BUILD SIMUL (2022) 15: 291–312

https://doi.org/10.1007/s12273-021-0785-8

Unsteady RANS simulation of wind flow around a building shape

obstacle

Research Article
Sheikh Hassan1, Md. Mamun Molla1 (), Preetom Nag1, Nasrin Akhter2, Amirul Khan3

1. Department of Mathematics & Physics, North South University, Dhaka 1229, Bangladesh
2. Department of Mathematics, Dhaka University of Engineering & Technology, Gazipur 1708, Bangladesh
3. School of Civil Engineering, University of Leeds, Leeds LS2 9JT, UK

Abstract Keywords
This work aims to find the origin and connection of the surface, near-wake, and far-wake turbulent wake;
structures in the flow encompassing a high-rise building for a high Reynolds number. The origin vortex shedding;
and interconnection of the stream-wise tip vortices, with the other components of the wake, is flow structures;
analysed in this study for the current scenario. The Unsteady Reynolds Averaged Navier-Stokes tip vortexes;
equations (URANS) together with the realizable k- turbulence model have been used in this model building;
URANS;
investigation to study the turbulent wake flow following a ground-surface-attached square shape
turbulent boundary layer
building. A moderately big obstacle aspect ratio of 4, a Reynolds number of 12,000, and a thin
evolving boundary layer thickness have been used in the flow modeling. The designed flow
Article History
addresses the reversed-flows at the outlet during computation to improve the accuracy of the
Received: 20 November 2020
realizable k- model. The Reynolds stress components are retrieved using the Boussinesq approach.
Revised: 07 February 2021
The wake’s principal compositions, including span-wise-side eddies and area of high stream-wise Accepted: 18 February 2021
vorticity in the uppermost portion of the wake, are illustrated by both three-dimensional (3D)
representations and planner projections of the mean flow distributions. A braided vortex formation, © Tsinghua University Press and
composed of asymmetric hairpin vortexes, is witnessed in the far-wake area. The association of the Springer-Verlag GmbH Germany,
near-wake vortex structures with the far-wake and near-wall flow, which is associated with the part of Springer Nature 2021
flow strengths, is also discussed. In this investigation, few areas of large stream-wise vorticity
magnitude, like tip vortexes, are correlated to the 3D curving of the fluid motion, and tip vortices
did not continuously reach to the free end part of the building. The 3D fluid motion interpretation,
which combined several measurements of the flow distribution encompassing the cylinder, shows
that the time-averaged near-wake structures are formed of two segments of distinct source and
section of dominance. Furthermore, addressing reversed-flow during computation shows notable

Indoor/Outdoor Airflow
improvement in the results.

and Air Quality

1 Introduction on flow dynamics and the coherent form of squared cylinder

flow. An increase in AR values results in a more active
Turbulent flow considering a bluff body, is regularly generated shear layer on the obstacle’s top. The interaction
exercised and has a wide range of mechanical engineering between this developed shear layer and the Kármán vortexes
and environmental sciences implementations. Detailed generated by two vertical obstacle sides and vertical base
knowledge of turbulent flow fields around obstacles (for structures in the boundary layer is quite intense.
example, building structures) and fluid-obstacle interaction Additionally, patterns of turbulent wake behind the
characteristics are essential to engineers and designers for obstacle get more composite and widespread. This additional
any efficient designs or models. The square cylindrical design complexity requires an appropriate computational domain.
with an AR (aspect ratio) value of 4 has a significant impact Detailed solution results (spatial changes of eddy flows and

E-mail: [email protected]
292 Hassan et al. / Building Simulation / Vol. 15, No. 2

List of symbols

a speed of sound u Swirling Strength Criterion (SSC)

Cp pressure coefficient v vertical velocity
d building’s width w span-wise velocity
gi element of the gravitational vector in i-th direction x length in x-direction
Gk generation of turbulence kinetic energy y length in y-direction
Gb production of turbulence kinetic energy due to YM role of the variable dilatation in compressible
buoyancy turbulence to the in-general rate
h height of the building z length in z-direction
Mt turbulent Mach number δ boundary layer thickness
Prt turbulent Prandtl number for energy μ eddy viscosity
Re Reynolds number  fluid viscosity
S modulus of the mean rate-of-strain tensor σk turbulent Pr for k
t time σ turbulent Pr for 
u stream-wise velocity ω vorticity
U velocity magnitude ωk angular velocity
U¥ free stream velocity Ωij rate-of-rotation tensor

flow formations of an extensive range of turbulence scales) energetic categorization of the structures. This topological
can be captured correctly with an appropriate domain. Since investigation put forward a different explanation of the
new technologies are becoming more accessible for CFD mean stream-wise vorticity, which previously had been
and experimental studies, researchers have been focusing on assumed to initiate from structures generated at the leading
studying turbulent flow, considering a bluff body for the last edge of the free-end. A different justification was given: the
several decades with significant outcomes. Sumner (2013) stream-wise vorticity occurs for the reorientation of the
published a review of recent progress in experimental and primary vertical vortexes in the sidewall shear layers for the
numerical studies regarding surface-mounted finite-height free-end shear layer’s lateral-vortexes. The free-end shear
circular cylinder flows and examinations of the aspect ratio layer is seen to alter the initiation of development of the
(AR) value change effects on the vortex generation surround- primary vertical vortexes produced by the span-wise
ing the cylinder. A short review of the square cylinder flow shear layers’ furl. “Can it be supposed, then, that the mean
mentioning relevant previous works, experimental and stream-wise vorticity found near the free-end is due to the
numerical, is presented in the subsequent subsections. stream-wise reorientation of vorticity, or is the stream-wise
vorticity structures that exist independently of the vortex
1.1 Previous notable experiments shedding process as described by Etzold and Fielder (1976)?”
For answering these questions regarding the shed vortex
In 2011, Bourgeois et al. (2011) introduced the outline and topology and its effect on the statistical dynamics, an in-depth
dynamics of the large-scale vortex structure for quasi- phase averaged PIV research was conducted. Some planar
periodic shedding in the turbulent wake of a finite (h/d = 4) measurements were employed to recreate a 3D portrayal of
square-cross-section surface-mounted cylinder protruding the shed structures. A vortex detection method was utilized
from a thin boundary layer (δ/h = 0.18). The 3D large-scale on the corresponding phase-averaged information to
structure was developed from phase averaged x-z and x-y extract the 3D large scale vortex structure. This method was
planner data quantified with particle image velocimetry employed to learn the large scale structure’s relations on
(PIV). Their work exercised instantaneous measurements the turbulent flow field. Two questions were particularly of
of the surface pressure variation on both sides of the obstacle concentration of the research (i) What are the characteristics
to phase-align the PIV planner data. This work showed that of the phase-averaged flow structure in the wake? and (ii)
the developed structures’ topology bears a resemblance to What is the dynamical significance of the phase-averaged
alternating half-loop structure interconnecting to the near structure (i.e., how much fluctuating energy does this
ground surface plate. This structure’s time-averaging gave structure represent)? Measurements had been carried out in
rise to mean stream-wise vortexes, similar to those illustrated an open-test-section suction wind tunnel for experimental
in the literature for related geometrical configurations. The work. The flow went into the tunnel and traveled throughout
flow illustrated was of dipole type and was taken up this in total of three 20 mesh metal grid, a 24 and a 30 mesh
Hassan et al. / Building Simulation / Vol. 15, No. 2 293

metal grid, and in 80 mesh nylon screen ahead of going “A key enabler is a universal phase average having a feature
at the inlet nozzle. The flow then exited the functioning of a gradually changing near ground surface flow, changeable
segment throughout a bell mount, a diffuser sector and then oscillation amplitude, and elevated harmonics”. This work
through the power unit section. A bottom surface turbulent found that these universal phase averaging methods mentioned
boundary layer was produced on a pointed top-edging flat above, more specifically, a “key enabler”, decreases the instant
plate. The free-stream wind velocity, U¥ = 15 m/s, matched residual by 30% compared to a conventional phase averaging
up to a Reynolds number of Re = U¥d/ = 12,000 and a method. Additionally, the determined differences disclose
free-stream wind turbulence intensity of 0.8% was set for critical limitations of the mean flow and oscillation points,
this work. A LaVision FlowMaster high-frame-rate PIV for example, the mean-field paraboloid. Later on, 3D flow
system was employed to take data from several horizontal was created from 2D PIV information on orthogonal planes.
(x-z) and vertical (x-y) planes. These data were used to Velocity profiles were determined in a rectangular near-wake
recreate the phase averaged 3D flow. Acquired data sets were area by two thick orthogonal settings of parallel PIV planes.
examined to determine the phase of the shedding cycle for This method was illustrated for the turbulent vortex shedding
all the PIV vector fields. To resolve the instantaneous phase, in the wake of a wall-mounted square cylinder of finite height.
the pressure signal was smoothed, and the local maxima This flow is a well-established standard construction for
was figured out. All the information was initially examined, oscillatory logical structures amongst prominent 3D features.
not including any guess, and after that these information Experimentation was carried out in an open functioning
sets were reprocessed guessing anti-symmetry to reduce segment suction wind tunnel facility. Immobile air was
statistical improbability. Both the infinite and finite cylinder drawn in the course of an inlet and run through metal grids.
wakes were categorized with rambling turbulent energy Then, it went into the functioning sector on the course
formed by the large scale structure into the prime interior. of an inlet-area: narrow allocated nozzle. The functioning
This work illustrated the dynamical meaning of the determined segment of the tunnel was open given that a generally
large scale structures and unsolved fluctuating kinetic energy unvarying stream in which an aluminium sharp-leading-edge
in the wake. Both parts are accountable for approximately flat plate was mounted, and a near ground surface turbulent
50% of the mean kinetic energy. An argument of turbulence boundary layer flow inside which a finite square-cross-
creation, in light of the base flow that provides it with energy, section cylinder was positioned (aspect ratio, h/d = 4, where
was demonstrated. d = 0.0127 m and h = 4d = 0.0508 m). The coordinate
In 2013, Bourgeois et al. (2013) experimentally looked system origin was set alongside the obstacle’s axis with
into the 3D wake at the back of a finite wall-mounted y = 0 at the plate junction. The free-stream settings of the
square cylinder at Reynolds number, Re = 12,000 and AR flow were matched to a mean velocity of U¥ = 15 m/s, a
value of 4. The center of the study was put on the ground turbulence intensity level of 0.8% and a Reynolds number
surface flow and oscillatory variation. This work offered a of Re = U¥d/ = 12,000. The selected geometry is a
method to produce dominant oscillatory dynamics for a large thoroughly-examined point of reference in the wake of finite
set of turbulent flow, for instance, vortex shedding behind wall-mounted cylinders. The suggested method for universal
obstacles (Williamson 1996), Kelvin-Helmholtz vortexes in phase averaging and creating 3D velocity profiles from 2D
mixing layers (Liu 1989), Rossiter modes in cavities (Rowley PIV information are suitable for a large category of turbulent
and Williams 2006) and wavy structures in Taylor-Couette flows with oscillatory dynamics.
flow; just to name a few, from experimental 2D particle
PIV information set, and to create time-resolved 3D velocity 1.2 Previous numerical works
profile. At first, 2D velocity fields were filtered in the range
of a quantitative and logical structure narrative considering In 2011, Wang et al. (2011) employed the Reynolds Stress
some factors. Then, PIV snaps, unlinked to 2D data, were Model (RSM) and DES model to examine the turbulent
combined into 3D time-resolved flows by linking data flow surrounding a surface-mounted rectangular cylinder
time-resolving data from pressure sensors. Time-resolved of aspect ratio h/d = 4 at a Reynolds number of 13,041.
3D velocity fields were created from high-frame-rate particle RSM generated improved results for both mean-velocity and
image velocimetry (PIV), and at the same time, surface root-mean-square velocity profiles compared to DES. Later,
pressure data were also noted. A modal disintegration was in Wang et al. (2014) numerically investigated turbulent flow
exercise: applying characteristics of Navier-Stokes equations around a surface-mounted cylinder of aspect ratio h/d = 4,
to experimental disintegration from existing flow information. presenting more details. Calculations were performed
Phase averaging was made universal to add low-frequency around at Reynolds number of 13,041 using RSM. In 2014,
flows in the near ground surface flow, and changes were a DNS investigation was executed by Saeedi et al. (2014)
made in the oscillation level along with elevated harmonics. to figure out the turbulent wake characteristics around a
294 Hassan et al. / Building Simulation / Vol. 15, No. 2

