International Journal of Heat and Fluid Flow 83 (2020) 108569

Contents lists available at ScienceDirect

International Journal of Heat and Fluid Flow

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

Aerodynamic forces and three-dimensional flow structures in the mean wake T

of a surface-mounted finite-height square prism
Barbara L. da Silva , Rajat Chakravarty, David Sumner, Donald J. Bergstrom
⁎

Department of Mechanical Engineering, University of Saskatchewan, 57 Campus Drive, Saskatoon, Saskatchewan, S7N 5A9, Canada

ARTICLE INFO ABSTRACT

Keywords: Different flow models have been proposed for the flow around surface-mounted finite-height square prisms, but
Flow structures there is still a lack of consensus about the origin and connection of the streamwise tip vortices with the other
Near wake elements of the wake. This numerical study was performed to address this gap, in addition to clarifying the
Near-wall flow relationship of the near-wake structures with the far wake and the near-wall flow, which is associated with the
Tip vortices
fluid forces. A large-eddy simulation approach was adopted to solve the flow around a surface-mounted finite-
Surface-mounted finite square prism
Bluff body
height square prism with an aspect ratio of AR = 3 and a Reynolds number Re = 500. The mean drag and
Large eddy simulation normal forces and the bending moment for the prism were quantitatively compared in terms of skin-friction and
pressure contributions, and related to the near-wall flow. Both three-dimensional visualizations and planar
projections of the time-averaged flow field were used to identify, qualitatively, the main structures of the wake,
including the horseshoe vortex, corner vortices and regions of high streamwise vorticity in the upper part of the
wake. These features showed the same qualitative behavior as reported in high Reynolds number studies. It was
found that some regions of high streamwise vorticity magnitude, like the tip vortices, are associated with the
three-dimensional bending of the flow, and the tip vortices did not continuously extend to the free end of the
prism. The three-dimensional flow analysis, which integrated different observations of the flow field around
surface-mounted finite-height square prisms, also revealed that the mean near-wake structure is composed of
two sections of different origin and location of dominance.

1. Introduction Wang et al., 2004; Wang and Zhou, 2009; Kawai et al., 2012;
Sumner, 2013; McClean and Sumner, 2014; Sumner et al., 2015, 2017;
The cross-flow around surface-mounted finite-height obstacles has Beitel et al., 2019). The critical AR value is dependent on the relative
been a field of intense research in recent decades, with engineering boundary layer thickness of the ground plane measured at the position
applications ranging from the study of wind loadings on buildings and of the prism (δ/H or δ/D). It typically lies between AR = 2–5 for many
dispersion of pollution and contaminants in urban environments to the studies of prisms of circular or square cross-section (McClean and
cooling of electronic devices and enhancement of heat exchangers. This Sumner, 2014; Sumner et al., 2017). Fig. 1 presents a schematic of some
type of flow is composed of strongly three-dimensional structures. of the mean flow features around a surface-mounted square prism
These structures differ in their vortex wake formation mechanisms and partially immersed in a flat-plate boundary layer, where U∞ is the free-
interactions when compared to the classic alternate vortex shedding stream velocity and the prism is oriented with its front face normal to
past two-dimensional (2D or infinite) obstacles such as cylinders or the incident flow.
square prisms (Martinuzzi and Havel, 2004; Ozgoren, 2006), which are From this time-averaged perspective, the main flow structures that
the most commonly found cross-section geometries in engineering ap- can be found for a generic surface-mounted finite square prism with its
plications. front face normal to the incident flow (angle of incidence α = 0°) in-
The three-dimensionality of the flow field becomes even more re- clude the horseshoe vortex system, the tip and base counter-rotating
levant for surface-mounted prisms of small aspect ratio (AR), defined pairs of streamwise vorticity (the tip vortices and base vortices) and a
here as AR = H/D, where H is the height of the obstacle and D is its recirculation region downstream of the prism (Martinuzzi and
width. Below a critical aspect ratio, different flow behaviors and trends Tropea, 1993; Krajnović and Davidson, 2001; Yakhot et al., 2006;
for the fluid forces and vortex shedding frequency are observed Wang and Zhou, 2009; McClean and Sumner, 2014; Sumner et al.,
(Sakamoto and Arie, 1983; Okamoto et al., 1990; Sumner et al., 2004; 2017; Zhang et al., 2017). The mean flow field in the vertical (xy)

⁎
Corresponding author.

https://doi.org/10.1016/j.ijheatfluidflow.2020.108569
Received 15 July 2019; Received in revised form 23 January 2020; Accepted 22 February 2020
Available online 09 March 2020
0142-727X/ © 2020 Elsevier Inc. All rights reserved.
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 1. Schematic of the mean flow features around a surface-mounted finite-height square prism. Note the coordinate system, where x corresponds to the streamwise
direction, y to the vertical direction and z to the transverse direction.

symmetry plane (Fig. 1b) typically presents a main vortex with its focus Zhou (2009) throughout their considered range of AR = 3–7, with two
point downstream of the trailing edge, referred to as vortex Bt by different symmetric and antisymmetric modes. The antisymmetric
Krajnović (2011) and Sumner et al. (2017). It may also contain a second vortices were found to be more probable at the mid-span, while the
vortex near the junction of the prism with the ground surface de- downwash and upwash flows promoted the symmetric mode and in-
nominated vortex Nw. A saddle point demarks the zones of influence of creased its probability near the ends of the prism. The critical AR would
these two vortices in the recirculation region, and of the downwash and then be associated with a higher probability of the symmetric mode for
upwash of the flow away from the prism. The base streamwise vorticity lower AR and antisymmetric mode for higher AR, which is supported by
pair is associated with a strong upwash, but is not found in every study the early observations of Sakamoto and Arie (1983).
as its occurrence depends on the prism aspect ratio and on the relative The model of Wang and Zhou (2009) was later contested by
boundary layer thickness. For thin boundary layers the base vortices Bourgeois et al. (2011) and Sattari et al. (2012). These latter studies
tend to be absent, characterizing a dipole wake type, but for thick observed that the phase-averaged wake of a prism with AR = 4 was
boundary layers the base vortices are present and define a quadrupole always antisymmetric. There were occurrences of a co-existing sym-
wake type (Wang et al., 2006; Hosseini et al., 2013; Sumner et al., metric vortex pair in the formation region, but these were followed by
2017). A transition state between dipole and quadrupole wakes was the alternate shedding of vortices at the end of this region. For their
identified and designated a “six-vortices type” wake by case, the time-averaged wake was a dipole-type and a new half-loop
Zhang et al. (2017). structure topological model was proposed based on a phase-average
In early flow models, the vortex structures in Fig. 1 were thought to analysis. The half-loop vortical structure consists of a nearly vertical leg
be mostly independent. One of the first illustrations of this model with at the ground surface, which is the principal core and resembles a
the inclusion of the tip (or trailing) vortices was made by Kármán vortex, but with a predominantly streamwise connector strand
Kawamura et al. (1984), in this case for a circular cylinder. From an at the top of it that links it to the adjacent half-loops, as they are shed
instantaneous perspective, the near wake of the prism was proposed to alternately. For this model, the streamwise tip vortices were explained
present antisymmetric von Kármán vortex shedding along its span for to be a result of the time-averaging of the connector strands
aspect ratios above the critical threshold, except for regions close to the (Bourgeois et al., 2011, 2012). Hosseini et al. (2013) extended the
free end or the prism-wall junction where it was inferred that the tip model for a square prism with a quadrupole wake, describing a full-loop
and base vortices would suppress this mode (Wang et al., 2004). Below structure connected to the neighboring structures at both ends.
the critical AR, an arch-type or hairpin vortex was proposed to prevail, The relationship of the critical AR with this new flow model per-
which consists of two spanwise “legs” connected near the free end and taining to flow structure changes is not yet clear. However, as was
shed in a contiguous manner from the top and lateral surfaces mentioned previously, recent studies carried out for surface-mounted
(Sakamoto and Arie, 1983; Sakamoto, 1985; Kawai et al., 2012). square prisms of higher AR found a second and even a third critical AR
Wang et al. (2004) described the flow field for low AR as being domi- that caused changes in the frequency along the prism height.
nated by the downwash promoted by the tip vortices, which interacted Moreau and Doolan (2013) were one of the first to identify more than
with the base vortices. one spectral peak along the prism span for AR > 8.7, accompanied by a
This model did not, however, satisfactorily explain the formation of change in the trend of the overall sound pressure level for AR > 6.8.
the streamwise tip and base vortices and their interaction with the in- This behavior was later confirmed by Porteous et al. (2017), who found
stantaneous flow structures, as evidence such as a uniform frequency additional spectral peaks at even higher AR. The authors classified the
peak along the prism span suggested that these flow structures are not flow into four regimes according to the number of acoustic peaks pre-
isolated (Wang et al., 2004; Wang and Zhou, 2009; Sattari et al., 2012; sent in the spectrum (0–3), and identified the dominant structures re-
Hosseini et al., 2013; McClean and Sumner, 2014) (at least for AR sponsible for each frequency based on the phase coherence between the
<9~10 (Moreau and Doolan, 2013; Porteous et al., 2017)). A different signals of a hot-wire anemometer and a microphone. The R0 regime
model was proposed by Wang and Zhou (2009) based on instantaneous was found for AR < 2, the primary critical AR, which is the same value
particle image velocimetry (PIV) measurements, which consisted of an reported by Sakamoto and Arie (1983) for a surface-mounted square
arch-type structure with its connection, near the free end, and the prism. The flow in this regime had no acoustic peak in the spectrum,
bottom of the “legs”, near the ground surface, bent upstream by the and the dominant flow structure was described as mostly vertical vortex
downwash and upwash flows, respectively. The structure was also filaments, shed from the lateral surfaces of the prism, that are inclined
identified by Kawai et al. (2012) for a prism with AR = 2.7. In this downstream near the free end. The RI regime occurred for
model, the origin of the tip and base vortices was attributed to the 2 < AR < 10, with the sole peak in the spectrum caused by the half-
streamwise projection of the arch structure in the yz plane (following loop structures. The RII regime presented a second peak of higher fre-
the axes of Fig. 1). This structure was observed by Wang and quency for 10 < AR < 18. In this regime, the first peak remained, but

