See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267554310

Numerical Simulation of Vortex Shedding from an Inclined Flat Plate

Article  in  Engineering Applications of Computational Fluid Mechanics · November 2014

DOI: 10.1080/19942060.2010.11015342

CITATIONS READS

35 1,119

2 authors, including:

Kit Ming Lam

The Hong Kong University of Science and Technology
142 PUBLICATIONS   1,857 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Investigation of heavy particles behavior in turbulence View project

Interference mechanism of two tall buildings View project

All content following this page was uploaded by Kit Ming Lam on 12 July 2017.

The user has requested enhancement of the downloaded file.

 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4, pp. 569–579 (2010)

NUMERICAL SIMULATION OF VORTEX SHEDDING FROM

AN INCLINED FLAT PLATE
K. M. Lam* and C. T. Wei

Department of Civil Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China
* E-Mail: [email protected] (Corresponding Author)

ABSTRACT: Vortex shedding flow from a flat plate inclined to a uniform flow at an angle of attack between 20o
and 45o is simulated with a finite volume CFD code with RNG k- turbulence model. The unsteady flow simulation
at Re=2104 with RANS shows two trains of vortices shed from the two different edges of the plate forming a vortex
street in the wake of the plate. The computed results provide support to previous experimental observations that in
this asymmetric flow geometry, the two trains of vortices in the vortex street possess different vortex strengths.
There is further evidence that the vortex from the plate leading edge is actually shed from a location near the trailing
end of the plate. The computed flow at successive phases of a vortex shedding cycle show different development and
shedding mechanisms for the two trains of vortices. The study also explores the generation mechanism of the
fluctuating lift and drag on the plate and its relationship with the vortex shedding processes.
Keywords: CFD, vortices, inclined flat plate, circular cylinder

1. INTRODUCTION relatively small angle of attack, down to =30o

(e.g., Fage and Johansen, 1927; Perry and Steiner,
Vortex shedding flow from a bluff body has 1987; Knisely, 1990). In the wake of these
attracted many experimental and numerical inclined plates, a vortex street was observed with
investigations due to its rich fluid dynamical evidence of vortex shedding alternatively from
phenomena and important engineering the two edges of the plates. It was also found that
applications such as unsteady fluid loading of the vortex shedding frequency, f, scales with the
structures, flow-induced vibration and flow noises. projected width B’ of the plate normal to the free-
Most of the studies investigate flow over a two- stream. The Strouhal number is approximately
dimensional bluff body such as a circular cylinder, constant at St’=fB’/U0.15 for =30o to 90o, U
a square cylinder and a flat plate normal to the being the free-stream velocity. The issue of any
flow. From these bodies with a symmetric asymmetry of the vortex street was not targeted in
geometry, periodic vortices are shed alternatively these previous studies, although Knisely (1990)
in the form of two trains of opposite-sign but reported that as  becomes smaller than 30o, St’
equal-strength vortices. The mean flow is increases sharply and the wake becomes
symmetric about the wake centerline with zero dominated by vortices from the trailing edge of
mean lift and the fluctuating lift oscillates at the the plate.
vortex shedding frequency. Flow over a bluff It is the aim of the first author to study in details
body with an asymmetrical geometry generates any asymmetry in the dynamics of the two trains
lift and examples include a flat plate or an aerofoil of vortices shed from the different edges of the
inclined to the flow, an asymmetric aerofoil at inclined plate. Lam (1996) started to measure
zero angle of attack, or a rotating circular cylinder. some details of the vortex street behind a flat
The first author has been investigating how the plate inclined at =30o. Lam and Leung (2005)
periodic vortex shedding and the vortex street are further studied the vortex dynamics in the wake of
affected by the degree of wake asymmetry (e.g., an inclined flat plate at  between 20o and 30o.
Lam, 1996 and 2009; Lam and Leung, 2005). Phase-locked vortex patterns were obtained with
For the present problem of flow past an inclined velocity measurement by particle-image
flat plate, the mean flow is asymmetric about the velocimetry (PIV). The vorticity contours showed
wake centerline and there exists a non-zero mean that at same axial distances, the train of vortices
lift on the plate. A number of past experimental from the trailing edge of the plate possess higher
investigations studied the mean wake and vortex magnitudes of peak vorticity levels than the train
shedding from flat plates at normal incidence, that of leading edge vortices. This asymmetry in the
is =90o, to plates inclined to the flow at a vortex street behind a plate at high incidence is

