a r t i c l e i n f o a b s t r a c t
Article history: A square cylinder at zero angle of attack is known to exhibit a self-sustained response beyond the lock-in
Received 1 February 2018 region. This self-sustained response is characterized by a frequency that is lower than the vortex shedding
Accepted 1 May 2018
frequency and with large amplitudes. However, this self-sustained oscillation is not present for circular
Available online 10 May 2018
cylinders. Prior literature indicates that shape changes through altering the corner sharpness can help to
Keywords: reduce these self sustaining oscillations. In this work, shape changes are realized through altering the cur-
Vortex-induced vibration vature of the edge(side) of the square cylinder. For different edge curvature and rounded corner combina-
Self-sustained oscillations tions, characterization of these self-sustained oscillations is studied. Detailed flow analysis is performed
Boundary layer to study the effect of straight edge, curved edge, sharp corner and rounded corner on the pressure distri-
OpenFOAM bution. A proper combination of edge curvature and corner sharpness proves to be beneficial in reducing
self-sustaining oscillations.
© 2018 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.compfluid.2018.05.006
0045-7930/© 2018 Elsevier Ltd. All rights reserved.
274 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
Fig. 1. A schematic showing the definitions of (a) edge and (b) corner curves re-
spectively on a square cylinder.
1
CL = − pI + μ ∇ U + ( ∇ U )T · n · ny dS, (8)
1
2
ρU 2 D S
1
Grid convergence index for different grid levels.
CLp = − pI · n · ny dS, (9) Grid level Displacement amplitude GCI % GCI ratio
1
2
ρU 2 D S
1 1.36 2.68 5.76
1
2 1.17 1.81 3.79
CLμ = μ ∇ U + (∇ U ) .n . · ny dS.
T
(10) 3 1.05 0.81 1.005
1
2
ρU 2 D S 4 1.006 0.09
5 1.0 0 07
The mean lift CL, mean and drag CD, mean coefficients are computed
by time averaging CL and CD over ten cycles of oscillations. Vali-
dation and convergence studies are carried out in the next section
before advancing to the detailed analysis of the problem. on grid level k if the GCI ratios between grid levels k, k + 1 and
k − 1, k is close to unity. The GCI on grid levels k + 1 is defined as
3. Validation and convergence
Fs ||
In this section, verification and validation studies are carried
GCIk,k+1 = (11)
rp − 1
out to determine an appropriate grid size to be used for all VIV
cases. At first, grid convergence studies are carried out at Re = 200, where = (yk+1 − yk )/yk is the relative error in the peak displace-
mass ratio m∗ = 10, reduced velocity Ur = 40 and damping ratio ment, r is the refinement factor, pc is the observed order of con-
ζ = 0. A series of five unstructured grids with a refinement factor vergence and Fs a factor of safety(3 for two grids and 1.25 for more
of two is generated and is shown in Table 1. than two grids). Using the grid specifications listed out in Table 1,
The displacement of the cylinder and the force coefficients are numerical computations were carried out till a periodic displace-
computed for each grid resolution. The displacements are further ment is obtained. Fig. 3 shows the time history of the cylinder dis-
used in the computation of Grid Convergence Index(GCI) [11]. The placement for different grid levels. The Tiny mesh has the highest
GCI is a measure of the deviation of the computed value from the peak displacement and this reduces asymptotically when the mesh
asymptotic value([11]) when a series of grids are used on the same is refined. There exists a phase difference in the displacement sig-
problem. The value can be any arbitrary quantity of interest such nal for different meshes. This is due to the fact that the onset of
as lift, drag, or peak displacement of the square cylinder in the vortex shedding is different in each case and the mesh plays a role
present context. The GCI can be defined for different levels of grids, in triggering the vortex shedding instability. The GCI is computed
and solutions are said to be in the asymptotic range of convergence for each grid level and is shown in Table 2.
276 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
Fig. 4. (a)Temporal variation of transverse force and displacement (b) frequency spectra of transverse force at m∗ = 10, Ur = 40, Re = 200.
