Dimension reduced turbulent flow data from deep vector quantisers

In this paper, the effects of adiabatic and isothermal conditions on the statistics in compressible turbulent channel flow are investigated using direct numerical simulation (DNS). DNS of two compressible turbulent channel flows (Cases 1 and 2) are performed using a mixed Fourier Galerkin and B-spline collocation method. Case 1 is compressible turbulent channel flow between isothermal walls, which corresponds to DNS performed by Coleman et al. (1995). Case 2 is the flow between adiabatic and isothermal walls. The flow of Case 2 can be a very useful framework for the present objective, since it is the simplest turbulent channel flow with an adiabatic wall and provides ideal information for modelling the compressible turbulent flow near the adiabatic wall. Note that compressible turbulent channel flow between adiabatic walls is not stationary if there is no sink of heat. In Cases 1 and 2, the Mach number based on the bulk velocity and sound speed at the isothermal wall is 1.5, and the Reynolds number based on the bulk density, bulk velocity, channel half-width, and viscosity at the isothermal wall is 3000. To compare compressible and incompressible turbulent flows, DNS of two incompressible turbulent channel flows with passive scalar transport (Cases A and B) are performed using a mixed Fourier Galerkin and Chebyshev tau method. The wall boundary conditions of Cases A and B correspond to those of Cases 1 and 2, respectively. Case A corresponds to the DNS of Kim & Moin (1989). In Cases A and B, the Reynolds number based on the friction velocity, the channel half-width, and the kinematic viscosity is 150. The mean velocity and temperature near adiabatic and isothermal walls for compressible turbulent channel flow can be explained using the non-dimensional heat flux and the friction Mach number. It is found that Morkovin's hypothesis is not applicable to the near-wall asymptotic behaviour of the wall-normal turbulence intensity even if the variable property effect is taken into account. The mechanism of the energy transfers among the internal energy, mean and turbulent kinetic energiesis investigated, and the difference between the energy transfers near isothermal and adiabatic walls is revealed. Morkovin's hypothesis is not applicable to the correlation coeffcient between velocity and temperature fluctuations near the adiabatic wall.

Generating inflow fields for scale-resolving simulations of turbulent flow is crucial for a wide range of applications and is an active area of research. In this thesis, a method for inflow generation employing a precursor turbulent channel flow simulation is proposed. A procedure for determining the parameters of the precursor simulation based on the properties of the inflow is given. To evaluate the performance of the method, results from a simulation of a flat-plate zero-pressure-gradient turbulent boundary layer are analysed. The adaption length is quantified in terms of the development of integral quantities and the statistical moments of the velocity field. The performance is also compared with that of a state-of-the-art rescaling method for the generation of inflow data. It is shown that both approaches result in adaption lengths of comparable sizes, which makes the proposed method an attractive alternative due to its conceptual simplicity and robustness.As part of the work on inflow generation, a Python package, eddylicious, was developed. The purpose of the package is to be a framework within which various generation methods can be implemented. The package is available online under an open-source license. An overview of the architecture and currently implemented functionality of the package is given in this thesis.Furthermore, the results of a preparatory study on large-eddy simulation of wall-bounded turbulent flows are discussed. Fully-developed turbulent channel flow is used as a model problem, and the general-purpose computational fluid dynamics solver OpenFOAM is employed. The accuracy of the results with respect to the resolution of the computational mesh is analysed. Several modelling approaches for the subgrid scale stresses are considered.

High-order gas-kinetic scheme with parallel computation for direct numerical simulation of turbulent flows

This study proposes a novel deep-learning-based method for generating reduced representations of turbulent flows that ensures efficient storage and transfer while maintaining high accuracy during decompression. A Swin-transformer (ST) network combined with a physical constraints-based loss function is utilized to compress the turbulent flows with high compression ratios and then restore the data with underlying physical properties. The forced isotropic turbulence is used to demonstrate the ability of the ST-based model, where the instantaneous and statistical results show the excellent ability of the model to recover the flow data with a remarkable accuracy. Furthermore, the capability of the ST model is compared with a typical convolutional neural network-based auto-encoder (CNN-AE) by using the turbulent channel flow at two friction Reynolds numbers Reτ = 180 and 550. The results generated by the ST model are significantly more consistent with the direct numerical simulation data than those recovered by the CNN-AE, indicating the superior ability of the ST model to compress and restore the turbulent flow. This study also compares the compression performance of the ST model at different compression ratios (CR s) and finds that the model has low enough error even at very high CR. Additionally, the effect of transfer learning (TL) is investigated, showing that TL reduces the training time by 64% while maintaining high accuracy. The results illustrate for the first time that the Swin-transformer-based model incorporating a physically constrained loss function can compress and restore turbulent flows with the correct physics.