Received: 21 Jan 2010; Revised: 27 Apr. 2010; Accepted: 29 Jun. 2010

569
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

not very significant and has been shown so far in RANS but the main emphasis was on the drag on
experiments by the first author. To seek additional the plate. The most notably CFD study of flow past
evidence to the experimental observation of a an inclined flat plate is probably that of Breuer et al.
street of alternating vortices of unequal strengths, (2003). That study chose the inclined flat plate at
the present authors attempt a computational fluid =18o as a representative of high-lift aerodynamic
dynamics (CFD) study of flow over an inclined flows with a massive separation region. Thus, the
flat plate at a number of  between 20o and 45o. focus was on the large-scale separation mainly from
The focus is on the dynamics of vortices shed the leading edge of the plate and the prediction
from the two edges of the plate, any asymmetry in performance by RANS, LES and detached eddy
the vortex street and the relationship with lift simulation. There was little information on the
production. This paper is to report the CFD dynamics of the shed vortices in the downstream
results, a part of which have been included in a wake. As stated earlier, the main objective of the
conference paper (Lam and Wei, 2006) which is present RANS-based CFD study is to confirm
not widely available. whether there exist differences in the dynamics of
There have been numerical computation studies the two trains of vortices in the wake of the plate.
on vortex shedding from bluff bodies and this It is not intended to discuss the accuracy of the
problem is often used as a benchmark test for CFD approaches as in Breuer el al. (2003).
different CFD approaches. Traditionally, the As a closing note to this introduction, the
Reynolds-averaged Navier-Stokes (RANS) asymmetry in the vortex street behind an inclined
equations are the least computational demanding plate is illustrated by the picture of oil spill in the
approach to the simulation of engineering turbulent recent incident of a stranded ship in the Great
flow. Vortex shedding from circular and square Barrier Reef shown in Fig. A of the Appendix.
cylinders has been modelled with some success
with unsteady RANS solutions employing the 2. COMPUTATIONAL SCHEME
standard or modified k- turbulence model (e.g.,
Bosch and Rodi, 1998; Iaccarino, et al., 2003; Shao Flow computation was carried out for a number of
and Zhang, 2006). With the advance of flow cases including flow over an inclined flat
computational resources, the more powerful plate at four different angles of attack, =20o, 25o,
approaches of large-eddy simulation (LES) and 30o and 45o. The main flow parameters are
direct numerical simulation (DNS) are increasingly summarized in Table 1. Prior to the inclined plate
used in the predictions of engineering flows (e.g., flow, the validity of the CFD approach is studied
Breuer, 1998; Rodi, 2006). Compared to the square by modeling vortex shedding from a circular
and circular cylinders, there have been few CFD cylinder at different Reynolds numbers (Re) and
studies on flow over a flat plate. Lasher (2001) the computational cases are shown in Table 2. All
computed the flow over a normal flat plate with the flow cases in Tables 1 and 2 are two-

Table 1 Computed global wake characteristics of flow past an inclined flat plate at Re=2104 for two-dimensional
turbulent flow cases with 62103 computational cells.

 fB/U St FL CL CL’ CD CD’

2 U  B
1 2

20o 0.49 0.169 1.11 3.24 0.08 1.20 0.03

25o 0.39 0.165 1.31 3.11 0.20 1.47 0.09
30o 0.35 0.174 1.51 3.02 0.26 1.76 0.15
45o 0.23 0.166 1.75 2.47 0.32 2.48 0.30

Table 2 Computation of flow past a circular cylinder on effect of grid size and Re for two-dimensional flow cases.

Re Computation Grid fineness St CD CL’

(total no. of cells)
103 2D, laminar Coarser: 11103 0.23 1.35 0.81
103 2D, laminar Finer: 97103 0.23 1.65 1.06
103 2D, laminar Standard: 24103 0.23 1.64 1.05
104 2D, turbulent Standard: 24103 0.29 0.43 0.062
105 2D, turbulent Standard: 24103 0.28 0.29 0.072