Fig. 5. Temporal variation of (a) pressure and viscous transverse force (b)transverse displacement and low frequency component of transverse force at m∗ = 10, Ur = 40,
Re = 200.
It is observed that the GCI on the level 4 mesh is very small their characteristics are compared with theory of Hartog [3] is ap-
(0.09%) and that the GCI ratios between levels 3,4 and 4,5 is 1.005 plied in detail on the cylinders by varying edge curvature.
indicating that the Level 4 mesh is appropriate at Re = 200. Thus, Fig. 4shows temporal variation of transverse displacement and
for all subsequent computations reported in this paper, level 4 force at m∗ = 10 and Ur = 40. The displacement response is clearly
mesh is used. The converged displacement at mass ratio m∗ = 10 dominated by frequency flow ≈ 0.022 which is lower than the shed-
and reduced velocity Ur = 40 using the present numerical model ding frequency fsh ≈ 0.165. The corresponding frequency spectra of
is within 5% with the data of Joly et al. [6]. With the confidence the transverse load is plotted in Fig. 4(b). The decomposition of
gained from convergence and comparison studies, the next section forces into pressure and viscous components is shown in Fig. 5(a).
is dedicated to the detailed analysis of the results. The transverse pressure and viscous force coefficients are repre-
sented as CLp and CLμ respectively. It is observed that the viscous
4. Results and discussions forces are lower in magnitude in comparison with the pressure
forces.
The edge curvature effects on the flow and force characteris- The transverse force is represented as a sum of harmonic series
tics are discussed in the present section to understand the self- CL (t ) = i CLi (t ) = i CLci sin(2π fi t ). The low frequency force CL1 (t)
sustained oscillations with zero damping (ζ = 0). The forces and whose coefficients CLc1 = 0.105, fi = 0.0225, φi = −0.5883 and dis-
R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284 277
Fig. 6. Effect of angle of attack α (a) on the mean transverse load CL, mean (b) mean inline load CD, mean and
dCL,mean
dα
: filled symbols denote coefficient of drag CD and unfilled
dCL,mean
symbols denote slope dα
. Re = 200.
Fig. 7. Temporal variation of transverse force CL (t) and displacement y(t) at m∗ = 10, Ur = 40, Re = 200, (a) LCC = 4 (b) LCC = 2.
placement are plotted in the Fig. 5(b). The lower frequency force is 4.1. Edge curvature and corner radius effects
out of phase with the displacement response which indicates the
inertia dominated region. The effect of angle of attack on the force characteristics with
The condition for self-sustained oscillations at low angles of at- respect to different curvatures is shown in Fig. 6. Three cases
tack reported in [3] is LCC = ∞, 4, 2 are considered for the study with angles of attack
range 0◦ − 10◦ . For LCC = ∞ which is a square cylinder, the lift co-
efficient increases with increase in angle of attack up to AOA = 5◦
dCL,mean and decreases with further increases in angle of attack. The rate
> CD,mean . (12)
dα of change of lift due to angle of attack L,mean dC
and drag coeffi-
dα
cient CD, mean with respect to angle of attack are shown in Fig. 6(a).
The Eq. (12) is never satisfied for circular cylinder as the variation With increase in curvature, it is seen that the slope of lift coef-
of lift with respect angle of attack is zero owing to axis-symmetry ficient with respect to angle of attack reduces. The drag coeffi-
cross-sectional plane. Thus the component of drag always acts in a cient remains nearly constant with respect to curvature and an-
dCL,mean
direction opposite to the velocity of the self-sustained oscillations gle of attack. At LCC = ∞ and LCC = 4, dα
> CD,mean whereas for
of the circular cylinder, suppressing any low frequency oscillations LCC = 2,
dCL,mean
< CD,mean . Thus the self-sustained oscillations exist
dα
[3]. The effect of edge curvature varying from LCC = ∞ (square) to for LCC = ∞ and 4, but do not exist for LCC = 2. The temporal trans-
LCC = 0 (circle) will be investigated in the following sub-sections in verse displacement and force of the self-sustained oscillations for
connection to self-sustained oscillations.