surface-wall-mounted square cylinder having an AR value separate cubical model building with a rooftop vent. Bautista
of 4 and Re value of 12,000 considering a thickness value of (2015) focused on two of the major difficulties that the wind
developing boundary layer. This work concluded that the energy area faces to get excellent wind flow simulation
interactions of the flow with the squared-cylinder directly outcomes considering a complex terrain in the geometry.
dictate the major turbulent shear intensity; this turbulent The simulation outcomes demonstrated that the k-ω SST-
shear intensity (and intensity of generated vortexes and SIDDES turbulence model is capable of assuming practical
turbulence kinetic energy generation rate) reduces eventually wind characteristics considering a flat terrain and further
to the surroundings turbulence intensity level in the faraway complex scenarios. Hemmati et al. (2016) researched on
downstream areas. Saeedi and Wang (2016) then carried the wake of an unchanged thin flat plate for AR = 3.2
out a Large-Eddy Simulation (LES) to examine the turbulent concerning that to an infinite span (2D) plate at Re = 1200
wake behind a wall-mounted square cylinder. In the instant employing DNS method. In the work of Jadidi et al. (2016),
downstream area of the cylinder, the TKE profile creation they evaluated the potential of the ELES model to predict
rate showed a dual-peak pattern. Wang (2019) showed the flow around a model building. Zhang et al. (2017) used DNS
numerical simulation results of flow surrounding a base method to present a thorough analysis on the flow around
surface-mounted rectangular cylinder of AR = h/d = 7 at a finite square cylinder at AR = 4 and Re = 50, 100, 150, 250,
Re = 652 and 13,041. A dipole wake was attained at the upper 500, and 1000. It was shown that the mean stream-wise vortex
Re value, whereas a quadruple wake was attained for the formation also changes by changing Re values. Rastan et al.
lower Re value, representing that the Re robustly controlled (2017) employed DNS to study 3D unsteady flow features
the wake formation. In the far-wake area, a braided vortex around a finite wall-mounted square cylinder with AR = 7
formation produced by asymmetric hairpin vortexes was at Re values of 40–250. The outcome illustrated that the vortex
monitored for both Re values, and a novel wake topology was shedding structure emerges in the range of 75 < Re < 85.
offered for flows with similar geometry. A transitional flow emerged at Re = 150–200, where the
Davis et al. (2012) tested three RANS turbulence models wake volatilities were increased with higher Re, and the force
to evaluate the comparative accuracy of the established signal oscillation changed from a sinusoidal to a chaotic
models based on the experimental data of Bourgeois et al. form. Eventually, the wake flow turns into turbulent at Re >
(2011). Computationally, the realizable k- along with a 200. Zhang (2017) evaluated the outcomes of eight RANS
two-layer treatment found to be better compared to both two-equation turbulence models and two SGS LES methods
of the k- V2F (all y+ hybrid wall treatment) and the k-ω in the circumstances of unsteady flow surrounding a finite
models. Saha (2013) carried out a DNS study through a circular cylinder at AR = 1.0 with Re = 20,000. Amongst the
finite length square cylinder at a Reynolds number of 250. eight turbulence models studied, the K-Omega-SST model
The flow field was investigated by solving 3D unsteady (via. SST-V2003) showed the most excellent in general
Navier-Stokes equations utilizing second-order spatial and results considering the mean streamline profiles in several
temporal discretizations. Uffinger et al. (2013) calculated distinctive planes and the distribution of mean bed-shear-
3D flow fields surrounding three separate wall-mounted stress extension on the ground surface wall. Liu and Nei
cylinder geometries of finite length. A combinational method (2017) used the Lattice-Boltzmann method to produce the
of both experimental and numerical studies was applied flow around a finite cylinder. Their outcomes demonstrated
to present detailed data. Moazamigoodarzi et al. (2014) that the flow develops into an unsteady state when Re > 60.
presented a LES study of turbulent flow around a finite- Furthermore, the flow characteristics became further complex
height rectangular cylinder mounted to a base plane. Phase for greater Re value. Han et al. (2017) studied turbulent
averaging, considering the Strouhal number, showed that a flow surrounding a square cylinder mounted on a flat surface
wake formation with quasi-periodic characteristics greatly at a higher Re utilizing the LES method. This work showed
change from the formation seen in the mean vorticity that as the inlet wind velocity rises, the amplitude of the
field. In the work of Bazdidi-Tehrani and Jadidi (2014), peak in the vertical profile of mean turbulent momentum
they examined the efficiency of a dynamic one-equation flux rises as well because of the rise in velocity variation
sub-grid scale model for the LES method of atmospheric among the low-velocity and high-velocity flows. The com-
flow encompassing a separate cubic building. Joubert et al. parative performance of unsteady SAS turbulence model
(2015) utilized the IDDES turbulence model along with an was evaluated by Jadidi et al. (2018) for flow fields around
all-y+ wall treatment to computationally regenerate the flow a model building compared to other transient simulation
configurations around a square cross-sectioned beam of a such as LES and URANS. A good performance of SAS was
wind tunnel experiment at Re = 7.6×104. Bazdidi-Tehrani observed in the results in comparison with the SST k-ω. Leite
et al. (2015) examined the accuracy of several non-linear eddy et al. (2019) carried out a 3D DNS study at Re = 4.4×104,
viscosity turbulence models to study flow encompassing a after a detailed time-resolved 2D PIV experimental study
Hassan et al. / Building Simulation / Vol. 15, No. 2 295

of the wake of a square ground surface-mounted finite Important experimental and numerical studies discussed
cylinder at Re = 5.4×104 with AR= h/d = 3. In a work of above are outlined in Table 1.
Bazdidid-Tehrani et al. (2019), they studied the role of
various thermal stratification circumstances (stable, neutral 1.3 Present work
and unstable) in the airflow process encompassing a
high-rise non-isolated model building. Behera and Saha In this work, the aim will be to discover the origin and
(2020) conducted a LES study to computationally study flow connection of the surface, near-wake, and far-wake structures
characteristics of an elevated square jet in cross-flow (EJICF) in the flow encompassing a high-rise building with a thin
with Re = 20,000. This work demonstrated two dissimilar boundary layer for a high Reynolds number. The origin
modes of shedding, symmetric and anti-symmetric, near to and interconnection of the stream-wise tip vortices, with
the ground surface jet flow at VR = 0.5, which is similar to the other components of the wake, will be analyzed in this
the wake of a surface-mounted finite-size cylindrical flow. study for the current scenario. The unsteady version of
But the stack wake only demonstrated anti-symmetric modes realizable k- model will be used to study the flow physics
of shedding for VR = 1. A previous work by Bazdidi-Tehrani of the turbulent wake around a wall-mounted square
et al. (2020) investigated the influence of various shapes of cylindrical shape building with aspect ratio, AR = 4, in the
opening on cross-ventilation in an isolated building. circumstance of a thin developing boundary layer flow for

