The 15th International Conference on Wind Engineering

Beijing, China; September 1-6, 2019

EFFECT OF CORNER GEOMETRY MODIFICATIONS ON

WIND LOADING OF STRUCTURES
Kush Dwivedi a, Manuj Kushwaha a, Sabareesh G Rajasekharan a
a
BITS-Pilani, Hyderabad Campus, Jawahar Nagar, Hyderabad, India
Email: sabareesh@hyderabad.bits-pilani.ac.in

ABSTRACT: Wind flow around a structure leads to steady and unsteady forces on buildings.
Measurement and control of these forces are necessary for determining their effect on the structure.
To understand the effect of corner geometry modification on the building and abatement of the
forces acting on the walls of the structure, rounded corners with different radii (0.05 ≤ r/D ≤ 0.25)
are introduced in place of sharp corners of the geometry. In this study, computations are performed
for flow around a square cylinder at Re =1x106 with different corner configurations. Modifications
to the corner and their effects on the mean and fluctuating components of lift and drag forces
relative to the sharp-edged case is studied. The numerical study uses a finite volume method and
OUICK scheme. An Unsteady Reynolds averaged (URANS) based solver is used to simulate the
flow. This approach is validated against existing experimental and numerical results for a square
cylinder with sharp as well as rounded corners and the numerical predictions show a good
agreement with the available results. It was observed that a significant reduction in values of both
mean and fluctuating force components can be achieved by these modifications to sharp corner
configuration. For structures with rounded corners, the root mean square value of lift coefficient
(Clrms) increases as the radius at the corner increases as well as the values of both (Cdmean and Cdrms)
decreases with the increase in radius.

KEYWORDS: Wind loading, Force coefficients, Bluff body aerodynamics, Corner modifications,
Computational fluid dynamics, Rounded corners

1 INTRODUCTION

Wind loading is a critical factor in the design of buildings and research facilities housing sensitive
instruments. The unsteady forces lead to wind-induced oscillations. The effect can be
considerable even for buildings that are shorter in height. It is more pronounced for buildings
housing sensitive instruments having a low frequency of excitation [1]. The wind velocity and
direction vary depending on the region and topography. Selection of proper location and
orientation plays a crucial part in the development of critical research facilities [2].
Modification of external geometry of the structure can lead to a significant reduction in the
unsteady aerodynamic force and the vibration response [3]. The modifications may include a
change in the cross-sectional area like slotted corners, corner-cut, chamfered corners, and tapering.
Each method is suited to a particular flow environment and may be advantageous or
disadvantageous depending on the flow profile. The wind-induced response is mitigated better in
the across wind direction than in the along wind direction for a tapered building. The
modifications to the corners are effective in reducing the drag and lift by altering the
characteristics of the boundary layer and assisting in its reattachment while also reducing the width
of the wake region. Kwok et al. [4] performed wind tunnel tests of buildings with slotted and
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

chamfered corners and found that both lead to a reduction in the peak response but the reduction
in the case of chamfered corners was more substantial in the along wind and across wind directions.
The aerodynamic forces are more pronounced for a bluff body leading to areas of negative pressure
around the perimeter of the body [5]. It is evident from above that the effect of corner modification
on the flow behavior near the cylinder results in a reduction of aerodynamic loads relative to the
sharp corner case. It is also observed that the earlier studies were performed at a lower Reynolds
number and doesn’t discuss the effects of increasing rounded corner radius. The aim of the present
study is to predict the aerodynamic loads on a square cylinder in turbulent flow at high Reynolds
number in the supercritical regime. The aim is to use a much simpler two equation turbulence
model and an effective boundary condition to accurately model the mean flow behavior around
the cylinder. The focus is on understanding the effects of rounding the corners as well as increasing
the rounded corner radius on force coefficients as well as mean flow behavior. The work is
concerned with suggesting an optimum range of rounded corner radius that results in a maximum
reduction in aerodynamic loads while affecting the geometry to the minimum.