570
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

dimensional. shedding period. Truncation-induced flow

This study used the finite volume code FLUENT asymmetry in the computed flow is sufficient to
(Fluent, Inc., 2003) to solve the incompressible trigger initial instability and subsequent vortex
continuity and momentum equations in two shedding in the solution and the solutions became
dimensions (Patankar, 1980). For the turbulent periodic after a few vortex shedding periods.
flow cases, the equations were Reynolds-averaged Convergence of the unsteady solutions was
and the k- model was used for turbulence closure. normally achieved after about 3,000 iterations.
This study adopted the Re-Normalization Group Grid dependence was studied through
(RNG) extension of k- model (Yakhot and computation of flow over a circular cylinder at
Orszag, 1986). The RNG k- model has been Re=1,000 using three sets of grids (Table 2). The
shown by some workers to yield improvement Reynolds number is Re=UD/, where  is the
over the standard k- model for recirculating kinematic viscosity of air. In each of the CFD
flows and flows with strongly curving streamlines results, vortex shedding was modeled. The key
(e.g., O’Shea and Fletcher, 1994; Pagageorgak information of the flow, including the mean drag
and Assanis, 1999, Ferreira et al., 2002) and is coefficient, CD and the root-mean-square (rms)
often used in CFD for wind engineering problems lift coefficient, CL’ on the cylinder and the vortex
including the authors’ team (Stathopoulos, 2006; shedding frequency, in the form of Strouhal
Lam and To, 2006). Standard values were used number, St=fD/U, were calculated and used to
for model constants: C=0.0845, C1=1.42 and test grid independence. As shown in Table 2, the
C2=1.68. The flow equations and closure coarser grid with 11,000 meshes was not
equations were solved to obtain solutions of the sufficiently fine for grid independence. The finer
six flow variables, namely, pressure, three velocity grid with 97,000 meshes produced essentially the
components, k and . The solution scheme made same solutions of St, CD and CL’ as the standard
use of the SIMPLEC algorithm (Van Doormaal and grid of 24,000 meshes. Thus, the standard grid
Raithby, 1984) for pressure-velocity coupling and setting was judged to produce grid-independent
the QUICK scheme (Leonard, 1979) for convective solutions.
transport modelling. To study vortex shedding, In the standard grid setting for the circular
unsteady solutions of the equations were sought. cylinder, the cylinder circumference was divided
It can always be criticised that compared with the into 120 grid points, as compared to 90 or 240,
more advanced simulation by LES and DNS, the respectively, in the coarser or finer grid. The inlet
unsteady RANS method cannot accurately or outlet of the computational domain was
reproduce separated flow and vortex shedding. divided into 120 meshes and the length into 160
However, the main focus of the present study is a meshes. For flow over the inclined plate, an even
relative comparison between the features and denser computational mesh was used. There were
strengths of the two trains of vortices. With 60 grid points on one face of the plate. The two-
quantitative data of the flow of secondary concern, dimensional computation meshes were shown in
the more economical RANS computation was Fig. 1. The numbers of grids on the end and side
acceptable to meet the objective of the study. faces of the computational domain were 160 and
The two-dimensional flow was modelled with a 360, respectively and the total number of meshes
rectangular computational domain inside which a was 62,000.
circular cylinder of diameter D=10 cm or an
inclined flat plate of breadth B=15 cm was placed.
The width of the domain was 8D for the cylinder
flow and 5B for the inclined plate flow (Fig. 1).
The length of the domain was from 8D to 16D
or 6B to 14B, respectively. Uniform smooth flow
of air at velocity U entered the inlet of the
computational domain and all flow was made to
exit through the downstream end of domain. The
sides of the domains were set to permit no flow
across them. The surface of the cylinder or plate Fig. 1 Computational meshes for flow past inclined
was set as a solid wall with the no-slip condition flat plate. Total number of computational cells
and the standard wall function. Time-dependent =62103.
solutions were sought at different times. We used
a time step of 1/20 of the expected vortex

571
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

3. VALIDATION CASES OF FLOW OVER We can also observe in Fig. 2 the effect of Re on
CIRCULAR CYLINDER the vortex formation length. At Re=103, isolated
regions of vorticity concentration are detached
Validation of the present CFD approach is made from the attached separation shear layer as shed
by carrying out computation at three values of Re, vortices at x/D1.5. The vortex formation length
Re=103, 104 and 105. This is achieved by at Re=104 or 105 is much longer at x/D2 to 2.5.
changing U. At the lowest Re, the flow is Vorticity contours in Fig. 2 are at the instants of
computed assuming laminar flow. Fig. 2 shows peak upward and downward lift. At these instants,
the computed vorticity contours for the vortex the two regions of vorticity concentration at two
shedding flow behind the cylinder at the two sides of the cylinder are found to swing to their
lower Re, at the instants of peak lift in either extreme lateral positions. Nishimura and Yaniike
direction. For the turbulent flows, very similar (2001) investigated experimentally the
patterns are found at Re=105 as at Re=104, while relationship between vortex shedding and
the laminar flow at Re=103 shows a wider wake. generation of fluctuating lift and drag on a
Thus, results at the highest Re are not shown for circular cylinder. They suggested that shedding of
brevity. In all Re, the computed lift force on the a vortex leads to a rotation of both separation
cylinder exhibits sinusoidal fluctuations at the points on the cylinder surface away from the
vortex shedding frequency. There is a cylinder side where the vortex is shed. Essentially,
sinusoidally fluctuating component in the the two shear layers are pushed to the other side
computed drag force at double the vortex of the cylinder. This rotates the total force on the
shedding frequency. The values of St determined cylinder from the alongwind direction towards
from the lift curve have been listed in Table 2. one side. The lateral component of the total force
For the laminar flow at Re=103, the computed leads to an increasing lift while the drag, being
value of St=0.23 is in reasonable agreement with the alongwind component, drops. Alternative
the experimental observations (Fey et al., 1998). vortex shedding results in sideway oscillations of
At the higher Re, much higher values of St are the total fluid force about the mean flow direction
computed and these depart more from the (Drescher, 1956; Norberg, 2003). The lift thus
experimental results. Table 2 also shows that oscillates at the vortex shedding frequency while
higher drag coefficients are found for the laminar the drag fluctuates at twice that frequency. The
and lower Re turbulent flows. same sideways flipping of the shear layers is