Table 1 Outline of chosen experimental and CFD works of the surface-mounted finite square cylinder flow field characteristics
Authors Year Approaches Re AR δ/h Method(s)
Bourgeois et al. 2011 2011 Experiments 12,000 4 0.18 PIV
Bourgeois et al. 2013 2013 Experiments 12,000 4 0.18 PIV
RSM
Wang et al. 2011 2011 Simulations 13,041 4 n/a
DES
Davis et al. 2012 2012 Simulations 12,000 4 n/a RANS
Experiments LDA
Uffinger et al. 2013 2012 12,800 6 2.71 mm
Simulations SST, SAS, LES
Saha 2013 2013 Simulations 250 2, 3, 4, 5 n/a DNS
Wang et al. 2014 2013 Simulations 13,041 4 n/a RSM
Saeedi et al. 2014 2014 Simulations 12,000 4 0.18 DNS
Moazamigoodarzi et al. 2014 2014 Simulations 500 3 n/a LES
Experiments L/D=2.63 PIV
Joubert et al. 2015 2015 7.6×104 n/a
Simulations H/L=5 IDDES
Saeedi and Wang 2016 2016 Simulations 12,000 4 0.18 LES
Hemmati et al. 2016 2016 Simulations 1200 3.2 n/a DNS
50, 100, 150, 250,
Zhang et al. 2017 2017 Simulations 4 n/a DNS
500, 1000
Liu and Nie 2017 2017 Simulations 10 − 160 n/a n/a LBM
Rastan et al. 2017 2017 Simulations 40 − 250 7 n/a DNS
RANS
Zhang 2017 2017 Simulations 20,000 1 n/a
LES
Han et al. 2017 2017 Simulations n/a n/a n/a LES
652 0.052
Wang 2019 2019 Simulations 7 RSM
13,041 0.077
Experiments 5.4 × 104 0.375 PIV
Leite et al. 2019 2019 3
Simulations 4.0 × 104 0.1375 DNS
Behera and Saha 2020 2020 Simulations 20,000 7 n/a LES
URANS
Present 2020 Simulations 12,000 4 0.18
(realizable k-)
296 Hassan et al. / Building Simulation / Vol. 15, No. 2

Re = 12,000 (depending on the velocity of the free-stream

wind and the side length of the building). The test scenarios
based upon the wind-tunnel experimental work of Bourgeois
et al. (2011, 2013) to observe concerned vortices and other
properties of the flow. The wind tunnel experiment Bourgeois
et al. (2011, 2013) shows relatively higher energetic vortexes
and shear layer partitions in the wake region compared to
the cube flow case (AR = 1) experimented by Shah and
Ferzier (1997). These new characteristics of the flow further
make it hard to numerically simulate the case since it is Fig. 1 Schematic diagram of the computational domain (a benchmark
difficult to accurately capture energetic eddies when simulating computational domain (Saeedi et al. 2014; Saeedi and Wang 2016))
for the wall-mounted square cylinder flow (reprinted by permission
in a larger computational domain with adequate temporal
from Saeedi and Wang (2016), © Springer Nature)
and spatial resolutions. Additionally, a way of improving
the accuracy of the realizable k- turbulence model will be span-wise direction (z-axis), and 5d away from the top free-
explored. surface (y-axis). The distance of 23d is required between
the square cylinder and the outlet because it makes certain
2 Problem statement that wake dynamics are well captured without having any
influence of outlet boundary condition.
In this work, we first try to computationally reproduce The inlet boundary condition is set with awareness to
the experimental results of Bourgeois et al. (2011, 2013), replicate the same boundary-layer thickness at the square
which was done with a boundary layer wind tunnel. In the cylinder used in the wind-tunnel experiments (Bourgeois
experiment of Bourgeois et al. (2011, 2013) the free-stream et al. 2011; Sattari et al. 2012). To implement this phenomenon,
wind velocity was set at U¥ = 15 m/s together with a the following power-law (Saeedi et al. 2014; Saeedi and
turbulence intensity of 0.8% for a wall-mounted square Wang 2016) is used at the velocity inlet:
cylinder, which was designed to have a cross-section with
characteristics: side length d = 12.7 mm; height h = 4d. ì
ï y 0.16
ï <u>=U ¥(
ï ) if y < 0.15h
Theses design parameters set the Reynolds number at í 0.15h
ï
ï
Re = U¥d/ = 12,000. The square cylinder was placed at 4d î <u>=U ¥
ï if y ³ 0.15h
downwind of the leading edge of the flat plate. The boundary
The equation above makes sure that the boundary
layer thickness was thin. This interaction between the flow
condition at the inlet maintains the approximate thin
and the square cylinder was intense since the AR value,
developing boundary layer thickness to match with the
which was 4 in this case, and Reynolds number Re = 12,000
experimental setup. The turbulence intensity of 0.8% is added
were relatively high. These experimental study results showed
with the velocity profiles. The inflow turbulence generation
the alternating Kármán vortices on the cylinder’s two sides,
techniques have a significant influence on flow structures,
tip vortices by the cylinder’s top front edge, a strong shear
and it also influences the velocity fluctuations, recirculation
layer by the cylinder’s free-end, and the recirculation of the zones and plume patterns in the wake zone (Bazdidi-
flow on the cylinder’s leeward side. The demonstrated vortices Tehrani et al. 2016). As for the lateral and free surface
influenced the thin boundary layer’s flow structure and boundary, the boundary condition is symmetry, and the
production since these vortices were always changing and boundary conditions for all the solid surfaces are set as
were interacting with each other. no-slip condition (Wang et al. 2011; Wang et al. 2014;
To perform this numerical simulation, all the experimental Wang 2019).
design criteria are maintained. The adopted method for this ANSYS Fluent’s both Outflow, previously used in several
numerical work is URANS: realizable k- turbulence model. atmospheric flow design (Gromke et al. 2008; Buccolieri
For the realizable k- model, an appropriate computational et al. 2009; Salim et al. 2011), and Pressure-Outlet (with
domain is needed to match the simulation output with the 0 pressure), also used in several atmospheric flow design
experimental results. Thus, a benchmark computational (Wang et al. 2011; Wang et al. 2014; Saeedi et al. 2014; Saeedi
domain is used for this study (Saeedi et al. 2014; Saeedi and Wang 2016; Wang 2019), boundary conditions satisfy
and Wang 2016). Figure 1 shows the geometry of the com- the outlet boundary condition of the experimental work
putational domain. The square cylinder is at 4d downstream (Bourgeois et al. 2011, 2013; Fluent 2009, 2013). In the first
(on the x-axis) of the inlet, 23d upstream of the outlet (on case, the Outflow boundary condition is set at the outlet
the x-axis), 6.2d away from each lateral boundary in the boundary. However, this Outflow boundary condition does
Hassan et al. / Building Simulation / Vol. 15, No. 2 297

not have the facility to address possible back-flow/reversed-  the way to calculate turbulent viscosity;
flow at the outlet during numerical computation. Hence,  the turbulent Prandtl number values directing the turbulent
Pressure-outlet with 0.8% back-flow turbulence intensity is diffusion of k and ;
set for the second case in an attempt to improve the accuracy  the turbulence construction and destruction expressions

of the turbulence modeling, as advised in the User’s Guide in the  equation.

of ANSYS Fluent (Fluent 2009, 2013). It should be noted The momentum equations, the process of defining
that the realizable k- turbulence model has the feature of turbulent viscosity, and model constants are shown: the
addressing any possible back-flow/reversed-flow turbulence characteristics that fundamentally all models follow and the
intensity at the outlet boundary; not all turbulence models considering turbulent construction cause of shear buoyancy.
have this feature.
3.1.1 Transport equation for the realizable k-
3 Mathematical formulation The transport equations for k and  in the realizable k-
model are (Shaheed et al. 2019):
The governing equations for the incompressible unsteady
Reynolds Averaged Navier-Stokes equations (URANS) ¶ ¶ ¶ é μ ¶k ù
( ρk ) + ( ρku j ) = ê(μ + t )
êë
ú + Gk - ρ (3)
úû
turbulence model in a Cartesian coordinates system are ¶t ¶x j ¶x j σ k ¶x j
given below:
and
¶ui
=0 (1)
¶x i ¶ ¶ ¶ é μ ¶ ù
( ρ ) + ( ρu j ) = ê(μ + t )
êë
ú
úû
é ¶u ù ¶t ¶x j ¶x j σ  ¶x j
¶ ¶ ¶p ¶ ¶u
( ρui ) + ( ρui u j ) = - + ê ( μ + μ t )( i + j )ú
¶t ¶x j ¶xi ¶x j êë ¶x j ¶xi úû 2
+ ρC1S - ρC2 (4)
(2) k + 

where ρ is the density, p is the pressure, μ is the viscosity of here

the fluid, and the μt is the turbulent viscosity, which will be
é η ù k
determined by the realizable k- turbulent model that is C1 = max ê 0.43, ú, η = S
êë η + 5 úû 
described in the following subsections.
Gk represents the generation of turbulence kinetic energy,
3.1 Realizable k- model is modeled similarly for the standard, RNG, and realizable
k- methods. From the same formulation of the transport
Two major points make the realizable k- turbulence model of k, this can be described as:
(Shih et al. 1995) different from the standard k- model,
these are: (1) A newer equation is used for the turbulent ¶u j
Gk = -ρui¢u ¢j (5)
viscosity in this model; (2) The definite equation for the ¶xi
transport of the mean-square vorticity fluctuation is the
origin of the newly determined transport formulation for the To calculate Gk in a way consistent by means of the
dissipation rate . Certain mathematical limitations on the Boussinesq theory,
Reynolds stresses, consistent with the turbulent flow physics,
Gk = μ t S 2 (6)
are addressed by this model, hence the term “realizable”. The
standard k- and RNG k- models are not realizable. The here S is the modulus of the mean rate-of-strain tensor,
realizable k- model more correctly estimates the spreading termed as
rate of both planar and round jets (Fluent 2009, 2013),
which is an important advantage of this model. Excellent S º 2Sij Sij (7)
performance is very likely from this model for flows
concerning rotation, boundary layer under strong adverse Notably, the formulation for k (Eq. (3)) is the identical to
pressure gradients, separation and recirculation. that of the standard k- method and the RNG k- method,
Rest of this section explains the theorem regarding the except for the model constants. On the other hand, the
realizable k- turbulence model. All k- (standard, RNG, formula of the  is very altered as of those in the standard
and realizable) models have similar outlines, along with the and RNG-based k- methods. One feature worth mentioning
transport equations for k and . The main differences in is that the turbulence generation phrase in the formula for
these models are in the followings:  (the expression on the right part of the Eq. (4)) does not
298 Hassan et al. / Building Simulation / Vol. 15, No. 2

