You are on page 1of 12

This article was downloaded by: [Australian National University]

On: 09 January 2015, At: 03:34


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Hydraulic Research


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tjhr20

Effect of spur dike length on the horseshoe vortex


system and the bed shear stress distribution
a b
Mete Koken & Mustafa Gogus
a
Associate Professor, Department of Civil Engineering, Middle East Technical University,
Ankara, Turkey
b
Professor, Department of Civil Engineering, Middle East Technical University, Ankara,
Turkey
Published online: 21 Oct 2014.

Click for updates

To cite this article: Mete Koken & Mustafa Gogus (2014): Effect of spur dike length on the horseshoe vortex system and
the bed shear stress distribution, Journal of Hydraulic Research, DOI: 10.1080/00221686.2014.967819

To link to this article: http://dx.doi.org/10.1080/00221686.2014.967819

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of
the Content. Any opinions and views expressed in this publication are the opinions and views of the authors,
and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied
upon and should be independently verified with primary sources of information. Taylor and Francis shall
not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other
liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Journal of Hydraulic Research, 2014
http://dx.doi.org/10.1080/00221686.2014.967819
c 2014 International Association for Hydro-Environment Engineering and Research

Research paper

Effect of spur dike length on the horseshoe vortex system and the bed shear stress
distribution
METE KOKEN, Associate Professor, Department of Civil Engineering, Middle East Technical University, Ankara, Turkey
Email: mkoken@metu.edu.tr (author for correspondence)

MUSTAFA GOGUS, Professor, Department of Civil Engineering, Middle East Technical University, Ankara, Turkey
Downloaded by [Australian National University] at 03:35 09 January 2015

Email: mgogus@metu.edu.tr

ABSTRACT
Turbulent flow structures forming around isolated spur dikes in a horizontal channel are investigated in this study. In the analysis detached eddy
simulation is used under fully turbulent incoming flow conditions at a channel Reynolds number of 45,000. Changes in the structure of the horseshoe
vortex system, bed shear stress and pressure standard deviation on the bed are investigated for three different spur dike lengths. In all of the cases
the main horseshoe vortex undergoes bimodal oscillations, which leads to an amplification in the turbulence quantities such as turbulent kinetic
energy and pressure fluctuations along its axis. The main horseshoe vortex disappears over a much shorter distance in the flow direction for the
short spur dike than those of medium and long spurs. Large bed shear stress values and pressure standard deviation values observed around the tip of
the spur dike, beneath the upstream part of the main horseshoe vortex and beneath the separated shear layers increase with the increasing length of
the spur dike. Different from the long and medium spur dike, in the short spur dike case, it is shown that the secondary horseshoe vortex is as coherent
as the main horseshoe vortex and it contains bimodal oscillations together with the main horseshoe vortex.

Keywords: Bed shear stress; detached eddy simulation; horseshoe vortices; spur dike; turbulent flow

1 Introduction eddy simulation (LES) or detached eddy simulation (DES) pro-


vide a more detailed solution compared with the conventional
Spur dikes built along a river cause a very complex three- Reynolds averaged Navier Stokes (RANS) models, which are
dimensional flow field to occur within their proximity. As widely used.
the flow approaches a spur dike as a result of the blockage There are studies in literature that used eddy resolving tech-
effect, flow separates along the channel bed and horseshoe vor- niques such as LES and DES to investigate the flow dynamics
tices, which are necklace-like structures wrapping the spur, and the horseshoe vortex system around bluff bodies. Smith
form. These vortices play an essential role in the local scour & Foster (2007) investigated the flow field around a bottom
mechanism around the spurs. mounted short cylinder using LES at various channel Reynolds
Although there are various experimental studies on the pre- numbers based on flow depth (R = UD/ν) in the range of 300–
diction of the scour pattern, about the maximum scour depth 3900, where U is the average flow velocity, D is the flow depth
forming around different abutment and/or spur dike geome- and ν is the kinematic viscosity. Sohankar (2007) investigated
tries, and on the temporal development of the scour and scour the flow around a V-shaped vortex generator mounted on a
countermeasures (Kothyari & Raju, 2001; Kwan & Melville, channel bed with a flat surface using LES and direct numer-
1994; Melville, 1997; Morales, Ettema, & Barkdoll, 2008) ical simulation at various Reynolds numbers in the range of
not much work is done to investigate the coherent structures 200–2000. He found that kinetic energy and pressure fluctu-
and their effect in the scour process around spurs. In order to ations are high within the flow along the vortex cores. Paik,
make such investigations, numerical modelling is advantageous Sotiropoulos, & Porte-Agel (2009) investigated the flow around
compared with experimental modelling, as it provides a three- two wall-mounted cubes located in a tandem position on a flat
dimensional solution of a flow domain instead of point-wise surface using DES. Escauriaza & Sotiropoulos (2011) used DES
or planar measurements usually made in experiments. Among to investigate the effect of Reynolds number on the horseshoe
the turbulence models, eddy resolving techniques such as large vortex dynamics for flow around a bottom mounted cylinder.