Fig. 2 Computed vortex patterns at instants of peak upward and downward lift on circular cylinder. Re: (a) 103;
(b) 104. Left panels: peak upward lift; right panels: peak downward lift. Velocity vectors and contours of
non-dimensional vorticity shown at =D/U={±0.25, ±0.5, ±1, ±2, ±3, …}.

572
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

observed in our CFD results. In this case of a carried out at four angles of attack, =20o, 25o,
cylinder wake, the fluid force is contributed by 30o and 45o. The Reynolds number is
suction pressure caused by fluid entrainment from Re=UB/=2104, with the free-stream velocity at
both shear layers. This is different from the 2 m/s and B=15 cm. The Reynolds numbers of
mechanism of lift production on an inclined plate our previous experimental studies are Re=3104
where the flow separation points are fixed to the in Lam (1996) and Re=5.3103 in Lam and Leung
plate edges and the suction on the plate surface is (2005). At all four values of , periodic vortex
contributed almost entirely by the shear layer shedding is computed. Fig. 3 shows the computed
from the leading edge of the plate. The details vorticity fields at these four angles of attack at the
will be discussed in the next section. instants of maximum and minimum lift.
Alternating vortices of opposite senses are shed
4. FLOW PAST INCLINED FLAT PLATE from the plate in the form of a vortex street. The
sizes of the wake, vortices and the vortex street
Computation of flow past an inclined flat plate is

Fig. 3 Computed vortex patterns of flow past inclined flat plate. Angle of attack, : (a) 20o; (b) 25o; (c) 30o;
(d) 45o. Left panels: maximum upward lift; right panels: minimum lift. Velocity vectors and contours of
non-dimensional vorticity shown at =B1/U={±0.2, ±0.5, ±1, ±2, ±3, …}.

