International Journal of Heat and Fluid Flow 68 (2017) 269–280

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

journal homepage: www.elsevier.com/locate/ijhff

Direct numerical simulations of a wall-attached cube immersed in laminar MARK

and turbulent boundary layers
⁎
Carlos Diaz-Daniel , Sylvain Laizet, J. Christos Vassilicos
Department of Aeronautics, Imperial College London, London, SW7 2AZ, UK

A R T I C L E I N F O A B S T R A C T

Keywords: A wall-attached cube immersed in a zero pressure gradient boundary layer is studied by means of Direct
Wall-attached cube Numerical Simulations (DNS) at various Reynolds numbers ReH (based on the cube height and the free-stream
Boundary layer velocity) ranging from 500 to 3000. The cube is either immersed in a laminar boundary layer (LBL) or in a
Turbulence turbulent boundary layer (TBL), with the aim to understand the mechanisms of the unsteady flow structures
Strouhal number
generated downstream of the wall-attached cube. The mean locations of the stagnation and recirculation points
around the cube immersed in a TBL are in good agreement with reference experimental and numerical data, even
if in those studies the cube was immersed in a turbulent channel. In the TBL simulation, a vortex shedding can be
identified in the energy spectra downstream of the cube, with Strouhal number of . However, the frequency of
the vortex shedding is different in the LBL simulations, showing a significant dependence on the Reynolds
number. Furthermore, in the TBL simulation, a low frequency peak with St = 0.05 can be observed far away from
the boundary layer, at long streamwise distances from the cube. This peak cannot be identified in the LBL
simulations nor in the baseline TBL simulation without the wall-attached cube.

1. Introduction characteristics on the vortex-shedding frequency.

Kawamura et al. (1984) performed flow visualization experiments and
The turbulent flow around a wall-attached solid cube represents an measurements of surface pressure around a finite circular cylinder on a
interesting and complex problem from a fundamental point of view. flat plate, in order to study the main flow features close to the immersed
Additionally, this flow configuration is a simple model for the inter- objects. More recently, experimental research on the finite-length ef-
action between a boundary layer and complex bodies immersed in it. fects of wall-attached circular and cylinders using hot-wire anemometry
For instance, the wall-attached cube may represent a typical protu- was carried out by Park and Lee (2000). Additionally, the Particle
berance on the surface of aerodynamic vehicles, such as aircraft or Image Velocimetry (PIV) experiments of Wang and Zhou (2009);
vessels. The flow around low and high aspect-ratio square cylinders is Monnier et al. (2010); Wang et al. (2014a) provide further insight on
also very important for environmental applications, since it can model the flow structures generated by wall-attached circular and square cy-
the air movement around simplified buildings. linders. The wind tunnel experiments of Wang et al. (2006) studied the
In the last few decades, there has been extensive research on the effect of the inflow conditions on the interactions between a boundary
turbulent flow around wall-attached obstacles with high aspect ratios layer over a flat plate and flow around a wall-mounted finite-length
H/L ≫ 1 (where H is the obstacle height and L accounts for the base cylinder. Finally, the research of McClean and Sumner (2014);
side). Early experiments on the mean-flow characteristics and vortex Sumner et al. (2015), 2017) focused on the effect of the aspect ratio and
shedding of high aspect ratio wall-attached circular and square cylin- the incidence angle of wall-attached objects in a low-speed wind tunnel
ders were performed in the 70’s and 80’s. Corke et al. (1979) studied using PIV.
the flow near a building model in order to examine the response of the Direct Numerical Simulations (DNS) of square cylinders were per-
flow field to variations in the characteristics of the boundary layer. formed by Saeedi et al. (2014) with a study of the turbulent wake be-
Measurements of the vortex-shedding frequency behind a vertical rec- hind a wall-mounted square cylinder with aspect ratio 4.
tangular prism and a vertical circular cylinder attached to a plane wall Vinuesa et al. (2015) assessed the effect of inflow conditions by con-
were performed by Sakamoto and Arie (1983) to investigate the effects sidering a fully turbulent zero pressure gradient boundary layer and a
of the aspect ratio of these bodies and the boundary-layer laminar boundary layer. The evolution of various flow structures

⁎
Corresponding author.
E-mail addresses: [email protected] (C. Diaz-Daniel), [email protected] (S. Laizet), [email protected] (J. . Vassilicos).

http://dx.doi.org/10.1016/j.ijheatfluidflow.2017.09.015
Received 4 May 2017; Received in revised form 24 September 2017; Accepted 25 September 2017
Available online 15 October 2017
0142-727X/ © 2017 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Table 1
Summary of the simulation parameters (the momentum thickness is θ and the displacement thickness is δ*) for the present investigation.

Reθ Re δ * nx × ny × nz Lx Ly Lz Δywall/H Δytop/H ΔtU∞/H T/H

× ×
H H H

ReH = 500 LBL 68 175 357 × 129 × 192 35 × 15 × 8 0.02 0.68 0.007 10,000
ReH = 600 LBL 81 210 357 × 129 × 192 35 × 15 × 8 0.02 0.68 0.007 10,000
ReH = 750 LBL 101 263 357 × 129 × 192 35 × 15 × 8 0.02 0.68 0.007 10,000
ReH = 1100 LBL 149 385 357 × 129 × 192 35 × 15 × 8 0.02 0.68 0.007 10,000
ReH = 1700 LBL 230 596 357 × 193 × 192 35 × 15 × 8 0.015 0.41 0.005 10,000
ReH = 3000 LBL 406 1051 513 × 385 × 256 35 × 15 × 8 0.007 0.22 0.002 3000
ReH = 3000 TBL 750 1105 4, 097 × 513 × 256 320 × 27 × 10 0.0033 0.833 0.001 750