278 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
Fig. 8. Effect of edge curvature, rounded corner and angle of attack on (a) mean lift (b) mean drag.
Fig. 9. Instantaneous streamline contours at mean lift at α = 0◦ : (a) LCC = ∞, (b) LCC = 2, (c) LCC = ∞, f w = 0.1, (d)LCC = 2, f w = 0.1.
R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284 279
Fig. 10. Instantaneous streamline contours at mean lift at α = −6◦ : (a) LCC = ∞, (b) LCC = 2, (c) LCC = ∞, f w = 0.1, (d)LCC = 2, f w = 0.1.
dC
LCC = 4 is shown in Fig. 7(a). As the slope dαL for LCC = 4 is lower
than that of LCC = ∞, the amplitude of the self-sustained oscilla-
tions at LCC = 4 is lower than that of LCC = ∞. The low frequency
self-sustained oscillations found for LCC = ∞, 2 are not observed for
LCC = 2 as shown in Fig. 7(b). As the reduced velocity is in the in-
ertial range, the displacements are also very low. The reason for
dC
decrease in L,mean
dα
with increase in edge curvature will be investi-
gated using the flow contours in the following paragraph.
The present study is further augmented by rounding the sharp
corners along with the edge curvature effects. The parameter fw
representing the rounded corners is defined based on the distance
from the start of corner curve to the sharp corner of the square
as shown in Fig. 1(b). Two more cases are added for the study Fig. 11. Schematic of the bluff body with defined points. Point A: θ = 5π /4, Point
by rounding the sharp corners of the square cylinder LCC = ∞ and A1: θ = 5π /4 − 3π /180, Point A2: θ = 5π /4 − π /12, Point B2: θ = 3π /4 + π /12,
for the case with LCC = 2, the corner width fw is chosen to be 0.1. Point B1: θ = 3π /4 + 3π /180, Point B: θ = 3π /4, Point C: θ = π /4, Point C1: θ =
Fig. 8(a) shows the lift coefficient variation with respect to an- π /4 − π /12, Point D1: θ = −π /4 + π /12, Point D: θ = −π /4.
gle of attack with rounded corner effects. The lift coefficient for
both LCC = ∞, 2 with rounded corners is lower than its counter-
parts without rounded corner as shown in Fig. 8(a). The drag co- 4.2. Pressure contours and flow analysis
efficients are plotted in Fig. 8(b). The mean lift and mean drag co-
efficient are least for the case with edge curvature and rounded In this section detailed analysis of the flow is done to under-
corner. stand the effect of edge curvature and corner radius on the origin
of pressure force. Fig. 9 shows the streamlines of a bluff body with
280 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
Fig. 12. Comparison of coefficient of pressure distribution over the surface for dif- Fig. 14. Comparison of coefficient of pressure distribution over the surface for dif-
ferent bluff bodies at mean lift and angle of attack α = −6◦ for region R1. ferent bluff bodies at mean lift and angle of attack α = −6◦ for region R4.
LCC = ∞, LCC = 2, LCC = ∞, fw = 0.1 and LCC = 2, fw = 0.1 at mean The flow is also separated for LCC = 2 and LCC = ∞, fw = 0.1 on the
lift and angle of attack α = 0◦ . The flow separates at the sharp upper upstream corner as shown in Fig. 10(b) and (c).