573
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

4.0
smaller length scales (Fig. 3). Thus, the vortex
3.5 shedding frequency, f, as determined from the
CL   3.0
time history of the lift fluctuations, becomes
higher. This is shown by the value of fB/U in
2.5
Table 1. When scaled with the projected plate
2.0 width, the Strouhal number for vortex shedding
CD 1.5 shows similar values at St0.17±0.005 for
CD
different . Previous experiments found that St
1.0
has similar values near 0.15 for 30o and there is
0.5 a sharp increase in St at smaller  (Knisely, 1990).
0.0
Lam and Leung (2005) found that St=0.15 at
0 5 10 15 20 25 =25o and St=0.18 at =20o.
tU /B1
For the drag on the plate, the mean drag
Fig. 4 Fluctuating lift and drag coefficients on coefficient is found to increase with  even when
inclined flat plate. : solid line: 20o; broken the projected plate width is used to define the
line: 25o; +: 30o; : 45o. CL=CD at =45o. coefficient. The rms drag coefficient is also
higher at larger . One notable observation about
clearly scale with the projected plate width. In the the drag force is shown in Fig. 4. It is evident that
following sections, solutions of flow velocities, the drag oscillates at the vortex shedding
vorticity and pressure at successive phases of a frequency just as the lift. This is in contrast to the
vortex shedding cycle will be analysed to flow past a circular cylinder where the drag force
investigate the shedding process of the vortices oscillates at twice the vortex shedding frequency
and the relation to the oscillating lift. (which is also observed in our CFD results).
Furthermore, the fluctuations in drag and lift are
4.1 Strouhal number, lift and drag in phase, that is, the maximum drag occurs at the
coefficients same instant as the maximum lift. Fig. 4 also
shows that the fluctuating lift or drag starts to
The solution of the pressure field in the flow can
be used to compute the drag and lift forces on the develop a sub-harmonic when  increases to 30o
plate. For all four plates, there is a non-zero mean and this sub-harmonic is obviously observed at
lift plus a periodically fluctuating lift at the vortex =45o. An explanation for the above observations
shedding frequency. As expected, the mean lift will be suggested in the following section.
force on the plate, or the lift coefficient based on
the constant plate width B, is found to increase 4.2 Vortex shedding process and fluctuating
lift production
mildly with  (Table 1). For an inclined flat plate,
the projected plate width B1=B sin has been Fig. 5 shows in details the flow past the inclined
shown to be a better characteristic length for the plate at =30o at successive phases in one vortex
wake (Fage and Johansen, 1927; Knisely, 1990; shedding cycle, starting and ending with the
also see Fig. 3). Thus, unless specified otherwise, instant of maximum lift on the plate. The
B1 is used to calculate the Strouhal number for development of the clockwise-rotating vortex
vortex shedding, St=fB1/U, and the lift (and drag) from the trailing edge of the plate is relatively
coefficients, such as CL  FL ( 1 2 U  2 B1 ) (where simple as compared with the counterclockwise-
FL is the lift force on unit length of the plate). rotating vortex from the plate leading edge. Flow
Table 1 shows that the lift coefficient, based on separates from the trailing edge and the separation
the projected plate width, exhibits lesser shear layer rolls up into a vortex which grows on
being attached to the plate edge. The vortex
variations with  than the lift force itself. It is
grows to maturity just before the instant of
worth noting that CL increases with decrease in 
maximum lift (Fig. 5(f)) and is then shed from the
but the rms lift coefficient CL’ of the lift
plate trailing edge (Figs. 5(a) and (h)).
fluctuations (based on B1) decreases as  becomes An important observation is that associated with
smaller. When  decreases from 25o to 20o, CL’ this shedding is the movement of the separation
drops by more than one half. This suggests that shear layer from the plate leading edge towards
the strength of the vortex street, say at =45o, is the plate. The shear layer originates from flow
much larger than that at, say =20o. separation at the leading edge and extends over
As the angle of attack becomes smaller, the wake and beyond the plate length. As a new trailing
width becomes smaller and the vortices have edge vortex grows, say, from Fig. 5(b) onwards,

574
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

this separation shear layer from the plate leading space between the plate trailing edge and the
edge is observed to be pushed more away from trailing edge vortex. This breaks off the vortex
the upper surface of the plate. As the shear layer from its attachment to the plate trailing edge and
entrains fluid and thus produces suction pressure the lower shear layer, thus causing the vortex to
on the upper plate surface (Luo et al., 1994), its be shed in Fig. 5(h). This stage of vortex shedding
outward movement leads to lower suction mechanism has been well documented for vortex
pressure and consequently lower lift and drag shedding from a bluff body (e.g., Cantwell and
force on the plate. The minimum lift (and drag) Coles, 1983; Lam, 2009).
on the plate occurs at the instant between Figs. The formation and shedding of a vortex from the
5(c) and 5(d). After the instant of minimum lift, leading edge can be observed from Fig. 5(a)
the trailing edge vortex starts to move away from onwards. The vortex originates from the
the plate edge but remains attached to it. At the separation shear layer at the plate leading edge
same time, the middle part of the upper shear and it grows mainly in length. During the growth,
layer starts to move towards the plate upper it moves downwards to trigger the shedding of the
surface and makes its way into the lower shear trailing edge vortex (Fig. 5(a)) but it continues to
layer. This increases the suction pressure on the grow in length afterwards. In Fig. 5(b), a new
plate upper surface and raises the lift and drag. trailing edge vortex rolls up but the upper shear
Eventually, the upper shear layer intrudes into the layer continues to elongate. In Fig. 5(c), it extends

Fig. 5 Computed vorticity at successive phases of a vortex shedding cycle. Inclined flat plate at =30o. Phases
shown are in parts of a period of one shedding cycle from the instant of maximum lift. Velocity vectors and
vorticity contours shown at =B1/U={±0.25, ±0.5, ±0.75, ±1, ±1.25, …}.