associated with finite length cylinders immersed in a low Reynolds downstream of the cube. Data in the near-field of the cube are also
number boundary layer such as wakes, tip vortices, base vortices and validated against the reference data of Martinuzzi and Tropea (1993)
horse-shoe vortices were discussed by Saha (2013). A square rectan- and Yakhot et al. (2006b).
gular tall building was considered by Li et al. (2014) to investigate the
effects of turbulence integral length scale and turbulence intensity on
the building by means of Large Eddy Simulation (LES). Numerical in- 2. Computational setup
vestigation of the turbulent flow around a surface-mounted square cy-
linder of aspect ratio 4 were performed by Wang et al. (2014b) to get The results presented here have been obtained from high fidelity
detailed information about the flow structures around such a cylinder Direct Numerical Simulations (DNS) of zero-pressure gradient laminar
and to establish a suitable turbulent model that could yield accurate and turbulent boundary layers (LBL, TBL, respectively), with a solid
and reliable results for practical industrial applications. cube immersed in the computational domain. The baseline simulation
The flow around a wall-attached object with H = L is an important of the TBL case, which uses the same numerical domain without the
classical benchmark for simulations and experiments of bluff bodies. immersed wall-attached cube, was introduced and validated in a fun-
However, there is only a limited number of fundamental studies on the damental investigation on the wall shear-stress fluctuations by Diaz-
turbulence physics of this flow configuration. The investigation of Daniel et al. (2017). The local Reynolds number of the TBL covers the
Castro and Robins (1977) is among the first exhaustive experimental range Reθ = 270 − 2200, based on the momentum thickness θ and free-
studies on the turbulent flow around a wall-attached cube. The authors stream velocity U∞.
compared the effect of uniform and sheared turbulent incoming streams The computational flow solver, Incompact3d (Laizet and
at different Reynolds numbers. Since then, this flow configuration has Lamballais, 2009; Laizet and Li, 2011), uses sixth-order finite difference
been revisited, for instance, by the experimental work of schemes, with a spectral treatment for the pressure equation and a
Martinuzzi and Tropea (1993) at ReH = 40, 000, by semi-implicit time advancement for the viscous terms. The validation
Meinders et al. (1999) with 2750 < ReH < 4970 and by the Direct results of the TBL in Diaz-Daniel et al. (2017) include the computation
Numerical Simulation (DNS) of Yakhot et al. (2006b) at ReH = 1870 . of the budget terms of the mean turbulence kinetic energy equation.
The scalar concentration field behind a wall-attached cube has been The balance of the steady budget terms stays under 1% of the mean
studied experimentally by Ogawa et al. (1983), Li and Meroney (1983) dissipation rate in the entire computational domain. The statistics of
and Mavroidis et al. (2003) at high Reynolds numbers and computa- velocity and wall shear-stress are in excellent agreement with the re-
tionally by Rossi et al. (2010) at ReH = 5000, using DNS and Reynolds- ference data of Schlatter and Örlü (2010) and Jiménez et al. (2010) at
Averaged Navier Stokes (RANS) simulations. The recent study of a wall- equal Reynolds numbers.
attached cube by Hearst et al. (2016), at ReH = 1.8 × 106, suggested that The computational parameters of the present simulations are in-
different inflow conditions at high Reynolds numbers may not modify cluded in Table 1. The cube height is represented by H and the co-
the main shedding frequency or the mean position of the stagnation and ordinate variables in the streamwise, wall-normal and spanwise direc-
reattachment points but seem to affect the length of the turbulent wake tions are x, y, z, respectively. The coordinate system is shifted to a
behind the cube. streamwise position such that x = 0 is located at the front plane of the
The presence of a wall-immersed object in a boundary layer can cube. The computational domain is stretched in the wall normal di-
modify the flow properties in a noticeable way, even with a small rection using the metric described by Laizet and Lamballais (2009). In
blockage ratio. Its turbulent wake induces a momentum loss which the baseline TBL simulation, the mesh resolution, in wall viscous units
results in a rapid increase of the boundary layer thickness. Moreover, (at Reθ = 1470 ) is: Δx + = 10.2,Δz+ = 5.1,Δy+ = 0.42 at the wall and
despite of its relatively small size, the effect of a wall-attached body on Δy+ = 108.8 at the top of the domain. The stretching function para-
the energy spectra of the flow can persist at long distance from the meters guarantee that the wall-normal node spacing inside the
immersed object. However, there is little fundamental work published boundary layer is lower than Δy+ = 12 at the maximum Reynolds
on the influence of a wall-attached cube further downstream of its number Reθ ≈ 2200.
position and on the far-field fluctuations that it generates. On the other The inflow boundary condition in our simulations is a Blasius la-
hand, the far field dynamics generated by circular and square cylinders minar boundary layer profile prescribed at the inlet plane. In the TBL
are slightly better documented in literature, in particular by the recent simulation, the transition to turbulence is triggered via the random-
works of Becker et al. (2008), King and Pfizenmaier (2009), forcing method described in Schlatter and Örlü (2012). A streamwise
Porteous et al. (2013) and Moreau and Doolan (2013). An exhaustive convective equation is solved at the outlet and a no-slip condition is
review on the far-field dynamics has been recently compiled by imposed at the bottom wall. Periodic boundary conditions are used in
Porteous et al. (2014). the spanwise direction, effectively modelling an infinite array of cubes,
The present numerical study investigates the downstream signature and an homogeneous Neumann condition is imposed at the top
of a wall-attached cube, comparing situations where the cube is im- boundary.
mersed in a laminar and in a turbulent boundary layer. In particular, we The solid cube, of size H, is modelled with an immersed boundary
focus on the various peaks found in the energy spectra inside the method (see Laizet and Lamballais (2009) for the details). In the si-
boundary layer but also at large distances from the wall and far away mulation with an incoming TBL, the height of the cube, H, is equal to
0.42δ, where δ is the local boundary layer thickness, and the Reynolds

