Session H28: Computational Fluid Dynamics: LES, DNS, Hybrid RANS/LES I

Subgrid scale modeling of cavitation inception

Unresolved pressure fluctuations at the subgrid scale (SGS) level of any LES/RANS computation restrict the prediction of cavitation inception as SGS low pressures are ignored. Statistical relations of the SGS turbulence, the pressure fluctuations, and cavitating behavior of nuclei in such turbulence are used to build up a model of cavitation inception in the SGS. The fluctuating pressure history of Lagrangian nuclei of different sizes is used to solve the Rayleigh-Plesset equation to simulate the dynamics of bubbles. The evolution of the bubbles is analyzed, and a cavitation criterion is used to determine inception. Cavitation event statistics are built and tabulated for a range of SGS Reynolds numbers (????_??), nuclei size, TKE dissipation rate, and mean pressure. The table is used to estimate the cavitation event rate in each cell of the CFD solution. Two cases of homogeneous isotropic turbulence (HIT) at ????_??=240 and 324 have been studied where the pressure statistics predicted by the SGS model with LES are validated against DNS. A high ????_?? HIT flow is simulated using LES and cavitation events are compared against experimental data. The inception model successfully predicts the inception pressure and the cavitation rates in the flow.   

Eigenspace perturbations for subgrid modeling in large-eddy simulations.

The method of eigenspace perturbations has been developed to characterize structural uncertainty in turbulence closure models. In the framework of large eddy simulations, the eigenvalues and eigenvectors of the modeled subgrid stress tensor can be used to drive realizable perturbations towards the limiting states of turbulence anisotropy thus creating a family of models that represent solution envelopes around a baseline prediction. In the present work, we apply the principles of eigenspace perturbations and information from the resolved scales with the goal of improving the prediction of a baseline subgrid model. A target stress that is computable on the grid is identified; then, perturbations are applied to the modeled subgrid tensor towards the eigenspace of the target tensor, with a view to correcting the shape and orientation of the baseline subgrid tensor. The approach involves minimal computing overhead, and is self-contained, only using information already generated as part of the simulation.   

Not Participating

We developed an enrichment wall-model for the spectral element method (SEM) for wall-modeled large-eddy simulations (WMLES) of turbulent flows. Ordinarily, higher-order methods, such as SEM, require significant mesh refinement near the wall in order resolve the turbulent boundary layer without log-layer mismatch or spurious oscillations due to the large gradients in the boundary layer. To avoid this issue, as is done in traditional WMLES, we take the velocity at matching points away from the wall and fit it with an analytical wall function to compute the shear stress on the wall and apply it to the flow at the boundary. Then, in order to increase the fidelity of the model, we incorporate the wall function as an enrichment term in the solution representation. This allows the near wall behavior to be captured without unphysical oscillations or mismatches at the wall and improves the accuracy of the overall solution. We discuss the procedure for integrating the enrichment wall-model in the SEM and its implementation in the SEM CFD solver Nek5000. The method is demonstrated in the context of RANS and WMLES for canonical wall-bounded turbulent flows.   

Physics-Guided Neural Networks for Reconstructing High-resolution Turbulent Flows

Direct numerical simulation (DNS) of turbulent flows is compu- tationally expensive and cannot be applied to flows with large Reynolds numbers. Large eddy simulation (LES) is an alternative that is computationally less demanding, but is unable to capture all of the scales of turbulent transport accurately. Our goal in this work is to build a new data-driven methodology based on super- resolution techniques to reconstruct DNS data from LES predictions. We leverage the underlying physical relationships to regularize the relationships amongst different physical variables. We also introduce a hierarchical generative process and a reverse degra- dation process to fully explore the correspondence between DNS and LES data. We demonstrate the effectiveness of our method through a single-snapshot experiment, and a cross-time experi- ment. The results confirm that our method can better reconstruct high-resolution DNS data over space and over time.   

Leveraging LES data to investigate pressure peaks on high-rise buildings in neutral atmospheric boundary layers

High-rise buildings in a neutral Atmospheric Boundary Layer (ABL) can experience significant negative pressure peaks near the upper-rear corner of their leeward facade. Several wind tunnel experiments have investigated these peak phenomena through high-resolution pressure measurements, and the associated pressure magnitude has been characterized successfully. However, the physics behind the formation of these peaks is not yet fully understood.

This work aims to validate an LES simulation of the flow about a high-rise building against experimental data and then leverage it to understand the physics driving pressure peak events. A synthetic, divergence-free turbulence generator is used to achieve an incoming ABL for the LES that matches experimental data. The pressure predicted over the surface of the building by the LES agrees well with the experimental data in terms of pressure mean, rms, and peak values.

The validated LES simulation is finally used to study peak events for the case of an incoming ABL with a wind direction of 20°. The instantaneous flow field in the vicinity of peak events is visualized and it is shown how eddies sheddeded from the shear layers on top of the roof of the building interact with edge separations to form the observed pressure peaks.   

LES Scale Enrichment in spatially-decaying isotropic turbulence

High Reynolds number flows, common in engineering applications, limit the bandwidth of scales available to modern computers when solving the full nonlinear governing equations. We have shown that second-order subgrid scale statistics are accurately reconstructed using spatially- and spectrally- local Gabor modes. The method relies on a quasi-homogeneous assumption and expands the sub-grid velocity field as a local sum over Gabor modes which evolve dynamically with the large-scale field.