2. COMPUTATIONAL DETAILS
2.1 Governing Equations
For solving the Navier Stokes equations an unsteady Reynolds averaged (URANS) approach is
used. The governing equations for the conservation of mass and momentum for an incompressible
fluid can be written as follows:
𝜕
(𝑈 ) = 0 (1)
𝜕𝑥𝑗 𝑗
𝜕𝑈𝑖 𝜕𝑈𝑖 𝜕 𝜕𝑈𝑖 1 𝜕𝑝
+ 𝑈𝑗 = (𝜈 − ̅̅̅̅̅)
𝑢𝑖̇ 𝑢𝑗̇ − (2)
𝜕𝑡 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜕𝑥𝑗 𝜌 𝜕𝑥𝑖
Where, 𝑈𝑖 is the mean component of velocity, 𝑢𝑖̇ is the fluctuating component of velocity, 𝜌 is
the density, 𝜈 is the kinematic viscosity and 𝑝 is the mean pressure. The Reynolds stresses (𝑢
̅̅̅̅̅)
𝑖̇ 𝑢𝑗̇
in equation 2 are appropriately modeled by employing Boussinesq hypothesis and thus relating
them to the mean velocity gradients as follows:
𝜕𝑈𝑖 𝜕𝑈𝑗 2
− ̅̅̅̅̅
𝑢𝑖̇ 𝑢𝑗̇ = 𝜈𝑡 ( + ) − 𝛿𝑖𝑗 𝑘 (3)
𝜕𝑥𝑗 𝜕𝑥𝑖 3
Where 𝜈𝑡 is the eddy viscosity and 𝑘 is the turbulence kinetic energy. The eddy viscosity (𝜈𝑡 )
is obtained by the 𝑘 − 𝜔 shear stress transport (SST) model given by Menter[6], which effectively
blends the formulation of 𝑘 − 𝜔 model in the near wall region with the 𝑘 − 𝜀 model in the far field
thus achieving robust and accurate formulation near the wall. The turbulence kinetic energy (𝑘)
and the specific dissipation rate (𝜔) are obtained from the following equations.
𝜕 𝜕 1 𝜕 𝜕𝑘
(𝑘) + (𝑘𝑈𝑖 ) = ( (Γ𝑘 ) + 𝐺̃𝑘 − 𝑌𝑘 ) (4)
𝜕𝑡 𝜕𝑥𝑖 𝜌 𝜕𝑥𝑗 𝜕𝑥𝑗
𝜕 𝜕 1 𝜕 𝜕𝜔
(𝜔) + (𝜔𝑈𝑖 ) = ( (Γ𝜔 ) + 𝐺̃𝜔 − 𝑌𝜔 + 𝐷𝜔 ) (5)
𝜕𝑡 𝜕𝑥𝑖 𝜌 𝜕𝑥𝑗 𝜕𝑥𝑗
Where 𝐺̃𝑘 and 𝐺̃𝜔 represents the generation terms for turbulence kinetic energy (𝑘) and specific
dissipation (𝜔), Γ𝑘 and Γ𝜔 represents the effective diffusivity for 𝑘 and 𝜔 respectively. 𝑌𝑘 and 𝑌𝜔
represents dissipation due to turbulence for 𝑘 and 𝜔 and 𝐷𝜔 represents the cross diffusion term.
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

For the present study simulations were performed using a commercial CFD code, FLUENT. It
uses a finite-volume discretization for the governing equations and the SIMPLE (semi implicit
pressure linked equation) pressure correction technique to solve the discretized equations. The
spatial discretization is done using QUICK scheme (Hayase et al. [7]) whereas for time marching
a second order implicit scheme has been adopted. The more detailed description for the K-omega
SST model may be found in the Fluent theory guide [8].