Received 29 August 2013; accepted 20 August 2014/Currently open for discussion.

ISSN 0022-1686 print/ISSN 1814-2079 online


http://www.tandfonline.com
1
2 M. Koken and M. Gogus Journal of Hydraulic Research (2014)

At both Reynolds numbers that they studied (2.0 × 104 and scheme. This is needed to minimize the level of numerical dis-
3.9 × 104 ) the presence of bimodal oscillations in the main sipation away from solid boundaries. All other terms in the
horseshoe vortex was observed. momentum and pressure-Poisson equations are approximated
Koken & Constantinescu (2008a, 2008b, 2009) and Koken using second-order central differences. The discrete momentum
(2011) investigated the structure and the unsteady flow dynam- (predictor step) and turbulence model equations are integrated in
ics of the flow around an isolated spur dike both for flat bed and pseudo-time using the alternate direction implicit (ADI) approx-
equilibrium scour bed conditions at various channel Reynolds imate factorization scheme. A general description of the code
numbers ranging from 18,000 to 500,000. They used both used in this study is given by Constantinescu & Squires (2004).
LES and DES models in their investigations. Similar to the In the SA-based DES model, a transport equation is solved for
experimental findings of Devenport & Simpson (1990), who the modified eddy viscosity, ν̃:
introduced the bimodal nature of the main horseshoe vortex
that forms around a wing-shaped body, they observed that the ∂ ν̃ ∂ ν̃ 1
+ u j j = Cb1 S̃ ν̃ + [∇((ν + ν̃)∇ ν̃)
main horseshoe vortex undergoes bimodal aperiodic oscillations ∂t ∂ξ σ
 2
especially at locations close to the tip of the spur dike. These ν̃
oscillations were causing the turbulent kinetic energy and pres- + Cb2 (∇ ν̃)2 ]Cw1 fw (1)
d
Downloaded by [Australian National University] at 03:35 09 January 2015

sure fluctuations inside the core of the main horseshoe vortex


to be at least one order of magnitude larger than the surround- where ν is the kinematic viscosity, u j is the contravariant
ing flow. The presence of these bimodal oscillations for flow resolved velocity, t is time, d is the turbulence length scale,
around abutments and piers is also shown by many authors and ξ j is the curvilinear coordinate in the j direction. The other
(Bressan, Ballio, & Armenio, 2011; Escauriaza & Sotiropoulos, variables and parameters are:
2011; Kirkil & Constantinescu, 2010; Paik, Escauriaza, &
Sotiropoulos, 2007; Simpson, 2001). S̃ ≡ S + (ν̃/κ 2 d2 )fv2 (2)
Koken (2011) investigated the effect of approach flow angle
on the coherent structures forming around a similar spur geome- where S is the magnitude of the vorticity, κ is the Von Karman
try used in the present study. He found that the horseshoe vortex constant, which is 0.41 and
system is strongest and the magnitude of the bimodal oscilla-  
tions of the main horseshoe vortex is largest when the spur is 1
fv2 = 1 − ν̃/ + ν̃fv1 (3)
perpendicular to the flow direction. R
Within the present study, with an increase in the spur dike
length the changes in the coherent structures forming around The eddy viscosity νt is obtained from
the spurs and their possible impacts on the scour process is
investigated in flat-bed conditions. Specifically, changes in the νt = ν̃fv1 (4)
dynamics of the horseshoe vortex system is investigated. Three
spur dike lengths are used within this study, which will be where
named as case short spur (SS), case medium spur (MS) and case
long spur (LS) throughout the rest of the text. fv1 = χ 3 /(χ 3 + C3v1 ) (5)
ks
χ = ν̃/ν + 0.5 (6)
d
2 Numerical model  1/6
1 + C6w3
fw = g (7)
g 6 + C6w3
A Spalart-Almaras (SA) based DES model is used in the present
study. The reason for selecting this model is its superior capa- g = r + Cw2 (r6 − r) (8)
bility in capturing the massively separated flows compared with ν̃
conventional RANS-based models. In a DES model, unsteady r≡ (9)
S̃κ 2 d2
RANS mode is active close to the solid boundaries whereas
away from the walls the model acts like a LES. No wall func- The model constants in the above equations are: Cb1 = 0.135,
tions are used in the present code, i.e. the viscous sublayer is Cb2 = 0.622, σ = 0.67, Cv1 = 0.71, Cw2 = 0.3, Cw3 = 2.0 and
directly resolved with sufficiently small grid spacing near all Cw1 = Cb1 /κ 2 + (1 + Cb2 )/σ . In the DES model, the distance
solid walls. The 3D incompressible Navier-Stokes equations to the nearest wall, d, which is used in the destruction term
are integrated using a fully-implicit fractional-step method. The of the transport equation of ν̃, is replaced with a new length
governing equations are transformed into generalized curvilin- scale dDES = min(d, CDES ), where the model parameter CDES
ear coordinates on a non-staggered grid. Convective terms in is equal to 0.65 and  is the local grid size. This will result in an
the momentum equations are discretized using a blend of fifth- eddy viscosity which is proportional to the mean rate of strain
order accurate upwind biased scheme and a second-order central and 2 as in an LES which employs a Smagorinsky model.
Journal of Hydraulic Research (2014) Effect of spur dike 3