575
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

to a length of almost 2B without showing any associated negative pressure region still cannot
eminent breaking. It is until some growth of the produce a noticeable force on the plate due to its
trailing edge vortex which is about to intrude into location at the plate trailing edge and thus
the elongated upper shear layer that the shear downstream of the plate (Fig. 6(d)). Instead, its
layer eventually breaks it into two parts (Figs. fluid circulation induces some reverse flow to hit
5(e)-(g)), with the detached part becoming a shed back onto the plate surface (Fig. 6(c)) and causes
leading edge vortex. The actual shedding of the some pressure recovery. The presence of the
leading edge vortex takes place not near the plate vortex also pushes the upper shear layer from the
edge but is at a location similar to that of the plate leading edge farther away from the plate.
trailing edge vortex. This is in agreement with the The upper plate surface is thus under low negative
experimental findings of Lam and Leung (2005) pressure, leading to the smallest normal force on
and the LES simulation of Breuer et al. (2003). the plate. Fig. 6 suggests that the flow and
Lam and Leung (2005) discussed that for the pressure on the lower side of the plate do not
development and shedding of the leading edge experience large changes during a cycle of vortex
vortex, the upper surface of the inclined plate acts shedding.
like an after body part of a bluff body and thus the
shedding location is not from the fixed separation
point of the plate leading edge. On the contrary,
the plate surface has a much weaker effect on the
development of the trailing edge vortex and the
singular plate edge condition leads to a simpler
shedding location. This discussion is supported by
the present CFD result in Fig. 5. It should,
however, be noted that the actual formation of the
leading edge vortex involves large-scale flow
separation which cannot be well modelled by the
present RANS computation. The LES simulation
of Breuer et al. (2003) showed a more complex
process but the focus of this study is on the vortex
street downwards and the production of fluid
force.
Fig. 6 shows the computed flow and distributions
of pressure coefficient above the inclined plate at
the instants of maximum and minimum lift. It is
evident that the maximum lift occurs when the
trailing edge is shed from the plate trailing edge
(Figs. 6(a) and (b)). Although the vortex is
associated with negative pressure (suction), its
location is completely downstream of the plate
such that the negative pressure does not
contribute to fluid force on the plate. Instead, the
upper separation shear layer from the plate
leading edge moves to the closest distance from
the upper plate surface. The region above the
plate is entirely under the effect of this shear layer
which entrains air from the region. There is
recirculating flow in this region with uniformly
negative pressure. This produces the largest Fig. 6 Computer flow vectors and pressure
normal force on the plate within the vortex distribution around inclined flat plate at
shedding cycle. Since both the plate lift and drag =30o. (a) instant of maximum lift;
are components of this force, they reach their (b) instant of minimum lift. Contours of
maximum values together (Fig. 4). pressure coefficient shown at Cp={±0.2, ±0.4,
±0.6, …}. Solid contours for negative
About half a cycle later, the next trailing edge
pressure; broken contours for positive
vortex grows to maturity. Being remained pressure.
attached to the plate, it occupies the largest area
above the trailing edge of the plate. However, the