2.2 Computational Model

The computations are performed for a two-dimensional square cylinder with width D and various
corner radius (r) as shown in Figure 1, where 0.05 ≤ r/D ≤ 0.25 with a Δr/D of 0.05. An additional
case with r/D = 0.167 is also performed for validation with the earlier studies. To simplify the
calculation major dimensions are referenced to the cylinder width (D). The length of the
computational domain is taken to be 36D and the width is taken as 24D (Figure 2). The dimensions
were also used by Younis and Pruzlj [9] to yield solutions that are not affected by the blockage
effects. The center of the domain coincides with the center of the square cylinder. The blockage
ratio (Bf) defined as the ratio of cylinder width to domain width at the inlet for the computational
domain was 4.17%.

Figure 1 Schematic diagram representing square cylinder with sharp corner and rounded corner (r) configurations.

For the present simulations, a structured hexahedral mesh is used. The meshing of the model
was divided into 3 zones. The inner region which is 5D x 5D zone centered at the origin, the wake
region which is a 5D x 6D zone situated after the inner region in Figure 2 and the outer region. For
the inner zone, the grid expansion ratio is 1.01. The first cell height for the cells in contact with
the cylinder is taken to be 2.5 x 10-5 D to satisfy the requirements y+ ≤ 1 for the SST 𝑘 − 𝜔 model.
To accurately capture the details of the rounded corners and to keep the aspect ratio in a nominal
range, about 800 grid points were taken on the cylinder wall. For the rounded corner case (r/D =
0.167) the computational mesh that resulted in a grid independent solution was found to have 3.2
million grid points. To assess the accuracy of the numerical solution, the grid convergence index
(GCI) procedure as described by Celik et al. [10] has been performed on a rounded corner case
(r/D =0.167). By making use of the Richardson extrapolation method on three different grid sizes
(N1, N2, N3) with N1 being the finest grid, numerical discretization error is estimated. The ratio of
the number of grid cells in each refinement is taken as 1.35. It can be observed from Table 1 that
21
for fine grid (N1) the numerical uncertainty is under 0.3% of the 𝐺𝐶𝐼𝑓𝑖𝑛𝑒 value for all the three
variables (root-mean-square coefficients (𝐶𝑙𝑟𝑚𝑠 and 𝐶𝑑𝑟𝑚𝑠 ) and the mean drag coefficient
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

(𝐶𝑑𝑚𝑒𝑎𝑛 )) and hence N1 grid has been used. It should also be noted that for 𝐶𝑑𝑟𝑚𝑠 oscillatory
convergence has been observed.
To estimate the effect of time discretization and to quantify the sensitivity of the numerical
solution to the choice of time step size (Δt), computations were performed with three different time
step sizes corresponding to three different non-dimensional time step sizes (Δt*) on the rounded
corner case (r/D =0.167). It can be noticed from Table 2 that the values of lift and drag coefficients
(𝐶𝑙𝑟𝑚𝑠 , 𝐶𝑑𝑟𝑚𝑠 and 𝐶𝑑𝑚𝑒𝑎𝑛 ) show no direct correlation to the time step size. Thus in order to
accurately capture the boundary layer flow non-dimensional time step (Δt*) was set equal to 0.03.
The relaxation factors for all the variables were set to be 0.7 and the bi-conjugate gradient
stabilized method (Barrett et al. [11]) was used in order to improve the convergence of AMG
(Algebraic multigrid) solver and stabilization of solution in the initial state. The convergence
criterion for the iterative process was also set to get all the residuals below a value of 10−4.
Table 1. The GCI method assessment of discretization error
Variable/coefficients Φ = 𝐶𝑙𝑟𝑚𝑠 Φ = 𝐶𝑑𝑚𝑒𝑎𝑛 Φ= 𝐶𝑑𝑟𝑚𝑠

N 1 , N2 , N 3 3.2 x 106 , 2.37 x 106 , 1.75 x 106

φ1 0.807 0.88 0.116
φ2 0.805 0.875 0.115
φ3 0.754 0.874 0.120
r 1.35
p 10.79 5.36 5.36
21
Φ𝑒𝑥𝑡 0.80708 0.8812 0.11625
𝑒𝑎21 0.0025 0.0057 0.0086
21
𝐺𝐶𝐼𝑓𝑖𝑛𝑒 0.00013 0.0018 0.0027

