Session R08: Computational Fluid Dynamics: RANS Modeling (5:00pm - 5:45pm CST)

LES-informed RANS Predictions

Within the context of data-driven turbulence modeling and uncertainty quantification for Reynolds-averaged Navier-Stokes (RANS) simulations, several studies have used a framework that introduces perturbations to the magnitude, orientation, and anisotropy of the Reynolds stress tensor. The aim of the present work is to investigate if the number of perturbation parameters in this framework could be reduced, by verifying if using accurate knowledge on the Reynolds stress orientation and anisotropy in the transport equations for a RANS turbulence model automatically provides a better prediction for the Reynolds stress magnitude. We focus on a bluff body flow representative of the flow around a high-rise building, and solve the transport equations for the k-$\omega$ SST model using a normalized Reynolds stress anisotropy tensor and mean velocity field from a large-eddy simulation (LES) as input. We analyze the discrepancies between the predicted and LES turbulence kinetic energy, and we investigate calibration of a double-scale version of the k-$\omega$ SST model to reduce the discrepancies.   

Variational Data Assimilation for Incompressible RANS Closure Models

Flow simulations based on the Reynolds-averaged Navier-Stokes (RANS) equations have attractively low requirements for computational cost, but only provide averaged quantities and depend on turbulence modelling. Recently, data driven closure models have been proposed to overcome some weaknesses of existing methods. Our approach is based on performing variational data assimilation (DA) on the eddy viscosity using sparsely distributed velocity reference data. In particular, we enhance the result of existing turbulent viscosity models through DA. The discrete adjoint method is used for DA since its computational cost is independent of the large number of parameters. A gradient based optimizer is then applied to modify the eddy viscosity, such that the reference data is matched as close as possible. A fully coupled solver is extended to solve the adjoint equations. Some simplifications are applied when computing the adjoint gradient to reduce the computational effort. The novelty of our approach is the way of computing the adjoint gradient without a costly finite difference approach. Due to the high dimensionality of the problem, many evaluations of the forward problem can thus be avoided. We demonstrate our method by applying it to the incompressible stationary periodic hill case using the k-epsilon model and literature LES reference data. An emphasis is put on the analysis of different reference data types, reference data distributions, and regularizations.   

The Response of Modified Turbulent Pipe-Flow Targeting Distinct Fourier Modes at a Range of Reynolds Numbers

The response and recovery of turbulent pipeflow undergoing modification of its cross-sectional structure is examined using Reynolds-Averaged-Navier-Stokes (RANS) realizable k - $\varepsilon $ model at Re$=$500-158600. The three-dimensional modification of the pipe cross-section was by targeting three distinct Fourier modes following the experiments of Van Buren et al. (2017). The reference case involved flow in a circular cross-section pipe with a length of 220D. The inserts are added to the pipe for the three cases based on a modified cross-section with an azimuthal Fourier mode of m$=$3 (Case I), m$=$15 (Case II), and m$=$3$+$15 (Case III). Preliminary results show that there exists a long-lasting effect on the flow, and the flow recovery is slow with the equilibrium not achieved at least 20D downstream of the insert at high Re. This reduces as the flow Re drops. Moreover, the combined Fourier mode insert (Case III) delays the flow recovery substantially, which is also accompanied by a smaller pressure drop compared to the other inserts. The effect of Reynolds number on the response and recovery is further examined for Case III. This study will expand into the effects on flow symmetry and unsteady behaviour for the dominant inserts.   

Evaluation of entropy production in a vacuum objective supersonic ejector via unsteady Reynolds-averaged Navier-Stokes simulation

Unsteady Reynolds-averaged Navier-Stokes simulations are performed to investigate entropy production in a supersonic air ejector operating with three area ratios and at two stagnation pressure ratios. Local entropy generation is dominated by viscous dissipation in the flow at locations corresponding to the flow unsteadiness, shear layer instabilities, recirculation zones and shock structures. Particularly, the flow structure inside the primary nozzle is found to have a significant impact on the ejector irreversibility. As the area ratio is increased, the nozzle flow transmutes from under-expanded to over-expanded. The over-expanded nozzle generates significantly higher entropy via a shock-induced flow separation inside the primary nozzle with the major production at the shock front when compared with the under-expanded nozzle wherein the major entropy production is near the flow separation zones. A higher-pressure ratio if dominated by the unsteady flow separation zones in the diffuser generate more irreversibilities when compared with the lower stagnation pressure ratio wherein the unsteady flow features are absent during the steady operation. These findings provide insight into identifying the main sources of loss and perhaps their minimization.   

Normalized Reynolds Stress Closure

The Reynolds Averaged Navier-Stokes (RANS-) equation together with a closure model for the Reynolds stress govern the mean velocity field of a Newtonian fluid in inertial and in non-inertial frames. Mixing and production of the turbulent kinetic energy and the turbulent dissipation depend on the Reynolds stress and the Cauchy stress. Over the past 143 years, J. Boussinesq and many others have assumed that the Reynolds stress and the Cauchy stress are both frame insensitive. Unfortunately, multiple versions of the eddy viscosity model are still being used to regress experimental data. It is noteworthy that the Coriolis Theorem associated with equivalent motions predicts that the strain rate is frame insensitive; and, the normalized Reynolds stress is frame sensitive. This presentation will review the underlying physical and mathematical principles that define the Cauchy stress and the Reynolds stress. The impact of the Coriolis acceleration on the anisotropy of the turbulent kinetic energy cascade for homogeneous decay was unexpected.   

Numerical Study of the Particles Dispersion Expelled during Breathing, Coughing, and Sneezing in Public Indoor Environments

Transmission of most respiratory diseases is mainly attributed to the airborne dispersion of viruses through the droplets that are produced in the respiratory system of infected individuals. Understanding how these droplets disperse in indoor environments is key for the detection and cleaning processes of living spaces and would be the input to identify which areas have the highest concentration of contaminated particles. This work aims to numerically study the saliva droplet dispersion and transport in public indoor environments by using Computational Fluid Dynamics (CFD). Unsteady RANS simulations with the Euler-Lagrange approach were implemented for three different time-dependent velocity profiles that represent real human breathing, coughing, and sneezing. The Lagrange approach was used to track the saliva droplets through the flow field. The results show a considerable impact of human body heat flux on the exhaled droplets dynamics and the local velocity field close to the human body. The heat plumes appeared to prevent small saliva droplets from depositing human body surfaces. Besides, the strength of air conditioning velocity may have a significant effect on the saliva droplets trajectories.