270
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

number based on H and U∞ is ReH = 3000 . The cube is placed at a simulation at ReH = 3000, a recirculation point can be found at
streamwise distance of 72H from the inlet, where the local Reynolds xD = 1.63H , yD = 1.11H (point D in the bottom plot of Fig. 1), in addi-
number is Reθ = 750 . At the cube’s top face location, y = H, the wall- tion to the stagnation point D′. For Re ≤ 3000, no recirculation region
normal stretching function satisfies Δy+ = 0.82 . In the simulations with can be found over the top surface of the cube but the mean-flow
an incoming LBL, the cube has a height H = δ and is located at a dis- streamlines are strongly curved over the cube due to a strong backflow
tance 9H from the inlet. Six different Reynolds numbers are simulated, in the streamwise direction.
ReH = 500, 600, 750, 1100, 1700 and 3000. The value of δ/H guaran- The upstream stagnation point (point F in Fig. 1) moves farther from
tees that the ratio between the local displacement thickness of the the front face with increasing Reynolds numbers. On the other hand, its
boundary layer and the cube height is similar in the LBL and TBL si- streamwise position xF has been reported to be approximately constant
mulations, being respectively δ */ H = 2.86 and δ */ H = 2.7 . The blockage when the flow around the cube becomes fully turbulent (Yakhot et al.,
ratio of the cube, based on the frontal area of the obstacle and the total 2006b). An empirical correlation for the position of this stagnation
area occupied by δ, is σ = 4.2% in the TBL simulation and σ = 12.5% in point in the range 300 < ReH < 1500 was proposed by Hwang and
the LBL simulations. Yang (2004), xF / H = −0.77 log(ReH ) + 0.564, measured on the x − z
The statistics presented in this study have been averaged over a time plane y = 0.006H . In our simulations, the location of the upstream
period T indicated in Table 1, after letting the simulations run for a stagnation point measured at the plane y = 0.025H (xF in Table 2), also
sufficiently long initial transient period. For the computation of the follows a logarithmic trend for 500 < ReH < 3000,
energy spectra, the time signals were split and averaged over windowed xF / H = −1.24 log(ReH ) + 1.77 (see Fig. 2(a). The 10–15% difference
intervals (using a Hanning window) with 50% overlap. The number of with the correlation predictions of Hwang and Yang (2004) can be at-
windowed intervals is 40 in the LBL simulations up to ReH = 1700, 20 in tributed to the different set-up (channel flow of size 2H versus boundary
the LBL simulation at ReH = 3000 and 2 in the TBL simulation. The non- layer).
dimensional power spectral density (PSD) has been defined as The horseshoe vortex system observed just upstream of the cube is
−1 −1
PSDui = Eui ui U∞ H , where Eui ui is the temporal energy spectrum of the stable for all our laminar simulations. The work of Baker (1979) in-
velocity component ui. vestigated the horseshoe vortex system around high-aspect ratio cy-
linders for different flow conditions, and suggested that its topology and
3. Wall-attached cube under laminar upstream conditions stability depend mostly on the Reynolds number and the ratio D/δ*,
where D is the cylinder diameter and δ* is the boundary layer dis-
The focus in this section is on the coherent structures generated by a placement thickness. Depending on the pair of dimensionless numbers
cube under laminar upstream conditions for Reynolds numbers ranging {ReD, D/δ*} (ReD is based on the cylinder diameter), the horseshoe
from 500 to 3000. According to the results of Meinders et al. (1999) and vortex system may be either stable with 2, 4 or 6 vortices, or unstable
Yakhot et al. (2006b), the mean flow topology and dynamics seem to be following a quasi-periodic behaviour. The stability map obtained from
approximately Reynolds number independent for ReH > 2000 when the the experiments on cylinders by Baker (1979) is presented in Fig. 2(b).
incoming boundary layer is fully turbulent. Therefore, the results at Our simulations have been included in this map for reference, based on
ReH = 3000 under TBL upstream conditions are expected to be re- their values of {ReH, H/δ*}. The different vortex systems from our si-
presentative of higher Reynolds numbers cases. mulations can be identified in the streamline visualisations of Figs. 1
(stable vortices) and 6(a) (unstable vortex, see Baker (1979) for a more
3.1. Mean flow topology detailed description).
In Fig. 2, the position of the stagnation point is plotted for the
Previous computational studies of a wall-attached object immersed present simulations and for the work of Hwang and Yang (2004), in-
in a channel for Reynolds numbers ranging from 0.01 to 3500 suggested dicating the number of steady horseshoe vortices found for each case.
that the mean-flow topology around a wall-attached cube under LBL The ratio H/δ* is different in both investigations, with a value of 2.8 in
conditions is strongly dependent on the Reynolds number, at least up to our study and a value of 1 in Hwang and Yang (2004). According to the
ReH < 2000 (Liakos and Malamataris, 2014; van Dijk and de Lange, stability map of Baker (1979) in Fig. 2(a), the horseshoe vortex system
2007; Hwang and Yang, 2004). The location of the main mean-flow around wall-attached cylinders should consist of 2 vortices with a low
features obtained in our simulations are summarised in Table 2. The last ratio H/δ* and ReH between 500 and 1700. However, in the present
row presents the results for a cube under TBL conditions (see Section 4 simulations, the horseshoe vortex system contains 4 steady vortices
for a detailed discussion). approximately for 300 < ReH < 1000, 6 vortices for
For Re ≤ 1700, the mean-flow streamlines behind the cube, shown 1000 < ReH < 3000 and 8 vortices for ReH ≥ 3000 and a similar be-
in Fig. 1, do not create a closed recirculation region. However, there haviour is inferred from the results of Hwang and Yang (2004). At
exists a stagnation point, marked D′ in Fig. 1, which is located at closer ReH = 3000, the horseshoe system remains stable even if some unsteady
distance x D′ from the cube for increasing Reynolds numbers. In the fluctuations are noticeable. This suggests that the stability limits can be
significantly different depending on the geometry of the wall-attached
Table 2 objects.
Positions of the mean flow features of a wall-attached cube with a laminar incoming In contrast to the Reynolds number dependence of the topological
boundary layer. In the last row, we have added, for comparison, the results of the si- features discussed in the previous paragraph, the position of the stag-
mulation with incoming TBL at ReH = 3000, described in Section 4.
nation point on the cube front face (A in Fig. 1) is approximately
Front Stag. Stag. Stag. Recirc. Recirc. constant for the range ReH = 500 − 3000, at yA = 0.82. Finally, the
face. point point y D′ pointxF point xD point yD mean-flow streamlines suggest that the wake behind the cube becomes
stag. yA x D′ wider for increasing Reynolds numbers. In the low Reynolds number
simulations, ReH = 500, 600, 750, the streamlines in the cube wake are
Re = 500 0.82 3.48 0.17 − 1.6 − −
Re = 600 0.82 3.22 0.21 −1.7 − − almost parallel to the streamwise direction for |z| > 1.4H, x > 6H. This
Re = 750 0.82 2.98 0.24 −1.8 − − suggests that the wake width at low Reynolds numbers is approximately
Re = 1100 0.82 2.19 0.145 −2.0 − − constant, with value Wwake = 2.8H .
Re = 1700 0.81 2.15 0.141 −2.25 − −
Re = 3000 0.81 1.93 0.181 −2.57 1.63 1.11
− − −1.4
3.2. Dynamic structures
Re = 3000 0.67 1.45 0.87
(turb.)
Instantaneous visualisations from the LBL simulations at

271
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 1. Mean flow streamlines and time-averaged streamwise velocity contours around a wall-attached cube at different Reynolds numbers. Left: x − y plane z = 0 (the colourmap for u
ranges from − 0.1U∞, dark blue, to 1.1U∞, dark red). Right: x − z plane y = 0.025 . (colourmap for u from − 0.1U∞, dark blue, to 0.2U∞), dark red. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

ReH = 500 − 3000, using the Q criterion (defined by Hunt et al. (1988)), edges induce vortical motions, which presumably interact with the
are presented in Figs. 3 and 4. These visualisations suggest that the shear layer created over the cube and this may lead to flow instability.
coherent velocity fluctuations may be associated with a periodic gen- This mechanism creates a primary street of symmetric hairpin vortices,
eration of hairpin vortices from the top of the cube. The two upper side which are detached from the wall. For low Reynolds numbers

272
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 2. a) Position of the upstream stagnation point generated by a wall-attached cube under laminar inflow conditions. In Hwang and Yang (2004), it is measured in a channel at a
distance y = 0.006H from the wall, while in the current simulation it was measured in an LBL at y = 0.025H from the wall. The value nHS indicates the number of steady horseshoe vortex
found at each Reynolds number. b) Dependency of the horseshoe system dynamics on the parameters ReD and D/δ* for the wall-attached cylinder experiments of Baker (1979), indicated
with blue cross symbols. The continuous lines indicate the empirical threshold between the 2-vortex, 4-vortex, 6-vortex and unstable horseshoe systems obtained by these authors. The
triangles and the square represent the pairs ReH and H/δ* in our cube simulations under LBL and TBL upstream conditions, respectively. Note that the numbers of horseshoe vortices found
in these simulations are different than those predicted by the diagram for cylinders proposed by Baker (1979). (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

2
Fig. 3. Simulation of a wall-attached cube at different Reynolds numbers with a laminar incoming boundary layer. a,b,c) Isocontours of Q = 0.1U∞ / H2, coloured by streamwise velocity
2 2
(from − 0.5U∞, in dark blue, to 1.1U∞, in dark red). d) Isocontours of Q = 0.2U∞ / H2 and e,f) isocontours of Q = 0.28U∞ / H2 . In subfigure (a), the low opacity surface represents the
2
isocontour Q = 0.003U∞ / H2 . (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

