work, we propose combining the direct and indirect sparse data to train neural network-based turbulence models. The backpropagation technique and the observation augmentation approach are used to train turbulence models with different observation data in a unified ensemble-based framework. These two types of observation data can explore synergy to constrain the model training in different observation spaces, which enables learning generalizable models from very sparse data. The present method is tested in secondary flows in a square duct and separated flows over periodic hills. Both cases demonstrate that combining direct and indirect observations is able to improve the generalizability of the learned model in similar flow configurations, compared to using only indirect data. The ensemble-based method can serve as a practical tool for model learning from different types of observations due to its non-intrusive and derivative-free nature.

physics.flu-dyn↗

arXiv2023

Learning neural-network-based turbulence models for external transonic flows using ensemble Kalman method

This paper presents a neural network-based turbulence modeling approach for transonic flows based on the ensemble Kalman method. The approach adopts a tensor basis neural network for the Reynolds stress representation, with modified inputs to consider fluid compressibility. The normalization of input features is also investigated to avoid feature collapsing in the presence of shock waves. Moreover, the turbulent heat flux is accordingly estimated with the neural network-based turbulence model based on the gradient diffusion hypothesis. The ensemble Kalman method is used to train the neural network with the experimental data in velocity and wall pressure due to its derivative-free nature. The proposed framework is tested in two canonical configurations, i.e., 2D transonic flows over the RAE2822 airfoils and 3D transonic flows over the ONERA M6 wings. Numerical results demonstrate the capability of the proposed method in learning accurate turbulence models for external transonic flows.

physics.flu-dyn↗

arXiv2022

Ensemble Kalman method for learning turbulence models from indirect observation data

In this work, we propose using an ensemble Kalman method to learn a nonlinear eddy viscosity model, represented as a tensor basis neural network, from velocity data. Data-driven turbulence models have emerged as a promising alternative to traditional models for providing closure mapping from the mean velocities to Reynolds stresses. Most data-driven models in this category need full-field Reynolds stress data for training, which not only places stringent demand on the data generation but also makes the trained model ill-conditioned and lacks robustness. This difficulty can be alleviated by incorporating the Reynolds-averaged Navier-Stokes (RANS) solver in the training process. However, this would necessitate developing adjoint solvers of the RANS model, which requires extra effort in code development and maintenance. Given this difficulty, we present an ensemble Kalman method with an adaptive step size to train a neural network-based turbulence model by using indirect observation data. To our knowledge, this is the first such attempt in turbulence modelling. The ensemble method is first verified on the flow in a square duct, where it correctly learns the underlying turbulence models from velocity data. Then, the generalizability of the learned model is evaluated on a family of separated flows over periodic hills. It is demonstrated that the turbulence model learned in one flow can predict flows in similar configurations with varying slopes.

physics.flu-dyn↗

arXiv2021

On the wavenumber-frequency spectra of wall pressure fluctuations in turbulent channel flows

The characteristics of the wavenumber-frequency spectra of the rapid, slow and total wall pressure fluctuations are investigated using direct numerical simulation (DNS) of turbulent channel flow up to $\Rey_τ\approx 1000$. For the wavenumber-frequency spectra of the total wall pressure fluctuations, a valley-like behavior of contour lines in the sub-convective region is found, which may be linked to the Kraichnan-Phillips theorem. For the decomposition of the wall pressure spectra, it is commonly assumed in previous studies that the cross spectral density (CSD) between the rapid and slow components of the wall pressure fluctuations can be neglected. Yet no experimental or numerical evidence is available for either confirming or disproving this assumption. In this paper, we use DNS data to quantitatively evaluate this assumption. Emphasizes are put on the error in decibel scale caused by neglecting the CSD between the rapid and slow components of the wall pressure fluctuations. It is found that this assumption is approximately accurate for one- and two-dimensional spectra, but causes a large magnitude of error in the three-dimensional wavenumber-frequency spectra. An error of 5dB is observed in the sub-convective region and such a large error is observed for a wide range of Reynolds numbers ($180\le\Rey_τ\le 1000$). The analyses show that the assumption of neglecting the CSD needs to be applied carefully at the scales falling in the sub-convective region.