corners at the leading side of the square and a circulation blob is At zero angle of attack, the bluff bodies are symmetric with re-
attached to the sides of the square cylinder as shown in Fig. 9(a) spect to the flow direction which results zero mean lift. As the an-
for α = 0◦ . The flow separation is absent for the bluff bodies with gle of attack increases the magnitude of mean lift increases and
edge curvature, corner radius as shown in Fig. 9(b)–(d). The circu- reaches a maximum around angle of attack |α| = 6◦ for all the
lation blob on the top and behind of the bluff body for LCC = ∞ is cases as shown in Fig. 8. The pressure distribution and its rela-
larger than the remaining cases. tion to the flow contours at α = −6◦ is analysed in the rest of this
The streamline detailed analysis of the flow is done to under- section. The direction of mean lift at this angle of attack is nega-
stand the effect of edge curvature and corner radius on the origin tive. In-order to understand the effect of sharp corners, edge cur-
of pressure force. Fig. 9 shows the streamlines of a bluff body with vature and the corner radius, the bluff body is divided into four
LCC = ∞, LCC = 2, LCC = ∞, fw = 0.1 and LCC = 2, fw = 0.1 at mean parts as shown in Fig. 11. The upstream upper corner is denoted
lift and angle of attack α = 0◦ . The flow separates at the sharp by A, upstream down corner by B, downstream upper corner by C
corners at the leading side of the square and a circulation blob is and downstream down corner by D whose location at zero angle of
attached to the sides of the square cylinder as shown in Fig. 9(a) attack are C, θ = π /4, B, θ = 4π /4, A, θ = 5π /4, D, θ = 7π /4. The
for α = 0◦ . The flow separation is absent for the bluff bodies with upstream edge region R1 is defined as the projected surface of the
edge curvature, corner radius as shown in Fig. 9(b)–(d). The circu- line between point A and B. Similarly the upper edge region R2,
lation blob on the top and behind of the bluff body for LCC = ∞ is lower edge region R3 and downstream edge region R4 are defined
larger than the remaining cases. as projected surface of the line BC, AD and CD respectively. The
The streamline over the bluff bodies for the cases considered at flow physics is analysed in detail based on these regions.
angle of attack α = −6◦ , is shown in Fig. 10. For LCC = ∞ as shown Table 3 shows the contribution of pressure forces to the mean
in Fig. 10(a), the separation region is extended from the upstream lift at angle of attack α = −6◦ for all the bluff bodies. The contribu-
upper corner to the downstream upper corner dragging the circu- tion to the lift from upstream edge region R1 to the overall mean
lation blob into it. Thus the circulation blob at the downstream up- lift is large for LCC = ∞ followed by LCC = 2 and LCC = ∞, fw = 0.1
per corner for LCC = ∞ , α = 0◦ has moved towards the upstream and LCC = 2, fw = 0.1. The pressure coefficient distribution on the
upper corner along the upper edge for α = −6◦ . The separation re- region R1 for all the bluff bodies considered at angle of attack
gion on the lower edge is reduced for α = −6◦ compared to α = 0◦ . α = −6◦ is shown in Fig. 12. The differences in the pressure distri-
For the remaining cases other than LCC = ∞, the circulating blob is butions of region R1 among the four bluff bodies is related to the
present at the downstream upper corner as seen in Fig. 10(b)–(d). boundary layers at points A2, B2, A1, B1 which are shown in the
Fig. 13. Comparison of coefficient of pressure distribution over the surface for different bluff bodies at mean lift and angle of attack α = −6◦ for regions: (a) R2, (c) R3.
R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284 281
Fig. 15. Comparison of non-dimensional boundary layer profiles at angle of attack α = −6◦ at (a)A, (b) B, (c) A1, (d) B1, (e) A2, (f) B2; Refer Fig. 11 for definition of points.
Fig. 11. At point A2, the pressure for LCC = ∞ is larger followed by point A and is on the corner curve for bluff bodies LCC = ∞, fw 0.1
LCC = ∞, fw = 0.1 and LCC = 2 and LCC = 2, fw = 0.1. The boundary and LCC = 2, fw 0.1. The flow is accelerated due to the corner curve
layer at this point A2 is plotted in the Fig. 15(e). The lower slope which is reflected in the boundary layer as shown in Fig. 15(c) re-
of the boundary layer indicates more acceleration of the fluid. The sulting in lower pressures at A1 for bluff bodies with corner radius.
slope of boundary layer at point A2 for curved edges are lower A similar explanation holds good for the points B2 and B1. The cor-
than that of straight edges. Curvature of the bluff body allows the responding boundary layers are shown in Fig. 15(e) and (d). The
flow to accelerate resulting in lower pressure compared to that of contribution to the overall mean lift from region R1 for LCC = ∞ is
a straight edge. large compared to other cases due to higher pressure acting on it.
