This study introduces a novel approach to investigate the Reynolds analogy in complex flow scenarios. It is shown that the total mechanical energy $\mathit {B}$, viz. the sum of kinetic energy and pressure work, and the field $\Gamma =\theta ^2/2$ (where $\theta$ is the transported passive scalar) are governed by two equations that are similar in form, when time-averaged for statistically stationary flows. For fully developed channel flows the integral energy balance links the mean bulk velocity and scalar with the volume averages of the respective dissipation rates, allowing the assessment of the Reynolds analogy in terms of the dissipation fields. This approach is tested on direct numerical simulation data of rough-wall turbulent channel flow at two different roughness Reynolds numbers, namely $k^+=15$ and $k^+=90$. For a unit Prandtl number, the same qualitative behaviour is observed for the mean wall-normal distributions of the budget-equation terms of $B$ and $\Gamma$, the latter being larger than the corresponding terms in the mechanical-energy budget. The Reynolds decomposition of the flow into temporal mean and stochastic parts reveals that roughness primarily affects the mean-flow dissipation. For the $k^+=90$ case, the analysis shows that attached-flow and high-shear regions dominate the integral mean scalar and momentum transfer and exhibit the greatest differences between the mean mechanical and scalar dissipation rates. In contrast, well-mixed regions, sheltered by large roughness elements, contribute similarly and minimally to the integral scalar and momentum transfer.

The large-eddy simulation technique was used to investigate the dynamics of unsteady flow separation on a flat-plate turbulent boundary layer. The unsteadiness was generated by imposing an oscillating, wall-normal velocity profile at the top of the computational domain, and a range of reduced frequencies ($k$), from a very rapid flutter-like motion to a slow quasi-steady oscillation, was studied. Ambrogi et al. (J. Fluid Mech., vol. 945, 2022, A10) showed that the reduced frequency greatly affects the transient separation process, and at a frequency $k=1$, the separation region became unstable and was advected periodically out of the domain. In this paper, we discuss the causes of the observed advection process and the effects of the unsteadiness on the second moments. The time evolution of turbulent kinetic energy, for instance, reveals that an advection-like phenomenon is also present at a very low reduced frequency, but its dynamic behaviour is completely different from that of the intermediate frequency ($k=1$). At the intermediate frequency the entire recirculation region is advected downstream, keeping its shape. The advected structure is rotational in nature, and moves at constant speed. In contrast, in the low-frequency case the advected fluid originates at the reattachment point, and the structure is shear-dominated. Particle pathlines reflect the fact that the flow at the low frequency is quasi-steady-state, but show peculiar differences at the intermediate frequency, in which the flow response to the freestream forcing depends on the particle positions in the wall-normal direction.

Separating turbulent boundary layers over smooth and rough flat plates are studied by large-eddy simulations. A suction–blowing velocity distribution imposed at the top boundary of the computation domain produces an adverse-to-favourable pressure gradient and creates a closed separation bubble. The Reynolds number based on the momentum thickness and the free-stream velocity before the pressure gradient begins is 2500. Virtual sand grain roughness in the fully rough regime is modelled by an immersed boundary method. Compared with a smooth-wall case, streamline detachment occurs earlier and the separation region is substantially larger for the rough-wall case, due to the momentum deficit caused by the roughness. The adverse pressure gradient decreases the form drag, so that the point where the wall stress vanishes does not coincide with the detachment of the flow from the surface. A thin reversed-flow region is formed below the roughness crest; the presence of recirculation regions behind each roughness element also affects the intermittency of the near-wall flow, so that upstream of the detachment point the flow can be reversed half of the time, but its average velocity can still be positive. The separated shear layer exhibits higher turbulent kinetic energy (TKE) in the rough-wall case, the growth of the TKE there begins earlier relative to the separation point, and the peak TKE occurs close to the separation point. The momentum deficit caused by the roughness, again, plays a critical role in these changes.