Figure 2 Computational domain along with grid arrangement depicting the three different mesh refinement zones
(inner region, wake region and outer region) and boundary conditions at the inlet, outlet and symmetry.
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

Table 2. Variation of lift and drag coefficients with time-step size

Time step size (Δt) Nondimensional 𝐶𝑙𝑟𝑚𝑠 𝐶𝑑𝑚𝑒𝑎𝑛 𝐶𝑑𝑟𝑚𝑠 St
time step (Δt*)
0.001025 0.015 0.892 0.862 0.1195 0.269
0.00205 0.03 0.807 0.88 0.116 0.271
0.003075 0.045 0.983 0.856 0.1203 0.2145

2.3 Boundary Conditions

For both sharp corner case and rounded corner cases at inlet uniform stream-wise velocity
(𝑈 = 𝑈0 , 𝑉 = 0) is taken according to the Reynolds number (Re) of 1 x 106 . The turbulence
values (turbulence kinetic energy (𝑘 = 𝑘0 ) and specific dissipation ( 𝜔 = 𝜔0 )) at the inlet for
the external aerodynamic application with two-equation turbulence models as recommended by
Spalart and Rumsey[12] are used. At the inlet 𝑘0 /𝑈02 = 1 x 10−6 (or turbulent intensity = ̃ 0.1%)
and 𝜔0 𝑙/𝑈0 = 5, where 𝑙 = characteristic length (width of the cylinder (D)). These values yield a
viscosity ratio of 0.2 for the Reynolds number used. Also as suggested these values were used as
both inflow as well as floor value. At the outlet, the gradients of all the dependent variables in the
flow directions are taken to be zero according to equation 6. No slip condition representing a solid
boundary was used at the cylinder wall. At the side boundaries, symmetry conditions simulating a
frictionless wall were used as shown in equation 7.
𝜕𝛷
=0 (6)
𝜕𝑥

𝜕𝛷
𝑉 = 0, =0 (7)
𝜕𝑦

3. RESULTS AND DISCUSSION

Unsteady RANS simulations were performed to study the effect of corner modifications such as
rounded corners on mean and fluctuating components of the force coefficients as well as the effect
of changing the corner radius (r) on the behavior of the same coefficients. Prior to those results
comparison between present simulations at Re = 1 x 106 for sharp corner case as well as a rounded
corner case with (r/D= 0.167) with published studies are performed. For comparison, RMS and
mean drag (𝐶𝑑𝑟𝑚𝑠 and 𝐶𝑑𝑚𝑒𝑎𝑛 ) and RMS lift (𝐶𝑙𝑟𝑚𝑠 ) coefficients were computed. The Strouhal
(St) numbers were calculated from respective fluctuating lift data. The time- and spanwise-
averaged pressure coefficients are also used for the comparison. The time-averaged quantities were
calculated over 30 shredding cycles. The summary of the comparison is presented in table 3. It can
be observed that for the sharp corner case, mean force coefficients and St shows good match with
the numerical studies and experiments performed on a similar cylinder by Sohankar [13] and
Delany and Sorensen [14]. For the rounded corner case, the fluctuating component 𝐶𝑑𝑟𝑚𝑠 and St
shows good agreement with numerical studies done by Cao and Tamura [15]. The other force
coefficients for the case show similar trends as observed in other studies but the agreement in
values is not particularly close but it should be noted that the studies themselves show similar
degree of variation upon changing the parameters involved.
Figure 3(a) represents the time-averaged pressure contour for the sharp corner case. It can be
observed that the minimum value of the pressure coefficient (Cp) behind the cylinder lies at around
1D from the center of the cylinder with a value of (-2.1). The Cp value at the center of recirculating
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