(a) (b)

Figure 1 (a) Computational domain; (b) structured grid for case LS

It allows the energy to cascade down to length scales of the 3 Validation


Downloaded by [Australian National University] at 03:35 09 January 2015

size of the grid. The eddy viscosity predicted by DES in the


LES regions goes to zero if the local grid size decreases to zero, The code used in the present study is validated for various flow
which is the analogue to classical LES. conditions related to fluvial hydraulics, such as flow in mean-
The length scale is selected to be the flow depth dering channels and flow around bridge piers (Constantinescu,
(D = 0.135 m) whereas the mean approach velocity (U = Koken, & Zeng, 2011; Kirkil & Constantinescu, 2010). In all
0.335 m s−1 ) is used as the velocity scale. The flow depth and those cases a good agreement was achieved between the exper-
velocity magnitude are selected to be the same as a previous imental and numerical results. Specifically the code is validated
experimental study conducted under clear water scour condi- for the present geometry comparing experimental and numerical
tions for a median sediment size of d50 = 1.5 mm (Gogus & results by Koken (2011), using the mean flow velocities and the
Koken, 2011). All of the variables used in the text are either non- turbulent kinetic energy values at different sections around the
dimensionalized or expressed using flow depth and/or approach spur dike. All those quantities showed a very good agreement
flow velocity. The length and width of the domain are selected to with the experimental data where the position and magnitude of
be 48D and 11.11D, respectively. Spur dike lengths, including the amplifications in the turbulent kinetic energy values inside
the curved toe, are 1.11D, 1.85D and 2.59D for cases SS, MS the separated layers were reasonably captured. No additional
and LS, respectively, where the spur dike is located 8D from discussion will be made related to the validation here, one can
the inlet section. Structured meshes having 432 × 192 × 52 refer to the mentioned references for further details.
nodes in the streamwise, spanwise and vertical directions are
used in the simulations. A minimum grid spacing of y + =
4 Results
yuτ /ν ∼ 1 wall unit in the normal direction to all the solid
boundaries is achieved with a tanh-type stretching function
4.1 Horseshoe vortex system
(Fig. 1). Additional grid refinement is done within the horse-
shoe vortex region and along the separated shear layers where Coherent structures forming around the spur dike are visualized
the mesh size varies between 30–150 wall units, which enables for the mean flow using Q criterion for cases SS, MS and LS
capturing the vortical structures that are present around the in Fig. 2. Note that separated shear layers are covered within
spur dike. this figure to better emphasize the horseshoe vortices forming
Turbulent inflow conditions are obtained from a LES simu- around the spur. In all the cases, the horseshoe vortex system
lation from a periodic channel, simulation data of which were consists of three main vortices that wrap around the spur. These
stored in a file. These are fed into the inflow section of the vortices, ordered with respect to their closeness to the spur are:
DES simulation, which contains the spur dike, in a time-accurate the primary horseshoe vortex (HV1) the secondary horseshoe
fashion. The free surface is modelled as a rigid lid. A no slip vortex, (HV2) and the tertiary horseshoe vortex (HV3). For all
boundary condition is used on all solid surfaces. At the out- spur lengths HV1 forms close to the junction line formed by
flow a convective boundary condition is used that allows the the channel sidewall and the upstream spur face. It follows the
exit of coherent structures from the domain in a time accu- upstream spur face and then curves strongly in the flow direction
rate way without producing unphysical oscillations. The time once it gets close to the spur tip. Unlike HV1, HV2 and HV3
step used in the simulations was 0.025 D/U, which yields a form at a further upstream position of this junction line. This
Courant number of 0.3. The simulations were run for approx- causes a much smoother bending in the flow direction close to
imately 10,000 iterations on 12 processors of a PC cluster. The the spur tip compared with HV1.
parallelization of the code is done through a message passing Similar to the findings of Koken & Constantinescu (2008a),
interface (MPI). in all of the cases there is a coherent corner vortex, CV, at the
4 M. Koken and M. Gogus Journal of Hydraulic Research (2014)