When hovering over sandy terrain, the rotor of helicopters generates a downward jet that induces resuspension of dust and debris. We investigate the mechanisms that govern particle resuspension in such flow using an Eulerian–Lagrangian approach based on large-eddy simulation of turbulence. The wake generated by the helicopter is modelled as a vertical impinging jet, to which a sequence of periodically forced azimuthal vortices is superposed. The resulting flow field provides a unique range of flow scales with which the particles can interact. Downstream of the impingement region, layers of negative azimuthal vorticity (secondary vortices) form on the upwash side of the primary azimuthal (large-scale) vortices. These layers then detach from the surface together with the near-wall (small-scale) vortices populating the wall-jet region. We show how the dynamics of sediments is governed by its interaction with these structures. After initial lift off from the impingement surface, particles accumulate in regions where near-wall vortices roll around the impinging azimuthal vortex, forming rib-like structures that either propel particles away from the azimuthal vortex or entrap them in the shear layer between the azimuthal and secondary vortices. We demonstrate that these trapped particles are more likely to reach the outer flow region and generate a persistent cloud of airborne particles. We also show that, in a time-averaged sense, particle resuspension and deposition fluxes balance each other near the impingement surface.

We propose a new length scale as a basis for the modelling of subfilter motions in large-eddy simulations (LES) of turbulent flow. Rather than associating the model length scale with the computational grid, we put forward an approximation of the integral length scale to achieve a non-uniform flow coarsening through spatial filtering that reflects the local, instantaneous turbulence activity. Through the introduction of this grid-independent, solution-specific length scale it becomes possible to separate the problem of representing small-scale turbulent motions in a coarsened flow model from that of achieving an accurate numerical resolution of the primary flow scales. The formulation supports the notion of grid-independent LES, in which a prespecified reliability measure is used. We investigate a length-scale definition based on the resolved turbulent kinetic energy (TKE) and its dissipation. The proposed approach, which we call integral length-scale approximation (ILSA) model, is illustrated for turbulent channel flow at high Reynolds numbers and for homogeneous isotropic turbulence (HIT). We employ computational optimization of the model parameter based on various measures of subfilter activity, using the successive inverse polynomial interpolation (SIPI) and establish the efficiency of this route to subfilter modelling.

The physics of the roughness sublayer are studied by direct numerical simulations (DNS) of an open-channel flow with sandgrain roughness. A double-averaging (DA) approach is used to separate the spatial variations of the time-averaged quantities and the turbulent fluctuations. The spatial inhomogeneity of velocity and Reynolds stresses results in an additional production term for the turbulent kinetic energy (TKE) – the ‘wake production’; it is the excess wake kinetic energy (WKE), generated from the work of mean flow against the form drag, that is not directly dissipated into heat, but instead converted into turbulence. The wake production promotes wall-normal turbulent fluctuations and increases the pressure work, which ultimately leads to more homogeneous turbulence in the roughness sublayer, and to the increase of Reynolds shear stress and the drag on the rough wall. In the fully rough regime, roughness directly affects the generation of the wall-normal fluctuations, while in the transitionally rough regime, the region affected by roughness is separated from the region of intense generation of these fluctuations. The budget of the WKE and the connection between the wake and the turbulence suggest strong interactions between the roughness sublayer and the outer layer that are insensitive to the variation of the outer-layer conditions. Furthermore, the present results may have implications for the relationship between the roughness geometry and the flow dynamics in the region directly affected by roughness.

We performed large-eddy simulations of the flow over a series of three-dimensional (3D) dunes at laboratory scale. The bedform three-dimensionality was imposed by shifting a standard two-dimensional (2D) dune shape in the streamwise direction according to a sine wave. The turbulence statistics were discussed in Part 1 of this article (Omidyeganeh & Piomelli, J. Fluid Mech., vol. 721, 2013, pp. 454–483). Coherent flow structures and their statistics are discussed concentrating on two cases with the same crestline amplitudes and wavelengths but different crestline alignments: in-phase and staggered. The present paper shows that the induced large-scale mean streamwise vortices are the primary factor that alters the features of the instantaneous flow structures. Wall turbulence is insensitive to the crestline alignment; alternating high- and low-speed streaks appear in the internal boundary layer developing on the stoss side, whereas over the node plane (the plane normal to the spanwise direction at the node of the crestline), they are inclined towards the lobe plane (the plane normal to the spanwise direction at the most downstream point of the crestline) due to the mean spanwise pressure gradient. Spanwise vortices (rollers) generated by Kelvin–Helmholtz instability in the separated shear layer appear regularly over the lobe with much larger length scale than those over the saddle (the plane normal to the spanwise direction at the most upstream point of the crestline). Rollers over the lobe may extend to the saddle plane and affect the reattachment features; their shedding is more frequent than in 2D geometries. Vortices shed from the separated shear layer in the lobe plane undergo a three-dimensional instability while being advected downstream, and rise toward the free surface. They develop into a horseshoe shape (similar to the 2D case) and affect the whole channel depth, whereas those generated near the saddle are advected downstream and toward the bed. When the tip of such a horseshoe reaches the free surface, the ejection of flow at the surface causes ‘boils’ (upwelling events on the surface). Strong boil events are observed on the surface of the lobe planes of 3D dunes more frequently than in the saddle planes. They also appear more frequently than in the corresponding 2D geometry. The crestline alignment of the dune alters the dynamics of the flow structures, in that they appear in the lobe plane and are advected towards the saddle plane of the next dune, where they are dissipated. Boil events occur at a higher frequency in the staggered alignment, but with less intensity than in the in-phase alignment.