regions on the sides of the cylinder is about (-2.25). These recirculating regions behind the cylinder
as well as on the sides were also observed by Sohankar [13] at Re =1 x 106 with similar values of
Cp (-1.94 and -2.15 respectively).
Table 3. Comparison of force coefficients and Strouhal number (St) with available results.
Case (Re=1x106 ) 𝐶𝑙𝑟𝑚𝑠 𝐶𝑑𝑚𝑒𝑎𝑛 𝐶𝑑𝑟𝑚𝑠 St
Sharp corner case 1.98 2.23 0.6 0.137
Sohankar [13] (sharp corner) 1.5 2.25 ~0.2 0.127
Delany and Sorensen [14] (r/b0 =0.021) - ~2 - -
Rounded corner case (r = 0.167D) 0.807 0.88 0.116 0.281
Cao and Tamura [15] (r = 0.167D) 0.41 0.58 0.11 0.253
Delany and Sorensen [14] (r/b0 =0.167) - ~0.56 - -

Figure 3 (a) Time averaged pressure coefficient (Cp) contours for sharp corner case, (b) Angular distribution of
pressure coefficient (Cp) for both sharp and rounded corner (r/D =0.167) case.

Figure 3(b) shows the angular distribution of time-averaged pressure coefficients for both sharp as
well as rounded corner case (r/D=0.167). As the domain is symmetric about midline the <Cp> – θ
curve for both the upper and lower halves of the cylinder are identical so only upper half is used
for comparison. It can be observed that the distribution of mean-pressure coefficient (<Cp>) has
two valleys at the frontal and leeward corners one at θmin1 = 50.6deg and the other at θmin2 =
126.56deg respectively for the rounded corner case and for the sharp corner case two points of
sudden drop or increment at the corners of the cylinder one being at θmin1 = 46deg and the other at
θmin2 = 133deg with little to no change in the values of <Cp> between them was observed. Similar
observations with respect to the circumferential distribution of time-averaged pressure were made
by Cao and Tamura [15] and the value of θmin1 and θmin2 are compared in Table 4. For the case
with rounded corners, there is an increment in the <Cp> values with respect to the corner case in
the side wall region. It can be observed that the introduction of rounded corners delays the flow
separation and cause the flow separation angle to move further downstream.
Table 4. Comparison of the Angular distribution of time-averaged pressure coefficient with previous studies.
Case (Re=1x106) θf θl θmin1 θmin2
Sharp corner case 46 136 46 133
Rounded corner case (r = 0.167D) 33.7-56.3 123.7-146.3 50.6 126.56
Cao and Tamura [15] (r = 0.167D) 33.7-56.3 123.7-146.3 51 126
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

These observations are similar to that predicted by Dai et al. [16] for flow past rounded corner
cylinder at high Reynolds number. Overall all these comparative studies with the available results
suggest that the current setup using URANS is effective in predicting the mean flow behavior as
well as relative behavior of fluctuating parameters to an acceptable accuracy.
On comparing the force coefficients of rounded corner case to that of the sharp corner case it
was observed that the rounded corners cause the mean and fluctuating components of lift and drag
coefficients to decrease (Figure 4(a)). It was also observed that the percentage reduction in the lift
coefficient is lesser than that of drag coefficients (Figure 4(b)). Figure 4(a) shows behavior of lift
and drag coefficients (Clrms, Cdmean and Cdrms) with an increase in rounded corner radius(r). It was
observed that the value of Clrms increases with an increase in radius but the percentage reduction
with successive increment goes down (Figure 4(b)). For Cdrms it was observed that there is not a
significant change in the values with an increase in radius and this is evident from a constant value
of percentage reduction in Figure 4(b). It was observed for Cdmean that its value progressively
decreases with an increase in radius and percentage reduction also increases. This is due to the fact
that the increase in radius causes flow separation angle to move downstream thus increasing the
base pressure at the back and reducing the drag due to lower suction pressure.

Figure 4 (a) Distribution of mean and root mean squared lift and drag coefficients (Clrms, Cdmean and Cdrms ) with
respect to increase in rounded corner radius (r/D), (b) Percentage reduction in mean and root mean squared lift and
drag coefficients (Clrms, Cdmean and Cdrms) with respect to sharp corner case.