The next point where the distinct difference in the pressure dis- The effect of a rounded corner in the decrement of pressure with
tribution is observed at point A1. The point A1 is near to the corner respect to square cylinder is greater than that of bluff body with
282 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
Grid R1 R2 R3 R4
LCC =∞ −0.064961 1.0543 −1.0970 −0.086396 Fig. 17. Effect of reduced velocity on the amplitude of vibration for various bluff
LCC =2 −0.037991 1.0574 −0.99572 −0.077585 bodies at Re = 200.
LCC = ∞, f w = 2 −0.029838 0.98856 −0.92837 −0.075458
LCC = 2, f w = 2 −0.014079 0.99572 −0.93041 −0.070317
point C for LCC = ∞ makes the fluid to accelerate more compared
to the other bluff body cases, which reflects in low pressure for
straight edge for region R1. Thus the contribution of lift from the LCC = ∞ at point C1 as shown in Fig. 14. Thus the contribution to
R1 of LCC = 2 is greater than LCC = ∞, fw = 0.1. With corner radius the overall lift from region R4 is largest for LCC = ∞ and lowest
on the bluff body, the pressure reduction due to the edge curvature LCC = 2, fw = 0.1 as seen in Table 3.
near the stagnation point A2 is larger than the pressure reduction The curvature near the stagnation point and upstream round
near the corner A1. Thus the contribution of lift from the region R1 corners allows the flow to accelerate resulting in lower pressures
for LCC = 2, fw = 0.1 is the lowest of all the cases considered. at those regions, which act against the overall mean lift. There is a
In upper edge region R2 the edge curvature has very small in- mismatch in the flow acceleration at the upstream sharp corners
fluence on the pressure force contribution to the lift as seen in the with straight edge resulting in large pressure differences at the
Table 3. The suction pressure observed near point B for sharp cor- upper and lower upstream corners. The edge curvature and sharp
ners is large compared to the rounded corner in the region R2 as corner reduces this mismatch in flow acceleration at the upstream
shown in Fig. 13(a). Fig. 15(b) shows the boundary layer variation corners, resulting in approximately the same pressure at the up-
at point B for the bluff bodies considered. For sharp corners, the stream corners. This minimizes lift contribution from upper and
variation of tangential velocity is plotted along the radial direction. lower edges to overall instantaneous lift for bluff body with curved
At the sharp corners, the flow acceleration is large compared to the edges and rounded corners. The flow separates at the upstream
rounded corners leading large suction pressures. Thus the magni- sharp corners and a large circulation region is formed on the up-
tude of lift force from region R2 is large for bluff body with sharp per straight edge. The circulation region due to separation on the
edges compared to the one with round corners. The direction of upper edge drags the circulation blob which was near the down-
lift from region R2 is opposite to the overall mean lift direction. stream corner at α = 0◦ into it, allowing the flow to accelerate on
The lift from lower edge R3 is larger in magnitude and oppo- the downstream edge. The accelerated flow results in large suc-
site in sign compared to R2 for the bluff body LCC = ∞. This is tion pressure on downstream straight edge with sharp corner con-
majorly due to the large suction pressure force near the point A tributing to lift. For bluff body with edge curvature and rounded
compared to point B as shown in Fig. 13(a) and (b). Thus this dif- corner, the separated circulation region due to the upstream cor-
ference contributes to the overall lift for the bluff body LCC = ∞. ner is absent. The circulation blob near to the downstream upper
For all the remaining cases, the lift from R3 is lower in magnitude corner does not allow the flow to accelerate on the downstream
and opposite in direction compared to region R2. The acceleration edge. Thus the downstream curved edge and rounded corner con-
of the fluid near the corner point A is lower due to edge curvature tributes to lift which is lower in magnitude than that of a bluff
and the corner radius compared to a sharp corner as shown in the body with sharp corners and straight edge. It is seen that adding a
Fig. 15(a). curvature to the edge and rounding the corner are both effective in
The circulation blob which was near point C for α = 0◦ is forced decreasing the overall mean lift which is then responsible in sup-
upstream by the separated circulation region on R2 for the bluff pressing the self-sustained oscillations. The next section is dedi-
body LCC = ∞ at α = −6◦ as shown in the Fig. 10(a). Where as for cated to vortex-induced vibrations of the bluff bodies with edge
the remaining cases the circulation blob stays near to the point C curvature and rounded corners.