2
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

with a lower frequency. The authors showed that the low frequency dimensional visualizations will be used to make a comparison with the
peak was caused mostly by vortical structures near the tip of the prism, planar projections of the mean flow field, that are commonly used in
deformed upstream near the free end, and the high frequency peak was the literature.
caused by vortex filaments close to the mid-span. These same structures
occur at regime RIII for AR > 18, with the addition of a third peak at a 2. Computational models and methods
frequency intermediate to the other two, and associated with down-
stream-inclined vortex filaments near the base of the prism. Note that A large-eddy simulation (LES) turbulence approach was adopted for
the critical AR thresholds are expected to be influenced by the relative the present numerical analysis. The filtered Navier-Stokes equations
boundary layer thickness. which describe the large-scale and incompressible flow, for a
Although significant effort has been made to improve the under- Newtonian fluid, are given by
standing of the near-wake structures of the flow based on time-average,
phase-average or instantaneous analyses, little attention has been given ui
= 0 and
to their relationship with the near-wall flow on the prism and ground xi (1)
plane surfaces. The near-wall flow was experimentally investigated by
means of oil-film visualization, but usually with emphasis on the u¯ i u¯ i u¯ j 1 p¯
SGS
ij
+ (u¯ i u¯ j ) = + ,
horseshoe vortex on the ground surface or in the occurrences of se- t xj xj xj xi xi xj (2)
paration and reattachment on the surfaces of the prism (Martinuzzi and
Tropea, 1993; Nakamura et al., 2001; El Hassan et al., 2015). Near-wall where ui or (u, v, w) are the filtered velocity components along the
flow investigations for surface-mounted square prisms were also per- Cartesian coordinates xi or (x, y, z), ρ is the fluid density and p̄ is the
formed with particle image velocimetry (PIV) by Depardon et al. (2005) filtered pressure. The subgrid-scale (SGS) stress tensor ijSGS accounts for
and Sumner et al. (2017). The use of computational fluid dynamics the effects of the small scales, and it was calculated using the dynamic
(CFD) techniques facilitates the analysis of the relationship between Smagorinsky subgrid-scale model of Germano et al. (1991), with the
near-wall and wake flows by providing the complete flow field around stability modifications proposed by Lilly (1992). A specific description
the obstacle. Although simulations are usually restricted to low Rey- of the model is omitted here for brevity, and the reader is referred to the
nolds numbers Re = DU∞/ν (where ν is the fluid kinematic viscosity aforementioned references.
and the characteristic length D is adopted as the surface-mounted finite- The finite-volume method was used to discretize the filtered Navier-
height prism width), they typically present the same behaviors ob- Stokes equations. The problem was solved using an in-house code with
served at higher Reynolds numbers (Behera and Saha, 2019). This point the explicit second-order Adams-Bashforth method as part of a two-step
is also mentioned by Rastan et al. (2017). Sau et al. (2003) investigated fractional-step procedure. A pressure correction method was im-
vorticity cross-cancelation effects by doing a direct numerical simula- plemented and solved with a four-level multigrid scheme to ensure
tion (DNS) of the flow around a surface-mounted prism with AR = 1.7 mass conservation. The central differencing scheme was used for the
and Re = 225–500, showing the wake and surface flow topologies. spatial discretization of advective and viscous terms, except for the low-
Saha (2013) performed DNS for the flow past surface-mounted square turbulence region upstream of the prism where the QUICK scheme was
prisms with AR = 2–5 and Re = 250 to evaluate the influence of AR on used for the advective terms, to stabilize the dynamic Smagorinsky
the mean and instantaneous flow field. The near-wall flow on the prism subgrid-scale model. The in-house code and LES model implementation
with AR = 4 was presented and the spanwise variation of the drag were extensively validated for channel flows (Wang and
coefficient was evaluated, but it was not directly linked to the flow Bergstrom, 2005; Yin et al., 2007; Wang et al., 2008; Yin et al., 2008)
structures. Zhang et al. (2017) studied the effects of different values of and for the flow around an infinite square prism in proximity of a wall
Re = 50–1000 (and, as a consequence, different δ/H = 0.08–0.29) for (Samani and Bergstrom, 2015). The flow around a surface-mounted
the flow around a surface-mounted square prism with AR = 4. The finite-height square prism based on the in-house code has not yet been
topological structure of the mean streamwise vortices and the in- fully documented in a journal paper, therefore a careful comparison of
stantaneous structures were highlighted, while the mean near-wall flow the present results with DNS results from the literature at similar con-
on the prism surfaces was compared for their different cases. A more ditions is carried out in Section 3.
detailed investigation of the near-wall flow was reported by The computational domain is the one presented in Fig. 1a, for a
Cao et al. (2019), using an implicit large-eddy simulation (LES) ap- surface-mounted finite-height square prism of D = 0.02 m and AR = 3
proach for prisms with AR = 3 and 4 and Re = 5 × 104. Krajnović and (the figure does not represent the true extent of the domain in the x and
Davidson (2001, 2002) and Yakhot et al. (2006) provided a more y directions). The axes origin is located at the junction of the prism
thorough description of the flow structures, but limited to a surface- center with the ground plane, with the y axis oriented in the vertical
mounted cube, using LES with Re = 40,000 and DNS with Re = 1870, direction. The domain extends 3.5D upstream and 12.5D downstream of
respectively. the prism center, totaling 16D in the streamwise (x) direction, 6D in the
Considering the information in the literature about the flow around vertical or spanwise (y) direction and 7D in the transverse (z) direction.
surface-mounted finite-height square prisms, two main points moti- The corresponding blockage ratio is 7.1%, calculated as the cross-sec-
vated the present study. The first is the lack of agreement and con- tional area occupied by the prism divided by the cross-sectional area of
sistency on the origin of the tip vortices, their development, and their the domain, normal to the streamwise direction. This domain size was
relationship with the other structures in the prism wake. A second chosen based on the validated LES of the flow around a surface-
motivation lies in the connection between the three-dimensional mounted finite-height circular cylinder with AR = 2.5 of Fröhlich and
structures observed in the near and far wake with the near-wall flow on Rodi (2004). A non-uniform hexahedral grid (Fig. 2a and b) was created
the prism and ground plane surfaces. This issue has not yet been clearly with 128, 144 and 96 control volumes in the x, y and z directions,
addressed in the literature, even though it is of great importance in respectively, using hyperbolic-tangent functions to refine it near the
engineering design to relate the flow field with the fluid forces on the prism and ground plane surfaces. The prism edges contain 32 control
prism. The present study aims to describe the time-averaged three-di- volumes in the x and z directions, and 96 in the y direction, which is
mensional wake flow structures around a surface-mounted finite-height comparable with the resolution chosen by Saha (2013) in their grid
square prism with AR = 3 and Re = 500, including their connection independence test, and is higher than the resolution adopted in the DNS
with the near-wall flow on the ground plane and prism surfaces. In of Sau et al. (2003). The values of the dimensionless wall distance y+
addition, the forces caused by the flow around the prism will be ac- ranged between 0.01 and 0.88 on the ground plane and between 0.01
counted for, and both two-dimensional projections and three- and 1.74 on the prism walls. The highest values were located near the

3
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 2. Computational grid: (a) top xz-plane view, and (b) symmetry xy-plane view. LES grid resolution evaluation parameters in the xz-plane at y/H = 0.5: (c) ratio
of the maximum local grid dimension and the Kolmogorov length scale, and (d) ratio of the subgrid-scale viscosity and fluid viscosity (flow from left to right).

leading edges of the prism, where separation occurs. A uniform flow velocity of U∞ = 0.375 m/s was adopted at the inlet
The grid resolution has an important role in determining the range boundary, upstream of the prism, which gave a Reynolds number of
of scales that are resolved or delegated to the subgrid-scale model. Re = 500. This condition resulted in a laminar flat plate boundary layer
Therefore, the quality of the LES was assessed by comparison of the with a zero pressure gradient, which gave a relative boundary layer
local maximum grid dimension Δ with the Kolmogorov length scale η, thickness of δ/H = 0.13, based on the prism height H (or δ/D = 0.3), a
estimated as η = (ν3/ε)1/4 (Pope, 2000). The turbulence dissipation rate displacement thickness of δ*/H = 0.035 (or δ*/D = 0.105), a mo-
ε is given by the sum of both resolved and subgrid-scale contributions, mentum thickness of θ/H = 0.16 (or θ/D = 0.49) and a shape factor
following the definition in Meneveau and Katz (2000). The mean Kol- equal to 2.14 at the location of the prism. The prism and ground plane
mogorov length scale ranged from η/D = 0.005 to 0.32 in the vicinity surfaces were defined by a no-slip condition. A convective outflow
of the prism and wake regions. It gave a maximum ratio of Δ/η = 19, boundary condition was specified for the domain outlet, at the
staying below 4 near the prism except by the leading edges, where Δ/η boundary downstream of the prism, and a free-slip condition was
reached a maximum of 10. Therefore, the present LES is resolving scales adopted for the top and side boundaries of the domain.
with the same order of magnitude or one order higher than the Kol- A variable time step in the order of 1 × 10−4 s was used to ensure a
mogorov scale, which is considered sufficient to ensure a high quality maximum CFL number of 0.3 throughout the simulation. After the flow
simulation of the flow. Another criterion to evaluate the quality of the became fully developed, statistics of the velocity field were taken
LES is the ratio between the subgrid-scale viscosity and the fluid visc- during 50,000 time steps, which correspond to approximately 10
osity, νSGS/ν. It showed a maximum of 3.7, located at the mixing regions shedding cycles.
downstream of the surface-mounted prism, which further confirms that
the contribution of the subgrid-scales to momentum transport is 3. Results and discussion
minimum. Fig. 2c and d illustrate the described grid resolution eva-
luation criteria in the xz-plane at y/H = 0.5. In this section, the fluid force and bending moment coefficients of