(ReH = 500 − 750 ), vortex generation starts farther downstream of the at ReH = 500 is St = 0.17. The unsteady structures appear as a single
cube than for higher Reynolds numbers. sharp and intense peak in the turbulence energy spectra, which is
2
At ReH = 500, the isocontours of Q = 0.1U∞ / H 2 Fig. 3(a) do not show shown for different downstream locations in Fig. 4(a). These energy
any unsteady structure since the velocity fluctuations are very low. If spectra have been averaged over 9 equidistant spanwise locations be-
2
the threshold for Q is relaxed down to Q = 0.003U∞ / H 2 (drawn with low tween z = −1.5H and z = 1.5H from the cube centre plane. The ob-
opacity), weakly unsteady structures are revealed, which develop into tained value St = 0.17 is in good agreement with the Strouhal number
hairpin vortices at x > 15H. Therefore, this suggests that the critical obtained by the DNS of Yanaoka et al. (2007) at ReH = 500 (St = 0.159).
Reynolds number for flow unsteadiness may be close to ReH = 500 . The coherent velocity fluctuations are significantly stronger in
According to the flow visualisations, the hairpin vortices might be re- Yanaoka et al. (2007), where symmetry conditions were used in the
lated to an instability mechanism of the steady streamwise vortices spanwise direction and a slip condition on the domain top plane, lo-
which are generated at the cube top edges. cated at y = 10H from the wall. These authors reported that the flow
The Strouhal number associated with the hairpin vortex structures solution in their simulation at ReH = 450 is stable, which supports our

273
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 4. Simulation of a wall-attached cube at different Reynolds numbers with a laminar incoming boundary layer. Energy spectra of the streamwise velocity component u at different
−1 −1
positions x, y. The power-spectral density (PSD) has been non-dimensionalised as PSDu = Euu U∞ H .

previous statement suggesting that ReH = 500 may be close to the cri- with a lower Strouhal number St = 0.1. This peak can be associated
tical value for unsteady flow. with a new phenomenon since the harmonics found at ReH = 600 all
In the present simulations, the magnitude of the peak found in the have higher frequencies than the main peak. The interaction between
energy spectra is maximum around x = 15H , y = H from the cube, after the new peak at St = 0.1 and the primary peak at St = 0.21 also gen-
the shear layer becomes unstable. When moving farther downstream at erates the harmonic St = 0.1 + 0.21 = 0.31. Interestingly, the velocity
y = H, the peak intensity is reduced, but can still be detected up to the fluctuations at St = 0.1 for this Reynolds number seem to be less am-
domain outlet. The magnitude of the peak in the power spectra reported plified than in the ReH = 600 case. At ReH = 750, the primary peak
by Yanaoka et al. (2007) (measured at y = H ) also decreases for in- reaches its maximum magnitude around x = 4H and decreased down-
creasing x > 6H. stream of this location.
For ReH = 600, the contours of Q suggest that another two streets of The vortex visualisations presented in Fig. 3(d) suggest that, at
hairpin vortices, which are attached to the wall, are generated on the ReH = 1, 100, the flow structures are much more complex than in the
sides of the primary structures, possibly from a secondary interaction previous cases, but it is still possible to distinguish the vortical struc-
between the cube flow structures and the wall. These secondary vortex tures described before. The primary hairpin vortex street can be found
streets are symmetrically separated by a distance of approximately 1.2H close to the cube, but after a short distance away from the cube, the
from the cube centre plane z = 0 . Their generation may be associated interaction with other flow structures becomes very strong and the
with the instabilities caused by the vortical motion of the horseshoe vortices break down into less organised motions. The secondary streets
vortex legs. In particular, the flow region around a horseshoe vortex is of wall-attached hairpin vortices on the sides of the cube can be iden-
fundamentally similar to a quasi-streamwise vortex from the near-wall tified as well.
region of a turbulent wall-bounded flow. Therefore, it might be rea- The near-cube coherent structures are shed with higher frequency
sonable to expect similar hairpin-vortex structures to those discussed by than at lower Reynolds numbers and the Strouhal number of the main
Adrian (2007) for the buffer and log layers. peak in the energy spectrum computed at x = 4H , y = 0.75 is equal to
The Strouhal number of the primary top vortices, St = 0.19, is St = 0.32 . However, secondary peaks at St = 0.22 and St = 0.12 are also
slightly increased in comparison to the ReH = 500 case. The secondary identified, which might be the signature of the coherent fluctuations
wall-attached structures are shed at the same frequency, but the in- found at lower Reynolds numbers. The energy spectra suggest that the
teraction between the different structures seems to result in harmonic flow interactions at ReH = 1100 are non-linear and that the flow may
peaks at St = 0.38,St = 0.57 and higher multiples of the main Strouhal become turbulent further downstream of the cube. At this Reynolds
number St = 0.19. The interaction with the wall and between the vortex number, the turbulent kinetic energy is distributed in a broad band
streets may also be responsible for an amplification of the primary peak range of frequencies and no peaks can be easily identified for x > 20H
in the energy spectra, since its magnitude keeps increasing when and y < H. It is interesting to note that a low frequency peak, with
moving away from the cube. St = 0.05, can be found in the energy spectrum at x = 1.5H . It seems
At ReH = 750, stronger flow interactions between the cube and the that this low frequency peak is only detected inside the backflow region
wall produce a higher number of secondary structures and a more behind the cube (see Fig. 1), suggesting that the flow dynamics may be
disorganised distribution of them. The Strouhal number of the main different here. The peak at St = 0.05, weak in comparison with the other
shedding further increases to St = 0.21. The vortex interaction and ad- ones, is not present in the spectrum at x = 4H , y = 0.75H nor for higher
ditional flow instabilities behind the cube generate a secondary peak streamwise positions.

