arXiv subjects

Jianchun Wang

Publications and source records attributed to Jianchun Wang.

48 records · Page 3Linked to original sources

arXiv2022

Linear attention coupled Fourier neural operator for simulation of three-dimensional turbulence

Modeling three-dimensional (3D) turbulence by neural networks is difficult because 3D turbulence is highly-nonlinear with high degrees of freedom and the corresponding simulation is memory-intensive. Recently, the attention mechanism has been shown as a promising approach to boost the performance of neural networks on turbulence simulation. However, the standard self-attention mechanism uses $O(n^2)$ time and space with respect to input dimension $n$, and such quadratic complexity has become the main bottleneck for attention to be applied on 3D turbulence simulation. In this work, we resolve this issue with the concept of linear attention network. The linear attention approximates the standard attention by adding two linear projections, reducing the overall self-attention complexity from $O(n^2)$ to $O(n)$ in both time and space. The linear attention coupled Fourier neural operator (LAFNO) is developed for the simulation of 3D turbulence. Numerical simulations show that the linear attention mechanism provides 40\% error reduction at the same level of computational cost, and LAFNO can accurately reconstruct a variety of statistics and instantaneous spatial structures of 3D turbulence. The linear attention method would be helpful for the improvement of neural network models of 3D nonlinear problems involving high-dimensional data in other scientific domains.

physics.flu-dyn↗

arXiv2022

High-order Gas-kinetic Schemes with Non-compact and Compact Reconstruction for Implicit Large Eddy Simulation

High-order gas-kinetic scheme (HGKS) with 5th-order non-compact reconstruction has been well implemented for implicit large eddy simulation (ILES) in nearly incompressible turbulent channel flows. In this study, the HGKS with higher-order non-compact reconstruction and compact reconstruction will be validated in turbulence simulation. For higher-order non-compact reconstruction, 7th-order normal reconstruction and tangential reconstruction are implemented. In terms of compact reconstruction, 5th-order normal reconstruction is adopted. Current work aims to show the benefits of high-order non-compact reconstruction and compact reconstruction for ILES. The accuracy of HGKS is verified by numerical simulation of three-dimensional advection of density perturbation. For the non-compact 7th-order scheme, 16 Gaussian points are required on the cell interface to preserve the order of accuracy. Then, HGKS with non-compact and compact reconstruction is used in the three-dimensional Taylor-Green vortex (TGV) problem and turbulent channel flows. Accurate ILES solutions have been obtained from HGKS. In terms of the physical modeling underlying the numerical algorithms, the compact reconstruction has the consistent physical and numerical domains of dependence without employing additional information from cells which have no any direct physical connection with the targeted cell. The compact GKS shows a favorable performance for turbulence simulation in resolving the multi-scale structures.

physics.flu-dyn↗

arXiv2022

Temporally sparse data assimilation for the small-scale reconstruction of turbulence

Previous works have shown that the small-scale information of incompressible homogeneous isotropic turbulence (HIT) is fully recoverable as long as sufficient large-scale structures are continuously enforced through temporally continuous data assimilation (TCDA). In the current work, we show that the assimilation time step can be relaxed to values about 1 $\sim$ 2 orders larger than that for TCDA, using a temporally sparse data assimilation (TSDA) strategy, while the accuracy is still maintained or even slightly better in the presence of non-negligible large-scale errors. The one-step data assimilation (ODA) is examined to unravel the mechanism of TSDA. It is shown that the relaxation effect for errors above the assimilation wavenumber $k_a$ is responsible for the error decay in ODA. Meanwhile, The errors contained in the large scales can propagate into small scales and make the high-wavenumber ($k>k_a$) error noise decay slower with TCDA than TSDA. This mechanism is further confirmed by incorporating different levels of errors in the large scales of the reference flow field. The advantage of TSDA is found to grow with the magnitude of the incorporated errors. Thus, it is potentially more beneficial to adopt TSDA if the reference data contains non-negligible errors. Finally, an outstanding issue raised in previous works regarding the possibility of recovering the dynamics of sub-Kolmogorov scales using direct numerical simulation (DNS) data at Kolmogorov scale resolution is also discussed.