4
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Table 1
Mean drag and normal force coefficients at individual prism faces and Strouhal number.
Prism surface AR Re δ/H St C̄D C̄N
Skin-friction Pressure Skin-friction Pressure
Front (x-) – 0.789 0.055 –
Back (x+) – 0.561 0.020 –
Side (z-) −0.010 – 0.013 –
Side (z+) −0.010 – 0.015 –
Top −0.004 – – 0.670
Net contributions −0.024 (−1.8%) 1.350 (101.8%) 0.102 (13.3%) 0.670 (86.7%)
Total (present) 3 500 0.13 0.092 1.33 0.77
ESDU (1978)a 3 1 × 104 – 1 × 106 < 0.2 – 1.25 0.70
b
Sakamoto and Oiwake (1984) 3 1.88 × 104 – 9.38 × 104 0.67 0.1 1.22 –
Sakamoto (1985)c 3 104 0.13 0.1 1.60 –
Uffinger et al. (2013)d 6 1.28 × 104 0.02 0.1 1.41 –
Uffinger et al. (2013)e 6 1.28 × 104 0.02 0.11 1.62 –
Saha (2013)f 3 250 << 1 0.114 1.16 –
McClean and Sumner (2014)b 3 7.3 × 104 0.5 0.10 1.29 –
Zhang et al. (2017)f 4 500 0.11 0.113 1.298 –
Behera and Saha (2019)f 7 250 << 1 0.115 1.24 –
Cao et al. (2019)g 3 5 × 104 6.7 0.094 1.23 –
Heng (2019)a 3 6.5 × 104 0.28 0.11 – 0.82
Heng (2019)b 3 1.1 × 105 0.28 – 1.45 0.17

Method of force determination:

a
Integrated from pressure data.
b
Force measurement.
c
Correlation for δ/H = 0.13.
d
Calculated from momentum balance from laser Doppler velocimetry data.
e
Scale-adaptive simulation (SAS).
f
DNS.
g
LES.

the surface-mounted finite-height square prism (Re = 500, AR = 3, δ/ the mean drag force located at ypoa/H = 0.48 or ypoa/D = 1.44. There is
H = 0.13) will first be evaluated, followed by the mean flow field a considerable spread of C̄D in the literature due to variations in AR, Re
discussions. These will be based on a planar projection analysis in- and δ/H (Table 1), but the present value agrees with other studies
cluding the near-wall flow in Section 3.2, and on three-dimensional performed at similar conditions, such as Zhang et al. (2017). The same
visualizations in Section 3.3. is observed for the Strouhal number St = 0.092, which is close to the
average value of 0.1 reported in the literature. The point of action of the
3.1. Force and bending moments coefficients total drag force in the present study is lower than the approximately
constant value reported for surface-mounted finite-height circular cy-
The contributions to the mean drag coefficient C¯D = 2F¯D /( U2 DH ) , linders, of ypoa/H = 0.54–0.55 (Taniguchi et al., 1981; Beitel et al.,
normalized by the projected surface area of the front of the square 2019), but the same value was reported in the Engineering Sciences
prism, and normal force coefficient C¯N = 2F¯N /( U2 D 2) , normalized by Data Unit 71,016 database (ESDU, 1978) for a surface-mounted prism
the surface area of the free end, were calculated for each of the surface- subjected to a “uniform flow distribution” (δ/H → 0). The integration of
mounted prism surfaces to account for their individual contribution to the spanwise distribution of C̄D obtained by Saha (2013) for a surface-
the total coefficient, be it due to skin-friction or pressure forces. Table 1 mounted prism with AR = 3 also gives ypoa/H = 0.47.
presents the coefficients, discriminated by source (skin-friction or While integrated pressure data are a reasonable approach to cal-
pressure) and surface, followed by the net contributions of each source, culate the mean drag coefficient for the prism, the same is not true for
and the total coefficient value including both sources and coefficients the mean normal force coefficient. For this prism with AR = 3, the skin-
reported by other authors for surface-mounted finite-height square friction forces represent 13.3% of the total normal force, which is
prisms. The Strouhal number St = fD/U∞, where f is the dominant flow possibly why the present value is 10% higher than the value reported in
frequency, was also included for comparison with the present study, as the ESDU (1978), which was based on pressure only. The present
well as the respective AR, Re and δ/H. The force coefficients, Strouhal pressure-only normal coefficient, however, agrees with the one re-
number and pressure and skin-friction coefficients, to be presented in ported in the ESDU (1978). All sources of the normal coefficient are
the next section, have been corrected for blockage following the positive, which means that there is a mostly upward flow close to all
method of Maskell (1963). vertical surfaces of the prism, especially its front face, and an overall
As expected, the drag coefficient caused by skin-friction forces is negative pressure at the top of the prism, as a result of the separated
almost negligible when compared with the pressure drag coefficient. flow. However, a different behavior was observed by Heng (2019), in
The skin-friction drag coefficient is negative due to the reverse flow which case the skin-friction normal force acted in the downward di-
taking place near the side and top surfaces of the prism, as will be rection, providing a total C̄N which is lower than the pressure-only
shown in the next section, and it represents 1.8% of the total drag, close contribution. This discrepancy may be related to the differences in Re
to the 1.5% fraction obtained by Diaz-Daniel et al. (2017) for a surface- and δ/H. The mean point of action xpoa/D of the present normal force
mounted cube (Re = 3000, δ/H = 2.38). The front surface is the one was of −0.04, with a pressure-only xpoa/D of −0.02 which is the same
subjected to the highest pressure drag force due to it being the leading center of pressure reported in the ESDU (1978).
surface, but the back of the prism also presents an almost as high a The mean drag and normal forces were used to compute the mean
contribution to the drag coefficient, which is related to the strong base bending moment coefficient C¯ Mz = 2M ¯ z /( U2 DH 2) , where M̄z is the
suction caused by the separated wake flow. mean bending moment about the z axis at the junction of the prism with
An overall drag coefficient of C̄D = 1.33 was obtained considering the ground plane, and with a clockwise sense of rotation having a po-
both skin-friction and pressure contributions, with the point of action of sitive value. The total C̄Mz was 0.644, resulting from the individual

5
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Table 2 on the back of the prism are very low, which lead to weak values of |C¯f |
Bending moment coefficient C̄Mz contributions from drag and normal forces, on most of the surfaces and the small skin-friction force coefficient
due to skin-friction and pressure. contributions shown previously in Table 1. The flow goes mostly up-
Source of contribution From drag forces From normal Total ward on all vertical surfaces, which causes the positive skin-friction
forces normal force coefficients discussed in the previous section.
A higher wall shear stress occurs at the front of the prism before the
From skin-friction −0.014 0.002 (0.3%) −0.012 (−1.9%)
incident flow separates at the three leading edges, causing the highest
(−2.1%)
From pressure 0.654 (101.5%) 0.002 (0.3%) 0.656 (101.9%)
values of |C¯f |. This behavior is also observed at the top and lateral or
Total 0.640 (99.4%) 0.004 (0.6%) 0.644 (100%) side surfaces, as the flow approaches the leading edges, but |C¯f | is
ESDU (1978) 0.600 0.014 0.614 continuously low near the other top edges, suggesting that the upward
flow remains attached to the prism. Another minor high-|C¯f | region
takes place at the back of the prism, near the lateral edges, at a height
contributions organized in Table 2 from drag or normal forces, due to close to the rear stagnation point.
skin-friction or pressure. Stagnation points N1 and N2 are observed on the front and rear
Table 2 shows that the main source of the bending moment about surfaces of the prism, located at y/H = 0.19 (y/D = 0.56) and y/
the z axis is the pressure drag force. The contribution from the normal H = 0.16 (y/D = 0.47), respectively. Both stagnation points are closer
forces is negligible due to the small respective lever arms, and the to the ground plane when compared with the values obtained experi-
bending moment coefficient caused by the skin-friction source is also mentally by Sumner et al. (2017) for a prism with the same AR, espe-
negligible due to the overall small magnitude of the skin-friction force cially the front one, which is an effect of the thinner boundary layer in
coefficients (Table 1). The force coefficients and centers of pressure of the present simulation (Zhang et al., 2017). While the front stagnation
the ESDU (1978) gave a 5% lower C̄Mz due to the magnitude of the drag point is at the symmetry plane of the prism, the one at the back is
coefficient. dislocated in the negative z direction. Saha (2013) also found in their
simulation that this unstable center was instantaneously asymmetrical.
3.2. Plane-projected flow field Their prism with AR = 4 showed two stable focus points on the top, but
these are absent in the present flow field, which agrees more with the
It is common practice to present time-averaged flow field results by outward-bent streamlines of Sumner et al. (2017) and Cao et al. (2019)
means of planar projections, so the same procedure will be taken as a for their prism with AR = 3. Reattachment does not occur on the side
first approach. Fig. 3 presents the mean wall shear stress on the prism surfaces as happened in the case of Cao et al. (2019), but a different
surfaces, expressed by the skin-friction coefficient components trend in the streamline distribution at the bottom of the side surfaces in
C¯fi = 2 wi/( U2 ) , where τwi is the wall shear stress caused by the shear Fig. 3 was also observed, matching the junction-influence region of
velocity in the i wall-parallel direction. The contours show the magni- Cao et al. (2019). This behavior is due to the interaction of the sepa-
tude of |C¯f |=(C¯ fi + C¯ fj )0.5 for each surface, accompanied by “stream-
2 2 rated flow with the boundary layer on the ground plane, and it will be
lines” of the skin-friction coefficient vector field. The separated and addressed further in this section.
recirculating flow at the sides and top of the prism is evident from the Given the small asymmetry in the rear stagnation point, the field
upstream-directed streamlines. The flow velocities in these regions and variables were mirrored and averaged with reference to the vertical
(xy) symmetry plane for further analyses, and to simplify the flow field
visualization.
Fig. 4a shows the mean pressure coefficient C¯p = 2(p p )/( U2 )
on the surfaces of the prism and the ground plane, together with
streamlines for the very-near-wall velocity field, for the negative-z half
of the domain. In the equation, p is the local pressure and p∞ is the
reference free-stream pressure (in this case, equal to zero). The
streamlines are continuous along the prism edges in the regions where
the flow remains attached to the walls, and discontinuities can be ob-
served when the flow separates, with the streamlines following se-
paration lines near the edges. As expected, the pressure coefficient is
positive and the highest on the front face of the prism, and it attains
negative values on the other surfaces due to the separated and re-
circulating flow. The most negative pressure occurs near the leading
edges, confirming the flow separation.
The very-near-wall streamlines and pressure coefficient on the
ground plane can be seen in more detail in Fig. 4b. Critical saddle (S),
nodal (N) and focal (F) points are also highlighted in the figure. A
horseshoe vortex system can be seen upstream of the prism, forming
due to the separation of the boundary layer and its subsequent rolling-
up. The system can be divided into two regions by the streamlines that
start collapsing close to point P1 at x/D = −1 in Fig. 4b, a behavior
also reported by other authors for a range of Re and δ/D
(Martinuzzi and Tropea, 1993; Martinuzzi and Havel, 2000; Hwang and
Yang, 2004; El Hassan et al., 2015; Kindree et al., 2018). The size and
strength of the horseshoe vortex scale with the boundary layer thick-
ness (Castro and Robins, 1977), but remain approximately the same for
varying AR (Sumner et al., 2017), which enables the comparison of the
Fig. 3. Mean skin-friction coefficient distribution on the surfaces of the prism. position of critical and notable points with other works of different AR.
The streamlines correspond to the skin-friction coefficient vector field. N1 and The saddle point of separation S1 of the present horseshoe vortex
N2 denote the nodal stagnation points on the front and rear surfaces. system is located at x/D = −1.5, where the first zero-crossing occurs