Two solution methods for compressible flows and their application in Large-Eddy Simulation (LES) of turbulent flows are discussed. After a summary of the basic equations and subgrid scale models, the applied discretization methods, solution schemes and boundary conditions with their implementation on high-performance computers are presented. The two discretization methods for the advective terms used in the simulations are a second-order mixed central-upwind discretization based on a AUSM formulation and a higher-order compact finite difference scheme. First, results of the second-order scheme are shown for the case of a turbulent channel and pipe flow, which are only weakly influenced by subgrid scale models, but are in good agreement with DNS and experimental data. Higher-order compact schemes are then used for both laminar and turbulent flows and the results are compared to those from simulations with the second-order scheme. It is found that for the laminar flow and the LES of plane turbulent jets the higher-order compact scheme is not of principal advantage. This is primarily related to the sensitivity of the compact scheme to stretched and curvilinear grids, which requires explicit filtering steps to avoid unphysical oscillations in the solution. Further results are then presented, which are obtained with the computationally cheaper second-order scheme. The results include turbulent flows in pipe bends with different Reynolds and Dean numbers. The solutions show swirl switching with a distinct low frequency peak. Finally, turbulent flows in non-trivial geometries are considered. This includes a first step towards the LES of the flow in an internal combustion engine with four valves and a realistic geometry.

Direct simulation of turbulent channel flow with finite difference schemes on non-staggered grids

Lattice Boltzmann simulations of turbulent shear flow between parallel porous walls

Large-eddy simulation of turbulent channel flows with conservative IDO scheme

Typical engineering rough surfaces show only limited resemblance to the artificially constructed rough surfaces that have been the basis of most previous fundamental research on turbulent flow over rough walls. In this article flow past an irregular rough surface is investigated, based on a scan of a rough graphite surface that serves as a typical example for an irregular rough surface found in engineering applications. The scanned map of surface height versus lateral coordinates is filtered in Fourier space to remove features on very small scales and to create a smoothly varying periodic representation of the surface. The surface is used as a no-slip boundary in direct numerical simulations of turbulent channel flow. For the resolution of the irregular boundary an iterative embedded boundary method is employed. The effects of the surface filtering on the turbulent flow are investigated by studying a series of surfaces with decreasing level of filtering. Mean flow, Reynolds stress and dispersive stress profiles show good agreement once a sufficiently large number of Fourier modes are retained. However, significant differences are observed if only the largest surface features are resolved. Strongly filtered surfaces give rise to a higher mean-flow velocity and to a higher variation of the streamwise velocity in the roughness layer compared with weakly filtered surfaces. In contrast, for the weakly filtered surfaces the mean flow is reversed over most of the lower part of the roughness sublayer and higher levels of dispersive shear stress are found.

Detached direct numerical simulations of turbulent two-phase bubbly channel flow

To investigate the mechanism of drag reduction by dilute addition of polymer in turbulent flows, a direct numerical simulation (DNS) of turbulent channel flow including discrete-element models has been carried out. The influence of polymers is approximated by beads-spring-dashpot model. The number of grid point for fluid turbulence is 64×64×64 and the number of discrete-elements is 250000. The DNS result achieved 15% drag reduction. The viscosity represented by a dashpot, rather than the elongated elasticity by a spring, contributed to the drag reduction. Experimental evidence such as enhancement of streamwise fluctuation, reduction of Reynolds shear stress, increase of streak spacing in the near-wall turbulence was consistently reproduced in our simulation.

We apply the dynamic subgrid-scale model to a large-eddy simulation of turbulent channel flow with a square rib mounted on one wall. The Reynolds number Re is 3.21 x 10 3 based on the mean velocity above the obstacle and the obstacle's height. Near-wall structures are resolved with the no-slip boundary condition. The results show better agreement with direct numerical simulation than large-eddy simulation with a fixed model constant, verifying the value of the dynamic subgrid-scale model for simulating complex turbulent flows. ARGE-EDDY simulation (LES) is an accurate method of simulating complex turbulent flows in which the large flow structures are computed while small scales are modeled. The rationale behind this method is based on two observations: most of the turbulent energy is in the large structures, and the small scales are more isotropic and universal. Therefore, LES may be more general and less geometry dependent than Reynolds-averaged modeling, although it comes at higher cost. Even though LES has been used by many investigators, most research has been limited to flows with simple geometry. In engineering applications, however, one encounters more complicated geometries. Here we shall consider a rectangular parallelepiped mounted on a flat surface. Related flows are those over surfaces protruding from submarines (conning towers or control fins), wind flows around buildings, and airflows over computer chips, among others. The most distinctive features associated with these flows are three dimensionality, flow separation due to protruding surfaces, and large-scale unsteadiness. As a model flow, we consider a plane channel flow in which a two-dimensional obstacle is mounted on one surface (see Fig. 1). This relatively simple geometry contains flow separation and reattachment. Flow in this geometry has been studied by Tropea and Gackstatter1 for low Re and Werner and Wengle2 and Dimaczek et al.3 for high Re, among others. Recently, Germano et al. 4 suggested a dynamic subgridscale model (DSGSM) in which the model coefficient is dynamically computed as computation progresses rather than specified a priori. This approach is based on an algebraic identity between the subgrid-scale stresses at two different filter levels and the resolved turbulent stresses. They applied the model to transitional and fully turbulent channel flows and showed that the model contributes nothing in laminar flow and exhibits the correct asymptotic behavior in the nearwall region of turbulent flows without an ad hoc damping function. This is a significant improvement over conventional subgrid-scale modeling.