We performed numerical simulations of dissolved oxygen (DO) transfer from a turbulent flow, driven by periodic boundary-layer turbulence in the intermittent regime, to underlying DO-absorbing organic sediment layers. A uniform initial distribution of oxygen is left to decay (with no re-aeration) as the turbulent transport supplies the sediment with oxygen from the outer layers to be absorbed. A very thin diffusive sublayer at the sediment–water interface (SWI), caused by the high Schmidt number of DO in water, limits the overall decay rate. A decomposition of the instantaneous decaying turbulent scalar field is proposed, which results in the development of similarity solutions that collapse the data in time. The decomposition is then tested against the governing equations, leading to a rigorous procedure for the extraction of an ergodic turbulent scalar field. The latter is composed of a statistically periodic and a steady non-decaying field. Temporal averaging is used in lieu of ensemble averaging to evaluate flow statistics, allowing the investigation of turbulent mixing dynamics from a single flow realization. In spite of the highly unsteady state of turbulence, the monotonically decaying component is surprisingly consistent with experimental and numerical correlations valid for steady high-Schmidt-number turbulent mass transfer. Linearly superimposed onto it is the statistically periodic component, which incorporates all the features of the non-equilibrium state of turbulence. It is modulated by the evolution of the turbulent coherent structures driven by the oscillating boundary layer in the intermittent regime, which are responsible for the violent turbulent production mechanisms. These cause, in turn, a rapid increase of the turbulent mass flux at the edge of the diffusive sublayer. This outer-layer forcing mechanism drives a periodic accumulation of high scalar concentration levels in the near-wall region. The resulting modulated scalar flux across the SWI is delayed by a quarter of a cycle with respect to the wall-shear stress, consistently with the non-equilibrium state of the turbulent mixing.

We performed large-eddy simulations of flow over a series of three-dimensional dunes at laboratory scale (Reynolds number based on the average channel depth and streamwise velocity was 18 900) using the Lagrangian dynamic eddy-viscosity subgrid-scale model. The bedform three-dimensionality was imposed by shifting a standard two-dimensional dune shape in the streamwise direction according to a sine wave. The statistics of the flow are discussed in 10 cases with in-phase and staggered crestlines, different deformation amplitudes and wavelengths. The results are validated qualitatively against experiments. The three-dimensional separation of flow at the crestline alters the distribution of wall pressure, which in turn may cause secondary flow across the stream, which directs low-momentum fluid, near the bed, toward the lobe (the most downstream point on the crestline) and high-momentum fluid, near the top surface, toward the saddle (the most upstream point on the crestline). The mean flow is characterized by a pair of counter-rotating streamwise vortices, with core radius of the order of the flow depth. However, for wavelengths smaller than the flow depth, the secondary flow exists only near the bed and the mean flow away from the bed resembles the two-dimensional case. Staggering the crestlines alters the secondary motion; the fastest flow occurs between the lobe and the saddle planes, and two pairs of streamwise vortices appear (a strong one, centred about the lobe, and a weaker one, coming from the previous dune, centred around the saddle). The distribution of the wall stress and the focal points of separation and attachment on the bed are discussed. The sensitivity of the average reattachment length, depends on the induced secondary flow, the streamwise and spanwise components of the channel resistance (the skin friction and the form drag), and the contribution of the form drag to the total resistance are also studied. Three-dimensionality of the bed increases the drag in the channel; the form drag contributes more than in the two-dimensional case to the resistance, except for the staggered-crest case. Turbulent-kinetic energy is increased in the separated shear layer by the introduction of three-dimensionality, but its value normalized by the plane-averaged wall stress is lower than in the corresponding two-dimensional dunes. The upward flow on the stoss side and higher deceleration of flow on the lee side over the lobe plane lift and broaden the separated shear layer, respectively, affecting the turbulent kinetic energy.