6
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 4. Mean pressure coefficient contours and streamlines for the time-averaged velocity field: (a) very near the surface of the prism and the ground plane, and xz-
planes at (b) y/H = 0.002 (near the ground plane), (c) y/H = 0.16 and (d) y/H = 0.83. Note that the direction of the flow is from right to left in (b), (c) and (d).

1.5D from its center. In agreement with the oil-film visualization of

El Hassan et al. (2015) and the pressure contours of Sakamoto (1985),
the horseshoe vortex arms converge to a region of low pressure
downstream of the prism, constituting the stable focus point F1 where
separation takes place. At a higher plane located at y/H = 0.16
(Fig. 4c), this focus point is shown to develop and expand into a large
vortex inside the recirculation region. This large vortex, typically as-
sociated with the rolling-up of the free shear layer at the side of the
prism, will be shown in the next section to originate from the flow near
the ground plane. On the other hand, at y/H = 0.83 (Fig. 4d) the vortex
disappears, indicating that, at this first glance, the legs at each side of
the domain may be connected forming an arch-type structure, in
agreement with the one found by other authors for similar AR
(Kawai et al., 2012; Wang and Zhou, 2009).
Other notable points in Fig. 4b are the saddle point of separation S2
at x/D = 1.1, the unstable node of attachment N3 at x/D = 4.8 and z/
Fig. 5. Mean friction coefficient in the x direction along the symmetry line on
the ground plane.
D = −0.4, the saddle point of attachment S4 at x/D = 5.3 and the
neutral saddle point S3 at x/D = 4.3 and z/D = −1.4. The referred
neutral saddle point S3 marks the position where the wake is the nar-
for C̄fx along the symmetry line on the ground plane, as confirmed by rowest before its growth due to the reattachment of the shear layer and
the distribution in Fig. 5. Although variations would be expected, this the consequent fluid entrainment from its surroundings (Martinuzzi and
same position was found in the literature for δ/D ranging from 0.07 to Tropea, 1993). The saddle point of attachment S4 corresponds to the
an atmospheric-like distribution, for Re in the order of 104 reattachment of the shear layer on the symmetry plane, which is also
(Martinuzzi and Tropea, 1993; Martinuzzi and Havel, 2000; Saeedi and the maximum mean recirculation length LR, as shown by Fig. 6. The
Wang, 2015). Considering studies performed at lower Reynolds num- mean recirculation length is here defined as the streamwise distance
bers, the separation point of the horseshoe vortex was located at x/ from the center of the prism to the last point where the mean stream-
D = −1.85 in the simulation of Hwang and Yang (2004) (Re = 350 wise velocity is equal to zero.
and δ/D = 1), at x/D = −1.6 for Diaz-Daniel et al. (2017) (Re = 500 Fig. 6 also presents mean z-vorticity contours and velocity stream-
and δ/D = 1) and at x/D = −1.71 for Yakhot et al. (2006) (Re = 1870 lines on the symmetry plane. Vortices Bt and Nw, following the no-
and δ/D = 1.5). menclature of Krajnović (2011) and Sumner et al. (2017), can be found
Another zero-crossing occurs in front of the prism front wall as in the wake. Vortex Bt represents the connection between the vortex
shown by Fig. 5, at x/D = −0.58. This nodal point of attachment was legs, visible as the highlighted nodal point N4 in Fig. 4d. The position of
also identified by Hwang and Yang (2004), and it can be seen in Fig. 4b its center is lower and farther from the prism than in the case of
although it is not highlighted. The minimum C̄fx at x/D = −0.84 in Sumner et al. (2017), showing a behavior closer to the low-Re simula-
Fig. 5 is associated with an intense wall shear stress that takes place tions of Saha (2013) and Zhang et al. (2017), with a thin boundary
near the center of the main horseshoe vortex, which can be seen in the layer. The present vortex Nw has its region of influence delimited by the
xy symmetry plane in Fig. 6. nodal point N2 and saddle point S2 in Fig. 4. It is smaller than in the case
Fig. 4b shows that the horseshoe vortex arm extends along the side of Saha (2013) with AR = 3, but of a size similar to their prism with
of the prism with the outer separation line reaching a distance of up to AR = 4. On the other hand, vortex Nw was absent for the prism with

7
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 6. Mean transverse vorticity ( ¯ z D / U ) contours and streamlines for the time-averaged velocity field in the symmetry xy-plane. The mean recirculation length LR
along the prism height is highlighted by the dotted line, with the mean recirculation length at mid-height marked as reference.

AR = 3 of Sumner et al. (2017), but its position for their prism with
AR = 5 is close to the one found in the present study. This occurrence
indicates that the present AR is above the critical one for the thin
boundary layer of δ/H = 0.13. Vortex Nw also appeared in simulations
for the channel flow past surface-mounted cubes under δ/H = 1–1.5
(Krajnović and Davidson, 2001; Yakhot et al., 2006).
A distinctive feature of the time-averaged near wake in Fig. 6 is that
the downwash prevails along most of the prism span, causing a max-
imum value of LR = 5.3D and with only a weak upwash taking place
near the ground plane, demarked by the saddle point S8 at x/D = 4.7
and y/H = 0.02. This same behavior was found for all AR investigated
by Saha (2013) and in the low-Re simulation results of
Zhang et al. (2017) for AR = 4, for the thinner boundary layer cases.
The downwash for the prism with Re = 250 and δ/H = 0.145 in
Zhang et al. (2017) did not reach the ground plane, presenting a saddle
point at y/H = 0.22, but it did so for their case with Re = 500 and δ/
H = 0.11. The present study conditions and results (δ/H = 0.13 and
saddle point at y/H = 0.02) fit between these two cases.
To aid in the analysis of the different behavior observed for the very
near-wall streamlines on the side of the prism near the ground plane in
Figs. 3 and 4, Fig. 7 presents the mean y-vorticity ( ¯ y ) and streamlines
in xz-planes in this region. A corner vortex is formed at the leading
corner, and it is shifted downstream until about y/H = 0.09, when it is
convected into the wake. Above y/H = 0.1, which is below the
boundary layer thickness, no evidence of this vortex is seen in the wake.
Okuda and Taniike (1993) studied the corner or side vortex structure
for a square prism of AR = 4 under δ/H = 0.09 and 0.52. For the
thinner boundary layer case, the authors reported a flow field like the
one over a two-dimensional flat plate, but for δ/H = 0.52 they ob-
served an inverted conical vortex that extended along most of the prism
span. It is likely from the streamline distribution in Fig. 7 that an in- Fig. 7. Mean spanwise vorticity ( ¯ y D / U ) contours and streamlines for the
verted conical vortex exists in the present results, but it is soon dis- time-averaged velocity field in xz-planes. (a) y/H = 0.002; (b) y/H = 0.04; (c)
persed due to the boundary layer being thinner than in the case of y/H = 0.05; (d) y/H = 0.07; (e) y/H = 0.09 and (f) y/H = 0.1.
Okuda and Taniike (1993).
A topic of great interest in recent decades related to the flow past prism (Wang et al., 2004), from the bending of the arch-vortex
surface-mounted finite-height square prisms is the origin of the so- (Wang and Zhou, 2009) or from the connector strands of half-loop or
called “tip vortices”, which are the high-magnitude streamwise vorti- full-loop structures (Bourgeois et al., 2011; Hosseini et al., 2013). Fig. 8
city regions that were proposed to form from the top corners of the shows plane-projected views of the mean x-vorticity in xy-planes in

8
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 8. Mean streamwise vorticity ( ¯ x D / U ) contours in: (a) the xy-plane with z/D = −0.25; (b) the xy-plane with z/D = −0.5; (c) the yz-plane with x/D = 3.6 and
(d) the yz-plane with x/D = 10. Streamlines for the time-averaged velocity field are included in the xy planes.