This work focuses on the development of efficient computational tools for the simulation of turbulent multiphase polydispersed flows. In terms of methodologies we focus here on the use of Large Eddy Simulation (LES) and Quadrature-Based Methods of Moments (QBMM). In terms of applications the work is finalised, in order to be applied in the future, to particle production processes (precipitation and crystallisation in particular). An important part of the work concerns the study of the flow field in a Confined Impinging Jets Reactor (CIJR), frequently used in particle production processes. The first part is limited to the comparison and analysis of micro Particle Image Velocimetry (μPIV) experiments, carried out in a previous work, and Direct Numerical Simulation (DNS), carried out in this thesis. In particular the effects of boundary and operating conditions are studied and the numerical simulations are used to understand the experimental predictions and demonstrate the importance of unavoidable fluctuations in the experimental inlets. This represents a preparatory work for the LES modelling of the CIJR. Before investigating the accuracy of LES predictions for this particular application, the model and the implementation are studied in a more general context, represented by a well-known test case such as the periodic turbulent channel flow: the LES model implementation in TransAT, the code used in this work, is compared with DNS data and with predictions of other codes. LES simulations for the CIJR, provided with the proper boundary conditions obtained by the previous DNS/μPIV study, are then performed and compared with experiments, validating the model in a more realistic test case. Since particle precipitation and crystallization often result in complex interactions between particles and the continuous phase, in the second part of the work particular attention has been paid in the modelling of the momentum transfer and the resulting velocity of the particles (relative to the fluid). In particular the possibility of describing poly-disperse fluid-solid systems with QBMM together with LES and Equilibrium Eulerian Model (EEM) is assessed. The study is performed by comparing our predictions with DNS Lagrangian data in the turbulent channel flow previously described, seeded with particles corresponding to a realistic Particle Size Distribution (PSD). The last part of the work deals with particle collisions, extending QBMM to the investigation of non-equilibrium flows governed by the Boltzmann Equation with a hard-sphere collision kernel. The evolution of the particle velocity distribution is predicted and compared with other methods for kinetic equations such as Lattice Boltzmann Method (LBM), Discrete Velocity Method (DVM) and Grad's Moment Method (GM). The overall results of this thesis can be extended to a broad range of other applications of single-phase, dispersed multiphase and non-equilibrium flows

A statistical model is developed to reconstruct space–time energy spectra in turbulent flows from a non-extensive dataset comprising a time series of velocity fluctuations at a finite number of measurement points. This model is based on a higher approximation of energetic flow structures and developed by using local modulated waves. As a result, it can correctly predict the mean wavenumbers and spectral bandwidths. In contrast, Taylor’s frozen-flow hypothesis incorrectly predicts the spectral bandwidths to be zero, and the local wavenumber model significantly under-predicts the spectral bandwidths. An analytical example is formulated to illustrate the present model, and datasets from direct numerical simulations of turbulent channel flows are used to validate this model. The present statistical model is also discussed in terms of the dominating processes of temporal decorrelation in turbulent flows.

Turbulent flow in channels with gap is still a challenge for engineering. The velocity fluctuations and the appearance of coherent structures have been playing a major role in forced convective heat transfer process between the warm tube walls and the cooling fluid. In this paper numerical simulation of non-isothermal turbulent flow in a rectangular channel containing only one tube was performed. Unsteady Reynolds Navier-Stokes along with the energy equation were applied to model the problem. In order to overcome the closure problem the turbulence was modeled applying SAS model. The simulated geometry consists of a heated cylindrical tube, placed 10 mm from the bottom wall of the rectangular duct. The fluid flows externally to the heated tube along the mainstream direction. The channel’s length was based on the tube diameter, yielding L/D equal 80.