physics.flu-dyn↗

arXiv2020

A wall model based on neural networks for LES of turbulent flows over periodic hills

In this work, a data-driven wall model for turbulent flows over periodic hills is developed using the feedforward neural network (FNN) and wall-resolved LES (WRLES) data. To develop a wall model applicable to different flow regimes, the flow data in the near wall region at all streamwise locations are grouped together as the training dataset. In the developed FNN wall models, we employ the wall-normal distance, near-wall velocities and pressure gradients as input features and the wall shear stresses as output labels, respectively. The prediction accuracy and generalization capacity of the trained FNN wall model are examined by comparing the predicted wall shear stresses with the WRLES data. For the instantaneous wall shear stress, the FNN predictions show an overall good agreement with the WRLES data with some discrepancies observed at locations near the crest of the hill. The correlation coefficients between the FNN predictions and WRLES predictions are larger than 0.7 at most streamwise locations. For the mean wall shear stress, the FNN predictions agree very well with WRLES data. More importantly, overall good performance of the FNN wall model is observed for different Reynolds numbers, demonstrating its good generalization capacity.

physics.flu-dyn↗

arXiv2019

Formation of Graphene atop a Si adlayer on the C-face of SiC

The structure of the SiC(000-1) surface, the C-face of the {0001} SiC surfaces, is studied as a function of temperature and of pressure in a gaseous environment of disilane (Si2H6). Various surface reconstructions are observed, both with and without the presence of an overlying graphene layer (which spontaneously forms at sufficiently high temperatures). Based on cross-sectional scanning transmission electron microscopy measurements, the interface structure that forms in the presence of the graphene is found to contain 1.4 - 1.7 monolayers (ML) of Si, a somewhat counter-intuitive result since, when the graphene forms, the system is actually under C-rich conditions. Using ab initio thermodynamics, it is demonstrated that there exists a class of Si-rich surfaces containing about 1.3 ML of Si that are stable on the surface (even under C-rich conditions) at temperatures above about 400 K. The structures that thus form consist of Si adatoms atop a Si adlayer on the C-face of SiC, with or without the presence of overlying graphene.

cond-mat.mes-hall↗

arXiv2014

Formation of a Buffer Layer for Graphene on C-face SiC{0001}

Graphene films prepared by heating the SiC(000-1) surface (the C-face of the {0001} surfaces) in a Si-rich environment are studied using low-energy electron diffraction (LEED) and low-energy electron microscopy (LEEM). Upon graphitization, an interface with rt(43) x rt(43)-R7.6 degree symmetry is observed by in situ LEED. After oxidation, the interface displays rt(3) x rt(3)-R30 degree symmetry. Electron reflectivity measurements indicate that these interface structures arise from a graphene-like "buffer layer" that forms between the graphene and the SiC, similar to that observed on Si-face SiC. From a dynamical LEED structure calculation for the oxidized C-face surface, it is found to consist of a graphene layer sitting on top of a silicate (Si2O3) layer, with the silicate layer having the well-known structure as previously studied on bare SiC(000-1) surfaces. Based on this result, the structure of the interface prior to oxidation is discussed.

cond-mat.mes-hall↗

arXiv2014

Friedel Oscillation-Induced Energy Gap Manifested as Transport Asymmetry at Monolayer-Bilayer Graphene Boundaries

We show that Friedel charge oscillation near an interface opens a gap at the Fermi energy for electrons with wave vectors perpendicular to the interface. If the Friedel gaps on two sides of the interface are different, a nonequlibrium effect - shifting of these gaps under bias - leads to asymmetric transport upon reversing the bias polarity. The predicted transport asymmetry is revealed by scanning tunneling potentiometry at monolayer-bilayer interfaces in epitaxial graphene on SiC (0001). This intriguing interfacial transport behavior opens a new avenue towards novel quantum functions such as quantum switching.