(a) (b)
Downloaded by [Australian National University] at 03:35 09 January 2015

(c)

Figure 2 Vortical structures around the spur dike for the mean flow visualized by Q criterion for: (a) case SS; (b) case MS; (c) case LS

upstream recirculation region of the spur dike. This vortex orig- SS the recirculating flow cannot provide as much additional
inates from the free surface and goes into the deeper flow depths momentum to HV1. This difference is evident from the turbu-
where it bends and eventually merges with HV1. Therefore, cor- lent kinetic energy contours within this region (see Fig. 8 later),
ner vortex (CV) convects additional fluid and momentum from which will be discussed in Section 4.2. As a result HV2 and
the regions close to the free surface into the core of the primary HV3 form at a closer distance to the spur. This has an important
necklace vortex HV1. consequence on the turbulent characteristic of HV2 and will be
Although there are similarities in the orientation of the horse- discussed later.
shoe vortices for all of the cases investigated, there are important Figure 3 shows the non-dimensional circulation (vorticity
differences related to their coherence and turbulence characteris- integrated over area) along the cores of HV1 and HV2 at four
tics. HV1 is underdeveloped and its size and coherence is much different sections for the three spur dike lengths investigated. In
smaller in case SS compared with case LS (see also the circula- cases LS and MS, HV1 is more coherent than HV2 as expected
tion values in Fig. 3 and streamline patterns in Figs. 4 and 5). For (Koken & Constantinescu, 2008a, 2009). On the other hand for
case SS, HV1 loses its coherence at approximately 1D distance case SS, HV2 appears to be more coherent than HV1. For all
downstream of the spur axis. On the other hand, this distance the spur lengths tested it was observed that both HV1 and HV2
is approximately 17D for case LS where the main horseshoe have the maximum coherence at section ii close to the upstream
vortex completely diffuses within the flow. This is related to part of the spur tip where the scour is initiated.
both the relatively smaller flow blockage due to the shorter spur Figures 4 and 5 show the mean non-dimensional out-of-plane
length and to the smaller and weaker recirculating flow at the vorticity contours, ωn D/U, with streamlines superimposed and
upstream of the spur. Compared with cases LS and MS, in case non-dimensional pressure standard deviation, psd / (ρU2 ), on
Journal of Hydraulic Research (2014) Effect of spur dike 5

(a) (b)
Downloaded by [Australian National University] at 03:35 09 January 2015

Figure 3 Mean non-dimensional circulation for vortices HV1 (a) and HV2 (b) that is obtained at sections i–iv for cases SS (square), MS (triangle)
and LS (circle)

Figure 4 Mean non-dimensional out of plane vorticity contours together with the streamline patterns and resolved pressure standard deviation in
representative vertical sections for case LS (see inset in frame i)

some representative vertical planes around the spur, which cuts the surrounding flow. This is because of the bimodal aperiodic
through the horseshoe vortex system for cases LS and SS, oscillations of HV1, where it oscillates in between two modes
respectively. Flow depth, D, and the mean approach flow veloc- namely zero flow mode and back flow mode. The two-peaked
ity, U, in the channel are used to non-dimensionalize vorticity. distribution of the pressure standard deviation at the core of HV1
The positions of these vertical sections are given as insets in on planes i, ii and iii in Fig. 4 is as a result of these oscillations.
both figures. One can identify the position of the horseshoe Note that at section iv, which is approximately 3.4D downstream
vortices on these planes easily by considering the amplified of the spur axis, the double peaked pressure standard deviation
vorticity magnitudes and from the streamline patterns. distribution is replaced by a single peaked distribution along the
In case LS, at the core of HV1, pressure standard deviation is legs of HV1 where the bimodal oscillations are not present any
amplified for more than one order of magnitude with respect to more.
6 M. Koken and M. Gogus Journal of Hydraulic Research (2014)
Downloaded by [Australian National University] at 03:35 09 January 2015