The Strouhal number is obtained by using fluctuating lift coefficients and performing a fast
Fourier transformation. It can be observed from Figure 5(a) that the introduction of rounded
corners results in a decrease in strength of shed vortices as well as an increase in the value of
dominant frequency and thus Strouhal number. This can be due to the delay in flow separation
caused by the introduction of rounded corners and this delayed separation reduces the energy
content of shed vortices. With an increase in the value of corner radius, it was observed that there
is not much difference in the value of energy content. Figure 5(b) shows the angular distribution
of the mean pressure coefficient (<Cp>) around the cylinder and it can be observed that the peaks
at the upstream corner shifts downstream thus showing a delayed flow separation. Also the
magnitude at the peaks go down indicating a decrease in base pressure at the upstream corner.
After the first peak, <Cp> has a constant value on the side wall with an increase in negative
magnitude as the radius increases. The second peak exists in the upstream part of the leeward
rounded corner where the second separation of boundary layer occurs, it can be observed that with
an increase in radius the magnitude of the peak goes down and thus increasing the base pressure.
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

This results in lower values of suction pressures and hence the reduction in mean drag value as the
radius increases is observed in Figure 4(a). Based on these results it is possible to obtain an
optimum value of rounded corner radius for maximum reduction in aerodynamic loads, but for
that, the area that gets affected by these modifications should also be taken into considerations.
Thus a value of rounded corner radius (r/D) between 0.05 and 0.10 is suggested as the optimum
value as there is not much difference in the percentage reduction between all the different values
of rounded corner radius but the area affected by them significantly increases as r/D increases.

Figure 5 (a) Comparison of power spectrum and strouhal number for sharp corner and rounded corner case with
different radii (r/D), (b) Angular distribution of pressure coefficient (Cp) for different rounded corner modifications.

Figure 6 shows the field of normal Reynolds stress. In Figure 6(a), for the sharp corner case, it
can be observed that the flow separates from the upstream corner and doesn’t reattach to side faces.
Thus two large recirculation regions on the side of the cylinder appear as well as a pair of vortices
form behind the body. The velocity fluctuations near the side walls are even larger than the one
behind the body in near wake region this is due to the separated shear layers. In contrast for the
rounded corner cases, these fluctuations in velocity are eliminated by the attached boundary layers
and peaks are located behind the body in the near wake. For r/D =0.05 (Figure 6(b)) it is observed
that the flow separation happens at the upstream corner and it’s reattached to the side face and then
at the leeward corner the second separation occurs. It was observed that with an increase in the
radius of the rounded corner the flow separation gets delayed and the flow separates at lower
separation angles at the leeward rounded corner. It is clearly visible from 6(b) and 6(d) when the
radius of the rounded corner (r/D) goes from 0.05 to 0.15 the flows separation moves downstream
thus resulting in a narrower separated wake.

(a) Sharp corner case (b) r/D = 0.05 (c) r/D = 0.1
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

(d) r/D = 0.15 (e) r/D = 0.20 (f) r/D = 0.25

Figure 6 (a) Contours for span-averaged fluctuating stream-wise velocity for sharp corner case, (b) Contours for span-
averaged fluctuating stream-wise velocity for rounded corner case with r/D = 0.05, (c) r/D = 0.1, (d) r/D =0.15,
(e) r/D =0.2, (f) r/D = 0.25.

4. CONCLUSION

The work is concerned with understanding the effect of rounded corners and an increase in rounded
corner radius on the aerodynamic loads on a square cylinder in turbulent flow at high Reynolds
number in the supercritical regime. The simulations were performed using k-ω shear stress
transport (SST) model with effective inflow boundary conditions and floor values to avoid
contaminations due to high values of eddy viscosity. The benchmark flow past a square cylinder
with sharp corners as well as rounded corner case (r/D =0.167) is performed and used to validate
the implementation of the model. The results obtained are in accordance with data obtained in
other computations as well as experiments. Thereafter computation on flow past rounded corner
case with multiple values of r/D is performed. Comparison of force coefficients showed that
rounded corner results in a reduction of at least 50% in fluctuating lift coefficient as well as a
reduction of (55% and 75%) in mean and fluctuating drag coefficients respectively. It was also
observed that depending on the radius of the rounded corner a further reduction in these values can
be obtained thus predicting an optimum value of rounded corner radius. A rounded corner radius
(r/D) value of 0.05 to 0.10 is suggested as the optimum value. Overall the present study suggests
that the mean flow behavior, as well as effects of rounded corners, can be predicted by using an
URANS approach to acceptable accuracy provided turbulence modeling and boundary conditions
are chosen properly.