cond-mat.mes-hall↗

arXiv2013

Lift Enhancement by Dynamically Changing Wingspan in Forward Flapping Flight

Stretching and retracting wingspan has been widely observed in the flight of birds and bats, and its effects on the aerodynamic performance particularly lift generation are intriguing. The rectangular flat-plate flapping wing with a sinusoidally stretching and retracting wingspan is proposed as a simple model of biologically-inspired dynamic morphing wings. Direct numerical simulations of the low-Reynolds-number flows around the flapping morphing wing in a parametric space are conducted by using immersed boundary method. It is found that the instantaneous and time-averaged lift coefficients of the wing can be significantly enhanced by dynamically changing wingspan in a flapping cycle. The lift enhancement is caused not only by changing the lifting surface area, but also manipulating the flow structures that are responsible to the generation of the vortex lift. The physical mechanisms behind the lift enhancement are explored by examining the three-dimensional flow structures around the flapping wing.

physics.flu-dyn↗

arXiv2013

Temporal decorrelations in compressible isotropic turbulence

Temporal decorrelations in compressible isotropic turbulence are studied using the space-time correlation theory and direct numerical simulation. A swept-wave model is developed for dilatational components while the classic random sweeping model is proposed for solenoidal components. The swept-wave model shows that the temporal decorrelations in dilatational fluctuations are dominated by two physical processes: random sweeping and wave propagation. These models are supported by the direct numerical simulation of compressible isotropic turbulence, in the sense of that all curves of normalized time correlations for different wavenumbers collapse into a single one using the normalized time separations. The swept-wave model is further extended to account for a constant mean velocity.

physics.flu-dyn↗

arXiv2012

Ensemble Averaging for Dynamical Systems under Fast Oscillating Random Boundary Conditions

This paper is devoted to provide a theoretical underpinning for ensemble forecasting with rapid fluctuations in body forcing and in boundary conditions. Ensemble averaging principles are proved under suitable `mixing' conditions on random boundary conditions and on random body forcing. The ensemble averaged model is a nonlinear stochastic partial differential equation, with the deviation process (i.e., the approximation error process) quantified as the solution of a linear stochastic partial differential equation.

math.PR↗

arXiv2012

Formation of Graphene on SiC(000-1) Surfaces in Disilane and Neon Environments

The formation of graphene on the SiC(000-1) surface (the C-face of the {0001} surfaces) has been studied, utilizing both disilane and neon environments. In both cases, the interface between the graphene and the SiC is found to be different than for graphene formation in vacuum. A complex low-energy electron diffraction pattern with rt(43) x rt(43)-R\pm7.6° symmetry is found to form at the interface. An interface layer consisting essentially of graphene is observed, and it is argued that the manner in which this layer covalently bonds to the underlying SiC produces the rt(43) x rt(43)-R\pm7.6° structure [i.e. analogous to the 6rt(3) x 6rt(3)-R30° "buffer layer" that forms on the SiC(0001) surface (the Si-face)]. Oxidation of the surface is found to modify (eliminate) the rt(43) x rt(43)-R\pm7.6° structure, which is interpreted in the same manner as the known "decoupling" that occurs for the Si-face buffer layer.

cond-mat.mes-hall↗

arXiv2012

Quantifying Model Uncertainties in the Space of Probability Measures

Due to lack of scientific understanding, some mechanisms may be missing in mathematical modeling of complex phenomena in science and engineering. These mathematical models thus contain some uncertainties such as uncertain parameters. One method to estimate these parameters is based on pathwise observations, i.e., quantifying model uncertainty in the space of sample paths for system evolution. Another method is devised here to estimate uncertain parameters, or unknown system functions, based on experimental observations of probability distributions for system evolution. This is called the quantification of model uncertainties in the space of probability measures. A few examples are presented to demonstrate this method, analytically or numerically.

math.PR↗

arXiv2012

Graphene formed on SiC under various environments: Comparison of Si-face and C-face