Figure 5 Mean non-dimensional out of plane vorticity contours together with the streamline patterns and resolved pressure standard deviation in
representative vertical sections for case SS (see inset in frame i)

In case LS, compared with HV1, HV2 and HV3 do not oscillates together with HV1. In case SS at section iv, which is
impose significant amounts of amplification on pressure stan- positioned at approximately 1.5D downstream of the spur, HV2
dard deviation (Fig. 4). On the other hand, in case SS, HV2 loses its coherence where the decay in pressure standard devia-
imposes approximately the same level of pressure standard devi- tion inside its core is approximately 40%. At this section HV1
ation amplifications as HV1 (Fig. 5). At sections i–iii, close to is not present as it has already diffused within the flow. As dis-
the tip of the spur, a double peaked distribution is observed cussed, both HV1 and HV2 forms at a very close distance to the
inside the core of HV1 and HV2. This indicates the presence of spur in case SS. At the downstream part of the abutment, HV1
large scale bimodal oscillations both for HV1 and HV2. In Fig. 6 is surrounded by the separated shear layers and the coherent
instantaneous non-dimensional out-of-plane vorticity contours HV2 within a narrow region. Therefore, it is thought to interact
are plotted on plane ii for case SS at two different instants in both with HV2 and the separated shear layers. These interac-
time, which describes the two modes, namely zero flow mode tions might be responsible for its losing of coherence at a very
and back flow mode in between which HV1 and HV2 oscillates short distance from the abutment axis. The pressure standard
aperodically. Because of the downflow at the upstream face of deviation has a one peaked distribution inside the core of HV2
the spur, a jet-like flow occurs close to the bed in the opposite at this section indicating that the large scale bimodal oscilla-
direction to the flow. If this jet like flow contains low momentum tions are not present anymore inside the legs of HV2. Note that
fluid that comes from deeper water levels, an early separation the large psd values observed on the right-hand side at section iv
occurs close to the bed (see white dashed lines in Fig. 6a) and are due to the separated shear layers.
zero flow mode occurs. On the other hand, if it contains higher For both spur lengths the pressure standard deviations are
momentum fluid that comes from upper water levels close to largest close to the spur tip at section ii where HV1 is strongest.
the free surface, the separation becomes late (see white dashed In case LS, amplification inside the core of HV1 is approx-
lines in Fig. 6b) and back flow mode occurs. To the best of the imately 5 times larger than that is observed in case SS. This
authors’ knowledge this is the first time where the bimodal oscil- indicates that the strength of HV1 is much larger in case LS.
lations are observed in the secondary horseshoe vortex together
with the primary horseshoe vortex. These bimodal oscillations
4.2 Separated shear layers
are thought to be occurring because of the closeness of HV2 to
the spur. In order to quantify, at its closest position to the spur, Non-dimensional out-of-plane vorticity contours and the
HV2 is at 0.55D distance from the spur in case SS whereas it streamline patterns at the free surface are given for the mean
is at 1.20D away from it in case LS. As forming at a close dis- flow for all of the cases studied in Fig. 7. In cases SS and MS,
tance to the obstruction, HV2 is affected by the jet like flow and there is one clockwise rotating eddy RC1 at the downstream of
Journal of Hydraulic Research (2014) Effect of spur dike 7

(a) and 1.53L, respectively. For case MS, the recirculating bubble
has a length of 10.70L and a width of 1.57L. Results are in a rea-
sonable agreement for case SS and SM suggesting that the size
of the recirculation bubble is scaling with the spur length. How-
ever, for the LS case the length and width values are 32.7D and
4.6D, that is 12.6L and 1.8L. This means that once expressed in
(b)
terms of the abutment length, the width and length of the recir-
culation bubble get larger both in length and width by 20% for
case LS.
Non-dimensional turbulent kinetic energy values, k/U2 , at
the mid flow depth and at a 0.05D depth above the chan-
Figure 6 Instantaneous non-dimensional out of plane vorticity con- nel bed are given in Fig. 8. Non-dimensional turbulent kinetic
tours together with the velocity vectors on plane ii (see inset in Fig. 5) energy values smaller than 0.06 are eliminated from this figure
for case SS at: (a) zero flow mode; (b) back flow mode to better emphasize on the amplification regions. For all of
the cases, large k/U2 values are observed along the SSL and
Downloaded by [Australian National University] at 03:35 09 January 2015

