Session HF: Separated Flows II

Experimental study of the flow over a backward-facing rounded ramp

The backward-facing rounded ramp (BFR) is a very simple geometry leading to boundary layer separation, close to the backward facing step (BFS) flow. The main difference with the BFS flow is that the separation location depends on the incoming flow while it is fixed to the step edge for the BFS flow. Despite the simplicity of the geometry, the flow is complex and the transition process still has to be investigated. In this study we investigate the BFR flow using time-resolved PIV. For Reynolds number ranging between 300 and 12 000 we first study the time averaged properties such as the positions of the separation and reattachment, the recirculation length and the shear layer thickness. The time resolution also gives access to the characteristic frequencies of the time-dependant flow. An appropriate Fourier filtering of the flow field, around each frequency peak in the global spectrum, allows an investigation of each mode in order to extract its wavelength, phase velocity, and spatial distribution. We then sort the spectral content and relate the main frequencies to the most amplified Kelvin-Helmholtz instability mode and its harmonics, the vortex pairing, the low frequency recirculation bubble oscillation and the interactions between all these phenomena.   

Forcing the shear layer of a backward-facing step flow using DBD plasma actuator

The Kelvin-Helmholtz convective instability is the key mechanism in a backward facing step (BFS) flow. The natural flow selects the most amplified mode below the cutoff wavenumber. We introduce time-dependant perturbations inside the shear layer using a DBD plasma actuator. The perturbation is a pulsed low velocity spanwise jet, parallel to the mean flow. It is introduced inside the boundary layer, just upstream separation. Using time-resolved visualizations we show that forcing with a given frequency can have a dramatic effect on the spectral content of the flow and its global properties. The modification of the BFS flow depends strongly on the value of the frequency with respect to the natural frequency. Using 4Hz Particle Image Velocimetry (PIV) we study the evolution of time-averaged properties such as the recirculation length or the shear layer thickness. Recording the forcing perturbation signal together with the PIV acquisition signal we achieve phase-average reconstruction based on the forcing frequency. The phase-averaged time series give access to the wavelength, phase velocity and the spatial distribution of the vortex shedding. It also allows us to build the stability diagram of the shear layer.   

Sensitivity of an Asymmetric 3D Diffuser to Vortex Generator Induced Inlet Condition Perturbations

Experiments were performed to investigate the flow in an asymmetric 3D diffuser that is highly sensitive to inlet condition perturbations. Previous velocity field measurements showed that in its standard configuration this diffuser develops a stable three-dimensional separation bubble. However, weak secondary flows induced by tunable dielectric barrier discharge plasma actuators in the inlet of the diffuser resulted both in dramatic improvements and degradations in the diffuser's performance. Two configurations of vortex generators were selected based on their having analogous effects on the diffuser's pressure recovery. These vortex generators were placed in the inlet of the diffuser and magnetic resonance velocimetry was used to obtain three-dimensional velocity data. The data reveal markedly different separation bubble structures, with the improved pressure recovery corresponding to a reduced reversed flow area. Additionally, the vortex generators which improve the diffuser's performance create a more uniform velocity profile at the end of the expanding section, while the other configuration facilitates the persistence of a high velocity core through the diffuser's outlet.   

Low-frequency dynamics in supersonic shock turbulent boundary layer interactions

Supersonic shock/turbulent boundary layer interactions are studied using a large LES database. Consistent with previous experimental and numerical findings, the simulations indicate the occurrence of low-frequency dynamics, mainly related to the oscillation of the reflected shock foot. In the proximity of the latter, the wall pressure spectra exhibit substantial spectral content at frequencies at least two orders of magnitude smaller that those typical of the incoming boundary layer. Such trend becomes more pronounced for the strong interactions, with significant flow separation. The analysis of the low- and high-pass filtered flow fields allows to isolate two basic `modes': i) a low-frequency `breathing' mode of the separation bubble/reflected shock system, which is weakly coupled with the upstream flow; and ii) a high-frequency mode, which is dominated by the system response to the incoming boundary layer turbulence. The analysis of the Koopman modes (eigenmodes of the Koopman operator associated with snapshots of the flow field) confirms such scenario, and shows that the wall pressure signature in the low-frequency range is primarily affected by the breathing mode.   

Global stability analysis of supersonic shock/turbulent boundary layer interactions

The global stability of supersonic shock/turbulent boundary layer interactions is investigated in a wide range of shock intensities. The analysis relies on linearization of the governing equations about the mean turbulent flow obtained from LES calculations. The global stability analysis relies on the Arnoldi iterative method, whose convergence requires suitable preliminary smoothing of the base flow. Results of two-dimensional stability analysis highlight the occurrence of a single exponentially growing, zero-frequency mode, also observed by previous investigators. The dominance of such mode in the linearized flow dynamics is confirmed by three-dimensional calculations with slightly disturbed initial conditions. However, the global stability analysis also shows the occurrence of a marginally stable oscillatory mode, whose characteristic frequency is close to those found in LES. Such mode becomes less stable as the incident shock strength increases, and it features a typical `breathing' motion of the recirculation bubble, whereby fluid is periodically entrapped and released, also driving the motion of the reflected shock.   