contain the generation of k; i.e., it contains different Gk A0 = 4.04, As = 6 cos 

expression from the other k- methods. It is assumed that
the current expression better signifies the spectral energy here
transportation. A better characteristic is that the destruction
1 Sij , S jk , Ski
expression (the phrase subsequentto the last phrase on the  = cos-1 ( 6W ), W =
right part of Eq. (4)) does not maintain any singularity; i.e., 3 S3
1 ¶u j ¶ui
its denominator never disappears, even when k disappears S = Sij Sij , Sij = ( + )
or turns lesser than 0. This characteristic is the difference 2 ¶xi ¶x j
from the other k- methods, they comprise of a singularity
cause of the k in the denominator. Its observed that Cμ is related to the mean strain and
The realizable k- method had been broadly authenticated rotation rates, the angular velocity of the system rotation
considering an extensive variety of flows (Shih et al. 1995; and the turbulence fields (k and ). Cμ in Eq. (8) retains the
Kim et al. 1999), as well as rotating uniform shear flows, standard value of 0.09 for an inertial sub-layer in a balanced
free flows together with jets and mixing layers, channel and boundary layer.
boundary layer flows, and separated flows. All these flows 3.1.3 Model constants
showed significantly improved results for the realizable k-
method than that of the standard k- method. Specifically, The constants C2, σk and σ have been set to make sure that
the realizable k- method determines the round-jet anomaly; the method executes perfectly for some canonical flows. The
i.e., it calculates the distribution rate for axisymmetric jets model constants for this turbulence model are:
including planar jets.
C2 = 1.9, σk = 1.0, σ = 1.2
3.1.2 Modeling the turbulent viscosity
4 Numerical solution approach
Similar to other k- methods, the eddy viscosity is calculated
from
In this work, computational geometry is designed to be
k2 28d × 9d × 13.5d (seen in Figure 1), and the numerical
μt = ρC μ (8)
 solution is generated at Re = 12,000, based on the value of d
and U¥. Here d is the width of the square obstacle. The mesh
The distinction of the realizable k- method compared
has denser distribution surrounding the squared cylinder
to the standard and RNG k- methods is that Cμ is variable
obstacle walls and in the area covering 4d height from the
and is calculated from
ground surface. The number of cells and nodes, maximum
1 volume, minimum volume and average orthogonal quality
Cμ = (9) for each grid are presented in Table 2. In order to further
kU *
A0 + As improve the mesh quality, FLUENT’s Adapt-Smooth feature

is used (Fluent 2013). Figure 2 shows the results of the grid
here independence test. Results show only moderate differences
ij Ω
ij between Grid 1 and Grid 2. In the Grid 1, the number of
U * º Sij Sij + Ω (10)
cells are 7,540,910 and the number of nodes are 7,632,302,
and it is used for this work.
and
ANSYS Fluent 15.0 (Fluent 2013) is used to solve this
ij = Ωij - 2ijk ωk , Ωij = Ωij - ijk ωk
Ω problem. The pressure-based solver is utilized with velocity
formulation, and transient time is selected to make the flow
here Ωij represents the mean rate-of-rotation tensor seen unsteady and enable the time statistics. Near-wall treatment
in a rotating reference frame by means of the angular velocity is modeled using standard wall functions (Wang et al. 2011;
ωk. The model constants A0 and As are Wang et al. 2014; Joubert et al. 2015; Wang 2019). The

Table 2 Mesh information

Maximum volume Minimum volume Average orthogonal quality
Mesh Cells Nodes (m3) (m3) (Lachance-Barrett and Alexander 2018) Δy+
Grid 1 7, 540, 910 7, 632, 302 4.91 × 10−9 3.85 × 10−12 0.99 (outstanding) 2.120
−8 −12
Grid 2 6, 359, 104 6, 420, 292 1.48 × 10 9.91 × 10 0.99 (outstanding) 3.471
Hassan et al. / Building Simulation / Vol. 15, No. 2 299

Fig. 2 Grid independence test: velocity distributions at y/d = 2.5, (a,b,d) z/d = 0 and (c) x/d = 0

additional turbulent kinetic energy and dissipation rate normalized average stream-wise velocity on the x-z plane
equations are used for modeling the turbulent viscosity. Air are presented in Figures 3(a)–(b) at different altitudes y/d =
is used as the fluid with the constant viscosity (kg/(m·s)) 0.03 and y/d = 3, respectively. A horseshoe vortex is developed
of 1.5875×10−5. As for the solution methods semi-implicit encompassing the cylinder, quite near to the ground surface
pressure linked equations (SIMPLE) algorithm is applied for wall (at y/d = 0.03), visible in Figure 3(a). Bluff body flows
the pressure-velocity coupling. The second-order scheme have these features of the horseshoe vortex about a lower
spatially discretizes pressure, momentum, turbulent kinetic part of the obstacle. A prior computational investigation by
energy, and turbulent dissipation rate equations, and for Saeedi and Wang (2016) had the presence of the saddle and
the transient formulation, the second-order implicit solution nodal points in the horseshoe vortex upon the development
method is used (Wang et al. 2011; Uffinger et al. 2013; Saeedi of this vortex encompassing an obstacle. The air grows idle
et al. 2014; Want et al. 2014; Saeedi and Wang 2016; Wang because of the flow’s interaction with the cylinder about
2019; Leite et al. 2019; Behera and Saha 2020). Convergence the nodal point (discovered near to the close location of the
is assumed when the maximum residuals are less than 10−5 windward surface of the obstacle). Hence, a backward flow
for all the equations. The time step size (s) is 0.000155. The on the saddle point is developed, forming an arc-formed
maximum iteration per time step is 20. Approximately 55,000 reversed flow region in the idle area’s face. A couple of
of time steps in total are taken to obtain the converged acceleration regions likewise emerge symmetrically on
results. both sides of the cylinder. The stream-wise average velocity
develops in a drastic amount because of the obstruction
5 Results and discussions forces within the acceleration regions. It is apparent that the
flow compositions encompassing the solid object change by
The airflow patterns and statistics over a model square the altitude by examining Figure 3(a) and Figure 3(b). The
building with aspect ratio of four has been simulated for horseshoe vortex does not appear anymore in Figure 3(b)
a fixed Reynolds number, Re = 12,000 using the URANS on a higher altitude (for y/d = 3), but a couple of large
approach with realizable k- model. The detailed results re-circulating vortexes appear about the rear area of the
and discussions are given in the following subsections. obstacle, and a couple of small vortexes appear next to the
span-wise side surfaces. The small vortexes’ presence indicates
5.1 Vortex structures I the obstructing impact of the obstacle and the evolution
and separation of the boundary layers near both obstacle
The time-averaged stream-traces and contours of the sides, like a couple of acceleration regions in Figure 3(a).
300 Hassan et al. / Building Simulation / Vol. 15, No. 2

Fig. 3 Time-averaged stream-traces and contour of normalized Fig. 4 Time-averaged stream-traces and contour of normalized
average stream-wise velocity <u>/U∞ in x-z plane at (a) y/d = 0.03 average stream-wise velocity <u>/U∞ in x-z plane at (a) y/d = 1
and (b) y/d = 3 and (b) y/d = 3.5

Figures 4(a)–(b) show the time-averaged stream-traces on

two more horizontal planes (y/d = 1.0 and y/d = 3.5). Two
equivalent vortexes with opposed circulations are seen on
the lower elevation of y/d = 1.0 (Figure 4(a)) for the current
Reynolds number, this was previously evident for Re = 652
and 13,041 (Wang 2019). Then, two equivalent vortexes with
opposed circulations are still seen on the higher elevation
of y/d = 3.5 (Figure 4(b)), this was previously evident for Re =
652 (Wang 2019). Furthermore, two corner vortexes are seen
in this study for both elevations of y/d = 1 and y/d = 3.5.
Figure 5 illustrates time-averaged stream-traces of the
averaged stream-wise velocity on x-y plane (at z/d = 0) to
visibly show the extensive recirculation zone following the
obstacle. The formation of the extended recirculating eddy
following the obstacle is certainly seen in this illustration.
The recirculation zone’s extent prolonged from soaring to
more profound altitudes correlated to the down-wash of the Fig. 5 Contour and stream-traces of the time-averaged stream-wise
wind flow following the obstacle. Furthermore, some of the velocity <u>/U∞ in x-y plane at z/d = 0
Hassan et al. / Building Simulation / Vol. 15, No. 2 301

mean wind flow characteristics are displayed in Figure 5.