along some part of the mean position of HV1. In case SS,


the spur. On the other hand, in addition to RC1 there is also a there is no considerable amount of amplification in the k/U2
second counterclockwise rotating eddy, RC2, just at the down- value at the upstream recirculation region close to the free sur-
stream part of the spur in case LS. Separated shear layers (SSL) face, whereas in cases MS and LS, there is a notable amount
can be identified from the amplified vorticity contours in Fig. 7. of amplification at the same region suggesting that the recir-
In case SS, SSL is almost parallel to the flow direction whereas culating flow at the upstream of the spur is much stronger.
with increasing spur length it is tilted towards the main channel In general, similar levels of amplification are observed in the
(see the white arrows in Fig. 7). SSL makes approximately a 15◦ k/U2 values for all of the cases where the maximum values
angle with the flow direction in case LS. This is related to the are observed along SSL. In fact, the decay of k/U2 along SS
momentum transfer from RC2 to SSL. Such a deviation in the in the streamwise direction is much faster in SS compared
direction of SSL results in a longer and wider recirculation bub- with LS and MS at both depths. Close to the channel bed,
ble at the downstream of the spur dike even if defined in terms of large k/U2 values induced by the horseshoe vortex system can
the spur dike length. The size of the recirculation bubble should be observed. It is clear that as the length of spur increases,
be proportional to the Reynolds number, which is based on spur large values of k/U2 induced by HV1 are observed along a
dike length, L (RL = UL/ν). Since in our simulations U and longer patch. In case LS, the k/U2 value induced by HV1 is
ν are constants it is expected to be directly proportional to the approximately 20% larger than that imposed in case SS. These
spur dike length, L. In cases SS, the recirculation bubble has a are some additional indicators that HV1 is much stronger in
11.7D length and 1.7D width, which corresponds to of 10.54L case LS.

(a)

(b)

(c)

Figure 7 Mean non-dimensional out of plane vorticity contours and the streamline patterns at the free surface for: (a) case SS; (b) case MS; (c) case
LS
8 M. Koken and M. Gogus Journal of Hydraulic Research (2014)

(a)

(b)
Downloaded by [Australian National University] at 03:35 09 January 2015

(c)

Figure 8 Non-dimensional turbulent kinetic energy contours at the mid-depth (left) and close to channel bed (right) for: (a) case SS; (b) case MS;
(c) case LS. (k/U2 values smaller than 0.06 are blanked within the figure, position of the horseshoe vortices are shown in (b) with dashdot lines)

4.3 Shear stress and pressure standard deviation on the bed (a)
Distribution of the normalized mean bed shear stress, τw /τwo ,
is given in Fig. 9 for cases SS, MS and LS. Here, τwo is the
mean bed shear stress observed at the entrance of the channel.
The shear stress amplifications larger than 5 times correspond-
ing to the value on the incoming flow is shown with solid lines in
this figure where amplifications larger than five times fall inside
these solid lines. These regions will be called large bed shear
stress regions. In all of the cases, similarly shear stress values
(b)
are amplified along the flow acceleration region near the tip of
the spur, where the scour is initiated, beneath the mean position
of the horseshoe vortices (especially beneath HV1) and along
the upstream part of SSL. Although the maximum normalized
mean bed shear stress values are comparable in all of the cases
(13 in case LS and 11 in cases SS and MS), which are observed
close to the tip of the spur, there are major differences in the
spread of the normalized bed shear stress distribution.
The decay rate of the bed shear stress values within the flow (c)
direction is very different. In case SS, there is one large bed shear
stress region, which is observed only in the close proximity of
the spur and the bed shear stress values decay very rapidly in the
flow direction. This region is formed as a combined effect of the
flow acceleration, presence of HV1, and SSL and it extends up
to 1D distance from the spur axis in the streamwise direction. As
the spur length increases, a similar large bed shear stress region
is observed; however, the decay rate is much slower. Bed shear
Figure 9 Normalized mean bed shear stress contours for: (a) case
stress values that are five times larger than τwo are observed at
SS; (b) case MS; (c) case LS. The black solid contour line in frames
a distance up to 7.5D away from the spur axis in case LS. This corresponds to τw /τwo = 5
Journal of Hydraulic Research (2014) Effect of spur dike 9