Direct numerical simulations (DNS) of bypass transition due to high-amplitude free-stream turbulence (FST) are carried out for a flat-plate boundary layer. The computational domain begins upstream of the plate leading edge and extends into the fully turbulent region. Thus, there is no ad hoc treatment to account for the initial ingestion of FST into the laminar boundary layer. We study the effects of both the FST length scale and the disturbance behaviour near the plate leading edge on the details of bypass transition farther downstream. In one set of simulations, the FST parameters are chosen to match the ERCOFTAC benchmark case T3B. The inferred FST integral length scale L11 is significantly larger (RL = UL11/ν = 6580) than that employed in previous simulations of bypass transition (RL ≃ 1000). An additional set of simulations was performed at RL = 1081 to compare the transition behaviour in the T3B case with that of a smaller value of FST length scale. The FST length scale is found to have a profound impact on the mechanism of transition. While streamwise streaks (Klebanoff modes) are observed at both values of the FST length scale, they appear to have clear dynamical significance only at the smaller value of RL, where transition is concomitant with streak breakdown. For the T3B case, turbulent spots form upstream of the region where streaks could be detected. Spot precursors are traced to quasi-periodic spanwise structures, first observed as short wavepackets in the wall-normal velocity component inside the boundary layer. These structures are reoriented to become horseshoe vortices, which break down into young turbulent spots. Two of the four spots examined for this case had a downstream-pointing shape, similar to those found in experimental studies of transitional boundary layers. Additionally, our simulations indicate the importance of leading-edge receptivity for the onset of transition. Specifically, higher fluctuations of the vertical velocity at the leading edge of the plate result in higher levels of streamwise Reynolds stress inside the developing boundary layer, facilitating breakdown to turbulence.

The interaction between a zero-pressure-gradient turbulent boundary layer and a pair of strong, common-flow-down, streamwise vortices with a sizeable velocity deficit is studied by large-eddy simulation. The subgrid-scale stresses are modelled by a localized dynamic eddy-viscosity model. The results agree well with experimental data. The vortices drastically distort the boundary layer, and produce large spanwise variations of the skin friction. The Reynolds stresses are highly three-dimensional. High levels of kinetic energy are found both in the upwash region and in the vortex core. The two secondary shear stresses are significant in the vortex region, with magnitudes comparable to the primary one. Turbulent transport from the immediate upwash region is partly responsible for the high levels of turbulent kinetic energy in the vortex core; its effect on the primary stress 〈u′v′〉 is less significant. The mean velocity gradients play an important role in the generation of 〈u′v′〉 in all regions, while they are negligible in the generation of turbulent kinetic energy in the vortex core. The pressure-strain correlations are generally of opposite sign to the production terms except in the vortex core, where they have the same sign as the production term in the budget of 〈u′v′〉. The results highlight the limitations of the eddy-viscosity assumption (in a Reynolds-averaged context) for flows of this type, as well as the excessive diffusion predicted by typical turbulence models.

Direct and large-eddy simulations of the interaction between a laminar boundary layer and a von Kármán vortex street behind a circular cylinder are carried out for three values of the Reynolds number based on the cylinder diameter and free-stream velocity: $\hbox{\it Re}_D\,{=}\,385$, 1155 and 3900. Rapid, bypass-like transition to turbulence is observed in the two higher-Reynolds-number cases. Flow statistics in the transitional and turbulent regions are examined, followed by an investigation of the underlying transition mechanism. Qualitative similarities between wake-induced transition and bypass transition due to free-stream turbulence are discussed and the challenges of predicting boundary-layer transition in this complex environment are pointed out.

