Vous êtes sur la page 1sur 11

Journal of Hydraulic Research Vol. 44, No. 5 (2006), pp.

663–673
© 2006 International Association of Hydraulic Engineering and Research

Scour due to a horizontal turbulent jet: Numerical and experimental investigation


Affouillement dû à un jet turbulent horizontal: Recherche numérique et
expérimentale
CLAUDIA ADDUCE, Post-Doc, Dipartimento di Scienze dell’Ingegneria Civile, Università Roma Tre, via Vito Volterra 62,
00146, Rome, Italy. Tel.: +390655173468; fax: +390655173441; e-mail: adduce@uniroma3.it. (author for correspondence)

GIAMPIERO SCIORTINO, Professor, Dipartimento di Scienze dell’Ingegneria Civile, Università Roma Tre, via Vito Volterra 62,
00146, Rome, Italy. Tel.: +390655173448; fax: +390655173441; e-mail: sciorti@uniroma3.it.

ABSTRACT
In this paper both numerical and experimental investigations of local scour downstream of a sill followed by a rigid apron are presented. Nine laboratory
experiments were carried out in clear water scour conditions, with different values of discharge. At the end of each run, velocity measurements both
on the apron and on the scour hole were performed by ultrasonic Doppler velocimetry. A mathematical-numerical model was developed, simulating
Downloaded by [IAHR ] at 07:23 17 January 2012

local scour downstream of a sill followed by an apron. The model uses information related both to the measured velocity fields and to the physical and
mechanical properties of the sand constituting the mobile bed. The mathematical structure of the model consists of a second order partial differential
parabolic equation whose unknown is the shape of the mobile bed. The numerical integration of this nonlinear equation, with suitable boundary
conditions, is in agreement with the measured scour profiles at the end of the run. Upon comparing experimental and numerical data, a similar
temporal evolution of the maximum scour depth is observed.
RÉSUMÉ
Cet article présente des recherches numériques et expérimentales sur l’affouillement local en aval d’un seuil prolongé par un tablier rigide. Neuf
expériences de laboratoire ont été effectuées en eaux claires, avec des débits différents. À la fin de chaque essai, des mesures de vitesse sur le
tablier et sur le trou affouillé ont été exécutées par vélocimétrie Doppler ultrasonique. Un modèle mathématique numérique a été développé, simulant
l’affouillement local en aval d’un seuil prolongé par un tablier. Le modèle utilise à la fois l’information des champs de vitesses mesurées et les
propriétés physiques et mécaniques du sable constituant le lit mobile. La structure mathématique du modèle se compose d’une équation parabolique
aux dérivées partielles du second ordre dont l’inconnue est la forme du lit mobile. L’intégration numérique de cette équation non-linéaire, avec des
conditions aux limites appropriées, est en accord avec les profils d’affouillement mesurés après les essais. La comparaison des données numériques
et expérimentales, montre une évolution temporelle semblable de la profondeur maximum d’affouillement

Keywords: Scour, velocity measurements, UDV, 2D horizontal jet, mathematical model.

1 Introduction aim was the development of empirical formulae predicting the


maximum (or equilibrium) scour depth and length. Schoklitsch
Both turbulent plunging and horizontal jets can arise when a flow (1932), Kotulas (1967) and Bormann and Julien (1991) pro-
interacts with hydraulic structures such as underflow gates, cul- posed formulae for scour due to plunging jets, Qayoum (1960),
verts, large dams and stilling basins (Farhoudi and Smith, 1985; Altinbilek and Basmaci (1973), Farhoudi and Smith (1985),
Hassan and Narayanan, 1985; Fritz and Hager, 1998). When Breusers and Raudkivi (1991), Hoffmans (1997) and Chatterjee
a jet impinges on a mobile bed it lifts the sediments, which et al. (1994) suggested relations for scour downstream of 2D hor-
are transported downstream, and a scour hole is formed. Local izontal jets. A comparative review of the existing formulae can
scour downstream of hydraulic structures is a relevant prob- be found in Hoffmans and Verheij (1997). Empirical relations are
lem, given its significant practical value. Scour can endanger based mainly on small-scale laboratory studies. They have “to be
the stability of structures and cause the risk of failure if the foun- viewed with caution” if applied to prototypes (Graf, 1998). The
dations are not designed taking into account the maximum scour possibility of developing mathematical-numerical models pre-
depth. The study of a local scour process is a complex problem, dicting scouring processes is becoming more and more attractive
given the numerous variables related to both the heterogeneity and several authors (see Hogg et al., 1997; Istiarto, 2001; Jia
of the eroded soil and the turbulent flow producing the erosion. et al., 2001; Karim and Ali, 2001; Salehi Neyshabouri et al.,
Many experimental investigations have been carried out whose 2003) have recently focused on the numerical simulation of local

Revision received July 26, 2005/Open for discussion until October 31, 2007.