576
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

6 by a constant adjustment factor of value 2. The

: Trailing edge vortex. unsteady RANS solutions exhibit almost perfectly
▲: leading edge vortex.
Solid symbols: present CFD data. periodic flow patterns but jitter among cycles was
Open symbols: PIV data of Lam and
Leung (2005) with adjustment.
present due to random turbulent fluctuations in
4
the experiments. The jitter resulted in reduced
' values of the phase-averaged peak vorticity levels.
This is why an adjustment is necessary to bring
2 the CFD data and experimental data to similar
levels. The lower peak vorticity levels in the
experimental data are also believed to be caused
by the much larger spatial resolution of the
0
velocity vector data from PIV as compared to
0 4 8 12
x/B' those in the fine computational grid.
Notwithstanding this adjustment, the two sets of
Fig. 7 Magnitudes of peak vorticity levels of leading data show very similar trends of vorticity change
and trailing edge vortices behind inclined flat
with axial distance. The CFD results for the plate
plate at =30o.
inclined at other values of  show the same
Fig. 4 shows that at the largest 30o, the lift (and observation of different strengths between the two
drag) experiences a short reduction after it reaches trains of vortices. The details are not shown for
the maximum. It have been discussed in Fig. 5 brevity.
that the production of fluctuating normal force on
5. CONCLUSIONS
the upper plate surface is affected by the sideways
swaying movement of the upper separation shear
The vortex shedding flow past an inclined flat
layer in relation to the growth and shedding of the
plate at Re=2104 is simulated with unsteady
trailing edge vortex. At =30o (Fig. 5) and
RANS computation. The primary objective is to
smaller values of  (Fig. 3), the movement of the provide evidence to the previous experimental
upper shear layer does not appear to reach the finding of two trains of unequal-strength vortices
plate surface even in the extreme instant of its behind a flat plate at high incidence (Lam, 1996;
intrusion into the lower shear layer to shed the Lam and Leung, 2005). Computation is
trailing edge vortex. However, the CFD results
performed at four values of =20o, 25o, 30o and
for =45o (e.g., Fig. 3) suggest that upon 45o and vortex shedding in the form of a vortex
shedding of the trailing edge vortex, the intrusion street is observed from each edge of the inclined
of the upper shear layer does reach the plate. The global features of the wake from the
downstream part of the upper plate surface. It is computational results are in agreement with
believed that there is some partial pressure previous experimental investigation. These
recovery associated with this flow hitting the include Strouhal number and the scaling of
plate surface and this is responsible for the sub- vortices and wake on the projected plate width.
harmonic variation in the lift curve at =45o (Fig. The computed vorticity field shows different
4). formation and shedding stages between the
vortices from the leading edge and the trailing
4.3 Strengths of vortices in vortex street edge of the plate. While these main findings are in
One objective of the present CFD study is to agreement with the experiments, the additional
provide evidence to the experimental observations information of flow-induced pressure and force in
of Lam and Leung (2005) that the trailing edge the CFD results shows the relationship between
vortices possesses higher vorticity levels than the vortex shedding and the generation of fluctuating
leading edge vortices at the same axial locations. lift and drag forces on the plate. The trailing edge
Fig. 7 shows the magnitudes of peak vorticity vortex develops from the rollup of the separation
inside the leading and trailing edge vortices in Fig. shear layer at the plate trailing edge. Although it
5 plotted against the axial locations of the vortex remains attached to the plate during its growth, it
centres after shedding. The computed non- does not contribute much to the fluid force on the
dimensional vorticity, =B1/U, is shown in the plate due to its location downstream of the plate.
figure. It is evident that the CFD results also show The fluctuating force on the plate is largely
higher peak vorticity levels for the trailing edge caused by the pressure on the upper plate surface
vortex. The experimental data of (x) are which is governed by the location, length and
included in Fig. 7 but they have been multiplied fluid entrainment of the upper shear layer from