274
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

At ReH = 1700, the main peak with St = 0.32 found at ReH = 1100 4. Wall-attached cube under turbulent upstream conditions
and x = 4H can also be identified, but its Strouhal number increases to
St = 0.37 . While the peak at St = 0.22 is much weaker at ReH = 1700, the 4.1. Mean-flow features in the near-cube region
peak at St = 0.14 has a greater magnitude than the peak at St = 0.12 at
ReH = 1100 . Since the Strouhal number is the same as the one identified The mean flow features of our TBL simulation at ReH = 3000 are
in the simulation with a turbulent incoming boundary layer in compared with published experimental and simulation data
Fig. 10(a), the peaks found at St = 0.14 for ReH = 1700 and ReH = 3000 (Martinuzzi and Tropea, 1993; Yakhot et al., 2006b). In those studies,
might possibly be associated with the same flow structures. Hwang and the Reynolds number ReH is similar to ours, but the cube is immersed in
Yang (2004) stated that the mechanism which generates the dominant a turbulent channel instead of a turbulent boundary layer. The mean
peak with St ≈ 0.12 − 0.14 at high Reynolds numbers is not well un- flow streamlines in Fig. 6(a) and (b) show the time averaged structures
derstood, since the main coherent structures are shed with higher around the cube. In the centre plane z = 0, the stagnation point A is
Strouhal numbers. However, these authors suggested that the peak at located at yA / H = 0.67, the reattachment point E at xE / H = 2.5, the
St ≈ 0.13 actually dominates the force coefficient of the total spanwise front vortex C has its centre at x C / H = 0.48 and the horizontal location
loading on the cube. While the horseshoe vortex is still stable at of the rear recirculation centre D is xD / H = 1.45. Those spatial positions
ReH = 1700, the isocontours of Fig. 3(e) show a strong generation of are in good agreement with the numerical results of Yakhot et al.
hairpin vortices around its legs. The shedding of hairpin vortices can be (2006b), with differences of less than 3%. On the other hand, the lo-
related to a new peak with St = 0.75 found in the energy spectra, since cation of the top recirculation bubble B, xB / H = 0.65 and yB / H = 1.13,
these are the only coherent structures found to be shed at such high and the vertical position of the rear recirculation D, yD / H = 0.87, have a
frequencies. At higher downstream distances from the cube, the energy 10–15% relative error with respect to the values found in Yakhot et al.
spectra at St = 1700 does not predict any dominant peak, only broad- (2006b). This can be explained by the different top boundary condition,
band fluctuations. since the upper wall in the channel configuration constraints the flow in
At ReH = 3000, a single peak with St = 0.1 can be identified in the the vertical direction.
energy spectra of Fig. 4(f) at x = 1.5H and x = 4H , but it is no longer The results are summarized in Table 2 and the comparison with
detected far downstream. The horseshoe vortex system is stable, and Fig. 1 shows important disparities in the location and size of the mean-
the instantaneous and averaged streamlines in front of the cube are very flow features between simulations under LBL and TBL conditions for the
similar to each other. However, Fig. 5(a) shows that the largest same Reynolds number. For instance, the distance from the mean
horseshoe vortices are not steady and they generate weak velocity stagnation point and the cube front face, xF, is 45% lower for the TBL
fluctuations which can be associated to a peak in the energy spectra simulation and the location of the stagnation point A on the cube front
found at approximately St = 0.085, as seen in Fig. 5(b). The St value is face is lower by 17% with yA = 0.67 . The authors in
relatively similar to the one reported in Yakhot et al. (2006a) for the Vinuesa et al. (2015) have also previously reported that the inflow
unstable horseshoe vortex system of a cube under turbulent upstream conditions can have an important influence on the main flow features
conditions (St ≈ 0.08), suggesting that the velocity fluctuations in the around a high aspect-ratio square cylinder. It was suggested that, while
two cases might be related to the same physics. the Strouhal number of the main shedding is approximately the same
Finally, the far-field velocity signature of the cube at ReH = 3000 is (St = 0.1) under incoming turbulent and laminar upstream conditions,
also presented in Fig. 5(b). The turbulence energy spectra suggests that the upstream horseshoe vortex dynamics and the downstream wake
the signature of the coherent motions detected near the cube is not parameters may be significantly different.
noticeable far downstream of the cube location. At y = 3H and The TBL simulation at ReH = 3000 exhibits an unstable horseshoe
x > 20H, the energy is spread in a broad bandwidth of frequencies, vortex system in front of the cube, as confirmed by the mean-flow
centred around St = 0.15 − 0.2, and no sharp peak can be observed. The streamlines in Fig. 6(a), and in agreement with the results reported by
results suggest that the main frequency of the excited region of the Yakhot et al. (2006a). It seems that the dynamics of this flow feature are
energy spectra may decrease for increasing values of x/H. strongly dependent on the turbulence upstream conditions, as sug-
gested by Baker (1979) with a dependence with the parameters ReH and
δ*/H only. The LBL and TBL simulations at ReH = 3000 have more or
less the same value of δ*/H ≈ 2.8 but the horseshoe vortex dynamics
and the main mean-flow features around the cube are fundamentally

0.06

0.05

0.04

0.03

0.02

0.01

0
10 -2 10 -1 10 0

(a) (b)
Fig. 5. Simulation of a wall-attached cube at ReH = 3000 with a laminar incoming boundary layer: a) Instantaneous contours of the streamwise velocity component and instantaneous
streamlines on the x − y plane z = 0 . Detail of the front horseshoe vortex. b) Energy spectra of the streamwise velocity component u at the front horseshoe vortex position and in the cube
−1 −1
far-field. The non-dimensional PSD is defined as PSDu = Euu U∞ H .

275
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 6. a) Mean velocity streamlines in the spanwise plane z = 0 . Coloured contours by velocity magnitude (from − 0.25, blue, to 1.1U∞, red). b) Mean velocity streamlines in the wall-
normal plane y / H = 0.0045 . Coloured contours by streamwise velocity (from − 0.1U∞, blue, to 0.2U∞, red). (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.)

different. On the other hand, similar results obtained in the present recover a canonical state for the inner and buffer regions of the TBL and
simulation and in Yakhot et al. (2006b) suggest that the effect of the top the influence of the cube is mostly concentrated on the inertial and
boundary condition (TBL versus turbulent channel) is not that im- wake layers. The comparison between the span-averaged turbulent
portant for this flow feature. fluctuation profiles at Reθ = 1, 000 (22.3H downstream of the cube),
Periodic boundary conditions in the spanwise direction are model- presented in Fig. 7(b), shows a significant increment for span-averaged
ling an infinite array of cubes and one can average the flow variables streamwise fluctuations expressed in wall units urms+
in the cube simu-
over [−Lz /2, Lz /2] to estimate the effect of the cube on the boundary + +
lation between y ≈ 80 and y ≈ 300, while the inner part of boundary
layer statistics. The interaction with the cube increases the span-aver- layer remains unaltered. Thus, the effect of the immersed cube on the
aged momentum thickness by a constant Δθ, which is reflected in the span-averaged turbulence statistics is mostly concentrated around its
Reynolds number, as shown in Fig. 7(a). By using the physical meaning upper edges, located at y+ ≈ 132 .
2
of the momentum thickness, D = U∞ Lz Δθ , the drag coefficient of the
ΔRe L
cube can be related to Δθ as Cd = 2ΔθLz / H 2 = 2 Re θ Hz = 0.7 . The ob-
H 4.2. Energy spectra inside the boundary layer
tained value of the drag coefficient, Cd = 0.7, is in good agreement with
the result obtained by integrating the surface forces, equal to Cd = 0.72
In the near-field flow around the cube, for y < H, top, rear and
(the contribution of the pressure forces on the front and rear faces to the
lateral recirculations shed unsteady vortices, producing a dominant
form drag is 0.642 and 0.09 respectively, and the skin friction drag only
peak in the velocity spectra. Previous studies have reported a shedding
contribute as −0.011, a 1.5% of the total). Differences in the drag
frequency with a Strouhal number St = fH / U∞ = 0.08 − 0.15 (Yakhot
coefficient with the experiments of Martinuzzi and Havel (2004)
et al., 2006b; Porteous et al., 2014; Martinuzzi and Havel, 2004). In our
(Cd ≈ 0.95) can be attributed to different incoming flow conditions
simulation, close to the rear wall of the obstacle
which affect the cube wake characteristics (in their study, a Blasius
( x = 4.7H , y = 0.73H , z = 0 ), it is possible to identify a peak in the
laminar profile with δ / H = 0.07 was prescribed upstream of the cube).
turbulence spectra with St = 0.14 as seen in Fig. 8(a). The peak fre-
The experimental studies of Sakamoto et al. (1982) and Sakamoto and
quency lies within the range of values found in the literature and is in
Oiwake (1984) suggest that the drag coefficient of the cube strongly
good agreement with the empirical correlation of Wang and Lu (2012),
depends on the ratio H/δ. These authors obtained a value around
based on experimental results. Further downstream, the peak in the
Cd ≈ 0.4 for H / δ = 0.15, a value around Cd = 0.6 for H / δ = 0.5 and
energy spectra of the streamwise component u is masked by the
Cd = 0.95 for H / δ = 1.5.
boundary layer turbulence and cannot be detected for y < δ (see
Moving further downstream, the span-averaged velocity profiles
Fig. 8(a) for x = 36H ). The spectra of the spanwise component w also

Fig. 7. a) Effect of the cube on the span- averaged momentum thickness. b) Effect of the solid cube in the span-averaged streamwise fluctuations, at Reθ = 1000 . This Reynolds number
location can be found at 111H in the unperturbed boundary layer simulation and at 95H in the cube simulation (22.3H downstream of the cube).