663
664 Adduce and Sciortino

scour. These models, tested by laboratory experiments, may lead whose unknown is the shape of the mobile bed. The coupling
to a better understanding of this complex phenomena. Hogg et al. between the shape of the eroded bed and the characteristics of
(1997) developed an analytical model simulating local scour of the 2D horizontal jet make this problem highly complex. The
an initially flat bed of grains by a two-dimensional turbulent wall developed model requires some empirical assumptions concern-
jet. They calculated the steady-state profile by applying critical ing the modelling of the bed shear stress. In particular, we assume
conditions along the bed surface for the incipient motion of a that the source of the jet momentum is close to the bed and the
particle. The temporal evolution of the scour hole was obtained bed shear stress can be described by a Gaussian-like law (see
by integrating a sediment-volume conservation equation. The Hogg et al., 1997), when the shape of the bed is not flat. Numeri-
predicted profiles were compared with experimental studies by cal integration of this nonlinear equation with suitable boundary
Rajaratnam (1981). Karim and Ali (2001) applied the FLUENT conditions, namely absence of solid discharge both at the inlet
CFD package, testing three different closure models (Standard and at the end of the mobile bed agrees with the laboratory mea-
k–M model, the Reynolds Stress Model and the renormalization sured scour profiles. The absence of solid discharge at the end of
Group Theory-Based Model) to simulate the flow field gener- the mobile bed was experimentally observed in all the tests. A
ated by a turbulent water jet impinging on a rigid horizontal similar temporal evolution of the maximum scour depth is also
bottom and scoured beds. Istiarto (2001) developed a 3D numeri- observed when the laboratory and numerical data are compared.
cal model to simulate the flow around a cylinder on a scoured bed.
The Reynolds-averaged Navier–Stokes and continuity equations
for incompressible flow were used, together with a k–ε turbu- 2 Experimental set-up
lence closure model. Jia et al. (2001) simulated the local scour
due to a 2D plane impinging jet in a plunge pool with a loose The experiments were conducted at the Hydraulics Laboratory
of the University of Rome “RomaTre”, in a 17 m long, 1 m
Downloaded by [IAHR ] at 07:23 17 January 2012