This work studies the momentum and energy transport mechanisms in the corner between a free surface and a solid wall. We perform large-eddy simulations of the incompressible fully developed turbulent flow in a square duct bounded above by a free-slip wall, for Reynolds numbers based on the mean friction velocity and the duct width equal to 360, 600 and 1000. The flow in the corner is strongly affected by the advection due to two counter-rotating secondary-flow regions present immediately below the free surface. Because of the convection of the inner eddy, as the free surface is approached, the friction velocity on the sidewall first decreases, then increases again. A similar behaviour is observed for the surface-parallel Reynolds-stress components, which first decrease and then increase again very close to the surface. The budgets of the Reynolds stresses show a strong reduction of all terms of the dissipation tensor in both the inner and outer near-corner regions. They exhibit a reduction in both production and dissipation towards the free surface. Very close to the solid boundary, within 15–20 viscous lengths of the sidewall, the turbulent kinetic energy production and the surface-parallel fluctuations rebound in the thin layer adjacent to the free surface. The Reynolds-stress anisotropy appears to be the main factor in the generation of the mean secondary flow. The multi-layer structure of the boundary layer near the free surface is also discussed.

Large-eddy simulations of temporally evolving turbulent mixing layers have been carried out. The effect of the initial conditions and the size of the computational box on the turbulent statistics and structures is examined in detail. A series of calculations was initialized using two different realizations of a spatially developing turbulent boundary-layer with their free streams moving in opposite directions. Computations initialized with mean flow plus random perturbations with prescribed moments were also conducted. In all cases, the initial transitional stage, from boundary-layer turbulence or random noise to mixing-layer turbulence, was followed by a self-similar period. The self-similar periods, however, differed considerably: the growth rates and turbulence intensities showed differences, and were affected both by the initial condition and by the computational domain size. In all simulations the presence of quasi-two-dimensional spanwise rollers was clear, together with ‘braid’ regions with quasi-streamwise vortices. The development of these structures, however, was different: if strong rollers were formed early (as in the cases initialized by random noise), a well-organized pattern persisted throughout the self-similar period. The presence of boundary layer turbulence, on the other hand, inhibited the growth of the inviscid instability, and delayed the formation of the roller–braid patterns. Increasing the domain size tended to make the flow more three-dimensional.

A three-dimensional shear-driven turbulent boundary layer over a flat plate generated by moving a section of the wall in the transverse direction is studied using large-eddy simulations. The configuration is analogous to shear-driven boundary layer experiments on spinning cylinders, except for the absence of curvature effects. The data presented include the time-averaged mean flow, the Reynolds stresses and their budgets, and instantaneous flow visualizations. The near-wall behaviour of the flow, which was not accessible to previous experimental studies, is investigated in detail. The transverse mean velocity profile develops like a Stokes layer, only weakly coupled to the streamwise flow, and is self-similar when scaled with the transverse wall velocity, Ws. The axial skin friction and the turbulent kinetic energy, K, are significantly reduced after the imposition of the transverse shear, due to the disruption of the streaky structures and of the outer-layer vortical structures. The turbulent kinetic energy budget reveals that the decrease in production is responsible for the reduction of K. The flow then adjusts to the perturbation, reaching a quasi-equilibrium three-dimensional collateral state. Following the cessation of the transverse motion, similar phenomena take place again. The flow eventually relaxes back to a two-dimensional equilibrium boundary layer.

The spatial evolution of the disturbances that lead to boundary-layer transition on a swept wedge is computed by large-eddy simulations (LES). Stationary and travelling crossflow-vortex disturbances are generated using steady and random-amplitude suction and blowing on the wedge. For a fixed initial amplitude of the stationary vortex and low-amplitude unsteady disturbances, the LES show the evolution of stationary-dominated crossflow disturbances similar to previous simulations and experiments: linear amplification is followed by vortex roll-over and doubly inflectional velocity profiles just prior to transition. A high-frequency secondary instability is associated with the double inflection points in the velocity profiles. The harmonic modes of the primary disturbance were found to be amplified, while no energy was found in any subharmonic mode. The physical phenomena were significantly different when the stationary and travelling vortices have comparable initial amplitudes: in this case, the vortex roll-over does not occur and transition is dominated by the travelling-wave component.