The morphology of graphene on SiC {0001} surfaces formed in various environments including ultra-high vacuum, 1 atm of argon, and 10^-6 to 10^-4 Torr of disilane is studied by atomic force microscopy, low-energy electron microscopy, and Raman spectroscopy. The graphene is formed by heating the surface to 1100 - 1600 C, which causes preferential sublimation of the Si atoms. The argon atmosphere or the background of disilane decreases the sublimation rate so that a higher graphitization temperature is required, thus improving the morphology of the films. For the (0001) surface, large areas of monolayer-thick graphene are formed in this way, with the size of these areas depending on the miscut of the sample. Results on the (000-1) surface are more complex. This surface graphitizes at a lower temperature than for the (0001) surface and consequently the growth is more three-dimensional. In an atmosphere of argon the morphology becomes even worse, with the surface displaying markedly inhomogeneous nucleation, an effect attributed to unintentional oxidation of the surface during graphitization. Use of a disilane environment for the (000-1) surface is found to produce improved morphology, with relatively large areas of monolayer-thick graphene.

cond-mat.mtrl-sci↗

arXiv2011

Interface Structure of Graphene on SiC(000-1)

Graphene films prepared by heating the SiC(000-1) surface (the C-face of the {0001} surfaces) in vacuum or in a Si-rich environment are compared. It is found that different interface structures occur for the two situations. The former yields a well known 3x3 reconstructed interface, whereas the latter produces an interface with rt(43)xrt(43)-R\pm7.6 degrees symmetry. This structure is shown to contain a graphene-like layer with properties similar to the 6rt(3)x6rt(3)-R30 degrees "buffer layer" that forms on the Si(0001) surface (the Si-face).

cond-mat.mtrl-sci↗

arXiv2010

Comparison of Graphene Formation on C-face and Si-face SiC {0001} Surfaces

The morphology of graphene formed on the (000-1) surface (the C-face) and the (0001) surface (the Si-face) of SiC, by annealing in ultra-high vacuum or in an argon environment, is studied by atomic force microscopy and low-energy electron microscopy. The graphene forms due to preferential sublimation of Si from the surface. In vacuum, this sublimation occurs much more rapidly for the C-face than the Si-face, so that 150 C lower annealing temperatures are required for the C-face to obtain films of comparable thickness. The evolution of the morphology as a function of graphene thickness is examined, revealing significant differences between the C-face and the Si-face. For annealing near 1320 C, graphene films of about 2 monolayers (ML) thickness are formed on the Si-face, but 16 ML is found for the C-face. In both cases, step bunches are formed on the surface. For the Si-face, layer-by-layer growth of the graphene is observed in areas between the step bunches. At 1170 C, for the C-face, a more 3-dimensional type of growth is found. The average thickness is then about 4 ML, but with a wide variation in local thickness (2 - 7 ML) over the surface. The spatial arrangement of constant-thickness domains are found to be correlated with step bunches on the surface, which form in a more restricted manner than at 1320 C. It is argued that these domains are somewhat disconnected, so that no strong driving force for planarization of the film exists. In a 1-atm argon environment, permitting higher growth temperatures, the graphene morphology for the Si-face is found to become more layer-by-layer-like even for graphene thickness as low as 1 ML. However, for the C-face the morphology becomes much worse, with the surface displaying markedly inhomogeneous nucleation of the graphene. It is demonstrated that these surfaces are unintentionally oxidized, which accounts for the inhomogeneous growth.

cond-mat.mtrl-sci↗

arXiv1998

Statistics of Dissipation and Enstrophy Induced by a Set of Burgers Vortices

Dissipation and enstropy statistics are calculated for an ensemble of modified Burgers vortices in equilibrium under uniform straining. Different best-fit, finite-range scaling exponents are found for locally-averaged dissipation and enstrophy, in agreement with existing numerical simulations and experiments. However, the ratios of dissipation and enstropy moments supported by axisymmetric vortices of any profile are finite. Therefore the asymptotic scaling exponents for dissipation and enstrophy induced by such vortices are equal in the limit of infinite Reynolds number.

chao-dyn↗