bed. The sediment transport model of Nakagawa and Tsujimono


(1980) was modified by including the effects of shear stress and high and 0.8 m wide flume with a rectangular cross-section.
a fluctuating lift force, the latter was related to pressure fluctu- Scouring tests were carried out in a 0.3 m high, 0.8 m wide and
ations by using empirical functions. The flow field was solved L = 3 m long sediment recess section, positioned 7 m down-
by CCHE3D, a finite element-based unsteady 3D model, with a stream of the flume inlet and created artificially by raising the
k–ε turbulence closure model. Salehi Neyshabouri et al. (2004) flume bed. In order to have a mobile bed, a uniformly graded
presented a numerical simulation of local scour of a flat bed due sand of mean diameter d50 = 0.72 mm, d90 = 0.96 mm, a den-
to a 2D free falling jet. The momentum equations, the continuity sity ρs = 2650 kg/m3 and a porosity p = 0.44, was used to
fill the sediment recess section. The geometric standard devia-
equation and the k–ε equations for turbulent flows were applied. √
First the turbulent flow due to a free falling jet was computed, tion of the sand σg = d84 /d16 = 1.21 was smaller than the
then the distribution of the sand concentration was determined threshold proposed by Breusers and Raudkivi (1991) for the def-
and the scoured bed was predicted. inition of nonuniform grading (σg = 1.35). The same sand was
Making use of both laboratory and numerical experiments, glued both to the upstream and downstream fixed-bed sections,
the present paper investigates local scour downstream of a sill to produce a bed of homogeneous roughness. No sediments were
followed by a rigid apron. No sediments are supplied from the supplied from the upstream into the scour zone and the experi-
upstream into the scour zone and the experiments are conducted ments were conducted under clear water scour. A sill of height
under clear water scour. Nine laboratory experiments with local Dp = 0.15 m, followed by a Lp = 0.5 m long rigid apron, was
scour due to a 2D horizontal jet are conducted. Measurements of positioned upstream of the test section, as shown in Fig. 1. A con-
both the temporal evolution of the scour hole and its geometry at trol gate, positioned at the end of the flume, was used to vary the
the end of the run are performed. Ultrasonic Doppler velocime- downstream water depth. The water level in the flume was mea-
try is used to measure the velocity field on the rigid apron and sured by a resistive point gage of accuracy ± 0.10 mm, mounted
on the scour hole at the end of the run. A 1D mathematical- on a carriage which moved along the flume. The water discharge
numerical model is developed, simulating the temporal evolution at the inlet, controlled by an inlet valve, was measured with an
of the scour profile due to a 2D horizontal wall jet. The model electromagnetic flowmeter.
uses the equation of continuity for the local solid discharge and Before starting the experiments, the level of the mobile sand
physical and mechanical properties of the sand. The analysis bed, positioned in the sediment trap, was set to the same height
of the forces acting on a sand particle is performed taking into
account the angle of repose of the sand, the instantaneous shape
of the mobile bed and a model for the shear stress, following
an approach similar to that of Hogg et al. (1997). The proposed
model describes the progressive erosion of an initially flat bed
of cohesionless grains, but unlike the model proposed by Hogg
et al. (1997), the scour profile at the end of the run can be com-
puted only by the complete temporal reconstruction of the scour
profile. The mathematical model consists of a second order par-
tial integro-differential equation with a parabolic-like structure Figure 1 Definitions sketch for model tests: side view.
Scour due to a horizontal turbulent jet 665

a signal consisting of ultrasonic pulses travelling across the fluid


and backscattered by the targets, i.e. small air bubbles or particles,
moving with the flow. Evaluating the elapsed time between the
emission of the pulse and its reception the position of the target is
determined. The received backscattered pulses are shifted in fre-
quency, due to the Doppler effect, so the velocity component of
the flow along the probe axis can be evaluated by this frequency
shift. For a detailed description of the UDV technique see Takeda
(1995, 1999). In the present study pulses were repeated with a
pulse repetition frequency (PRF) of 2200 Hz and the number of
emissions used to compute one profile was 40, corresponding
Figure 2 Definition sketch for model tests: top view. to a final sampling frequency of about 15 Hz. The size of the
measurement volume depends on the characteristics of the probe
as the fixed bed. To avoid undesirable scour of the sediment bed, (emitting frequency and diameter) and increases far from the
the control gate was completely closed and the flume was slowly probe, due to the lateral spreading of the pressure wave. Probes
filled with water pumped from the laboratory water reservoir. with an emitting frequency of 2 MHz and with a diameter of
The water discharge was then slowly increased and set to the 14 mm were used, having a divergence angle equal to 1.83◦ . The
desired experimental value, to avoid any movement of the sand UDV allows a quasi-instantaneous measurement of the veloc-
bed. The experiments started when the downstream water depth ity in different volumes (gates), positioned along the axis of the
was set to the desired experimental value, by opening the con- probe. The distance between the centres of these gates in the
Downloaded by [IAHR ] at 07:23 17 January 2012