as shown in the Fig. 10(b)–(d). This is basically due to the absence
of strong circulation region due to separation at point B for these 5. Vortex-induced vibrations
cases. Large accelerations of the fluid at point C1 is found for LCC =
∞ which is followed by LCC = ∞, fw = 0.1 and LCC = 2 and LCC = Fig. 17 shows the
√ effect of reduced velocity on amplitude re-
2, fw = 0.1 as shown in Fig. 16. The presence of circulation region sponse Aymax,rms = 2Ay,rms for various bluff bodies, where Ay, rmsh
at point C acts as a hindrance for the fluid to convect along the is the root mean square transverse amplitude. The lock-in region
downstream region R4. Thus the absence of circulation region at for the circular cylinder is in the range 5 ≤ Ur ≤ 8, where the dis-
R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284 283
Fig. 18. Effect of reduced velocity at Re = 200 on the (a) frequency ratio f/fN , (b) transverse force CL, rms .
placements are large. The lock-in region for square cylinder LCC = corresponding to shape changes, LCC = ∞, 4, 2. As the edge cur-
∞ at α = 0 is narrower than that of a circular cylinder which was vature increases the slope of the lift curve with respect to an-
also observed by Sen and Mittal [12]. The amplitude of the square gle of attack decreases. For LCC = ∞ and LCC = 4 the condition
cylinder in the lock-in region is far smaller than the circular cylin- dCL,mean
dα
− CD,mean > 0 holds true, leading to a self-sustained re-
der. Beyond the lock-in region the amplitude response increases dC
sponse. For LCC = 4, the value of L,mean
dα
− CD,mean is lower than
and is larger than that of a circular cylinder. The lock-in regime is
that of LCC = ∞ leading to a smaller amplitude self-sustained re-
narrower for the cylinder LCC = 2 similar to that of square cylinder. dC
The amplitude response in the lock-in regime lies between circular sponse. For LCC = 2, the condition L,mean
dα
− CD,mean > 0 does not
and square cylinders. Beyond the lock-in region, the amplitude re- hold true, thereby suppressing the self-sustained response. When
sponse is lower than that of circular cylinder. As discussed in the the corners are rounded, similar effects are observed as that of
previous sections, the self-sustained response is being suppressed altering the edge curvature. A combination of rounding the cor-
due to edge curvature. For the remaining cases of cylinder with ners and altering the edge curvature is more effective in suppress-
LCC = ∞, fw = 0.1 and LCC = 2, fw = 0.1, a similar amplitude varia- ing the self-sustained oscillations. A detailed study of the effect of
tion with respect to reduced velocity is observed as that of LCC = 2. edge curvature and rounded corner on the lift is reported by com-
The lock-in region for circular cylinder is wider compared to the paring the flow physics at angle of attack corresponding to maxi-
remaining cases with respect to frequency ratio response as seen mum mean lift magnitude.
in the Fig. 18(a). The dominant response of the square at larger For VIV cases, it is seen that the lock-in region of the bluff body
reduced velocities Ur > 10 is a self-sustained frequency which is with altered edge curvature and with a rounded corner is similar
lower than the shedding frequency. The frequency ratio is approx- to that of square cylinder whose lock-in region is narrower com-
imately same for cylinder with LCC = 2 and LCC = ∞, fw = 0.1 and pared to that of a circular cylinder. As for the amplitudes of oscil-
LCC = 2, fw 0.1 and greater than that of circular cylinder. lations in the lock-in region, they are lower for the square cylinder
Below frequency ratio of one, the root mean square of lift force with altered edge curvature and with rounded corners than that of
coefficient CL, rms is increasing with increase in reduced velocity for a circular cylinder.