The horseshoe vortex and a recirculating flow zone
downstream of the cylinder are involved in the principal
flow formations that may be seen concerning a general
surface-mounted finite square cylinder with its leading
surface normal to the approaching wind flow (angle of
approaching flow is 0°), from aforementioned time-averaged
viewpoints (da Silva et al. 2020). The principal vortex
including its centre position downstream of the trailing
end, been noted as vortex Bt in previous works (Krajnović
2011; Sumner et al. 2017; da Silva et al. 2020), is typically Fig. 7 3D vortex formations in the wake zone and neighbouring
area of the obstacle ends reflected exercising low-pressure iso-surfaces
demonstrated in the time-averaged flow field on the vertical
at normalized pressure-coefficient Cp = −0.03.
(x-y) plane (Figure 5). A secondary vortex including the
ground facade designated vortex Nw (da Silva et al. 2020),
can also be contained close to the intersection of the cylinder. 5.2 Velocity statistics
The regions of the importance of these two eddies in the
5.2.1 Time-averaged velocity structure I
recirculating flow zone and the down-wash and up-wash of
the cylinder’s wind flow are demarcated via a saddle point. The time-averaged stream-wise and span-wise velocity profiles
Figure 6 illustrates the pressure field, which exhibits a retrieved from the realizable k- are examined against the
uniform and symmetrical vortex. This pattern of vortex experimental and previous numerical estimations in Figure 8
previously appeared in the work of Saaedi and Wang (2016). at two stream-wise positions (x/d = 2 and 3.5) at an altitude
The 3D vortical arrangements reflected plotting low-pressure of y/d = 3. The stream-wise velocities <u>/U¥ are negative
iso-surfaces, encompassing the obstacle, are presented in round z/d = 0 at the stream-wise position x/d = 2, presented
Figure 7. It is apparent from the illustration that there are in Figures 8(a)–(b), meaning a reversed occurring wind
a pair of alternating vortex tubes (marked tubes 1 and 2) flow in the recirculating zone following the obstacle. But,
advancing to the ground surface at a circuitous angle emitted the spot beyond the recirculation bubble, about x/d = 3.5,
from the obstacle. These vortex tubes were previously seen witnesses no negative velocity. The earlier analysis of
in the work of Saaedi and Wang (2016). Figure 7 exhibits big Figures 3–4 on the configuration and extent of the recirculation
complex alternating low-pressure vortex tubes profoundly bubble in the immediate downstream area of the cylinder
embedded in the turbulent wake downstream of the obstacle, has consistency with these results. Boundary layers evolve
which is initiated by interacting flow volatility small vortex and distribute toward both vertical sides of the obstacle
tubes developed spatially and emitted within the rear section constructing two robust shear layers in the downstream
of the obstacle. area while the wind flow crosses about the obstacle, similar
to the scenario seen earlier in Figures 3–4. The tip values of
Figures 8(a)–(d) intimates these two robust shear layers.
While the stream-wise range of the flow from the obstacle
progresses, the intensity of the shear layers and the affiliated
tip value of <u> reduces. Certainly, the association of
Figures 8(a, b) and (c, d) confirms that the tip values of
<u>/U¥ at x/d = 3.5 are almost 10% lesser than at x/d = 2.
The estimated and retrieved time-averaged span-wise
velocity profiles are analyzed in Figures 8(e)–(h) at the exact
positions. A couple of tip values are seen in the time-averaged
span-wise velocity profile, as presented in Figures 8(e, f),
through x/d = 2. The obstruction impact causes the big
tip values in the side areas of the obstacle, and the pair
of recirculating vortexes, following the obstacle, causes the
inner little tip values. The inside miniature tip values have
gone at x/d = 3.5, meaning that the recirculating vortexes
Fig. 6 Kármán vortex street displayed via the contours of nature of the counter-rotating bubble does not exist in
normalized pressure-coefficient (Cp) on the x-z plane at y/d = 1 this position anymore. The computational estimations of the
302 Hassan et al. / Building Simulation / Vol. 15, No. 2

Fig. 8 Time-averaged stream-wise and span-wise velocity profiles downstream of the cylinder at two stream-wise places (x/d=2 and 3.5)
on an altitude of y/d = 3: (a) <u>/U∞ at x/d = 2, −z direction; (b) <u>/U∞ at x/d = 2, +z direction; (c) <u>/U∞ at x/d = 3.5, −z direction (d)
<u>/U∞ at x/d = 3.5, +z direction; (e) <w>/U∞ at x/d = 2, −z direction; (f) <w>/U∞ at x/d = 2, +z direction; (g) <w>/U∞ at x/d = 3.5, −z
direction; (h) <w>/U∞ at x/d = 3.5, +z direction (PIV experiment by Bourgeois et al. (2011, 2013) and LES by Saeedi and Wang (2016))

time-averaged stream-wise and span-wise velocity profiles on the whole. <u>/U¥ has 0 value on the leeward cylinder
agree fairly with the wind tunnel experimental and previous facade (positioned on x/d = 0.5), as displayed in Table 3, of
numerical data in general, it is seen in Figure 8. The the obstacle then retains negative values into its immediate
time-averaged stream-wise velocity <u> in the side zones downstream area, indicating the counter-rotating wind
of the obstacles (for |z/d| > 1) nevertheless is somewhat flow outline in the circulating flow bubble following the
overestimated, as presented in Figures 8(a)–(d). The disparity obstacle (seen in Figures 3–5). The negative tip point of
in the obstruction proportions of the wind tunnel experiment the velocity is around <u>/U¥ = −0.21 at y/d = 1 along with
and the current simulation is largely accountable for the <u>/U¥ = −0.17 at y/d = 3. It is clear by evaluating the
over-estimation of <u>. The wind tunnel experiment had a outlines on altitudes y/d = 1 and 3 that the time on which
cross-sectional region of the analysis area of 1550d2 (the side the time-averaged stream-wise velocity evolves to the free-
width of the square-shaped analysis area was 0.5 m, and the stream point (that is, <u>/U¥ = 1) is noticeably earlier on
side width of the square-shaped obstacle was d = 0.0127 m) y/d = 3 compared to on y/d = 1. The area of the circulating
(Bourgeois et al. 2011, 2013). On the other hand, this flow section is wider on a lower altitude, which is connected
computational work has a cross-sectional area of 120.6d2 to the earlier argument. It is also carefully observed that
(13.4d × 9d), this is just around 8% of that of the analysis the realizable k- model better predicts the flow in the
region of the wind tunnel experiment (Saeedi et al. 2014; stream-wise region of x/d = 1.1 to x/d = 3.1 for y/d = 1.
Saeedi and Wang 2016). Thus, this computational work has Additionally, the result of Table 3 shows that the realizable
12.8 times higher obstruction rate compared to the wind k- model also performs better in the region of x/d = 0.5 to
tunnel experiment (3.3% vs. 0.26%). x/d = 2.7 for y/d = 3.
The estimated and calculated stream-wise time-averaged Two more dissimilar span-wise spots (seen in Figure 9)
velocity (<u>) profiles, in the mid x-y plane (z/d = 0), are stream-wise time-averaged velocity and span-wise time-
evaluated and displayed in Table 3 on two dissimilar altitudes averaged velocity are measured up to the experimental
for y/d = 1 and 3. The estimated and calculated values (Bourgeois et al. 2011, 2013) and previous numerical (Wang
concur convincingly with each other on these two altitudes 2019) data. Fair concurrences are seen in the calculated
Hassan et al. / Building Simulation / Vol. 15, No. 2 303

Table 3 Comparison of the time-averaged normalized stream-wise velocity <u>/U∞ with PIV (Bourgeois et al. 2011) and LES (Saeedi
et al. 2014). Values inside the parentheses indicate the relative error of the corresponding results with the PIV results
x/d 0.5 1.35 2.5 4.5 6.5 8.5 10.5 12.5
PIV −0.07621 −0.35196 −0.07206 0.34642 0.5515 0.68314 0.76074 0.8037
LES −0.00277 −0.33118 −0.22448 0.29515 0.5903 0.73995 0.82171 0.89099
(rel. err.) (0.964) (0.059) (2.115) (0.148) (0.070) (0.083) (0.080) (0.109)
y/d = 1 Pressure outlet −0.00277 −0.21478 −0.07206 0.42402 0.62771 0.75104 0.82864 0.88406
(rel. err.) (0.964) (0.390) (0.000) (0.224) (0.138) (0.099) (0.089) (0.010)
Outflow −0.00277 −0.20508 0.0679 0.4739 0.66513 0.7746 0.85081 0.90346
(rel. err.) (0.964) (0.417) (1.942) (0.368) (0.206) (0.134) (0.118) (0.124)
PIV −0.05304 −0.19009 0.17867 0.78869 0.90544 0.9459 0.96128 0.96202
LES −0.0034 −0.13679 0.17867 0.79496 0.95978 1.00965 1.02398 1.03936
(rel. err.) (0.936) (0.280) (0.000) (0.008) (0.060) (0.067) (0.065) (0.080)
y/d = 3 Pressure outlet −0.0034 −0.15978 0.10342 0.8608 1.00995 1.04623 1.05847 1.05921
(rel. err.) (0.936) (0.159) (0.421) (0.091) (0.115) (0.106) (0.101) (0.101)
Outflow −0.0034 −0.17023 0.17344 0.90783 1.02458 1.05041 1.05847 1.05921
(rel. err.) (0.936) (0.104) (0.029) (0.151) (0.131) (0.110) (0.101) (0.101)

Fig. 9 Evaluation of velocity profiles on y/d = 2: (a) time-averaged stream-wise velocity at z/d = −0.0877, (b) time-averaged stream-wise
velocity at z/d = −0.321, (c) time-averaged span-wise velocity at z/d = −0.0877, (d) time-averaged span-wise velocity at z/d = −0.321 (PIV
experiment by Bourgeois et al. (2011, 2013) and RSM by Wang (2019))

stream-wise time-averaged velocity and span-wise time- the stress components. The stream-wise distributions of the
averaged velocity compared to the measured wind tunnel stress component <u ¢¢v ¢¢> / U ¥
2
(Boussinesq approach vs.
experimental data. The current realizable k- model performs Reynolds Stress Transport Model) are analyzed in Figure 10
much better in the regions of x/d = 0.5 to 0.9 in Figure 9(a), on two separate altitudes (y/d = 1 and 3) on the mid x-y plane
x/d = 0.5 to 0.7 in Figure 9(b), x/d = 0.7 to 3.1 in Figure 9(c), (z/d = 0). The course of the distributions has been achieved
and x/d = 0.7 to 3.3 in Figure 9(d). by the simulation compared to the wind tunnel experimental
data, which is visible in the figure. The tip values of
5.2.2 Stress components
<u ¢¢v ¢¢> / U ¥
2
on these two altitudes have opposing signs,
This work uses Boussinesq approach (Hinze 1975) to retrieve which is a notable remark. The wind flow features within the
304 Hassan et al. / Building Simulation / Vol. 15, No. 2