5 Conclusion

Detached eddy simulation was used to investigate the changes


in the coherent structures forming around isolated spur dikes
of different lengths for flat-bed conditions. The possible scour
regions were identified by considering bed shear stress and
pressure standard deviation on the bed.
At all spur lengths three horseshoe vortices form wrapping
around the spur. The coherence of the main horseshoe vortex
was found to be very different for the three spur lengths. In the
short spur, it disappeared at approximately 1D downstream of
the spur axis whereas in the long spur it continued to be coher-
ent up to 17D downstream of the spur axis, where it completely
diffused within the flow. For all the cases investigated, similar
to the findings of Koken & Constantinescu (2008a, 2008b), the
Downloaded by [Australian National University] at 03:35 09 January 2015

main horseshoe vortex was found to undergo aperiodic bimodal


oscillations at locations close to the tip of the spur where tur-
bulence quantities such as turbulent kinetic energy and pressure
standard deviation are amplified along its mean position.
In the short spur case, pressure standard deviation contours
represented a double peaked distribution for both the main
and secondary horseshoe vortices suggesting that the aperiodic
bimodal oscillations were also present in the core of the sec-
ondary horseshoe vortex. Unlike the long and medium spur dike
cases, the secondary horseshoe vortex undergoes the same kind
of bimodal oscillations together with the main horseshoe vor-
Figure 10 Distribution of the non-dimensional pressure standard tex in the shorter spur case. This is thought to be related to the
deviation at the bed,psd /(ρU2 ) for: (a) case SS; (b) case MS; (c) case closeness of the secondary horseshoe vortex to the spur.
LS In all spur lengths, bed shear stress values are amplified at
positions close to the spur tip where flow is accelerating, beneath
the mean position of the horseshoe vortices (especially the main
is a consequence of the coherent HV1 that is present in case LS, horseshoe vortex) and along the upstream parts of the separated
which diffuses much slower within the flow. In case LS, the large shear layers. A single large bed shear stress region, where shear
bed shear stress region is observed within a wider lateral extent stress values are at least five times larger than the shear stress
than that of case SS. This is related to the size of HV1, which value of the incoming flow, is observed in the short spur case,
is larger in case LS. Different from case SS, in cases LS and whereas two separate large shear stress regions are observed in
MS there is also a second large bed shear stress region further the long and medium spur cases. Comparing three cases, in the
away from the spur. This region occurs because of the presence long spur case the total area of the large bed shear stress region
of HV2 and it follows the mean position of it. is approximately 30 and six times larger, while its streamwise
Pressure standard deviation on the bed is another important length is approximately six and three times longer than those of
quantity in sediment entrainment. The distributions of the non- short and medium spur cases, respectively.
dimensional pressure standard deviation on the bed for different Finally, for all spur lengths, large pressure standard devia-
spur lengths are given in Fig. 10. In all cases, large pressure tion values were recorded around the tip of the spur beneath
standard deviation values are observed along the upstream part the upstream part of the main necklace vortex and especially
of the separated shear layers and beneath a part of HV1 close beneath the separated shear layers. Compared with the short
to the spur. In case SS, beneath the upstream part of HV2 there spur case, pressure standard deviation values are approximately
is a considerable amplification in pressure standard deviation, two times larger throughout all the amplified regions in the long
which is comparable to the values that are observed beneath spur case.
HV1. This is a consequence of the bimodal behaviour of HV2.
Compared with case SS, in case LS the pressure standard devia-
tion is approximately two times larger on all the amplified region Acknowledgements
except the part beneath HV2. Similar to the bed shear stress dis-
tribution, the decay rate of the pressure standard deviation in the This study is supported by The Scientific and Technological
flow direction is faster with the decrease in the spur length. Research Council of Turkey (TUBITAK) under Project No:
10 M. Koken and M. Gogus Journal of Hydraulic Research (2014)

108M590 and the computational resources used in this study Constantinescu, S. G., & Squires, K. D. (2004). Numerical
were partly provided by TUBITAK ULAKBIM High Perfor- investigation of the flow over a sphere in the subcritical and
mance Computing Center, and the Department of Computer supercritical regimes. Physics of Fluids, 16, 1449–1467.
Engineering at Middle East Technical University, which are Constantinescu, S. G., Koken, M., & Zeng, J. (2011). The struc-
gratefully acknowledged here. ture of turbulent flow in an open channel bend of strong
curvature with deformed bed: Insight provided by Detached
Eddy Simulation. Water Resources Research, 47, W05515,
Notation doi:10.1029/2010WR010114.
Devenport, W. J., & Simpson, R. L. (1990). Time-dependent and
Cb1 , Cb2 , Cv1 , = Spalart Almaras model time-averaged turbulence structure near the nose of a wing-
Cw1 , Cw2 , Cw3 constants body junction. Journal of Fluid Mechanics, 210, 23–55.
D = flow depth Escauriaza, C., & Sotiropoulos, F. (2009). Lagrangian model
d = turbulence length scale of bed-load transport in turbulent junction flows. Journal of
dDES = DES length scale Fluid Mechanics, 666, 36–76.
d50 = median sediment size Escauriaza, C., & Sotiropoulos, F. (2011). Reynolds number
Downloaded by [Australian National University] at 03:35 09 January 2015