5. REFERENCES

[1] R. H. Ferahian and H. S. Ward, “Vibration environment in laboratory buildings,” National Research Council
Canada, 1970.
[2] P. Hanzlik, S. Diniz, A. Grazini, M. Grigoriu, and E. Simiu, Building Orientation and Wind Effects Estimation,
vol. 131. 2005.
[3] S. C., R. E. W. E., L. M. James, and S. A., “Steady and unsteady wind loading of buildings and structures,”
Philos. Trans. R. Soc. London. Ser. A, Math. Phys. Sci., vol. 269, no. 1199, pp. 353–383, May 1971.
[4] K. C. S. Kwok, P. A. Wilhelm, and B. G. Wilkie, “Effect of edge configuration on wind-induced response of
tall buildings,” Eng. Struct., vol. 10, no. 2, pp. 135–140, 1988.
[5] G. Solari, “Wind Loading of Structures : Framework , Phenomena , Tools and Codi fi cation,” vol. 12, no.
March, pp. 265–285, 2017.
[6] F. R. Menter, “Two-equation eddy-viscosity turbulence models for engineering applications,” AIAA J., vol.
32, no. 8, pp. 1598–1605, Aug. 1994.
[7] T. Hayase, J. A. C. Humphrey, and R. Greif, “A consistently formulated QUICK scheme for fast and stable
The 15th International Conference on Wind Engineering
Beijing, China; September 1-6, 2019

convergence using finite-volume iterative calculation procedures,” J. Comput. Phys., vol. 98, no. 1, pp. 108–
118, 1992.
[8] Fluent, “ANSYS Fluent Theory Guide,” ANSYS.Inc., vol. 15317, no. November, pp. 724–746, 2013.
[9] B. A. Younis and V. P. Przulj, “Computation of turbulent vortex shedding,” Comput. Mech., vol. 37, no. 5, p.
408, 2005.
[10] I. Celik, U. Ghia, P. J. Roache, C. Freitas, H. Coloman, and P. Raad, “Procedure for Estimation and Reporting
of Uncertainty Due to Discretization in CFD Applications,” J. Fluids Eng., vol. 130, no. 7, pp. 78001–78004,
Jul. 2008.
[11] R. Barrett et al., Templates for the Solution of Linear Systems: Building Blocks for Iterative Methods. 1994.
[12] P. R. Spalart and C. L. Rumsey, “Effective Inflow Conditions for Turbulence Models in Aerodynamic
Calculations,” AIAA J., vol. 45, no. 10, pp. 2544–2553, Oct. 2007.
[13] A. Sohankar, “Flow over a bluff body from moderate to high Reynolds numbers using large eddy simulation,”
Comput. Fluids, vol. 35, no. 10, pp. 1154–1168, 2006.
[14] B. N. K. Delany, N. E. Sorensen, and M. Field, “Low-speed drag of cylinders of various shapes,” Tech. Rep.,
vol. TN3038, no. NACA, 1953.
[15] Y. Cao and T. Tamura, “Supercritical flows past a square cylinder with rounded corners,” Phys. Fluids, vol.
29, no. 8, 2017.
[16] S. S. Dai, B. A. Younis, and H. Y. Zhang, “Prediction of Turbulent Flow Around a Square Cylinder With
Rounded Corners,” J. Offshore Mech. Arct. Eng., vol. 139, no. 3, pp. 031804-031804–9, 2017.