confines of the circulating flow bubble are largely responsible estimation data at the identical positions to the time-
for this: on a more elevated altitude of y/d = 3; the velocity averaged velocity distributions in Figure 8. The highest
distribution is governed by a forward-down-wash stream as values of stress components are found at z/d ≈ ±1 – here the
noted previously in Figure 5; on a lower altitude of y/d = 1. shear generation is the most robust because of the boundary-
It is, however, governed by a reverse-down-wash stream. layer evolution and detachment from two side walls (correlated
Close to the edge of the circulating flow bubble, turbulence to Kárrmán vortexes), and this is visible in Figure 11. In
motions are the most energetic. At the remote downstream the outward side areas of the obstacle (for |z/d| > 2), the
area, on the greater altitude of y/d = 3, <u ¢¢v ¢¢> / U ¥
2
displays turbulence intensity decreases notably – here the values of
inadequate values. The time-averaged velocity gradients <u ¢¢w ¢¢> / U ¥
2
have turned almost nonexistent.
(particularly, ∂<u>/∂y) are nevertheless high on the more
5.2.3 RMS velocity structure
profound altitude of y/d = 1, and the generation term supports
the stress component’s peak values (this is in balance to The normalized stream-wise and span-wise RMS velocities
the time-averaged velocity gradients). Hence, <u ¢¢v ¢¢> / U ¥ 2
are analysed in Figure 12 against the wind-tunnel measure-
shows the nontrivial output on the elevation of y/d = 1 ment data and LES data at stream-wise position of x/d = 2 on
within the remote downstream area of the obstacle. The the altitude of y/d = 3. The distribution of <urms>/U¥ shows a
prediction ability of the present realizable k- model is better clear dual-tip pattern at x/d = 2, visible in Figures 12(a)–(b).
in the region of x/d = 2 to 4 for y/d = 1 and x/d = 1 to 2 and This signifies that the firm unsteady shear layers profoundly
x/d = 2.6 to 5 for y/d = 3, as seen in Figure 10. impact the most intense stream-wise turbulent departed
The numerical predictions of the time-averaged stress from the obstacle’s sidewalls, which happens at z/d ≈ 1
component <u ¢¢w ¢¢> / U ¥ 2
(Boussinesq approach vs. Reynolds (about 0.5d off the obstacle sidewalls). The cross-stream
Stress Transport Model) are analysed in Figure 11 with distributions of the span-wise RMS velocity retrieved from

Fig. 10 Stream-wise distributions of the normalized time-averaged stress components (Boussinesq approach vs. Reynolds Stress Transport
Models) <u¢¢v ¢¢> / U ¥
2
on two dissimilar elevations at the mid x- y plane z/d = 0: (a) y/d = 1, (b) y/d = 3 (PIV experiment by Bourgeois
et al. (2011, 2013) and LES by Saeedi and Wang (2016))

Fig. 11 Cross-stream outlines of the normalized stress components (Boussinesq approach vs. Reynolds Stress Transport Models)
<u¢¢w ¢¢> / U ¥
2
at two stream-wise positions on the altitude of y/d = 3: (a) x/d = 2, −z direction; (b) x/d = 2, +z direction; (c) x/d = 3.5, −z
direction; (d) x/d = 3.5, +z direction (PIV experiment by Bourgeois et al. (2011, 2013) and LES by Saeedi and Wang (2016))
Hassan et al. / Building Simulation / Vol. 15, No. 2 305

Fig. 12 Cross-stream distributions of the time-averaged stream-wise and span-wise RMS velocities at x/d = 2 on the altitude of y/d = 3:
(a) <urms)/U∞, −z direction; (b) <urms>/U∞, +z direction; (c) <wrms>/U∞, −z direction; (d) <wrms>/U∞, +z direction (PIV experiment by
Bourgeois et al. (2011, 2013) and LES by Saeedi and Wang (2016)

the computational simulation are compared against the differences in the obstruction ration and spatial extent amid
experimental and LES data in Figures 12(c)–(d) at x/d = 2. the wind tunnel experiment and the computational simulation
The distribution of <wrms>/U¥ also displays a duel-tip pattern may correlate to the disparities in the estimated and acquired
within the circulating flow bubble, alike to the distribution values. The environment turbulence intensity is low in this
of <urms>/U¥, however with more profound degree, visible computational simulation, shown by the disparities in the
in Figures 12(c)–(d). The realizable k- estimation is in far side areas. The inflow state practiced in this computational
accordance with the wind tunnel experiment data for the simulation possibly associated with this. The peak values at
course of the stream-wise and span-wise RMS velocity z/d = −0.8 and 0.8, seen in Figures 12(a) and (b), are captured
distributions generally. Under-prediction, to their extent, is better by this current realizable k- model.
observed, nonetheless. As explained earlier in Figure 8, the Figure 13 illustrates RMS stream-wise and RMS span-wise

Fig. 13 Resemblance of velocity distributions on the altitude of y/d = 2: (a) RMS stream-wise velocity, <urms)/U∞ at z/d = −0.0877, (b) RMS
stream-wise velocity, <urms)/U∞ at z/d = −0.321, (c) RMS span-wise velocity, <wrms>/U∞ at z/d = −0.0877, (d) RMS span-wise velocity,
<wrms>/U∞ at z/d = −0.321 (PIV experiment by Bourgeois et al. (2011) and RSM by Wang (2019))
306 Hassan et al. / Building Simulation / Vol. 15, No. 2

velocity contrast with the experimental (Bourgeois et al.

2011, 2013) and previous numerical (Wang 2019) data at two
dissimilar span-wise positions. Calculated RMS span-wise
velocity is in excellent accord with measurements. The model
however under-predict the RMS stream-wise velocity at both
positions, as seen in Figures 13(a, b).

5.2.4 Time-averaged velocity structure II

Most velocity statistics show improved results for addressed

back-flow turbulence intensity in the outlet boundary con-
Fig. 14 Iso-surface of normalized time-averaged pressure coefficient
dition compared to traditional outlet boundary condition.
at <Cp> =−0.03
Overall, addressing the back-flow turbulence intensity during
computation has a significant influence on the computational
prediction. Hence, the results of addressed back-flow
turbulence intensity are used throughout the paper for
analysis.
The time-averaged contour of the stream-wise velocity
is also presented in Figure 5 in the mid x-y plane (positioned
at z/d = 0), to show the comprehensive recirculating
flow region following the obstacle. The comprehensive
recirculation zone, formed following the obstacle, is observed
in this illustration. The recirculating flow region following
the obstacle has the consistent position and extent compared
to the experiment (Bourgeois et al. 2011, 2013) and previous
numerical (Saeedi and Wang 2016; Wang 2019) results.
Correlated to the down-wash of the flow following the Fig. 15 Vortex core region using Swirling Strength Criterion at
obstacle, the recirculating flow zone’s extent continues from u = 0.0025
more leading to more profound altitudes. But, there is a
miniature recirculating flow region on the tip of the obstacle viewed in the near-wake area for the case considered in this
in this numerical simulation, exhibited in Figure 5, which investigation. Previous experimental studies by Bourgeois
was not experimentally witnessed. Rather it was noticed et al. (2011, 2013) (for h/d = 4 and Re = 12,000) and previous
in previous numerical works (Saeedi and Wang 2016; numerical studies by Wang (2019) (for h/d = 7 and Re =
Wang 2019). 13,041), which figured that the shed configuration structure
a half-loop form for the un-tripped case (thinner boundary
5.3 Vortex structures II layer), detected this half-loop shedding configurations. It is
mentionable that in their works, the λ2-criterion (Bourgeois
Figure 14 presents the recognised vortex cores for the time- et al. 2011, 2013) and Q-criterion (Wang 2019) were exercised
averaged flow (as normalized time-averaged pressure to discover vortex formations; on the other hand, in the
coefficient <Cp> = −0.03), including vortex cores recirculating current investigation, the vortex formation is effectively
in the core area of the cylinder also including down-wash detected by Swirling Strength Criterion (u) instead of
on y/d = 0. Configuration of a dipole wake is observed in traditional quantities (like λ2- and Q-criterion).
this iso-surface of the normalized time-averaged pressure- A braided vortex formation is also seen in the distant
coefficient for the current Re value. A dipole wake was downstream region for the current Re value, in addition
previously seen in the experimental investigation carried to the standard vortex formation discovered following the
out by Bourgeois et al. (2011, 2013) (for Re = 12,000) and obstacle (seen in Figure 15). A previous numerical work
in the numerical investigation carried out by Wang (2019) carried out by Wang (2019) (for h/d = 7 and Re = 13,041),
(for Re = 13,041), for thin boundary layer cases. It might in an investigation of the turbulent wake of a finite surface-
be anticipated that just dipole wakes are found since the mounted squared shape obstacle, published such a far-wake
current investigation has a thin developing boundary layer braided vortex formation. This vortex is made of several
thickness. knitted hairpin vortex formations. The overall vortex
Figure 15 illustrates the vortex core region using Swirling formation was not transparent while a different study by
Strength Criterion, u. The half-loop formations are Rastan et al. (2017) had inspected the specific hairpin vortex
Hassan et al. / Building Simulation / Vol. 15, No. 2 307

formations. Asymmetric hairpin vortex formations were Re = 12,000). The experimental investigations of Bourgeois
witnessed in a DNS study by Rastan et al. (2017) (for h/d = et al. (2011, 2013) (for h/d = 4 and Re = 12,000) have
7 and Re = 40–250). The asymmetric hairpin vortexes are previously witnessed this vortex.
seen in the current investigation to connect into a braided
vortex formation while directing in an earthward orientation. 5.4 Vortex structures III
Figure 16 shows the vortex topologies for the current case
(for h/d = 4 and Re = 12,000). The λ2-criterion (Bourgeois et al. 2011, 2013) and Q-criterion
There is a half-loop structure in the braided vortex seen (da Silva et al. 2020) were previously exercised to retrieve
following the obstacle. The primary hairpin vortex, which the topologies in Figures 17–19. Bourgeois et al. (2011, 2013)
already displays the asymmetrical form, is detected following (experiment, h/d = 4 and Re = 12,000) and da Silva et al.
the half-loop vortex formation. Next, the one-legged form (2020) (numerical, h/d = 3 and Re = 500) listed alike time-
is taken by the secondary hairpin vortex and every subsequent averaged vortex formations, where the horizontal “tube” that
hairpin vortexes. The ’Principle Vortex Core’ is seen in travels toward the distant wake is correlated to the stream-
addition following the obstacle. Such a vortex formation wise vorticity. The flow close to the cylinder is governed
is described for the first time in a numerical investigation by the formation surrounding the cylinder by its side and
of a turbulent wake of a finite surface-mounted squared free-end, and through other minor but compact zones. The
shape obstacle for the current scenario (for h/d = 4 and stream-wise vorticity is correlated to most of these small
velocity gradient zones in the neighboring wake. The
y-vorticity (Figure 18) showing the enveloping formation
developed because of the separating shear-layer about the
cylinder side. In contrast, each three vorticity elements
constitute complex structures over the free-end and close
to the ground surface.
The 3D stream-traces are used to display the corner
vortex in Figure 20. Similar to the one witnessed by da Silva
et al. (2020), this vortex formation is curved skyward and
downstream, inducing an area of high-magnitude clockwise
x-vorticity (further added in Figure 20). This vortex is initiated
by the flow that enters the front end of the cylinder, which is
then diverted earthward and divided close to the upstream
corner. Because of fluid entrainment from its neighbouring
environment, this flow is diverted skyward, although it waits
inside the boundary layer until it is transported toward the
wake. It is also noted that the corner vortex further curved
Fig. 16 Vortex core region using Swirling Strength Criterion at outward (off the cylinder), so the stream-traces encompass
u = 0.005 the time-averaged re-circulating flow zone.