fv1 , fv2 , fw = Spalart Almaras model parameters effects on the coherent dynamics of the turbulent horse-
g = gravitational acceleration shoe vortex system. Flow Turbulence and Combustion, 86,
psd = pressure standard deviation 231–262.
r = Spalart Almaras model parameter Gogus, M., & Koken, M. (2011). Effect of the bridge abut-
t = time ment length to the flow characteristics and to the scour
k = turbulent kinetic energy process. TUBITAK Project Report 108M590, Civil Engineer-
L = spur dike length ing Department, Middle East Technical University, Ankara,
R = Reynolds number based on flow depth Turkey.
RL = Reynolds number based on spur dike length Kirkil, G., & Constantinescu, S. G. (2010). Flow and tur-
S = Magnitude of vorticity bulence structure around an in-stream rectangular cylinder
S̃ = Spalart Almaras model parameter with scour hole. Water Resources Research, 46, W11549,
u = streamwise velocity doi:10.1029/2010WR009336.
U = average flow velocity Koken, M., (2011). Coherent structures around isolated spur
uτ = friction velocity dikes at various approach flow angles. Journal of Hydraulic
uj = contravariant velocity component Research, 49, 736–743.
x = streamwise Cartesian coordinate Koken, M., & Constantinescu, S. G. (2008a). An inves-
y = transversal Cartesian coordinate tigation of the flow and scour mechanisms around iso-
z = vertical Cartesian coordinate lated spur dikes in a shallow open channel 1: Conditions
χ = Spalart Almaras model parameter corresponding to the initiation of the erosion and depo-
y + = non-dimensional grid spacing sition process. Water Resources Research, 44, W08406,
 = local grid spacing doi:10.1029/2007WR006489.
= circulation Koken, M., & Constantinescu, S. G. (2008b). An investi-
κ = Von Karman constant gation of the flow and scour mechanisms around isolated
ν = kinematic viscosity spur dikes in a shallow open channel 2: Conditions cor-
ν̃ = modified eddy viscosity responding to the final stages of the erosion and depo-
νt = eddy viscosity sition process. Water Resources Research, 44, W08407,
ρ = density doi:10.1029/2007WR006491.
σ = Spalart Almaras model constant Koken, M., & Constantinescu, S. G. (2009). An investigation
τw = wall shear stress of the dynamics of coherent structures in a turbulent channel
τwo = average wall shear stress in the approach flow with a vertical sidewall obstruction. Physics of Fluids,
channel 21(8), 085104, doi:10.1063/1.3207859.
ωn = out-of-plane vorticity Kothyari, U. C., & Ranga Raju, K. G. (2001). Scour around spur
ξj = curvilinear coordinate in j direction dikes and bridge abutments. Journal of Hydraulic Research,
39, 367–374.
Kwan, R. T., & Melville, B. W. (1994). Local scour and flow
References measurements at bridge abutments. Journal of Hydraulic
Research, 32, 661–673.
Bressan F., Ballio F., & Armenio V. (2011). Turbulence around a Melville, B. W. (1997). Pier and abutment scour: integrated
scoured bridge abutment. Journal of Turbulence, 12(3), 1–24. approach. Journal of Hydraulic Engineering, 123, 125–136.
Journal of Hydraulic Research (2014) Effect of spur dike 11

Morales, R., Ettema, R., & Barkdoll, B. (2008). Large-scale tandem. International Journal of Heat and Fluid Flow, 30,
flume tests of riprap-apron performance at a bridge abut- 286–305.
ment on a floodplain. Journal of Hydraulic Engineering, 134, Simpson, R. L. (2001). Junction flows. Annual Review of Fluid
800–809. Mechanics, 33, 415–443.
Paik, J., Escauriaza, C., & Sotiropoulos, F. (2007). On the Smith, H. D., & Foster, D. L. (2007). Three-dimensional
bimodal dynamics of the turbulent horseshoe vortex sys- flow around a bottom mounted short cylinder. Journal of
tem in a wing-body junction. Physics of Fluids, 19(4), Hydraulic Engineering, 133, 534–544.
045107. Sohankar, A. (2007). Heat transfer augmentation in a rectangu-
Paik, J., Sotiropoulos, F., & Porte-Agel, F. (2009). Detached lar channel with a vee-shaped vortex generator. International
eddy simulation of flow around two wall mounted cubes in Journal of Heat and Fluid Flow, 28, 306–317.
Downloaded by [Australian National University] at 03:35 09 January 2015

You might also like