physics.flu-dyn↗

arXiv2022

Attention-Enhanced Neural Network Models for Turbulence Simulation

Deep neural network models have shown a great potential in accelerating the simulation of fluid dynamic systems. Once trained, these models can make inference within seconds, thus can be extremely efficient. However, they suffer from a generalization problem when the flow becomes chaotic and turbulent. One of the most important reasons is that, existing models lack the mechanism to handle the unique characteristic of turbulent flow: multi-scale flow structures are non-uniformly distributed and strongly nonequilibrium. In this work, we address this issue with the concept of visual attention: intuitively, we expect the attention module to capture the nonequilibrium of turbulence by automatically adjusting weights on different regions. We benchmark the performance improvement with a state of the art neural network model, the Fourier Neural Operator (FNO), on two-dimensional (2D) turbulence prediction task. Numerical experiments show that the attention-enhanced neural network model can generalize well on higher Reynolds numbers flow, and can accurately reconstruct a variety of statistics and instantaneous spatial structures of turbulence. The attention mechanism provides 40% error reduction with 1% increase of parameters, at the same level of computational cost.

physics.flu-dyn↗

arXiv2021

Skin friction and heat transfer in hypersonic transitional and turbulent boundary layers

The decompositions of the skin-friction and heat transfer coefficients based on the two-fold repeated integration in hypersonic transitional and turbulent boundary layers are analyzed to explain the generations of the wall skin friction and heat transfer. The Reynolds analogy factor slightly increases as the wall temperature decreases, especially for the extremely cooled wall. The integral analysis is applied to explain the overshoot behaviours of the skin-friction and heat transfer coefficients in hypersonic transitional boundary layers. The overshoot of the skin-friction coefficient is mainly caused by the drastic change of the mean velocity profiles, and the overshoot of the heat transfer coefficient is primarily due to the viscous dissipation. In the hypersonic turbulent boundary layers, the skin-friction and heat transfer coefficients increase significantly as the wall temperature decreases. The effects of the mean velocity gradients and the Reynolds shear stress contribute dominantly to the wall skin friction, and have weak correlations with the wall temperature, except for the strongly cooled wall condition. The strongly cooled wall condition and high Mach number can enhance the effect of the Reynolds shear stress, and weaken the impact of the mean velocity gradients. Furthermore, the magnitudes of the dominant relative contributions of the mean temperature gradients, pressure dilatation, viscous dissipation and the Reynolds heat flux to the heat transfer coefficient increase as the wall temperature increases in the hypersonic turbulent boundary layers.

physics.flu-dyn↗

arXiv2021

Artificial neural network approach for turbulence models: A local framework

A local artificial neural network (LANN) framework is developed for turbulence modeling. The Reynolds-averaged Navier-Stokes (RANS) unclosed terms are reconstructed by artificial neural network (ANN) based on the local coordinate system which is orthogonal to the curved walls. We verify the proposed model for the flows over periodic hills. The correlation coefficients of the RANS unclosed terms predicted by the LANN model can be made larger than 0.96 in an a priori analysis, and the relative error of the unclosed terms can be made smaller than 18%. In an a posteriori analysis, detailed comparisons are made on the results of RANS simulations using the LANN and Spalart-Allmaras (SA) models. It is shown that the LANN model performs better than the SA model in the prediction of the average velocity, wall-shear stress and average pressure, which gives the results that are essentially indistinguishable from the direct numerical simulation (DNS) data. The LANN model trained in low Reynolds number Re = 2800 can be directly applied in the cases of high Reynolds numbers Re = 5600, 10595, 19000, 37000 with accurate predictions. Furthermore, the LANN model is verified for flows over periodic hills with varying slopes. These results suggest that the LANN framework has a great potential to be applied to complex turbulent flows with curved walls.

physics.flu-dyn↗

arXiv2020

Deconvolutional artificial neural network models for large eddy simulation of turbulence