Fig. 17 Iso-surfaces of the normalized time-averaged stream-wise vorticity at (a) <ωx>d/U∞ = 0.7 and (b) ωx>d/U∞ = −0.7
308 Hassan et al. / Building Simulation / Vol. 15, No. 2

Fig. 18 Iso-surface of the normalize time-averaged vertical vorticity at (a) <ωy>d/U∞ = 1.8 and (b) <ωy>d/U∞ = −1.8

Fig. 19 Iso-surface of the normalized time-averaged span-wise vorticity at (a) <ωz>d/U∞ = 1.8 and (b) <ωz>d/U∞ = −1.8

The corner vortex, too, comprises each vorticity com-

ponent’s elements additional to just the stream-wise vorticity,
as seen in Figures 17–19. Figure 21 highlights the corner
vortex-dominated area, including details of vortex Nw by
plotting the z-vorticity iso-surfaces close to the cylinder
terminal with the ground surface. Vortex Nw is principally
restricted to the central area following the cylinder; this is
explained by the iso-surfaces and 3D stream-traces. With the
flow coming from the recirculation zone close to the ground
surface, this vortex disappears close to the span-wise edge;
more specifically, it is diverted to the side of the cylinder.
As portrayed in Figure 21, few of the diverted stream-traces
enter the corner vortex area or re-enter the recirculation
zone before exiting the recirculating flow area.
The iso-surface and 3D stream-traces in Figure 22
illustrate the areas of high x-vorticity in the top portion of
the cylinder and the −z section of the computing domain.
Sumner et al. (2017) previously noted the positive, counter-
Fig. 20 Iso-surface of the normalized time-averaged stream-wise clockwise x-vorticity by the span-wise side edge at the top
vorticity at <ωx >d / U ¥ = -1.4 and 3D stream-traces for the time- surface of the cylinder, which was considered to be started
averaged velocity at the leading edge of the free end on both span-wise sides
Hassan et al. / Building Simulation / Vol. 15, No. 2 309

flow is transported by the divided shear layer as the flow

shifts to the low-pressure area by the front end of the
cylinder, depicting a flow that includes each element of
vorticity components. Sumner et al. (2017) and da Silva
et al. (2020) have given a similar statement, i.e., the negative,
clockwise vorticity aloft the top surface of the cylinder is
correlated to the outward deflection of the time-averaged
motion. It is noted that this outward deflection is succeeded
with a positive x-vorticity, i.e., the counter-clockwise “tip
vortex” since this deflection disappears not so far off the
cylinder.
The current data determines that the time-averaged
x-vorticity in this area is a consequence of the three-
dimensional curving of the motion, here the tip vortexes
are similar to edge vortexes, enveloped by notable vagueness.
The flow departs from the span-wise sides of the cylinder
and encompasses the near-wake recirculating flow zone,
presented in Figure 22. da Silva et al. (2020), for surface
Fig. 21 Iso-surface of the time-averaged span-wise vorticity at attached finite-height squared shape cylinder, identified that
<ωz>d/U∞ = −1.5 and 1.5 and 3D stream-traces for the time- the fluid motion has a skyward element near the free-end.
averaged velocity close to the zone of vortex Nw However, beyond the wake recirculating flow zone, the
span-wise-side flow joins the down-wash from the cylinder’s
tip and shifts earthward in the far-wake zone. The shift
becomes more definite and developed when the extent of
the down-stream from the near wake rises, furthermore,
this shift becomes most intense following the recirculating
flow zone. The stream-wise vorticity is further determined
by Bourgeois et al. (2011, 2013), seen as connector stands
in a phase-averaged view, to be produced by the y- and
z-vorticities that occur from the parted shear layers from the
span-wise-sidewalls and free-end, sequentially. In summary,
the current data suggest that the flow formations above the
free-end disappear, and are succeeded by this large-scale
stream-wise vorticity because of the three-dimensional
curving of the span-wise-side motion by the down-wash.
Accordingly, as previously suggested by da Silva et al. (2020),
there is no straight link amid the “tip” vortex and the
vortex formations above the free end of the cylinder in this
time-averaged view.

Fig. 22 Iso-surface of the normalized time-averaged stream-wise 6 Conclusion

vorticity at <ωz>d/U∞ = −1 and 1 and 3D stream-traces for the
time-averaged velocity close to the areas of high vorticity Reproduction of the wind-tunnel experiment of Bourgeois
et al. (2011, 2013) was one of the purposes of this investigation
of the cylinder, being an “edge vortex.” The high vorticity to use their determined data to verify the acquired first-
area does not fundamentally resemble a vortex formation, and second-order turbulence statistics. The simulations have
as seen from the current outcomes. The positive stream- successfully replicated the central flow and 3D large-scale
wise vorticity of a large degree does not develop from the vortex formation in the wake of the cylinder.
edge in this scenario. Rather, it arises from the skyward Important remarks from this work are:
flow by the span-wise side surface of the cylinder and its (i) A horseshoe vortex is detected on a rather profound
entrainment toward the recirculating flow area that takes altitude near to the ground surface (on y/d = 0.03). The
place above the free end. Noticeable in Figures 17–19, the horseshoe vortex disappears as the altitude rises to y/d = 3,
310 Hassan et al. / Building Simulation / Vol. 15, No. 2

and two big counter-rotating vortexes in the rear section (viii) In the front intersection corners of the obstacle, a
of the cylinder and two tiny vortexes next to the span- vortex structure is observed by the ground surface
wise-sidewalls emerge. Two other altitudes, y/d = 1 and because of the distributed fluid motion interaction
y/d = 3.5, show these two big counter-rotating vortexes with the boundary layer. Identified as “corner vortex”,
and these two tiny vortexes. The incidence of Kármán this vortex is of tapering shape, curved skyward and
shedding has been examined using an outline of the outward, plus its plumb expansion is assumed to rely
pressure distribution. on the boundary layer width.
(ii) The distributions of the stress component <u ¢¢w ¢¢> / U ¥
2
(ix) Fluid motion from near the ground surface toward a
display an anti-symmetric dual-peak character and the low-pressure zone following the obstacle induces the
highest measure of <u ¢¢w ¢¢> / U ¥ 2
through z/d ≈ ±1. base flow formation. This rises in a 3D spiral shape,
The divided boundary layers from the obstacle span- up to around the obstacle mid-section, covering its
wise-side surfaces govern the fluid motion through span-wise-sides. The rolling-up of the distributed fluid
z/d ≈ ±1, producing high turbulent shear stresses. The motion channeled from the free-end traversing point
distributions of urms and wrms show symmetrical dual-peak induces the top flow formation. The central vortex Bt
characteristics compared to the characteristics of the in the recirculating flow zone results from this fluid
distribution of <u ¢¢w ¢¢> / U ¥2
inside the recirculating motion that grows to the outer zone to join the base
flow area because of the strong shear composition rate flow formation. However, the flow moves away from the
from span-wise-sides of the obstacle. neighboring wake before joining the separate structure’s
(iii) The up-wash fluid motion is weaker for this Reynolds dominance zone conversely.
number to that extent; the flow formation is governed Present information also aids in recognizing the facade’s
by the down-wash fluid motion and the span-wise root and interconnection, adjacent-wake, and faraway-
vortex shedding. wake formations in the fluid motion encompassing square
(iv) The alternating half-loop shedding with a dipole wake cylinder. Thicker boundary layers and influences of much
is seen in the wake structure following the obstacle. higher Reynolds numbers can be considered to investigate
Then, a braided vortex formation is developed in the 3D flow formations. Additionally, their characteristics,
far-wake area by hairpin vortexes. encompassing the cylinder of unconventional aspect-ratio,
(v) Swirl Strength Criterion (u) can be efficiently practiced would be of interest to expand this data and continue
for the analytical measure of the vortex formation the enrichment of this research to more prevailing flow
because of its simplistic usability. circumstances.
(vi) Furthermore, to explain the very near-wall fluid motion
on the facades of the cylinder and the ground surface, Acknowledgements
and its correlation with the near-wake and far-wake
flow formations, a time-dependent strategy is selected. The first three authors acknowledge the Grant for Advanced
Surface-projected and 3D points of view of the time- Research in Education (GARE), Bangladesh Bureau of
averaged flow formations are examined to describe Educational Information Statistics (BANBEIS), Ministry
these graphical resources. of Education, Bangladesh, for providing the financial
(vii) In the neighboring and faraway wake, a different support for this research gratefully (No. MS20191054). This
stream-wise vorticity zone forms, and these were work has also been funded by the Faculty Research Grant
introduced in earlier research being “tip vortexes”. (CTRG19/SEPS/9 CTRG19/SEPS/15), North South University
However, this investigation notes an important (NSU), Dhaka, Bangladesh.
vagueness in the source and evolution of these zones
References
of vorticity. An influential association to the other
flow formations atop the obstacle is not witnessed for
Bautista MC (2015). Turbullence modelling of the atmospheric
this vorticity. It develops because of the 3-dimensional
boundary layer over complex topography. PhD Thesis, École de
curving of the distributed span-wise-side fluid motion technologie supérieure, Canada.
instead since it encompasses the neighboring-wake Bazdidi-Tehrani F, Jadidi M (2014). Large eddy simulation of dispersion
recirculating flow area, induced by the down-wash. around an isolated cubic building: evaluation of localized dynamic
Following the recirculation zone, where the down- kSGS-equation sub-grid scale model. Environmental Fluid Mechanics,
wash is the most powerful, this vorticity is most robust, 14: 565–589.
and its intensity diminishes downstream in the faraway Bazdidi-Tehrani F, Mohammadi-Ahmar A, Kiamansouri M, et al.
wake region. (2015). Investigation of various non-linear eddy viscosity turbulence
Hassan et al. / Building Simulation / Vol. 15, No. 2 311