276
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 8. Energy spectra of a) streamwise and b) spanwise velocities in the near field of the wall-attached cube, at y = 0.73H and z = 0 . Comparison at x = 4.7H and x = 36H downstream
−1 −1
of the cube (points P and P′ in Fig. 9). The non-dimensional PSD is defined as PSDui = Eui ui U∞ H .

presents a peak with St = 0.14 which is shown in Fig. 8(b). While the the boundary layer edge. The frequency of the peak is very low and
magnitude of this peak decreases further downstream, it is still no- cannot be associated directly with the vortex shedding of the cube
ticeable at y / H = 36, as the background spanwise fluctuations of the measured closer to the wall. Note that in previous experimental studies
boundary layer are less intense than the streamwise ones. of round and square cylinders by Porteous et al. (2013) and
Porteous et al. (2014), a low-frequency peak was also detected in the
far-field with St = 0.07 . The authors associated this peak with the tip
4.3. Energy spectra outside the boundary layer flow shedding, occurring at a different frequency from the main vortex
shedding with St ≈ 0.15 − 0.2 . However, this low frequency peak was
In the free-stream, away from the boundary layer, an array of virtual only detected in the far-field spectra of high aspect ratio cylinders (H/
probes recorded the velocity signal as a function of time at different D > 9, where D is the cylinder diameter), which is not the case here.
streamwise positions and same distance from the wall, y / H = 4.7, as Therefore, it is reasonable to think that the far-field peak in our simu-
sketched in Fig. 9. In this region, the flow statistics have small varia- lations may not be related to a tip flow shedding but is otherwise
tions in the spanwise direction and, thus, the frequency spectra have connected to another physical phenomenon.
been averaged over 16 equally-spaced spanwise positions to improve Fig. 11 shows that the peak found in the energy spectra of the
statistical convergence. Far away from the boundary layer (at least streamwise velocity component can also be found for the wall-normal
y / H = 3 − 4 from the wall), a sharp peak with Strouhal number component v. Moreover, the magnitude of the peak is of the same order
St = 0.05 is found in the turbulence spectra for large distances down- for these two components. On the other hand, the energy spectra of the
stream of the cube, around x/H > 30 (Fig. 10(a)). The peak magnitude spanwise component does not seem to present a well-defined peak: a
is low since this position is far from the turbulent region, but it is over 5 single frequency with significantly higher power spectral density
times higher than the magnitude obtained in the simulation with no cannot be clearly identified. Indeed, the maximum value in the w-
cube at the exact same spatial locations. The shape of the far-field component spectra at x = 36H , y = 4.7H is about 15 times lower than
spectra generated by the baseline turbulent boundary layer is similar to the peak value found in the energy spectra of u and v. It suggests that
the one obtained experimentally by Favre et al. (1957), and may be the far-field velocity fluctuations are fundamentally two-dimensional in
explained by the theoretical model of irrotational fluctuations by our simulations, which may possibly be explained by either of the fol-
Philips (1955). The free-stream spectra of a TBL was briefly discussed lowing reasons or the combination of them:
again by Rodríguez-López et al. (2016).
Fig. 10(b) shows that, at an equal distance y / H = 4.7 from the wall, a) The large-scale fluctuations generated by the cube are two-dimen-
peak magnitudes in the cube simulation’s span-averaged streamwise sional themselves and might be unrelated to the shedding at
spectra increase with downstream distance, but the peak frequency St = 0.15 detected in the near-wall region, which generates spanwise
does not change. This suggests that the fluctuations created by the cube fluctuations (see Fig. 8(b)). The far-field fluctuations might be as-
may propagate and possibly amplify downstream and upwards. This sociated with spanwise-oriented vortices generated on top of the
effect could be associated with the boundary layer thickness growth, cube and/or with the turbulent interaction of the heads of the
but the value of the energy spectra peak measured at hairpin vortices observed behind the cube.
x = 20H , y = 4.7H , located at 2.3H from the boundary layer edge, is b) The scale of the far-field fluctuations is so large that it occupies the
lower than the peak value measured at x = 45H , y = 6.5H , at 2.9H from

Fig. 9. Probe positions for the frequency spectra in the wall-

attached cube simulation. Points P, P′: inside boundary layer,
x / H = 4.7, 36, y / H = 0.7 . Points A − D, free-stream far-field,
y / H = 4.7 and x / H = 20, 28, 36, 45 ; D′ at y / H = 6.5,
x / H = 45 .

277
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 10. a) Span-averaged energy spectra of the streamwise component u. Comparison with and without cube at x = 36H , y = 4.7H from the cube position (point P in Fig. 9). b) Span-
averaged temporal spectra of the streamwise velocity component, from the cube simulation, measured in the free-stream at several streamwise positions (points A − D and D′). The non-
−1 −1
dimensional PSD is defined as PSDu = Euu U∞ H .

entire spanwise extent of the computational domain. The time se- at this location.
paration of the structures associated with St = 0.05 is Δt ≈ 20H/U∞ The sharp peak described in this section is not observed in the si-
and if one assumes Taylor hypothesis (convection velocity equal to mulations of a cube immersed in a LBL. This suggest that the far-field
U∞), the scale associated with such structures would be 20H, larger structures responsible for this peak is only generated when the cube
that the spanwise extent Lz = 10H . interacts with an incoming turbulent boundary layer. The comparison
between these two configurations revealed that some flow structures in
The instantaneous streamwise velocity fluctuations, probed at the cube near-field are fundamentally different, even if the Reynolds
x = 45H , y = 6H , z = 0 and plotted in Fig. 12(a), show evident differ- number is the same (ReH = 3000 ). For instance, we previously discussed
ences between the two TBL simulations with and without the cube. The the differences between the mean-flow features around the cube and
time-signal from the immersed-cube simulation presents higher maxima mentioned that the horseshoe vortex system in front of the cube is
and minima and the separation between these peaks is relatively con- unstable only in the TBL simulation.
stant over time. This suggests that the cube is exciting or enhancing
free-stream fluctuations at a particular low frequency, consistent with 5. Conclusions
the peak location in the energy spectra.
To confirm this, the probability distribution function (PDF) of the The interaction between a turbulent boundary layer and a wall-at-
time-lapse between maxima of u′ (conditioned to u′ > 2 × 10−3 ) was tached cube generates a low-frequency sharp peak in the far-field en-
computed at x = 45H , y = 6H . This PDF shows that the events with time ergy spectra which persists for long downstream distances with a con-
such that Δt ≈ 20H/U∞, equivalent to the frequency St = 0.05, have a stant Strouhal number St = 0.05. This peak is not due to numerical
high probability peak of approximately 8% when the cube is present effects nor related to the background boundary layer turbulence, since
(Fig. 12(b)). The PDF of the time-lapse between local minima (condi- it was not identified in the unperturbed zero-pressure gradient turbu-
tioned to u′ < −2 × 10−3 ) does not show such high peaks at Δt ≈ 20H/ lent boundary layer. The Strouhal number of this peak does not cor-
U∞, supporting existing evidence of high skewness in the velocity signal respond to the vortex shedding detected close to the cube at St = 0.14,