Deconvolutional artificial neural network (DANN) models are developed for subgrid-scale (SGS) stress in large eddy simulation (LES) of turbulence. The filtered velocities at different spatial points are used as input features of the DANN models to reconstruct the unfiltered velocity. The grid width of the DANN models is chosen to be smaller than the filter width, in order to accurately model the effects of SGS dynamics. The DANN models can predict the SGS stress more accurately than the conventional approximate deconvolution method (ADM) and velocity gradient model (VGM) in a prior study: the correlation coefficients can be made larger than 99\% and the relative errors can be made less than 15\% for the DANN model. In an a posteriori study, a comprehensive comparison of the DANN model, the implicit large eddy simulation (ILES), the dynamic Smagorinsky model (DSM), and the dynamic mixed model (DMM) shows that: the DANN model is superior to the ILES, DSM, and DMM models in the prediction of the velocity spectrum, various statistics of velocity and the instantaneous coherent structures without increasing the considerable computational cost. Besides, the trained DANN models without any fine-tuning can predict the velocity statistics well for different filter widths. These results indicate that the DANN framework with consideration of SGS spatial features is a promising approach to develop advanced SGS models in the LES of turbulence.

physics.flu-dyn↗

arXiv2018

Model Reduction with Memory and the Machine Learning of Dynamical Systems

The well-known Mori-Zwanzig theory tells us that model reduction leads to memory effect. For a long time, modeling the memory effect accurately and efficiently has been an important but nearly impossible task in developing a good reduced model. In this work, we explore a natural analogy between recurrent neural networks and the Mori-Zwanzig formalism to establish a systematic approach for developing reduced models with memory. Two training models-a direct training model and a dynamically coupled training model-are proposed and compared. We apply these methods to the Kuramoto-Sivashinsky equation and the Navier-Stokes equation. Numerical experiments show that the proposed method can produce reduced model with good performance on both short-term prediction and long-term statistical properties.

cs.LG↗

arXiv2015

Testbeam studies of pre-prototype silicon strip sensors for the LHCb UT upgrade project

The LHCb experiment is preparing for a major upgrade in 2018-2019. One of the key components in the upgrade is a new silicon tracker situated upstream of the analysis magnet of the experiment. The Upstream Tracker (UT) will consist of four planes of silicon strip detectors, with each plane covering an area of about 2 m$^2$. An important consideration of these detectors is their performance after they have been exposed to a large radiation dose. In this article we present test beam results of pre-prototype n-in-p and p-in-n sensors that have been irradiated with fluences up to $4.0\times10^{14}$ $n_{\rm eq}$ cm$^{-2}$.

physics.ins-det↗

arXiv2011

Charged Particle Tracking with the Timepix ASIC

A prototype particle tracking telescope has been constructed using Timepix and Medipix ASIC hybrid pixel assemblies as the six sensing planes. Each telescope plane consisted of one 1.4 cm2 assembly, providing a 256x256 array of 55 micron square pixels. The telescope achieved a pointing resolution of 2.3 micron at the position of the device under test. During a beam test in 2009 the telescope was used to evaluate in detail the performance of two Timepix hybrid pixel assemblies; a standard planar 300 micron thick sensor, and 285 micron thick double sided 3D sensor. This paper describes a detailed charge calibration study of the pixel devices, which allows the true charge to be extracted, and reports on measurements of the charge collection characteristics and Landau distributions. The planar sensor achieved a best resolution of 4.0 micron for angled tracks, and resolutions of between 4.4 and 11 micron for perpendicular tracks, depending on the applied bias voltage. The double sided 3D sensor, which has significantly less charge sharing, was found to have an optimal resolution of 9.0 micron for angled tracks, and a resolution of 16.0 micron for perpendicular tracks. Based on these studies it is concluded that the Timepix ASIC shows an excellent performance when used as a device for charged particle tracking.

physics.ins-det↗

arXiv2003

Charmed-strange Mesons Experimental Results

Two new states in the charm strange sector, DsJ(2317) and DsJ(2460), have recently been discovered at electron positron collider experiments. The new states are first observed in the dominant Ds pi0 and Ds* pi0 modes respectively and are very narrow. They are consistent with 0+ and 1+ P-wave charm anti-strange mesons. The DsJ(2460) meson is also observed in Ds gamma and Ds pi+ pi- modes. A review of the discoveries and possible explanations is given.

hep-ex↗

arXiv2000

Study of the spatial resolution achievable with the BTeV pixel sensors

A Monte Carlo simulation has been developed to predict the spatial resolution of silicon pixel detectors. The results discussed in this paper focus on the unit cell geometry of 50 $μ$m x 400 $μ$m, as chosen for BTeV. Effects taken into account include energy deposition fluctuations along the charged particle path, diffusion, magnetic field and response of the front end electronics. We compare our predictions with measurements from a recent test beam study performed at Fermilab.

hep-ex↗