577
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

the plate leading edge. These activities of the for the circular cylinder in the range
upper shear layer are found to exhibit 47  Re  2  105 . Phys. Fluids 10(7):1547–
interrelationship with the growth of the trailing 1549.
edge vortex. The maximum lift occurs during the 9. Fluent Inc. (2003). Fluent 6.1 User’s Manual,
shedding of the trailing edge vortex when the www.fluent.com.
upper shear layer moves closest to the plate to 10. Iaccarino G, Ooi A, Durbin PA, Behnia M
trigger the shedding. On the other hand, when the (2003). Reynolds averaged simulation of
trailing edge vortex grows to maturity but remains unsteady separated flow. Int. J. Heat and
attached to the plate, the upper shear layer is Fluid Flow 24:147–156.
pushed farthest away from the plate and the plate 11. Knisely CW (1990). Strouhal numbers of
is under the minimum lift. rectangular cylinders at incidence: a review
The CFD results provide support to the and new data. J. Fluids Struct. 4:371–393.
experimental observation that the leading edge 12. Lam KM (1996). Phase-locked eduction of
vortex is formed from a break off of the upper vortex shedding in flow past an inclined flat
shear layer. The actual shedding location is near plate. Phys. Fluids 8(5):1159–1168.
the trailing end of the plate. A comparison 13. Lam KM (2009). Vortex shedding flow
between the computed peak vorticity levels at the behind a slowly rotating circular cylinder. J.
vortex centres shows again that at the same Fluids Struct. 25(2):245–262.
downstream locations, the trailing edge vortex has 14. Lam KM, Leung MYH (2005). Asymmetric
higher vorticity levels that the leading edge vortex. vortex shedding flow past an inclined flat
plate at high incidence. Euro. J. of Mech.-
REFERENCES B/Fluids 24:33–48.
15. Lam KM, To AP (2006). Reliability of
1. Bosch G, Rodi W (1998). Simulation of numerical computation of pedestrian level
vortex shedding past a square cylinder with wind environment around a row of tall
different turbulence models. Int. J. Numer. buildings. Wind Struct. 9:473–492.
Meth. Fluids 28:601–616. 16. Lam KM, Wei CT (2006). Characteristics of
2. Breuer M (1998). Large eddy simulation of vortices shed from a circular cylinder and an
the subcritical flow past a circular cylinder: inclined flat plate. Proc. 4th Int. Sym. Comput.
numerical and modeling aspects. Int. J. Wind Eng., Yokohama, July 2006:649–652.
Numer. Meth. Fluids 28:1281–1302. 17. Lasher WC (2001). Computation of two-
3. Breuer M, Jovicic N, Mazaev K (2003). dimensional blocked flow normal to a flat
Comparison of DES, RANS and LES for the plate. J. Wind Eng. Ind. Aerodyn. 89:493–513.
separated flow around a flat plate at high 18. Leonard BP (1979). A stable and accurate
incidence. Int. J. Numer. Meth. Fluids convective modelling procedure based on
41:357–388. quadratic upstream interpolation. Comp. Meth.
4. Cantwell B, Coles D (1983). An experimental in Appl. Mech. and Engrng. 19:59–98.
study of entrainment and transport in the 19. Luo SC, Yazdani MG, Chew YT, Lee TS
turbulent near wake of a circular cylinder. J. (1994). Effects of incidence and afterbody
Fluid Mech. 136:321–374. shape on flow past bluff cylinders. J. Wind
5. Drescher H (1956). Messung der auf Eng. Ind. Aerodyn. 53:375–399.
querangestr. omte Zylinder ausgeubten 20. Nishimura H, Taniike Y (2001).
zeitlich veranderten Drucke. Zeitschrift fur Aerodynamic characteristics of fluctuating
Flugwissenschaften und Weltraumforschung forces on a circular cylinder. J. Wind Eng. Ind.
4:17–21. Aerodyn. 89:713–723.
6. Fage A, Johansen FC (1927). On the flow of 21. Norberg C (2003). Fluctuating lift on a
air behind an inclined flat plate of infinite circular cylinder: review and new
span. Proc. R. Soc., London, Ser. A 116:170– measurements. J. Fluids Struct. 17:57–96.
197. 22. O’Shea N, Fletcher CAJ (1995). Prediction of
7. Ferreira AD, Sousa ACM, Viegas DX (2002). turbulent delta wing vortex flows using an
Prediction of building interference effects on RNG k   model. Lecture Notes in Physics
pedestrian level comfort. J. Wind Eng. Ind. 453:496–501.
Aerodyn 90:305–319. 23. Papageorgakis GC, Assanis DN (1999).
8. Fey U, Konig M, Eckelmann H (1998). A Comparison of linear and nonlinear RNS-
new Strouhal-Reynolds-number relationship based k   models for incompressible
turbulent flows. Num. Heat Transfer 35:1–22.

578
 Engineering Applications of Computational Fluid Mechanics Vol. 4, No. 4 (2010)

24. Patankar SV (1980). Numerical heat transfer

and fluid flow. Hemisphere Publ. Corp.,
Washington D.C.
25. Perry AE, Steiner TR (1987). Large-scale
vortex structures in turbulent wake behind
bluff bodies. Part 1. Vortex formation
processes. J. Fluid Mech. 174:233–270.
26. Rodi W (2006). DNS and LES of some
engineering flows. Fluid Dyn. Res. 38:145–
173.
27. Shao J, Zhang C (2006). Numerical analysis
of the flow around a circular cylinder using
RANS and LES. Int. J. Comput. Fluid Dyn.
20:301–307.
28. Stathopoulos T (2006). Pedestrian level winds
and outdoor human comfort. J. Wind Eng. Ind.
Aerodyn. 94:769–780.
29. Van Doormaal JP, Raithby GD (1984).
Enhancement of the SIMPLE method for
predicting incompressible fluid flows. Num.
Heat Transfer 7:147–163.
30. Yakhot V, Orszag SA (1986).
Renormalization-Group analysis of
turbulence. Phys. Rev. Lett. 57:1722–1724.

APPENDIX

In April 2010, a Chinese bulk carrier ran aground

near the Great Barrier Reef and caused oil spill
(http://en.wikipedia.org/wiki/2010_Great_Barrier
_Reef_oil_spill). It is interested to note that the
flow of ocean current past the stranded ship on
the sea surface bears close resemblance to the
flow over an inclined plate. The flow was clearly
visualized by the 3.7 km length of leaked oil slick
and detergents. There are video clips on the
patterns from the above quoted webpage and a
photograph is reproduced here in Fig. A. One can
clearly observe the difference between the masses
contained within the leading edge vortices and the
trailing edge vortices.

Fig. A Oil slick pattern from a stranded ship.

579

View publication stats