Search

Data-driven approaches offer novel opportunities for improving the performance of turbulent flow simulations, which are critical to wide-ranging applications from wind farms and aerodynamic designs to weather and climate forecasting. While conventional continuum Navier-Stokes solvers have been the subject of a significant amount of work in this domain, there has hitherto been very limited effort in the same direction for the more scalable and highly performant lattice Boltzmann method (LBM), even though it has been successfully applied to a variety of turbulent flow simulations using large-eddy simulation (LES) techniques. In this work, we establish a data-driven framework for the LES-based lattice Boltzmann simulation of near-wall turbulent flow fields. We do this by training neural networks using improved delayed detached eddy simulation data. Crucially, this is done in combination with physics-based information that substantially constrains the data-driven predictions. Using data from turbulent channel flow at a friction Reynolds number at $5200$, our simulations accurately predict the behaviour of the wall model at arbitrary friction Reynolds numbers up to $1.0 \times 10^6$. In contradistinction with other models that use direct numerical simulation datasets, our physics-informed model requires data from very limited regions within the wall-bounded plane, reducing by three orders of magnitude the quantity of data needed for training. We also demonstrate that our model can handle data configurations when the near-wall grid is sparse. Our physics-informed neural network approach opens up the possibility of employing LBM in combination with highly specific and therefore much more limited quantities of macroscopic data, substantially facilitating the investigation of a wide-range of turbulent flow applications at very large scale.

Despite remarkable achievements in video synthesis, achieving granular control over complex dynamics, such as nuanced movement among multiple interacting objects, still presents a significant hurdle for dynamic world modeling, compounded by the necessity to manage appearance and disappearance, drastic scale changes, and ensure consistency for instances across frames. These challenges hinder the development of video generation that can faithfully mimic real-world complexity, limiting utility for applications requiring high-level realism and controllability, including advanced scene simulation and training of perception systems. To address that, we propose TrackDiffusion, a novel video generation framework affording fine-grained trajectory-conditioned motion control via diffusion models, which facilitates the precise manipulation of the object trajectories and interactions, overcoming the prevalent limitation of scale and continuity disruptions. A pivotal component of TrackDiffusion is the instance enhancer, which explicitly ensures inter-frame consistency of multiple objects, a critical factor overlooked in the current literature. Moreover, we demonstrate that generated video sequences by our TrackDiffusion can be used as training data for visual perception models. To the best of our knowledge, this is the first work to apply video diffusion models with tracklet conditions and demonstrate that generated frames can be beneficial for improving the performance of object trackers.

In order to study the properties of SEP elements, we propose the concepts of one sided a_idempotent and one sided a_equivalent. Under the condition that an element in a ring is both group invertible and MP_invertible, some equivalent conditions of such an element to be an SEP element are given based on these two concepts, as will as based on projections and the second and the third power of some products of some elements., Comment: 11 pages

The synthetic turbulence generator (STG) lies at the interface of the Reynolds averaged Navier-Stokes (RANS) simulation and large eddy simulation (LES). This paper presents a STG for the multiple-relaxation-time(MRT) lattice Boltzmann method(LBM) framework at high friction Reynolds numbers, with consideration of near wall modeling. The Reichardt wall law, in combination with a force-based method, is used to model the near wall field. The STG wall-modeled(STG-WM) LES results are compared with turbulent channel flow simulations at $Re_{\tau}=1000,2000,5200$ at different resolutions. The results demonstrate good agreement with DNS, with the adaptation length of 6 to 8 boundary layer thickness. This method has a wide range of potentials for hybrid RANS/LES-LBM related applications at high friction Reynolds numbers.

Forests play an important role in influencing the wind resource in atmospheric boundary layers and the fatigue life of wind turbines. Due to turbulence, a difficulty in the simulation of the forest effects is that flow statistical and fluctuating content should be accurately resolved using a turbulence-resolved CFD method, which requires a large amount of computing time and resources. In this paper, we demonstrate a fast but accurate simulation platform that uses a lattice Boltzmann method with large eddy simulation on Graphic Processing Units (GPU). The simulation tool is the open-source program, GASCANS, developed at the University of Manchester. The simulation platform is validated based on canonical wall-bounded turbulent flows. A forest is modelled in the form of body forces injected near the wall. Since a uniform cell size is applied throughout the computational domain, the averaged first-layer cell height over the wall reaches to $\langle \Delta y^+\rangle = 165$. Simulation results agree well with previous experiments and numerical data obtained from finite volume methods. We demonstrate that good results are possible without the use of a wall-function, since the forest forces overwhelm wall friction. This is shown to hold as long as the forest region is resolved with several cells. In addition to the GPU speedup, the approximations also significantly benefit the computation efficiency.

The paper presents a synthetic turbulence generator (STG) for the lattice Boltzmann method (LBM) at the interface of the Reynolds averaged Naiver-Stokes (RANS) equations and the LBM Large Eddy Simulation (LES). We first obtain the RANS velocity field from a finite volume solver at the interface. Then, we apply a numerical interpolation from the RANS velocity field to the LBM velocity field due to the different grid types of RANS and LBM. The STG method generates the velocity fluctuations, and the regularized LBM reconstructs the particle distribution functions at the interface. We perform a turbulent channel flow simulation at Re_{\tau} = 180 with the STG at the inlet and the pressure-free boundary condition at the outlet. The velocity field is quantitatively compared with the periodic lattice Boltzmann based LES (LES-LBM) channel flow and the direct numerical simulation (DNS) channel flow. Both adaptation length and time for the STG method are evaluated. Also, we compare the STG-LBM channel flow results with the existing LBM synthetic eddies method (SEM-LBM) results. Our numerical investigations show good agreement with the DNS and periodic LES-LBM channel flow within a short adaptation length. The adaptation time for the turbulent channel flow is quantitatively analyzed and matches the DNS around 1.5 to 3 domain flow-through time. Finally, we check the auto-correlation for the velocity components at different cross-sections of the streamwise direction. The proposed STG-LBM is observed to be both fast and robust. The findings show good potential for the hybrid RANS/LES-LBM based solver on the aerodynamics simulations and a broad spectrum of engineering applications.