We have applied the method to spatially decaying isotropic turbulence. Our investigation is two-fold. First, we are interested in exploring the limit of infinite turbulence intensity at the inlet. This is difficult to achieve in the laboratory and, to our knowledge, no numerical investigation has been conducted. Initial results indicate that scaling exponents for the integral length scale and turbulence intensity are strong functions of inlet intensity due to the increased role of turbulent transport to balance energy dissipation. Second, we will evaluate the ability of the enrichment method to handle the high-turbulence intensity regime and represent the subgrid field consistently; we have seen good results for moderate intensities.   

Using Large-Eddy Simulation to Study Water Tunnel Confinement Effects for a Marine Propeller in Crashback

We discuss water tunnel confinement effects for a marine propeller in the off-design mode of operation known as crashback. In crashback, the propeller rotates in the reverse direction while the vessel moves in the forward direction, yielding highly unsteady loads. The experiments by Jessup et al. [25th Symposium of Naval Hydrodynamics (2004)] studied marine propeller David Taylor Model Basin (DTMB) 4381 in the 36-inch variable pressure water tunnel (VPWT). In the crashback mode of operation, these experiments acknowledge a large difference between the propeller loads measured in the water tunnel and previous tow tank (open water conditions) experiments. Water tunnel confinement effects in the open-jet test section are hypothesized to be one of the contributors. Large-eddy simulation is used to study the confinement effects of the 36--inch VPWT for the marine propeller (DTMB) 4381 in both forward and crashback modes of operation. The simulations use a novel, unstructured grid methodology, developed by Horne & Mahesh [J. Comput. Phys (2019) 376:585-596].   

Optimizing speed and accuracy of the Partially Averaged Navier-Stokes Method

The Partially Averaged Navier-Stokes (PANS) formulated by Girimaji et al. (2003) and Girimaji (2006) is being increasingly used to solve complex and high-Reynolds number industrial flows. This approach, which belongs to the bridging hybrid RANS-LES methods, is designed to resolve a part of the turbulence spectrum adjusting seamlessly from RANS to DNS (Direct Numerical Simulation).  Various researchers have contributed to the PANS method following two main paths: (a) applying the basic derivation rules on different RANS models e.g. k-w, k-e etc. thus providing the background PANS model and (b) developing calculation approaches for the main model resolution parameter, namely, the unresolved to total kinetic energy ratio (fk=ku/ktot). In this work, we analyse the performance of different ‘fk’ approaches for the external car aerodynamics on large computational meshes. Three different approaches for specifying fk are considered in this work: (1) using the constant value over the entire computational domain; (2) a dynamic update of fk per computational cell at every time step following the formula of Girimaji and Abdul-Hamid (2005); (3) using the most complex approach that solves additional equation for so called scale-supplying variable (Basara, Pavlovic, Girimaji 2018). Calculations are also performed by using a time activation ramp for different methods to balance speed and accuracy. Measurements, but also previous RANS calculations, are used as a reference point to the present calculations.   

Temporal Large-Eddy Simulation based on Direct Deconvolution

We propose an approach for Temporal Large-Eddy Simulation (TLES) with direct deconvolution. In this model the non-filtered fields are recovered using a direct deconvolution given by the differential form of the filter operator. Closure is obtained by an evolution equation of the temporal residual-stress tensor, which is analytically derived from the relation of the filtered and the non-filtered fields. A secondary regularization term based on selective frequency damping is employed. The Temporal Direct Deconvolution Model was implemented in the spectral element code Nek5000 to simulate different test cases such as homogeneous isotropic turbulence at Re_λ=190, turbulent channel flow at Re_τ=180, Taylor-Green vortex at Re=3000 and flow over a periodic hill at Re=10595. The following simulation data are discussed: Energy spectra of the homogeneous isotropic turbulence as well as mean flow, root-mean-square of the velocity fluctuations, and the Reynolds stresses of turbulent channel and periodic hill flows. The results demonstrate an improvement compared to no-model solutions, while the computational cost is reduced dramatically compared to direct numerical simulation. Furthermore, an analysis of the relation between temporal and spatial filtering is presented.   

Evolution of Coaxial Synthetic Jet with Different Velocity Ratios

In this study, we propose and demonstrate two synthetic jets which are arranged coaxially with 0^o orientation angle to form a coaxial synthetic jet (CSJ). Two diaphragms are mounted independently with two cavities such that their oscillating amplitude and frequency can be varied separately to achieve various velocity ratios (V_r), which is the ratio of spatial and time-averaged velocity at the inner orifice exit to that at the annular orifice exit. Both the diaphragms are oscillated at 50 Hz frequency, and the amplitude is varied such that various V_r of 0.5, 1, 2, and 4 can be achieved. The CSJ is associated with two vortex rings: (a) single toroidal inner vortex ring (VR_i) generated from the inner orifice, and (b) double toroidal annular vortex ring (VR_a) generated from the annular orifice. The effect of different V_r along with the effect of absolute velocity for a constant V_r on the evolution and interaction of these vortex rings, and the effect of this interaction on the flow field of the CSJ is studied. For V_r = 0.5, as the velocity output of the annular jet is twice the inner jet, the diameter of the outer ring of the VR_a is large as compared to V_r = 1, and 2 resulting in more azimuthal instability making the jet wider. Also, the breaking of the VR_i into streamwise vortices near the orifice exit contributes in widening the jet for V_r = 0.5. For V_r = 1, the VR_i advances faster in the downstream direction as compared to the VR_a due to the presence of low-pressure region at the core of the outer ring of the VR_a generated in the previous cycle. This makes the CSJ strong and, also the additional annular jet around the inner jet makes the CSJ wide as compared to a single SJ. Overall, the study briefs the nature of CSJ on the basis of different velocity ratios and absolute velocities, which can be helpful for effective utilization of the CSJ in flow control and heat transfer applications.