Fig. 8a and b and in yz-planes in Fig. 8c and d. vorticity region of high magnitude increases in size and decreases in
The evolution of vortex Bt can be seen in Fig. 8a and b as the vertical strength downstream, in agreement with Wang et al. (2004) and
xy symmetry plane is displaced in the negative z direction, although Wang and Zhou (2009). Its non-circular shape resembles that of prisms
vortex Bt’s relationship with the vertical legs cannot yet be clearly seen of higher AR (Wang et al., 2004; Wang and Zhou, 2009;
by the projected plots. Strong x-vorticity is produced from the separa- Unnikrishnan et al., 2017), although a similar mean wake structure was
tion of the flow at the top leading edge of the prism, with a negative also observed for a surface-mounted cube by Sumner et al. (2016).
(blue-colored) sign that corresponds to a clockwise rotation at a more Other streamwise vorticity regions of high magnitude take place in
elevated position and only at the beginning of the separated shear layer, the near wake and are highlighted in Fig. 8. In the xy-plane located at z/
and a positive or counter-clockwise x-vorticity (red-colored) that is D = −0.5, an additional focus point F3 appears at x/D = 3.4 and y/
found closer to the top surface of the prism. These two vorticity regions, H = 0.15 accompanied by a region of intense streamwise vorticity, also
highlighted in Fig. 8b above the top of the prism, have also been re- seen in Fig. 8c. This structure, as all other structures discussed up to this
ported by other authors (Wang et al., 2004; Sumner et al., 2017; point, is highly three-dimensional, and therefore a three-dimensional
Zhang et al., 2017). The positive x-vorticity remains along the sepa- visualization approach is necessary to clarify its features.
rated shear layer region, reaching its most intense magnitude at about
x/D = 3.6, which according to Fig. 6 is just downstream of the re-
circulation region. This vorticity region is the counter-clockwise vortex 3.3. Three-dimensional flow field visualization
of the pair of tip vortices, which is visible at z/D = −0.5 and also on
the yz planes shown by Fig. 8c and d. The clockwise tip vortex coun- A few features of the near wake could not be well-explained based
terpart would be visible in the other half of the domain. on the previous plane projection-based analysis. These include the
The mean streamwise vorticity distribution in Fig. 8c is almost the corner structures near the ground plane, the connection between the
same in shape and magnitude to that obtained by Zhang et al. (2017) mean near-wake vortex and the ground plane and vortex Bt, the tip
for a surface-mounted finite-height square prism with AR = 4 and vortices, the focus point F3, and the minor regions of high x-vorticity.
Re = 1000 at x/D = 5. Note that the authors’ plane at x/D = 5 and the This section will be dedicated to the better understanding of these time-
present plane at x/D = 3.6 are both located close to the mean re- mean structures, by using three-dimensional (3D) resources such as
circulation length at mid-span of each flow field. Fig. 8d presents the mean vorticity isosurfaces and 3D streamlines.
mean x-vorticity contours at x/D = 10, to allow for comparison with Vorticity-based methods have been questioned in the vortex iden-
the measurements of Unnikrishnan et al. (2017). In their case and for tification literature due to flow regions of poor swirling motion but high
the present plane at x/D = 3.6, the base vortices of negative vorticity shear also presenting vorticity – for example, in boundary layers
are absent, a characteristic of dipole wakes that is consistent with the (Jeong and Hussain, 1995; Cucitore et al., 1999). It also depends on the
strong downwash and thin boundary layer (Wang et al., 2006; specification of a threshold, but this point is also a difficulty en-
Hosseini et al., 2013; Zhang et al., 2017). However, at x/D = 10, a countered in other vortex identification methods, like the second in-
small region of negative (clockwise) vorticity forms due to the variant of the velocity gradient tensor (Q) criterion (Hunt et al., 1988)
streamwise evolution of the upwash, which leads to small regions of and the λ2 criterion (Jeong and Hussain, 1995). With this in mind,
negative x-vorticity starting from x/D = 7.5. These were also shown in vorticity is still a viable method for vortex visualization, with the ad-
the dipole wake of Zhang et al. (2017) and found by vantage that it allows for a separate analysis of the swirling motions in
Unnikrishnan et al. (2017) for their prism with AR = 5, but attributed each direction and in each rotation sense. As shown by Fig. 9, ap-
to induction from the tip vortices or the ground plane. The positive x- proximately the same structures are obtained by plotting the overlaid
vorticity components (Fig. 9d) and the second invariant of the velocity

9
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 9. Iso-surfaces of the individual mean vorticity components (a) ¯ x D / U , (b) ¯ y D / U and (c) ¯ z D / U , (d) overlaid mean vorticity components and (e) the Q
criterion for arbitrarily chosen thresholds.

gradient tensor Q (Fig. 9e), which is widely adopted for the analysis of vorticity other than the streamwise one. Fig. 11 presents z-vorticity
this type of flow (Krajnović and Davidson, 2001; Calluaud and isosurfaces near the prism junction with the ground plane, which
David, 2002; Wang and Zhou, 2009; Krajnović, 2011; highlights the corner vortex-dominated region and also details vortex
Unnikrishnan et al., 2017). The use of velocity streamlines to represent Nw. The isosurfaces and three-dimensional streamlines reveal that
vortices has the downside of not being Galilean invariant vortex Nw is mostly confined to the center region behind the prism. It
(Haller, 2005). Even so, they will be used here with a zero velocity disappears near the corner as the flow that comes from the recirculating
reference frame, which has been adopted so far in the analysis of the flow near the ground plane, to be described further in this section, is
plane-projected flow field and which is also commonly used in the lit- deflected toward the side of the prism. Some of the deflected stream-
erature of the flow around surface-mounted finite-height square prisms lines join the corner vortex or re-enter the recirculating flow, as de-
(e.g. El Hassan et al., 2015; Sumner et al., 2017; Zhang et al., 2017; picted in Fig. 11, before leaving the recirculation region.
Kindree et al., 2018; Cao et al., 2019). The regions of high x-vorticity in the upper part of the prism and the
The same topology obtained with Q in Fig. 9e was also obtained negative-z half of the domain are shown by means of isosurfaces and
with the λ2 criterion. This mean vortex structure is similar to the one three-dimensional streamlines in Fig. 12a and b. The positive, counter-
reported by Bourgeois et al. (2011) and Zhang et al. (2017), where the clockwise x-vorticity along the side edge on the top of the prism was
horizontal “tube” that goes into the far wake is associated with the referred to by Sumner et al. (2017) as an edge vortex, that was believed
streamwise vorticity. Near the prism, the flow is dominated by struc- to originate at the leading corner of the free end at both sides of the
tures that envelop the prism through its side and free-end, and by other prism. The present results show that the high vorticity region does not
smaller, but complex, regions. These minor Q regions in the near wake necessarily correspond to a vortex. In this case, the positive streamwise
are mostly associated with the streamwise vorticity. The enveloping vorticity of high magnitude does not come from the corner. Instead, it
structure corresponds to the y-vorticity (Fig. 9b) formed due to the originates from the upward flow along the side wall of the prism
separating shear layer on the prism side, while above the free-end and (shown in Fig. 3 and by the positive normal force coefficients in
near the ground plane complex flow structures are formed, composed of Table 1) and its entrainment into the recirculation region that takes
all three vorticity components. place over the free end. The flow turns toward the low pressure region
Fig. 10 shows the corner vortex by means of 3D streamlines. As at the leading edge of the prism and is carried away with the separated
indicated by Fig. 7, this structure is bent upward and downstream, shear layer, describing a motion that contains all three vorticity com-
causing a region of high-magnitude clockwise x-vorticity also included ponents, as visible in Fig. 9. The negative, clockwise vorticity over the
in the figure and similar to the one observed by Okuda and top of the prism agrees with the explanation given by
Taniike (1993). The flow that reaches the leading edge of the prism is Sumner et al. (2017), i.e. that it is associated with the outward de-
deflected downward and separates near the upstream corner, initiating flection of the mean flow. Note that this outward deflection fades not
this vortex. It is deflected upward due to the entrainment of fluid from very far from the prism (see Fig. 8) and is replaced by a positive x-
its surroundings, but it stays within the boundary layer until is it carried vorticity, namely the counter-clockwise tip vortex.
into the wake. Note that the corner vortex is also bent outward (away The tip vortices are, just like the edge vortices, surrounded by sig-
from the prism), so the streamlines encircle the mean recirculation nificant ambiguity, as the present evidence shows that the mean x-
region. vorticity in this region is a result of the three-dimensional bending of
As shown by Fig. 9, the corner vortex also contains components of the flow. Fig. 12c and d show from different angles that the flow

10
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 10. Iso-surfaces of the mean streamwise vorticity ¯ x D / U = −1.4 and 3D streamlines for the time-averaged velocity showing the corner vortex from different
view angles.

separates from the sides of the prism and encircles the near-wake re-
circulation region. It has an upward component closer to the free-end,
which was also detected by Kawamura et al. (1984) for surface-
mounted finite-height circular cylinders, but past the wake recircula-
tion length the side flow meets with the downwash from the top of the
prism and shifts downward in the far-wake region. This shift is the most
intense after the end of the recirculation region (see Fig. 6) and it be-
comes more subtle and spread out as the distance downstream from the
near wake increases. This behavior is shown in the mean x-vorticity
contours in Fig. 8d and is in agreement with the results of Wang and
Zhou (2009) and Zhang et al. (2017), for whom the highest x-vorticity
occurred near the closure of the wake recirculation region.
Bourgeois et al. (2011) also found that the streamwise vorticity, which
showed as connector strands in a phase-averaged perspective, was in-
duced by the y- and z-vorticities that arise from the separated shear
layers from the side walls and free-end, respectively. In summary, the
present results suggest that the flow structures over the tip (free end)
fade and are replaced by this large-scale streamwise vorticity due to the
three-dimensional bending of the side flow by the downwash. Conse-
quently, there is not a direct connection between this “tip” vortex and
the structures over the tip (free end) of the prism, in this time-averaged
perspective.
Fig. 11. Iso-surfaces of the mean transverse vorticity ¯ z D / U = −1.5 and 1.5 A different reference velocity frame was investigated to confirm this
and 3D streamlines for the time-averaged velocity near the location of vortex statement. It was extracted from the location of the maximum stream-
Nw. wise vorticity in the region of the “tip” vortex, in an attempt to have a
reference frame that is moving with the vortex. Fig. 13 shows the 3D

Fig. 12. Iso-surfaces of the mean streamwise vorticity ¯ x D / U = −1 and 1 and 3D streamlines for the time-averaged velocity located near the regions of high
vorticity.