models for simulating flow and pollutant dispersion on and around Jadidi M, Bazdidi-Tehrani F, Kiamansouri M (2018). Scale-adaptive
a cubical model building. Building Simulation, 8: 149–166. simulation of unsteady flow and dispersion around a model building:
Bazdidi-Tehrani F, Kiamansouri M, Jadidi M (2016). Inflow turbulence spectral and POD analyses. Journal of Building Performance
generation techniques for large eddy simulation of flow and Simulation, 11: 241–260.
dispersion around a model building in a turbulent atmospheric Joubert EC, Harms TM, Venter G (2015). Computational simulation
boundary layer. Journal of Building Performance Simulation, 9: of the turbulent flow around a surface mounted rectangular
680–698. prism. Journal of Wind Engineering and Industrial Aerodynamics,
Bazdidi-Tehrani F, Gholamalipour P, Kiamansouri M, et al. (2019). 142: 173–187.
Large eddy simulation of thermal stratification effect on convective Kim S-E, Choudhury D, Patel B (1999). Computations of complex
and turbulent diffusion fluxes concerning gaseous pollutant turbulent flows using the commercial code fluent. In: Salas MD,
dispersion around a high-rise model building. Journal of Building Hefner JN, Sakell L (eds), Modeling Complex Turbulent Flows.
Performance Simulation, 12: 97–116. ICASE/LaRC Interdisciplinary Series in Science and Engineering.
Bazdidi-Tehrani F, Masoumi-Verki S, Gholamalipour P (2020). Dordrecht, Netherlands: Springer.
Impact of opening shape on airflow and pollutant dispersion in a Krajnović S (2011). Flow around a tall finite cylinder explored by
wind-driven cross-ventilated model building: Large eddy simulation. large eddy simulation. Journal of Fluid Mechanics, 676: 294–317.
Sustainable Cities and Society, 61: 102196. Lachance-Barrett S, Alexander K (2018). FLUENT—Wind Turbine
Behera S, Saha AK (2020). Evolution of the flow structures in an Blade FSI (Part 1)—Mesh. Cornell University.
elevated jet in crossflow. Physics of Fluids, 32: 015102. Leite HF, Diógenes AN, Avelar AC (2019). Numerical and experimental
Bourgeois JA, Sattari P, Martinuzzi RJ (2011). Alternating half-loop investigation of a finite cylinder wake. Journal of the Brazilian
shedding in the turbulent wake of a finite surface-mounted Society of Mechanical Sciences and Engineering, 41: 240.
square cylinder with a thin boundary layer. Physics of Fluids, Liu JC (1989). Coherent structures in transitional and turbulent free
23: 095101. shear flows. Annual Review of Fluid Mechanics, 21: 285–315.
Bourgeois JA, Noack BR, Martinuzzi RJ (2013). Generalized phase Liu Y, Nie D (2017). Lattice boltzmann simulation of flow past a finite
average with applications to sensor-based flow estimation of the cylinder. IOP Conference Series: Materials Science and Engineering,
wall-mounted square cylinder wake. Journal of Fluid Mechanics, 224: 012021.
736: 316–350. Moazamigoodarzi N, Bergstrom DJ, Einian M, et al. (2014). Phase
Buccolieri R, Gromke C, di Sabatino S, et al. (2009). Aerodynamic effects average visualization of a finite cylinder wake as predicted by
of trees on pollutant concentration in street canyons. Science of large eddy simulation. In: Zhou Y, Liu Y, Huang L, et al. (eds),
the Total Environment, 407: 5247–5256. Fluid-Structure-Sound Interactions and Control. Lecture Notes
Davis P, Rinehimer A, Uddin M (2012). A comparison of RANS- in Mechanical Engineering. Berlin: Springer.
based turbulence modeling for flow over a wall-mounted square Rastan MR, Sohankar A, Alam MM (2017). Low-Reynolds-number
cylinder. In: Proceedings of the 20th Annual Conference of the flow around a wall-mounted square cylinder: Flow structures and
CFD Society of Canada. onset of vortex shedding. Physics of Fluids, 29: 103601.
Etzold F, Fiedler H (1976). The near-wake structure of a cantilevered Rowley CW, Williams DR (2006). Dynamics and control of high-
cylinder in a cross-flow. Zeitschrift fur Flugwissenschaften, 24: Reynolds-number flow over open cavities. Annual Review of
77–82. Fluid Mechanics, 38: 251–276.
Fluent A (2009). 12.0 Theory Guide. Canonsburg, PA, USA: Ansys Inc. Saeedi M, LePoudre PP, Wang B-C (2014). Direct numerical simulation
Fluent A (2013). Ansys FLUENT Theory Guide 15.0. Canonsburg, of turbulent wake behind a surface-mounted square cylinder.
PA, USA: Ansys Inc. Journal of Fluids and Structures, 51: 20–39.
Gromke C, Buccolieri R, di Sabatino S, Ruck B (2008). Dispersion Saeedi M, Wang B-C (2016). Large-eddy simulation of turbulent flow
study in a street canyon with tree planting by means of wind tunnel around a finite-height wall-mounted square cylinder within a thin
and numerical investigations—Evaluation of CFD data with boundary layer. Flow, Turbulence and Combustion, 97: 513–538.
experimental data. Atmospheric Environment, 42: 8640–8650. Saha AK (2013). Unsteady flow past a finite square cylinder mounted
Han B-S, Kwak K-H, Baik J-J (2017). Analysis on vortex streets behind a on a wall at low Reynolds number. Computers & Fluids, 88:
square cylinder at high Reynolds number using a large-eddy 599–615.
simulation model: Effects of wind direction, speed, and cylinder Salim SM, Buccolieri R, Chan A, et al. (2011). Numerical simulation
width. Atmosphere, 27: 445–453. (in Korean) of atmospheric pollutant dispersion in an urban street canyon:
Hemmati A, Wood DH, Martinuzzi RJ (2016). Effect of side-edge Comparison between RANS and LES. Journal of Wind Engineering
vortices and secondary induced flow on the wake of normal thin and Industrial Aerodynamics, 99: 103–113.
flat plates. International Journal of Heat and Fluid Flow, 61: Sattari P, Bourgeois JA, Martinuzzi RJ (2012). On the vortex dynamics
197–212. in the wake of a finite surface-mounted square cylinder. Experiments
Hinze J (1975). Turbulence. New York: McGraw-Hill Publishing. in Fluids, 52: 1149–1167.
Jadidi M, Bazdidi-Tehrani F, Kiamansouri M (2016). Embedded large Shah KB, Ferziger JH (1997). A fluid mechanicians view of wind
eddy simulation approach for pollutant dispersion around a engineering: Large eddy simulation of flow past a cubic obstacle.
model building in atmospheric boundary layer. Environmental Journal of Wind Engineering and Industrial Aerodynamics, 67–68:
Fluid Mechanics, 16: 575–601. 211–224.
312 Hassan et al. / Building Simulation / Vol. 15, No. 2

Shaheed R, Mohammadian A, Gildeh HK (2019). A comparison of cylinder geometries of finite length. Journal of Wind Engineering
standard k–ε and realizable k–ε turbulence models in curved and Industrial Aerodynamics, 119: 13–27.
and confluent channels. Environmental Fluid Mechanics, 19: Wang YQ, Jackson P, Sui J (2011). Simulation of flow around a
543–568. surfacemounted square-section cylinder of aspect ratio four. In
Shih T-H, Liou WW, Shabbir A, et al. (1995). A new k-ε eddy viscosity Proceedings of the 20th Annual Conference of the CFD Society
model for high Reynolds number turbulent flows. Computers & of Canada.
Fluids, 24: 227–238. Wang Y, Jackson PL, Sui J (2014). Simulation of turbulent flow
da Silva BL, Chakravarty R, Sumner D, et al. (2020). Aerodynamic around a surface-mounted finite square cylinder. Journal of
forces and three-dimensional flow structures in the mean wake Thermophysics and Heat Transfer, 28: 118–132.
of a surface-mounted finite-height square prism. International Wang Y (2019). Effects of Reynolds number on vortex structure
Journal of Heat and Fluid Flow, 83: 108569. behind a surface-mounted finite square cylinder with AR = 7.
Sumner D (2013). Flow above the free end of a surface-mounted Physics of Fluids, 31: 115103.
finite-height circular cylinder: a review. Journal of Fluids and Williamson CK (1996). Vortex dynamics in the cylinder wake.
Structures, 43: 41–63. Annual Review of Fluid Mechanics, 28: 477–539.
Sumner D, Rostamy N, Bergstrom DJ, et al. (2017). Influence of aspect Zhang D (2017). Comparison of various turbulence models for unsteady
ratio on the mean flow field of a surface-mounted finite-height flow around a finite circular cylinder at Re=20000. Journal of
square prism. International Journal of Heat and Fluid Flow, 65: Physics: Conference Series, 910: 012027.
1–20. Zhang D, Cheng L, An H, et al. (2017). Direct numerical simulation
Uffinger T, Ali I, Becker S (2013). Experimental and numerical of flow around a surface-mounted finite square cylinder at low
investigations of the flow around three different wall-mounted Reynolds numbers. Physics of Fluids, 29: 045101.