all the cases as shown in Fig. 18(b). The transverse load suddenly
drops down when the frequency ratio equals one for Ur > 5 for all
References
cases. The circular cylinder has the highest transverse load com-
pared to other cases. [1] Carassale L, Freda A, Marrè-Brunenghi M. Effects of free-stream turbulence and
The presence of edge curvature and rounded corners on a corner shape on the galloping instability of square cylinders. J Wind Eng Ind
Aerodyn 2013;123:274–80.
square cylinder result in lower lift forces. This causes the drag to
[2] Carassale L, Freda A, Marrè-Brunenghi M. Experimental investigation on the
act against the velocity suppressing the self-sustained oscillations. aerodynamic behavior of square cylinders with rounded corners. J Fluids Struct
It is also observed that the amplitude response of the bluff-body 2014;44:195–204.
with either edge curvature or rounded corners or both is lower [3] Hartog JPD. Transmission line vibration due to sleet. Trans Am Inst Electr Eng
1932;51(4):1074–6. doi:10.1109/T-AIEE.1932.5056223.
than the circular cylinder in the lock-in region. [4] Jaiman RK, Sen S, Gurugubelli PS. A fully implicit combined field scheme for
freely vibrating square cylinders with sharp and rounded corners. Comput Flu-
ids 2015;112:1–18.
6. Conclusions
[5] Jasak H, Jemcov A, Tukovic Z, et al. Openfoam: a c++ library for complex
physics simulations. In: International workshop on coupled methods in numer-
The effect of changing the edge curvature and corner round- ical dynamics, 10 0 0. Dubrovnik, Croatia: IUC; 2007. p. 1–20.
ness of a bluff body at zero angle of attack on vortex-induced vi- [6] Joly A, Etienne S, Pelletier D. Galloping of square cylinders in cross-flow at low
reynolds numbers. J Fluids Struct 2012;28:232–43.
brations is studied. Numerical simulations have been conducted [7] Leontini JS, Thompson M, Hourigan K. Three-dimensional transition in the
to verify the theory of Hartog [3] with respect to different cases wake of a transversely oscillating cylinder. J Fluid Mech 2007;577:79–104.
284 R. C. Mysa, D. D.J. Chandar / Computers and Fluids 170 (2018) 273–284
[8] Leontini JS, Thompson MC. Vortex-induced vibrations of a diamond cross-sec- [12] Sen S, Mittal S. Free vibration of a square cylinder at low Reynolds numbers. J
tion: sensitivity to corner sharpness. J Fluids Struct 2013;39:371–90. Fluids Struct 2011;27:875–84.
[9] Mandal A, Faruk G. An experimental investigation of static pressure distribu- [13] Williamson HK C, Govardhan R. Vortex-induced vibrations. Annu Rev Fluid
tions on a group of square or rectangular cylinders with rounded corners. J Mech 2004;36:413–55.
Mech Eng 2010;41(1):42–9. [14] Zhao J, Leontini JS, Jacono DL, Sheridan J. Fluid-structure interaction of a square
[10] Nemes A, Zhao J, Jacono DL, Sheridan J. The interaction between flow-induced cylinder at different angles of attack. J Fluid Mech 2014;747:688–721.
vibration mechanisms of a square cylinder with varying angles of attack. J [15] Zhao M, Cheng L, Zhou T. Numerical simulation of vortex-induced vibration of
Fluid Mech 2012;710:102–30. a square cylinder at a low Reynolds number. Phys Fluids 2013;25:1–25.
[11] Roache PJ. Verification and validation in computational science and engineer-
ing, 895. Albuquerque, NM: Hermosa; 1998.