11
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

Fig. 13. Iso-surfaces of the mean streamwise vorticity ¯ x D / U = −1 and 1

and 3D streamlines for the time-averaged velocity with the tip vortex reference
frame.

streamlines for the “tip” vortex reference frame. They describe a swir- Fig. 15. Semitransparent mean transverse vorticity ( ¯ z D / U ) contours in the
ling pattern around the region of high streamwise vorticity magnitude xy-plane at z/D = −0.5 and 3D streamlines for the time-averaged velocity field
showing sections of the mean near wake: (a) bottom and (b) top.
but, again, there is no evident connection with the structures over the
tip (free end) of the prism. The core of the swirling flow region is,
however, close to the focus point of vortex Bt, which suggests a con- Fig. 8b, at x/D = 3.4. All these streamlines ascend with a clockwise
nection between it and the streamwise vortices. sense along the y axis by spiraling around the low-pressure region. They
To comprehend the origin of the minor x-vorticity regions of high do so in a complex manner that causes a mostly positive z-vorticity,
magnitude (shown earlier in Fig. 8a), the mean structure in the near shown in the inset of Fig. 14, and both the positive and negative regions
wake is investigated based on the low-pressure region identified in of minor x-vorticity seen in Fig. 8 and also Fig. 12c and d. Therefore,
Fig. 4. Fig. 14 shows that this region could be associated with the leg of point F3 in Fig. 8b is related to the entrainment of fluid in the near wake
the arch vortex, and that it is composed of the fluid entrained from the and the spiraling flow, and the large vortex with the focal point F2 in
horseshoe vortex and the surrounding flow, near the ground plane, Fig. 4c is a plane-projected view of this structure. The entrainment of an
rather than from the separated shear layers. The flow that is still at- ascending spiraling flow was also found by Sau et al. (2003), who
tached on the ground plane separates close to the point F1 (see Fig. 4b). compared it with the upright vortices for a square jet in cross-flow, and
Other streamlines that are above the ground either reattach at point N3 the streakline patterns shown in the compilation of Nakayama (1988)
and then turn toward the separation point, or get entrained through the also reveal an inclined vortex core similar to the present flow behavior.
side of the low-pressure region, crossing the focus point F3 shown by The spiraling flow does not ascend, however, up to the position of

Fig. 14. Iso-surfaces of the mean pressure coefficient C̄p = −0.7 and 3D streamlines for the time-averaged velocity field near the low-pressure region. Spanwise and
transverse vorticities in this region are shown in the insets and the critical points from Fig. 4b are included for reference.

12
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

vortex Bt to become the arch vortex. Fig. 14 shows that the streamlines large-scale streamwise vorticity in the near and far wake is illustrated as
leave the near-wake recirculation region as they encounter the down- being due to the bending of the flow after it encircles the recirculating
wash, some of them traveling to the side of the prism before doing so. region behind the surface-mounted prism.
Indeed, this mean near-wake vortex structure contains two main sec- The present Re = 500 is compatible with engineering applications
tions, presented by Fig. 15. The bottom structure in Fig. 15a is the same like the cooling of electronic devices and compact heat exchangers, but
as presented in Fig. 14. It is composed mainly of fluid entrained from it is lower than in other applications such as wind loadings on build-
near the ground plane and represents an ascending spiraling motion, ings. Nonetheless, the surface-mounted finite-height square prism wake
bent upstream as it ascends along the bottom half of the prism. The top is fully turbulent. Based on previous studies and discussion, the present
structure is formed from the flow that separates at the free-end leading case is expected to represent high-Re flows in a qualitative sense
edge and curls behind the prism, resulting in vortex Bt close to the (Behera and Saha, 2019). Any limitations due to this condition should
symmetry plane. This vortex expands outward as if to connect with the be mostly related to the size of the flow structures for this type of bluff-
bottom structure, but the streamlines leave the recirculation region body (Lim et al., 2007): Zhang et al. (2017), for example, did not find
without going all the way to the bottom of the prism wake. significant changes in the shape of the mean flow structures when in-
The two sections, when seen together, resemble the vortex lines of creasing the Reynolds number from Re = 500 to Re = 1000.
Tanaka and Murata (1999) and the arch vortex model of Wang and Although several references state that the near-wake structures are
Zhou (2009). They are both part of the near wake, but the present re- highly coherent and shed with the same frequency, likely constituting a
sults indicate that these structures are of different origin and dominate single structure along the span of a surface-mounted finite square prism
different regions of the wake. The same behavior was observed in- (Wang and Zhou, 2009; Sattari et al., 2012; Hosseini et al., 2013;
stantaneously in the simulation of Wang et al. (2014), who referred to McClean and Sumner, 2014), there is also evidence of a different be-
the two sections as the upwash (bottom) and downwash (top) motions. havior near the free end. Kawai et al. (2012) identified, by means of a
The overall present structure is similar to the one reported by proper orthogonal decomposition (POD) analysis for a prism with
Porteous et al. (2017) for their RI mode, and compatible with the model AR = 2.7, a symmetrical mode that was the strongest near the top of
of Bourgeois et al. (2011) where there is an alternate shedding of half- the arch-vortex. This mode showed a low-frequency peak attributed by
loop structures. Following this model, the bottom section in Fig. 15a the authors to a flow instability over the top of the prism. More re-
would correspond to the time-averaged principal vortex core in the near cently, Kindree et al. (2018) showed that a low-frequency instability
wake. was present over the entire span of a finite square prism, but it was the
The schematic in Fig. 16 presents the mean flow structures around a most intense – even surpassing the dominant vortex-shedding frequency
surface-mounted finite-height square prism with a thin boundary layer, – close to the free end. Behera and Saha (2019) observed that the
according to the findings of the present study. This flow model is unique symmetric modes were associated to occurrences of strong downwash
in the sense that it describes three-dimensionally and solely the time- and upwash flows, which originate from the end effects of the surface-
mean structures, compared to past flow models which mixed mean and mounted finite prism. These studies and the present results based on a
instantaneous structures (Wang et al., 2004) or described the in- time-averaged analysis support the hypothesis that the mean near wake
stantaneous flow only (Wang and Zhou, 2009). The model includes the contains individual structures that do not fit the single interconnected
horseshoe vortex, one of the pair of corner vortices near the junction of vortex model, so far observed only for surface-mounted finite-height
the prism with the ground plane and the two sections of the mean near- square prisms of higher AR (Moreau and Doolan, 2013; Porteous et al.,
wake structure. Representative streamlines of the near-wall flow high- 2017).
light the recirculating flow on the prism surfaces, and the origin of the

Fig. 16. Schematic of the main flow structures which compose the mean flow field around a surface-mounted finite-height square prism with a thin boundary layer.