DNS of Turbulent Boundary Layer Subject Strong Adverse Pressure Gradient

Direct Numerical Simulations of spatially evolving turbulent boundary layers with prescribed strong adverse pressure gradients are performed. The driven force behind this investigation is to analyze the interaction between the inner and outer layers in adverse pressure gradient with eventual separation. A method for prescribing realistic turbulent velocity inflow boundary conditions is employed. The approach is based on the rescaling-recycling method proposed by Lund et al. (1998) and the dynamic multi- scale method developed recently by Araya et al. (2009). The standard rescaling process requires prior knowledge about how the appropriate velocity and length scales are related between the inlet and recycle stations (e.g. classic scaling laws). Here a dynamic approach is proposed in which such information is deduced dynamically by involving an additional plane located between the inlet and recycle stations. The approach also distinguishes between the inner and outer regions of the boundary layer and enables the use of multiple velocity scales. This flexibility allows applications to boundary layer flows with arbitrary pressure gradients.   

DNS study of a separation bubble in a turbulent boundary layer

Direct numerical simulations (DNSs) of a separated turbulent boundary layer have been performed in which blowing and suction are imposed at the upper boundary in order to produce a separation bubble. The inlet data are prescribed by DNSs of a zero-pressure gradient turbulent boundary layer with the rescaling-recycling method. The Reynolds numbers at the inlet are set to be $Re_\theta=300$ and 600 where $Re_\theta$ is the Reynolds number based on the free stream velocity and the momentum thickness. Particular attention is given to the difference between large and small separation bubbles and also the Reynolds-number dependence. The present results indicate that large-scale structures of velocity and pressure fluctuations are more dominant in a large separation bubble than in a small separation bubble, which becomes more apparent with increasing Reynolds number. The relationship between the large-scale structures and the oscillatory behavior of the detachment and reattachment regions will also be discussed.   

Study of interacting shear layers in the formation of grid turbulence

Grid turbulence has been studied for many years as a method of producing approximately homogeneous isotropic turbulence to test classical theories. These studies have concentrated on the region well downstream of the grid since this is where the turbulence is supposed to take its classical form. The way in which the turbulence develops to this state from what is essentially the merged wakes of a series of rods is not yet well understood. This has particular implications as to the extent to which the turbulence downstream is independent of the grid's geometry. A related question is to what extent inhomogeneity due to the grid persists downstream. In order to study these questions, we fully map the first 15 mesh lengths in the wake of two different static biplanar grids using 2D high-speed PIV. This is done with 3 cameras to cover a large field of view in the downstream direction with adequate resolution. Measurements are taken for each grid at three different Reynolds numbers ($Re_{M}$ up to 16 000) and on two different planes with respect to the grid: half way between two rods and exactly behind a rod.   

PIV Measurements of the Near-Wake behind a Fractal Tree

An experimental study of turbulent flow in the wake of a fractal-like tree has been carried out. Fractals provide the opportunity to study the interactions of flow with complicated, multiple-scale objects, yet whose geometric construction rules are simple. We consider a pre-fractal tree with five generations, with three branches and scale- reduction factor 1/2 at each generation. Its similarity fractal dimension is $D_s \sim 1.585$. Experiments are carried out in a water tunnel with the ability of index- matching, although current measurements do not utilize this capability yet. The incoming velocity profile is designed to mimic the velocity profile in a forest canopy. PIV measurements are carried out on 14 horizontal planes parallel to the bottom surface. Drag forces are measured using a load cell. Mean velocity and turbulence quantities are reported at various heights in the wake. Mean vorticity contours on the upper planes show signatures of the smaller branches, although the wakes from the smallest two branches are not visible in the data possibly due to rapid mixing. Interestingly, their signatures can be observed from the elevated spectra at small scales. Momentum deficit in the wake profiles and drag forces are compared. The results from this experiment also serve as database against which to compare computer simulations and models.   

PIV Analysis Comparing Aerodynamic Downforce Devices on Race Car in Water Tunnel

There have been claims that the rear wing on the NASCAR Car of Tomorrow (COT) race car causes lift in the condition where the car spins during a crash and is traveling backwards down the track at a high rate of speed. When enough lift is generated, the race car can lose control and even fly off of the track surface completely. To address this concern, a new rear spoiler was designed by NASCAR to replace the wing and prevent this dangerous condition. Flow characteristics of both the rear wing and the new spoiler are qualitatively analyzed using particle image velocimetry (PIV). The experiment is done in a continuous flow water tunnel using a simplified 10\% scale model COT. Flow structures are identified and compared for both the wing and spoiler. The same conditions are also reviewed when the car is traveling backwards as it might during a crash. The cause of the lift generated by the rear wing when in reverse is shown.