Two observations drawn from a thoroughly validated direct numerical simulation of the canonical spatially developing, zero-pressure gradient, smooth, flat-plate boundary layer are presented here. The first is that, for bypass transition in the narrow sense defined herein, we found that the transitional-turbulent spot inception mechanism is analogous to the secondary instability of boundary-layer natural transition, namely a spanwise vortex filament becomes a Λ vortex and then, a hairpin packet. Long streak meandering does occur but usually when a streak is infected by a nearby existing transitionalturbulent spot. Streak waviness and breakdown are, therefore, not the mechanisms for the inception of transitional-turbulent spots found here. Rather, they only facilitate the growth and spreading of existing transitional-turbulent spots. The second observation is the discovery, in the inner layer of the developed turbulent boundary layer, of what we call turbulent-turbulent spots. These turbulent-turbulent spots are dense concentrations of small-scale vortices with high swirling strength originating from hairpin packets. Although structurally quite similar to the transitional-turbulent spots, these turbulent-turbulent spots are generated locally in the fully turbulent environment, and they are persistent with a systematic variation of detection threshold level. They exert indentation, segmentation, and termination on the viscous sublayer streaks, and they coincide with local concentrations of high levels of Reynolds shear stress, enstrophy, and temperature fluctuations. The sublayer streaks seem to be passive and are often simply the rims of the indentation pockets arising from the turbulent-turbulent spots.boundary layer | transition | turbulence | direct numerical simulation T he zero-pressure gradient, smooth, flat-plate boundary layer (ZPGSFPBL) is the simplest viscous external flow. It serves as the idealized limiting case and calibration benchmark of atmospheric and oceanic planetary boundary layers as well as aeronautical, maritime, and automotive boundary-layer flows. For over 60 y since the work by Theodorsen (1), a central theme in fundamental fluid mechanics research has been the search for the constitutive coherent structure in the turbulent ZPGSFPBL, particularly inside the near-wall/inner layer less than ∼100 viscous units away from the plate where the production and dissipation of turbulence kinetic energy reach their peaks (2-6). When the nature of the inner-layer structure and dynamics is thoroughly understood, this understanding can be incorporated in turbulence theory and predictive modeling.Decades of research have produced an apparent consensus view (7-9) that the inner layer, which consists of the buffer layer and the viscous sublayer, of a turbulent ZPGSFPBL is populated by randomly distributed quasistreamwise vortices as well as elongated high-and low-momentum streaks. Streaks are thought to actively participate in a self-sustaining bursting cycle that includes streak generation, lift up, oscil...

A novel approach to identify internal interfacial layers, or IILs, in wall-bounded turbulent flows is proposed. Using a Fuzzy Cluster Method (FCM) on the streamwise velocity component, a unique and unambiguous grouping of the Uniform Momentum Zones is achieved, thus allowing the identification of the IILs. The approach overcomes some of the key limitations of the histogram-based IIL identification methods. The method is insensitive to the streamwise domain length, can be used on inhomogeneous grids, uses all the available flow field data, is trivially extended to three dimensions, and does not need user-defined parameters (e.g. number of bins) other than the number of zones. The number of zones can be automatically determined by an a priori algorithm based on a Kernel Density Estimation algorithm, or KDE. The clustering approach is applied to the turbulent boundary layer (experimental, planar PIV) and channel flow (numerical, DNS) at varying Reynolds numbers. The interfacial layers are characterized by a strong concentration of spanwise vorticity, with the outer-most layer located at the upper edge of the log-layer. The three-dimensional interface identification reveals a streak-like organization; we show that the organization of the IILs is correlated to the underlying wall-bounded turbulent structures.

The implementation and verification of real-fluid effects towards the high-fidelity large eddy simulation of rocket combustors is reported. The non-ideal fluid behavior is modeled using a cubic Peng-Robinson equation of state; a thermodynamically consistent approach is used to convert conserved into primitive variables. The viscosity is estimated by Chung et al.'s method 1, 2 in the supercritical gas phase. In the transcritical liquid phase, a simple, accurate and efficient method to estimate the viscosity as a function of temperature and pressure is proposed. The highly non-linear coupling of the primitive thermodynamic variables requires special consideration in regions of high-density gradients to avoid spurious numerical oscillations. The characterization of the non-linearity of the equation of state identifies the regions of high sensitivity. In these regions, small relative changes in the pressure lead to significant changes in density and/or temperature, therefore, numerical instabilities tend to be amplified in these regions. To avoid non-physical oscillations, a first-order and second-order essentially non-oscillatory (ENO) schemes are locally applied to the flux computation on the faces identified with a dual-threshold relative density sensor. The evaluation of the sensor and capabilities of the non-oscillatory schemes on canonical test cases are presented. Finally, these schemes are used to model two canonical cases.

In this paper, we discuss properties of supercritical and real fluids, following the overarching question: 'What is a supercritical fluid?'. It seems there is little common ground when researchers in our field discuss these matters as no systematic assessment of this material is available. This paper follows an exploratory approach, in which we analyze whether common terminology and assumptions have a solid footing in the underlying physics. We use molecular dynamics (MD) simulations and fluid reference data to compare physical properties of fluids with respect to the critical isobar and isotherm, and find that there is no contradiction between a fluid being supercritical and an ideal gas; that there is no difference between a liquid and a transcritical fluid; that there are different thermodynamic states in the supercritical domain which may be uniquely identified as either liquid or gaseous. This suggests a revised state diagram, in which low-temperature liquid states and higher temperature gaseous states are divided by the coexistence-line (subcritical) and pseudoboiling-line (supercritical). As a corollary, we investigate whether this implies the existence of a supercritical latent heat of vaporization and show that for pressures smaller than three times the critical pressure, any isobaric heating process from a liquid to an ideal gas state requires approximately the same amount of energy, regardless of pressure. Finally, we use 1D flamelet data and large-eddy-simulation results to demonstrate that these pure fluid considerations are relevant for injection and mixing in combustion chambers.