13
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

4. Conclusions structure comes from the rolling-up of the separated flow from the
free-end leading edge. It results in the main vortex Bt in the re-
The flow around a surface-mounted finite-height square prism of circulation region, that expands outward as if to connect with the
AR = 3 with a low Reynolds number Re = 500, plus a very thin bottom structure, but the flow leaves the near wake before entering
boundary layer at the location of the prism of δ/H = 0.13, was nu- the other structure's region of dominance and vice-versa.
merically investigated using large eddy simulation. A time-averaged
approach was adopted to describe the very near-wall flow on the sur- The mean structures observed in the present results are compatible
faces of the prism and on the ground plane, and its relationship with the with recently proposed flow models, and their description helps identify
near-wake and far-wake structures. Aerodynamic drag and normal force the origin and interconnection of the surface, near-wake and far-wake
coefficients were evaluated for the prism and the mean flow structures structures in the flow around surface-mounted finite-height square
were explored from both plane-projected and three-dimensional points prisms. Analyses of the three-dimensional flow structures, as well as of
of view, to relate these visualization resources. Even though this study their instantaneous behavior, around surface-mounted finite-height
was carried out at a relatively low Reynolds number, the wake pre- square prisms of different AR, for thicker boundary layers and including
sented the same features as other higher Re cases described in the lit- the assessment of quantitative effects of higher Reynolds numbers
erature, and the surface, near-wake and far-wake structures should be would be of interest to broaden this knowledge and extend the con-
at least qualitatively the same. The main findings and conclusions are tributions of this study to more general flow conditions.
summarized as follows:
CRediT authorship contribution statement
i The mean drag and normal force coefficients and points of action
agreed with the literature, given the thin boundary layer thickness Barbara L. da Silva: Validation, Formal analysis, Writing - original
and low Reynolds number of the present case. The relative con- draft, Visualization. Rajat Chakravarty: Conceptualization,
tribution of pressure and skin-friction was accounted for in the Methodology, Software, Investigation. David Sumner:
coefficients, and indicated that, while the skin-friction contribution Conceptualization, Writing - review & editing, Supervision, Funding
can be reasonably neglected in drag force measurements, it is of acquisition. Donald J. Bergstrom: Conceptualization, Software,
significant magnitude and must be considered in the evaluation of Resources, Writing - review & editing, Supervision, Funding acquisi-
the normal force coefficient. As for the bending moment coefficient, tion.
the pressure contribution from drag forces is its main source.
ii The mean surface pressure and skin-friction distribution, along with Declaration of Competing Interests
the location of critical saddle, node and focus points on the ground
and symmetry plane agreed with other results reported in the lit- The authors declare that they have no known competing financial
erature for surface-mounted finite-height square prisms. Downwash interests or personal relationships that could have appeared to influ-
prevailed in the present wake and almost reached the ground plane, ence the work reported in this paper.
a feature expected for the case of a low AR and small δ/H, and
consistent with dipole-type wakes. Acknowledgements
iii Three mean streamwise vorticity regions of high magnitude were
identified and their origins observed from a three-dimensional The financial support of the Natural Sciences and Engineering
perspective. The first is formed at the free-end leading edge and Research Council of Canada (Discovery Grants Program numbers
projected over the free end, but is soon dissipated. This region RGPIN 05103-2016 and 2018-03760) and the contributions of N.
corresponds to the initially outward-deflected flow of the separated Moazamigoodarzi in setting-up the simulation are gratefully acknowl-
shear layer, which is then deflected back inward in the near wake. edged.
The second region occurs from the entrainment of the fluid that
comes upward from the side wall of the prism into the free end References
region, that later ascends joining the first region described above.
The third streamwise vorticity region takes place in the near and far Behera, S., Saha, A.K., 2019. Characteristics of the flow past a wall-mounted finite-length
wake, and it was referred to in past literature as “tip vortices”, square cylinder at low Reynolds number with varying boundary layer thickness. J.
Fluids Eng. 141, 061204. https://doi.org/10.1115/1.4042751.
though the present study shows that there is significant ambiguity in Beitel, A., Heng, H., Sumner, D., 2019. The effect of aspect ratio on the aerodynamic
the origin and development of these regions of high streamwise forces and bending moment for a surface-mounted finite-height cylinder. J. Wind
vorticity magnitude. This vorticity does not have a strong connec- Eng. 186, 204–213.
Bourgeois, J.A., Sattari, P., Martinuzzi, R.J., 2012. Coherent vortical and straining
tion to the other structures over the tip (free end) of the prism. structures in the finite wall-mounted square cylinder wake. Int. J. Heat Fluid Flow 35,
Instead, it arises due to the three-dimensional bending of the sepa- 130–140. https://doi.org/10.1016/j.ijheatfluidflow.2012.01.009.
rated side flow as it encircles the near-wake recirculation region, Bourgeois, J.A., Sattari, P., Martinuzzi, R.J., 2011. Alternating half-loop shedding in the
turbulent wake of a finite surface-mounted square cylinder with a thin boundary
caused by the downwash. It is the strongest right after the re- layer. Phys. Fluids 23, 095101. https://doi.org/10.1063/1.3623463.
circulating flow region, where the downwash is the most intense, Calluaud, D., David, L., 2002. Backward projection algorithm and stereoscopic particle
and its strength decreases downstream in the far wake. image velocimetry measurements of the flow around a square section cylinder. In:
Proceedings of the 11th International Symposium on Applications of Laser
iv A vortex was formed at the leading junction corners of the prism
Techniques to Fluid Mechanics. Presented at the 11thInternational Symposium on
with the ground plane, due to the interaction of the separated flow Applications of Laser Techniques to Fluid Mechanics proceedings of the 11th inter-
with the boundary layer. This corner vortex is of conical shape, bent national symposium on applications of laser techniques to fluid mechanic, Lisbon,
upward and outward, and its vertical extension is presumed to de- Portugal. July 8-11, 2002.
Cao, Y., Tamura, T., Kawai, H., 2019. Investigation of wall pressures and surface flow
pend on the boundary layer thickness. patterns on a wall-mounted square cylinder using very high-resolution Cartesian
v The mean near-wake structure is composed of two main sections, mesh. J. Wind Eng. Industr. Aerodyn. 188, 1–18. https://doi.org/10.1016/j.jweia.
based on their origin and location. The bottom structure is caused by 2019.02.013.
Castro, I.P., Robins, A.G., 1977. The flow around a surface-mounted cube in uniform and
fluid entrainment from near the ground plane into a low-pressure turbulent streams. J. Fluid Mech. 79, 307–335.
region behind the prism, which ascends in a three-dimensional Cucitore, R., Quadrio, M., Baron, A., 1999. On the effectiveness and limitations of local
spiraling fashion up to about the prism mid-span, on both of its criteria for the identification of a vortex. Eur. J. Mech. B/Fluids 18, 261–282. https://
doi.org/10.1016/S0997-7546(99)80026-0.
sides. This fluid motion is responsible for minor regions of high Depardon, S., Lasserre, J.J., Boueilh, J.C., Brizzi, L.E., Borée, J., 2005. Skin friction
streamwise vorticity that show in transverse planes. The top