trol gate. The water discharge was kept constant for the entire experiments was 3.52 mm. The longitudinal (parallel to the probe
duration of the experiment. The evolution of the scour hole was axis) size used for the sampling volume was 2.96 mm, while its
recorded by a CCD camera connected to a digital videocassette lateral (orthogonal to the probe axis) size depended on the dis-
recorder. A detailed study of the temporal evolution of the scour tance from the transducer. For these experiments it was between
hole was made possible by an image analysis technique (see 1.4 mm (close to the transducer) and 20 mm (close to the bottom).
Adduce, 2004). The bed profile was measured in five different Three probes with the same emitting frequency were used, see
longitudinal sections of the flume (see Fig. 2) by a point gage of Longo et al. (2001) to measure the profiles of the horizontal and
accuracy ±0.10 mm. These five longitudinal profiles were called the vertical velocity component.
R, CR, CC, CL and L and were positioned at a distance from the Nine experimental tests were performed, characterized by dif-
right wall of the flume of 0, 20, 40, 60 and 80 cm, respectively, ferent values of the water flow-rate (Q), of the water depth on
as shown in Fig. 2. the sill (hs ) and of the water depth downstream of the sill (h0 ), as
We stopped the experiments when the measured maximum shown in Table 1. Table 1 also reports the maximum scour depth
values (Dsmax ), the maximum scour length (Lsmax ), Froude num-
scour depth, plotted versus log time was sensibly constant, as √
ber at the sill,Frs = Q/(bhs ghs ), downstream Froude number,
it will be described in the following. At the end of the exper- √
iment velocity measurements were performed by an ultrasonic Fr0 = Q/(bh0 gh0 ) and duration (T ) for each test.
Doppler velocimeter (UDV; Signal-Processing DOP2000). UDV,
first developed for medical studies on blood flow (Willemetz,
3 Laboratory measurements
1992), is today used to perform measurements both in transpar-
ent (Hurther, 2001; Ito et al., 2001) and optically opaque fluids
3.1 Scour measurements
and in high temperature fluids (Eckert and Gerbeth, 2002), giving
spatio-temporal velocity information. The UDV is constituted by Figures 3 and 4 show scour profiles measured at the end of the run
a piezoelectric probe, which works both as emitter and receiver of for two different water discharges, corresponding to test 6 and

Table 1 Experimental parameters

Test Q (l/s) hs (m) h0 (m) Dsmax (m) Lsmax (m) Frs Fr0 T (min)

1 11.0 0.022 0.126 0.042 0.350 1.39 0.10 500


2 12.3 0.023 0.129 0.044 0.375 1.46 0.11 580
3 14.2 0.024 0.134 0.050 0.525 1.53 0.12 500
4 16.2 0.026 0.137 0.061 0.600 1.56 0.13 480
5 18.4 0.028 0.143 0.067 0.630 1.61 0.14 580
6 20.5 0.029 0.145 0.069 0.650 1.66 0.15 660
7 24.8 0.032 0.151 0.077 0.750 1.77 0.17 570
8 26.3 0.033 0.156 0.082 0.800 1.79 0.17 450
9 31.2 0.036 0.163 0.107 0.950 1.82 0.19 540
666 Adduce and Sciortino

Figure 3 Scour profiles measured in different longitudinal sections


for test 6.

Figure 6 Dimensionless scour profiles measured at the end of each test.

maximum scour depth, and Lsmax , the maximum scour length,


following Gaudio and Marion (2003).

Figure 4 Scour profiles measured in different longitudinal sections 3.2 Mean velocity field
for test 9.
The mean velocity field, measured by UDV at the end of the run,
Downloaded by [IAHR ] at 07:23 17 January 2012

the bed profile and the free water surface profile, all measured in
test 9, respectively. In each figure five different longitudinal scour the centre of the flume, are shown in Fig. 7. The horizontal and
profiles are presented. Following the reference system shown in vertical distances are divided by the water depth on the sill, hs .
Fig. 2, the five longitudinal profiles are called R, CR, CC, CL Due to instrument configuration, the data cannot be obtained
and L, corresponding to distances from the right wall (looking in regions closer than 4 cm to the water surface. For the same
downstream) of the flume, along the y axis, of 0, 20, 40, 60 reason it was not possible to perform velocity measurements in
and 80 cm (left wall), respectively. The eroded bed is always a region located approximately 35 cm upstream of the end of
symmetrical with respect to the centre line of the flume (Figs 3 the rigid apron. Velocity profiles were measured in the centre of
and 4), while it is also quasi-2D for large discharges (Fig. 4). the flume in 16 different positions (x = −35, −25, −15, −5,
In Figure 5, the self-similarity at different times of the mea- 5, 15, 25, 35, 45, 55, 65, 75 and 85 cm) following the refer-
sured dimensionless scour profiles is shown for test 8. Six ence system as shown in Fig. 1. The jet created by the water
different instantaneous dimensionless scour profiles are shown flowing over the sill first impinges on the free surface and then
at 1, 4, 16, 66, 260 and 501 min, respectively. The vertical and moves onto the rigid apron, positioned immediately downstream
horizontal length scales used are Ds , the instantaneous maximum of the sill, with a wall jet-like velocity distribution, as observed
scour depth, and Xs , the instantaneous abscissa of the maximum by Fritz and Hager (1998). The wall jet-like velocity distribu-
scour depth. tion on the apron was also observed by several investigators in
In Fig. 6 the self-similarity for different runs of the dimension- different experimental configurations (see Chatterjee and Gosh,
less scour profiles, measured at the end of each run, is shown. 1980; Hassan and Narayanan, 1985). A backward flow close to
The vertical and horizontal length scales used are Dsmax , the the water surface and due to the presence of an hydraulic jump