10 -3 10 -4
2 1.2

1
1.5
0.8

1 0.6

0.4
0.5
0.2

0 0
10 -2 10 -1 10 0 10 -2 10 -1 10 0

(a) (b)
Fig. 11. a) Span-averaged energy spectra of the wall-normal component v. Comparison with and without cube at x = 36H , y = 4.7H from the cube position (point P in Fig. 9) and x = 36H ,
y = 6.5H . b) Span-averaged energy spectra of the spanwise component w. Comparison with and without cube, also at x = 36H , y = 4.7H . The non-dimensional PSD is defined as
−1 −1
PSDui = Eui ui U∞ H .

278
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

Fig. 12. a) Time signal of the fuctuating streamwise velocity u′, with and without the cube, at the position x = 45H , y = 6H . b) Probability distribution function (PDF) of the time lapse
between velocity maxima (conditioned to u′ > 0.002) and minima (conditioned to u′ < −0.002 ), at the position x = 45H , y = 6H , z = 0 .

hence it might be associated with an additional phenomenon. The peak Acknowledgement

has been observed in the energy spectra of the streamwise and vertical
fluctuating velocity components but not in the spectra of the spanwise This work was funded by the European research project Multisolve,
fluctuations, which suggests that the responsible flow structures might granted by the European Comission (FP7 Marie Curie project, grant
be essentially two-dimensional. agreement No. 317269). The authors would like to acknowledge the use
A series of simulations with incoming laminar boundary layer of the following HPC facilities: SuperMUC, thanks to the European re-
conditions investigated the coherent structures generated by the cube at search infrastructure PRACE; ARCHER, supported by the UK
different Reynolds numbers, in the range ReH = 500 − 3000 . The co- Turbulence Consortium under the ERC grant number EP/L000261/1,
herent structures are mainly organised in two distinct vortex streets: and BlueJoule, supported by the STFC Hartree Centre. J.C. Vassilicos
hairpin vortices shed from the top of the cube and wall-attached hairpin acknowledges support from ERC Advanced Grant 320560.
vortices generated at both sides of the cube wake. This study revealed
that there is a strong Reynolds number dependence for the peaks found References
in the energy spectra, but the far-field peak was not found in those
simulations with incoming laminar conditions. Adrian, R.J., 2007. Hairpin vortex organization in wall turbulence. Phys. Fluids 19 (4),
The turbulent upstream conditions are therefore related, directly or 041301.
Baker, C.J., 1979. The laminar horseshoe vortex. J. Fluid Mech. 95 (2), 347–367.
indirectly, to the far-field peak. For this reason, the origin of the far- Becker, S., Hahn, C., Kaltenbacher, M., Lerch, R., 2008. Flow-induced sound of wall-
field peak is most probably related to mechanisms involving the in- mounted cylinders with different geometries. AIAA J. 46 (9), 2265–2281.
teraction between a turbulent boundary layer and a cube. The energy Castro, I., Robins, A., 1977. The flow around a surface-mounted cube in uniform and
turbulent streams. J. Fluid Mech. 79 (02), 307–335.
spectra of the incoming boundary layer at y = 4.7H has a broad peak Corke, T., Nagib, H., Tan-Atichat, J., 1979. Flow near a building model in a family of
centred around St = 0.07 − 0.1. The interaction with the cube might surface layers. J. Wind Eng. Ind. Aerodyn. 5 (1), 139–158.
modulate or amplify the oscillations at St = 0.05 and this modulation Diaz-Daniel, C., Laizet, S., Vassilicos, J., 2017. Wall shear stress fluctuations: mixed
scaling and their effects on velocity fluctuations in a turbulent boundary layer. Phys.
could be responsible for the far-field peak, since the low frequency
Fluids 29, 055102.
velocity fluctuations are weakly dissipated by viscous stresses. van Dijk, A., de Lange, H., 2007. Compressible laminar flow around a wall-mounted cubic
As explained before, the flow dynamics around the cube present obstacle. Comput. Fluids 36 (5), 949–960.
Favre, A., Gaviglio, J., Dumas, R., 1957. Space-time double correlations and spectra in a
some fundamental differences between the LBL and TBL simulations at
turbulent boundary layer. J. Fluid Mech. 2, 313–342.
ReH = 3000 . For instance, the main mean-flow recirculations, behind Hearst, R.J., Gomit, G., Ganapathisubramani, B., 2016. Effect of turbulence on the wake
and on top of the cube, are not completely developed in the LBL si- of a wall-mounted cube. J. Fluid Mech. 804, 513–530.
mulation. Secondly, the horseshoe vortex system in front of the cube is Hunt, J.C.R., Wray, A.A., Moin, P., 1988. Eddies, stream, and convergence zones in tur-
bulent flows. Proceedings of the 1988 Summer Program (SEE N89-24538), NASA
stable with 8 vortices in the LBL simulation and unstable in the TBL Center for Turbulence Research, Stanford, USA. pp. 193–208.
simulation. Hwang, J.-Y., Yang, K.-S., 2004. Numerical study of vortical structures around a wall-
The authors of Yakhot et al. (2006a) suggested that the unstable mounted cubic obstacle in channel flow. Phys. Fluids 16 (7), 2382–2394.
Jiménez, J., Hoyas, S., Simens, M.P., Mizuno, Y., 2010. Turbulent boundary layers and
horseshoe vortex system has a similar dynamic as the inviscid-viscous channels at moderate Reynolds numbers. J. Fluid Mech. 657, 335–360.
interaction between a vortex and the high-vorticity region near the wall Kawamura, T., Hiwada, M., Hibino, T., Mabuchi, I., Kumada, M., 1984. Flow around a
in a junction flow. They described the unsteady mechanism in front of finite circular cylinder on a flat plate: cylinder height greater than turbulent
boundary layer thickness. Bull. JSME 27 (232), 2142–2151.
the cube as a quasi-periodic regeneration of the horseshoe vortex and a King, W., Pfizenmaier, E., 2009. An experimental study of sound generated by flows
low-momentum fluid ejection away from the wall. The dominant fre- around cylinders of different cross-section. J. Sound Vib. 328 (3), 318–337.
quency of this cycle is reported to be at St = 0.08. It is unclear whether Laizet, S., Lamballais, E., 2009. High-order compact schemes for incompressible flows: a
simple and efficient method with quasi-spectral accuracy. J. Comput. Phys. 228 (16),
the ejection of low-momentum fluid at such low frequencies might be 5989–6015.
related to the velocity fluctuations at St = 0.05 found in the far-field of Laizet, S., Li, N., 2011. Incompact3d : a powerful tool to tackle turbulence problems with
our simulation with a turbulent incoming boundary layer. Both phe- up to O(105) computational cores. Int. J. Numer. Meth. Fluids 67, 1735–1757.
Li, Q., Hu, G., Yan, B.-W., 2014. Investigation of the effects of free-stream turbulence on
nomena are presumably only found at high Reynolds numbers when the
wind-induced responses of tall building by large eddy simulation. Wind Struct. 18 (6),
flow around the cube is fully turbulent. 599–618.
Li, W.W., Meroney, R.N., 1983. Gas dispersion near a cubical model building. Part I. Mean
concentration measurements. J. Wind Eng. Ind. Aerodyn. 12 (1), 15–33.
Liakos, A., Malamataris, N.A., 2014. Direct numerical simulation of steady state, three