14
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

pattern analysis using near-wall PIV. Exp. Fluids 39, 805–818. https://doi.org/10. 56, 2291–2299 (in Japanese).
1007/s00348-005-0014-8. Okuda, Y., Taniike, Y., 1993. Conical vortices over side face of a three-dimensional square
Diaz-Daniel, C., Laizet, S., Vassilicos, J.C., 2017. Direct numerical simulations of a wall- prism. J. Wind Eng. Industr. Aerodyn. 50, 163–172. https://doi.org/10.1016/0167-
attached cube immersed in laminar and turbulent boundary layers. Int. J. Heat Fluid 6105(93)90071-U.
Flow 68, 269–280. https://doi.org/10.1016/j.ijheatfluidflow.2017.09.015. Ozgoren, M., 2006. Flow structure in the downstream of square and circular cylinders.
El Hassan, M., Bourgeois, J., Martinuzzi, R., 2015. Boundary layer effect on the vortex Flow Measur. Instrum. 17, 225–235. https://doi.org/10.1016/j.flowmeasinst.2005.
shedding of wall-mounted rectangular cylinder. Exp. Fluids 56, 33. https://doi.org/ 11.005.
10.1007/s00348-014-1882-6. Pope, S.B., 2000. Turbulent Flows. Cambridge University Press, Cambridge, UK.
ESDU, 1978. Fluid forces, Pressures and Moments On Rectangular blocks, ESDU 71016, Porteous, R., Moreau, D.J., Doolan, C.J., 2017. The aeroacoustics of finite wall-mounted
Amendment (C). ESDU International PLC, London, UK 01 Nov 1978. square cylinders. J. Fluid Mech. 832, 287–328. https://doi.org/10.1017/jfm.2017.
Fröhlich, J., Rodi, W., 2004. LES of the flow around a circular cylinder of finite height. 682.
Int. J. Heat Fluid Flow 25, 537–548. https://doi.org/10.1016/j.ijheatfluidflow.2004. Rastan, M.R., Sohankar, A., Alam, Md.M., 2017. Low-Reynolds-number flow around a
02.006. wall-mounted square cylinder: flow structures and onset of vortex shedding. Phys.
Germano, M., Piomelli, U., Moin, P., Cabot, W.H., 1991. A dynamic subgrid-scale eddy Fluids 29, 103601. https://doi.org/10.1063/1.4989745.
viscosity model. Phys. Fluids A 3, 1760–1765. https://doi.org/10.1063/1.857955. Saeedi, M., Wang, B.-.C., 2015. Large-eddy simulation of turbulent flow and dispersion
Haller, G., 2005. An objective definition of a vortex. J. Fluid Mech. 525, 1–26. https:// over a matrix of wall-mounted cubes. Phys. Fluids 27, 115104. https://doi.org/10.
doi.org/10.1017/S0022112004002526. 1063/1.4935112.
Heng, H.G.Y., 2019. Experimental investigation of the free-end pressure distribution for Saha, A.K., 2013. Unsteady flow past a finite square cylinder mounted on a wall at low
surface-mounted finite-height square prisms (Master's thesis). University of reynolds number. Comput. Fluids 88, 599–615. https://doi.org/10.1016/j.compfluid.
Saskatchewan, Saskatoon. 2013.10.010.
Hosseini, Z., Bourgeois, J.A., Martinuzzi, R.J., 2013. Large-scale structures in dipole and Sakamoto, H., 1985. Aerodynamic forces acting on a rectangular prism placed vertically
quadrupole wakes of a wall-mounted finite rectangular cylinder. Exp. Fluids 54, in a turbulent boundary layer. J. Wind Eng. Industr. Aerodyn. 18, 131–151. https://
1595. https://doi.org/10.1007/s00348-013-1595-2. doi.org/10.1016/0167-6105(85)90093-5.
Hunt, J.C.R., Wray, A.A., Moin, P., 1988. Eddies, streams, and convergence zones in Sakamoto, H., Arie, M., 1983. Vortex shedding from a rectangular prism and a circular
turbulent flows. In: Proceedings of the 1988 Summer Program. Presented at the cylinder placed vertically in a turbulent boundary layer. J. Fluid Mech. 126,
Studying Turbulence Using Numerical Simulation Databases. Stanford University, 147–165. https://doi.org/10.1017/S0022112083000087.
CA, USA. pp. 193–208 June 27-July 22, 1988. Sakamoto, H., Oiwake, S., 1984. Fluctuating forces on a rectangular prism and a circular-
Hwang, J.-.Y., Yang, K.-.S., 2004. Numerical study of vortical structures around a wall- cylinder placed vertically in a turbulent boundary-layer. J. Fluids Eng.-Trans. ASME
mounted cubic obstacle in channel flow. Phys. Fluids 16, 2382–2394. https://doi. 106, 160–166. https://doi.org/10.1115/1.3243093.
org/10.1063/1.1736675. Samani, M., Bergstrom, D.J., 2015. Effect of a wall on the wake dynamics of an infinite
Jeong, J., Hussain, F., 1995. On the identification of a vortex. J. Fluid Mech. 285, 69–94. square cylinder. Int. J. Heat Fluid Flow 55, 158–166. https://doi.org/10.1016/j.
https://doi.org/10.1017/S0022112095000462. ijheatfluidflow.2015.07.016.
Kawai, H., Okuda, Y., Ohashi, M., 2012. Near wake structure behind a 3D square prism Sattari, P., Bourgeois, J.A., Martinuzzi, R.J., 2012. On the vortex dynamics in the wake of
with the aspect ratio of 2.7 in a shallow boundary layer flow. J. Wind Eng. Industr. a finite surface-mounted square cylinder. Exp. Fluids 52, 1149–1167. https://doi.
Aerodyn. 104–106, 196–202. https://doi.org/10.1016/j.jweia.2012.04.019. org/10.1007/s00348-011-1244-6.
Kawamura, T., Hiwada, M., Hibino, T., Mabuchi, I., Kumada, M., 1984. Flow around a Sau, A., Hwang, R.R., Sheu, T.W.H., Yang, W.C., 2003. Interaction of trailing vortices in
finite circular cylinder on a flat plate. Bull. JSME 27, 2142–2151. the wake of a wall-mounted rectangular cylinder. Phys. Rev. E 68, 056303. https://
Kindree, M.G., Shahroodi, M., Martinuzzi, R.J., 2018. Low-frequency dynamics in the doi.org/10.1103/PhysRevE.68.056303.
turbulent wake of cantilevered square and circular cylinders protruding a thin la- Sumner, D., 2013. Flow above the free end of a surface-mounted finite-height circular
minar boundary layer. Exp. Fluids 59, 186. https://doi.org/10.1007/s00348-018- cylinder: a review. J. Fluids Struct. 43, 41–63. https://doi.org/10.1016/j.
2641-x. jfluidstructs.2013.08.007.
Krajnović, S., 2011. Flow around a tall finite cylinder explored by large eddy simulation. Sumner, D., Heseltine, J.L., Dansereau, O.J.P., 2004. Wake structure of a finite circular
J. Fluid Mech. 676, 294–317. https://doi.org/10.1017/S0022112011000450. cylinder of small aspect ratio. Exp. Fluids 37, 720–730. https://doi.org/10.1007/
Krajnović, S., Davidson, L., 2002. Large-eddy simulation of the flow around a bluff body. s00348-004-0862-7.
AIAA J. 40, 927–936. Sumner, D., Rostamy, N., Bergstrom, D.J., Bugg, J.D., 2017. Influence of aspect ratio on
Krajnović, S., Davidson, L., 2001. Large eddy simulation of the flow around a three-di- the mean flow field of a surface-mounted finite-height square prism. Int. J. Heat Fluid
mensional bluff body, in: 39th aerospace sciences meeting and exhibit. In: Presented Flow 65, 1–20. https://doi.org/10.1016/j.ijheatfluidflow.2017.02.004.
at the 39th Aerospace Sciences Meeting and Exhibit. Reno, NV, USA. American Sumner, D., Rostamy, N., Bergstrom, D.J., Bugg, J.D., 2015. Influence of aspect ratio on
Institute of Aeronautics and Astronautics, pp. 0432. https://doi.org/10.2514/6.2001- the flow above the free end of a surface-mounted finite cylinder. Int. J. Heat Fluid
432. January 8-11, 2001. Flow 56, 290–304. https://doi.org/10.1016/j.ijheatfluidflow.2015.08.005.
Lilly, D.K., 1992. A proposed modification of the Germano subgrid-scale closure method. Sumner, D., Unnikrishnan, S., Teng, M., Beitel, A., Das, A., Fulton, M., 2016. The mean
Phys. Fluids A 4, 633–635. https://doi.org/10.1063/1.858280. wake of low-aspect-ratio surface-mounted finite- height square prisms and the effects
Lim, H.C., Castro, I.P., Hoxey, R.P., 2007. Bluff bodies in deep turbulent boundary layers: of incidence angle. In: Presented at the 8th International Colloquium on Bluff Body
Reynolds-number issues. J. Fluid Mech. 571, 97. https://doi.org/10.1017/ Aerodynamics and Applications. Boston, Massachusetts, USA. June 7-11, 2016.
S0022112006003223. Tanaka, S., Murata, S., 1999. An investigation of the wake structure and aerodynamic
Martinuzzi, R., Tropea, C., 1993. The flow around surface-mounted, prismatic obstacles characteristics of a finite circular cylinder. JSME Int. J. Ser. B-Fluids Therm. Eng. 42,
placed in a fully developed channel flow. J. Fluids Eng. 115, 85. https://doi.org/10. 178–187.
1115/1.2910118. Taniguchi, S., Sakamoto, H., Arie, M., 1981. Flow around circular cylinders of finite
Martinuzzi, R.J., Havel, B., 2004. Vortex shedding from two surface-mounted cubes in height placed vertically in turbulent boundary layers. Bull. JSME 24, 37–44.
tandem. Int. J. Heat Fluid Flow 25, 364–372. https://doi.org/10.1016/j. Uffinger, T., Ali, I., Becker, S., 2013. Experimental and numerical investigations of the
ijheatfluidflow.2004.02.003. flow around three different wall-mounted cylinder geometries of finite length. J.
Martinuzzi, R.J., Havel, B., 2000. Turbulent flow around two interfering surface-mounted Wind Eng. Industr. Aerodyn. 119, 13–27. https://doi.org/10.1016/j.jweia.2013.05.
cubic obstacles in tandem arrangement. J. Fluids Eng. 122, 24. https://doi.org/10. 006.
1115/1.483222. Unnikrishnan, S., Ogunremi, A., Sumner, D., 2017. The effect of incidence angle on the
Maskell, E.C., 1963. A theory of the blockage effects on bluff bodies and stalled wings in a mean wake of surface-mounted finite-height square prisms. Int. J. Heat Fluid Flow
closed wind tunnel. Aeronautical Research Council, Reports and Memoranda3400. 66, 137–156. https://doi.org/10.1016/j.ijheatfluidflow.2017.05.012.
McClean, J.F., Sumner, D., 2014. An experimental investigation of aspect ratio and in- Wang, B.-.C., Bergstrom, D.J., 2005. A dynamic nonlinear subgrid-scale stress model.
cidence angle effects for the flow around surface-mounted finite-height square Phys. Fluids 17, 035109. https://doi.org/10.1063/1.1858511.
prisms. J. Fluids Eng. 136, 081206. https://doi.org/10.1115/1.4027138. Wang, B.-.C., Yee, E., Bergstrom, D.J., Iida, O., 2008. New dynamic subgrid-scale heat flux
Meneveau, C., Katz, J., 2000. Scale-invariance and turbulence models for large-eddy si- models for large-eddy simulation of thermal convection based on the general gradient
mulation. Annu. Rev. Fluid Mech. 32, 1–32. https://doi.org/10.1146/annurev.fluid. diffusion hypothesis. J. Fluid Mech. 604, 125–163. https://doi.org/10.1017/
32.1.1. S0022112008001079.
Moreau, D.J., Doolan, C.J., 2013. Flow-induced sound of wall-mounted finite length cy- Wang, H.F., Zhou, Y., 2009. The finite-length square cylinder near wake. J. Fluid Mech.
linders. AIAA J. 51, 2493–2502. https://doi.org/10.2514/1.J052391. 638, 453–490. https://doi.org/10.1017/S0022112009990693.
Nakamura, H., Igarashi, T., Tsutsui, T., 2001. Local heat transfer around a wall-mounted Wang, H.F., Zhou, Y., Chan, C.K., Lam, K.S., 2006. Effect of initial conditions on inter-
cube in the turbulent boundary layer. Int. J. Heat Mass Transf. 44, 3385–3395. action between a boundary layer and a wall-mounted finite-length-cylinder wake.
https://doi.org/10.1016/S0017-9310(01)00009-6. Phys. Fluids 18, 065106. https://doi.org/10.1063/1.2212329.
Visualized flow: fluid motion in basic and engineering situations revealed by flow vi- Wang, H.F., Zhou, Y., Chan, C.K., Wong, W.O., Lam, K.S., 2004. Flow structure around a
sualization. In: Nakayama, Y. (Ed.), Thermodynamics and Fluid Mechanics Series. finite-length square prism. In: Proceedings of the 15th Australasian Fluid Mechanics
Pergamon Press, Oxford. Conference. Presented at the 15th Australasian Fluid Mechanics Conference. Sydney,
Okamoto, S., Matsunaga, K., Sunabashiri, Y., 1990. Vortex shedding and turbulent wake Australia. December 13-17, 2004.
from a square cylinder of finite length placed on a ground plane. Trans. JSME Ser. B Wang, Y.Q., Jackson, P.L., Sui, J., 2014. Simulation of turbulent flow around a surface-

15
B.L. da Silva, et al. International Journal of Heat and Fluid Flow 83 (2020) 108569

mounted finite square cylinder. J. Thermophys. Heat Transf. 28, 118–132. https:// Combust. 81, 39–75. https://doi.org/10.1007/s10494-008-9135-5.
doi.org/10.2514/1.T3884. Yin, J., Wang, B.-.C., Bergstrom, D.J., 2007. Large-eddy simulation of combined forced
Yakhot, A., Liu, H., Nikitin, N., 2006. Turbulent flow around a wall-mounted cube: a and natural convection in a vertical plane channel. Int. J. Heat Mass Transf. 50,
direct numerical simulation. Int. J. Heat Fluid Flow 27, 994–1009. https://doi.org/ 3848–3861. https://doi.org/10.1016/j.ijheatmasstransfer.2007.02.014.
10.1016/j.ijheatfluidflow.2006.02.026. Zhang, D., Cheng, L., An, H., Zhao, M., 2017. Direct numerical simulation of flow around
Yin, J., Wang, B.-C., Bergstrom, D.J., 2008. Geometrical properties of the resolved-scale a surface-mounted finite square cylinder at low Reynolds numbers. Phys. Fluids 29,
velocity and temperature fields predicted using large-eddy simulation. Flow Turbul. 045101. https://doi.org/10.1063/1.4979479.

16