We study the boundary-layer turbulence and freestream turbulence interface (BTFTI), the turbulent spot and freestream turbulence interface (TSFTI), and the laminar boundary-layer and freestream turbulence interface (LBFTI) using direct simulation. Grid spacings in the freestream are less than 1 Kolmogorov length scale during transition. Probability density functions of temperature and its derivatives are used to select the interface identification threshold, corroborated by a vorticity-based method. The interfaces so detected are confirmed to be physical a posteriori by the distinctive quasi-step-jump behavior in the swirling strength and temperature statistics along traverses normal to the BTFTI and TSFTI. No interface-normal inflection is detected across the LBFTI for either swirling strength, temperature, vorticity magnitude, Reynolds shear stress, streamwise velocity, normal velocity, or turbulence kinetic energy. The present direct numerical simulation data thus cast serious doubts on the shear-sheltering hypothesis/theory, which asserts that a subset of freestream fluctuations is blocked by the LBFTI. In the early stage of transition, quasi-spanwise structures exist on the LBFTI. The TSFTI shape is dominated by head prints of concentrated hairpin vortices. Further downstream, the BTFTI geometry is strongly modulated by groves of hairpin vortices (the boundary layer large-scale motions) with a distinct streamwise preferential orientation. Streamwise velocity and turbulence kinetic energy only exhibit minor plateaus (rather than quasi-step-jump) across the BTFTI and the TSFTI. We emphasize that it is more meaningful and important to acquire reproducible and reliable interface-normal statistics prior to considering any plausible substructures and elusive transient dynamics of the BTFTI, TSFTI, and LBFTI.

We report on the implementation of the real fluid capabilities to CharlesX, the in-house, unstructured, large eddy simulation code used at the Center for Turbulence Research at Stanford University. A conceptually distinct implementation was needed for the puremixing and the flamelet/progress-variable (FPV) model combustion case. For the nonreacting simulations, a Newton-Raphson based iterative algorithm is used to determine the temperature from the transported density and energy. For the reacting simulations, an extended flamelet table is used that tabulates the departure functions as well as the compressibility factor. These tabulated parameters are used to correct the transported thermodynamic properties. The real fluid extension to CharlesX was used to investigate a non-reacting and a reacting case. In both of these cases, a second-order essentially nonoscillatory (ENO) schemes is locally applied to the flux computation on the faces identified with a dual-threshold relative density sensor. This avoids spurious oscillations of the numerical solution with limited numerical dissipation. This preliminary work illustrates the capability CharlesX to capture the important physics in a typical rocket engine configuration.

We have performed direct numerical simulations (DNS) of compressible turbulent channel flow at supercritical pressure with top and bottom isothermal walls kept respectively at a supercritical (T top > T pb ) and subcritical temperature (T bot < T pb ), where T pb is the pseudoboiling temperature. The DNS are conducted using a high-order discretization of the fully compressible Navier-Stokes equations in conservative form closed with the Peng-Robinsion (PR) state equation. Bulk density is adjusted to obtain a bulk pressure of approximately p b = 1.1p cr where p cr is the critical pressure of the working fluid. Top-to-bottom temperature differences investigated are ∆T = 5 K, 10 K, and 20 K, where T top/bot = T pb ± ∆T /2; buoyancy effects are neglected. Varying ∆T modifies the average location of pseudophase change from y pb /h = −0.23 (∆T = 5 K) to 0.89 (∆T = 20 K), where h is the channel half-height and y = 0 the centerline position. Real-fluid effects cause visible deviations from classical scaling laws in the mean velocity profile. Enstrophy generation due stretching and tilting decreases with ∆T . The proximity to the pseudotransitioning layer inhibits the intensity of the velocity fluctuations, while enhancing the density and temperature fluctuations. Conditional probability analysis reveals that the sheet of fluid undergoing pseudophase change is characterized by a dramatic reduction in the kurtosis of density fluctuations and becomes thinner as ∆T is increased. Instantaneous visualizations show dense fluid ejections from the pseudoliquid viscous sublayer, some reaching the channel core, causing positive values of density skewness in the respective buffer-layer region (vice versa for the top wall).

It remains an open problem as to whether droplets exist during the injection of liquid oxygen in a rocket combustion chamber at pressures higher than the oxygen critical pressure. While surface tension and thus droplets vanish in a pure fluid, a mixture state may exhibit a higher critical pressure, possibly reintroducing surface tension. In this paper, we address this problem by analyzing the non-premixed flamelet representation of combustion under liquid propellant rocket engine (LRE) conditions. The turbulent flames in a LRE can be thought of as being composed of elementary 1D laminar counterflow diffusion flames. The physically possible configurations for a given rocket operating condition, corresponding to the boundary conditions of the 1D flamelet problem, are captured by variation of the strain rate. For an exemplary supercritical operating condition (p = 7 MPa, Tin,LOX = 120 K, Tin,H2 = 295 K) we show that, despite local mixing, the fluid never reaches a multiphase state from equilibrium combustion to quenching. The transition from supercritical liquid oxygen to an ideal gas state is found to occur in what is essentially a pure fluid process; real fluid mixing only occurs among LOX and water with a water mass fraction < 3% before the ideal gas transition. Representing the mixing trajectories of each flamelet in a reduced pressure-reduced temperature diagram allows to capture all physical mixture states of a configuration in a single plot. This approach furthermore allows to intuitively assess changes in operating conditions with respect to critical-state conditions.