279
C. Diaz-Daniel et al. International Journal of Heat and Fluid Flow 68 (2017) 269–280

dimensional, laminar flow around a wall mounted cube. Phys. Fluids 26 (5), 053603. cylinder placed vertically in a turbulent boundary layer. J. Fluid Mech. 126,
Martinuzzi, R., Havel, B., 2004. Vortex shedding from two surface-mounted cubes in 147–165.
tandem. Int. J. Heat Fluid Flow 25 (3), 364–372. Sakamoto, H., Moriya, M., Taniguchi, S., Arie, M., 1982. The form drag of three-dimen-
Martinuzzi, R., Tropea, C., 1993. The flow around surface-mounted, prismatic obstacles sional bluff bodies immersed in turbulent boundary layers. J. Fluids Eng. 104 (3),
placed in a fully developed channel flow. J. Fluids Eng. 115, 85–92. 326–333.
Mavroidis, I., Griffiths, R., Hall, D., 2003. Field and wind tunnel investigations of plume Sakamoto, H., Oiwake, S., 1984. Fluctuating forces on a rectangular prism and a circular
dispersion around single surface obstacles. Atmos. Environ. 37 (21), 2903–2918. cylinder placed vertically in a turbulent boundary layer. .J Fluids Eng. 106 (2),
McClean, J., Sumner, D., 2014. An experimental investigation of aspect ratio and in- 160–166.
cidence angle effects for the flow around surface-mounted finite-height square Schlatter, P., Örlü, R., 2010. Assessment of direct numerical simulation data of turbulent
prisms. ASME J. Fluids Eng. 136 (8), 081206–081206–10. boundary layers. J. Fluid Mech. 659, 116–126.
Meinders, E., Hanjalic, K., Martinuzzi, R., 1999. Experimental study of the local con- Schlatter, P., Örlü, R., 2012. Turbulent boundary layers at moderate Reynolds numbers:
vection heat transfer from a wall-mounted cube in turbulent channel flow. J. Heat inflow length and tripping effects. J. Fluid Mech. 710, 5–34.
Transfer 121 (3), 564–573. Sumner, D., Rostamy, N., Bergstrom, D., Bugg, J., 2015. Influence of aspect ratio on the
Monnier, B., Neiswander, B., Wark, C., 2010. Stereoscopic particle image velocimetry flow above the free end of a surface-mounted finite cylinder. Int. J. Heat Fluid Flow
measurements in an urban-type boundary layer: insight into flow regimes and in- 56, 290–304.
cidence angle effect. Boundary Layer Meteorol. 135 (2), 243–268. Sumner, D., Rostamy, N., Bergstrom, D., Bugg, J.D., 2017. Influence of aspect ratio on the
Moreau, D.J., Doolan, C.J., 2013. Flow-induced sound of wall-mounted finite length cy- mean flow field of a surface-mounted finite-height square prism. Int. J. Heat Fluid
linders. AIAA J. 51 (10), 2493–2502. Flow 65, 1–20.
Ogawa, Y., Oikawa, S., Uehara, K., 1983. Field and wind tunnel study of the flow and Vinuesa, R., Schlatter, P., Malm, J., Mavriplis, C., Henningson, D.S., 2015. Direct nu-
diffusion around a model cube.-i. flow measurements. Atmos. Environ. (1967) 17 (6), merical simulation of the flow around a wall-mounted square cylinder under various
1145–1159. inflow conditions. J. Turbul. 16 (6), 555–587.
Park, C.-W., Lee, S.-J., 2000. Free end effects on the near wake flow structure behind a Wang, H.F., Cao, H.L., Zhou, Y., 2014. POD analysis of a finite-length cylinder near wake.
finite circular cylinder. J. Wind Eng. Ind. Aerodyn. 88 (23), 231–246. Exp. Fluids 55 (8), 1790.
Philips, O., 1955. The irrotational motion outside a free turbulent boundary. Math. Proc. Wang, H.F., Zhou, Y., 2009. The finite-length square cylinder near wake. J. Fluid Mech.
Cambridge Philos. Soc. 51, 220–229. 638, 453–490.
Porteous, R., Doolan, C.J., Moreau, D.J., 2013. Directivity pattern of flow-induced noise Wang, H.F., Zhou, Y., Chan, C.K., Lam, K.S., 2006. Effect of initial conditions on inter-
from a wall-mounted, finite length circular cylinder. Proceedings of Acoustics 2013, action between a boundary layer and a wall-mounted finite-length-cylinder wake.
Annual Conference of the Australian Acoustical Society,Victor Harbour, SA. Phys. Fluids 18 (6), 065106.
Porteous, R., Moreau, D.J., Doolan, C., 2014. A review of flow-induced noise from finite Wang, L., Lu, X.-Y., 2012. Flow topology in compressible turbulent boundary layer. J.
wall-mounted cylinders. J. Fluids Struct. 51, 240–254. Fluid Mech. 703, 255–278.
Rodríguez-López, E., Bruce, P.J.K., Buxton, O.R.H., 2016. On the formation mechanisms Wang, Y., Jackson, P., Sui, J., 2014b. Simulation of turbulent flow around a surface-
of artificially generated high reynolds number turbulent boundary layers. Boundary mounted finite square cylinder. J. Thermophys. Heat Transfer 28 (1), 118–132.
Layer Meteorol. 160, 1–24. Yakhot, A., Anor, T., Liu, H., Nikitin, N., 2006a. Direct numerical simulation of turbulent
Rossi, R., Philips, D., Iaccarino, G., 2010. A numerical study of scalar dispersion down- flow around a wall-mounted cube: spatio-temporal evolution of large-scale vortices.
stream of a wall-mounted cube using direct simulations and algebraic flux models. J. Fluid Mech. 566, 1–9.
Int. J. Heat Fluid Flow 31 (5), 805–819. Yakhot, A., Liu, H., Nikitin, N., 2006b. Turbulent flow around a wall-mounted cube: a
Saeedi, M., LePoudre, P.P., Wang, B.-C., 2014. Direct numerical simulation of turbulent direct numerical simulation. Int. J. Heat Fluid Flow 27, 994–1009.
wake behind a surface-mounted square cylinder. J. Fluids Struct. 51, 20–39. Yanaoka, H., Inamura, T., Kawabe, S., 2007. Turbulence and heat transfer of a hairpin
Saha, A.K., 2013. Unsteady flow past a finite square cylinder mounted on a wall at low vortex formed behind a cube in a laminar boundary layer. Numer. Heat Transfer Part
Reynolds number. Comput. Fluids 88, 599–615. A 52 (11), 973–990.
Sakamoto, H., Arie, M., 1983. Vortex shedding from a rectangular prism and a circular

280