Figure 5 Dimensionless instantaneous measured scour profiles for


test 8. Figure 7 Mean velocity field for test 4.
Scour due to a horizontal turbulent jet 667

is shown in Fig. 7. The backward flow becomes stronger as the


distance from the sill is decreased. The flow which moves over
the eroded bed, immediately downstream of the rigid apron, has
a free jet-like velocity distribution, due to a sudden change in the
bed depth. This free jet-like velocity distribution was observed
also by Hassan and Narayanan (1985), and the maximum veloc-
ity decreases as the distance from the apron increases. The flow
over the dune has a slightly upward movement, due to the upward
slope of the eroded-deposed bed.

4 Mathematical model
Figure 8 Sketch defining the unit vectors τf and nf .
Local scour downstream of a sill followed by a rigid apron is
simulated in this paper by a 1D mathematical model. The equa- τf has the same direction of the mean flow close to the bottom,
tion of continuity for the local solid discharge together with the see Fig. 8.
physical and mechanical properties of the sand are used to define Hence
the mathematical model. The experimental data on the spatial ∇(zb(x, t) − z) zbx i − k
development of the eroding wall jet are used to calibrate the nf = = (3)
|∇(zb(x, t) − z)| 1 + zbx2
mathematical model. In the analysis of the forces acting on a sand
i + zbx k
particle we considered the angle of internal friction, the instan- τf = 
Downloaded by [IAHR ] at 07:23 17 January 2012

(4)
taneous shape of the mobile bed and the shear stress obtained 1 + zbx2
by the hypothesis of a wall jet, according to the experimental where zbx is the space derivative of zb(x, t). In this case τcr is
results. During the local scouring process the wall jet-like pro- given by
file on the mobile bed changes (Hoffmans and Verheij, 1997), as Wg φ
does the shear stress related to it, as described by recent investiga- τcr = Wg · nf φ = (−Wg k) · nf φ =  . (5)
1 + zbx2
tions (Dey and Westrich, 2003). Hogg et al. (1997) supposed that
during the scouring process the bed shear stress changed accord- From equation (1)
ing to a Gaussian-like characteristics. This approach differs from τ̄cr
τcr =  (6)
the model proposed by Salehi Neyshabouri et al. (2003), who 1 + zbx2
described the evolution of the mobile bed by a direct resolution
where τ̄cr is the critical shear stress for a flat bottom. The stress
of the hydrodynamic field and solved the turbulence effects by a
acting on a grain is obtained by adding the hydrodynamic bed
k–ε model. Both our model and Salehi Neyshabouri et al. (2003)
shear stress and stress due to the gravity, τg as
model assume a bed shear stress dynamic described by the wall
log-layer. τ = τidr + τg = τidr + (−Wg k) · τf (7)
The condition of transport, for a horizontal bottom, can be
or
written as:
τ̄cr zbx
τ = τidr −  (8)
τidr > τ̄cr = Wg φ (1) φ 1 + zbx2
where The condition of incipient motion is in this case

φ = tan (ϕ) (2) τ = τidr + τg > τcr (9)

Wg is the submerged weight of the particle divided by a charac- or


teristic particle area ≈ D2 , τidr is the hydrodynamic bed shear τ̄cr (φ + zbx )
τidr > τcr − τg =  (10)
stress, τ̄cr = const · (γs − γ) · d50 (Shields, 1936), γs = ρs g, g φ 1 + zbx2
is the acceleration of gravity, γ = ρg, ρs is the sediment density, Equation (10) reduces to Eq. (1) for zbx → 0, i.e. a horizontal
ρ is the water density and ϕ is the angle of internal friction. In bottom. In Fig. 9 a sketch of stresses acting on a grain for an
the following the value of constant = 0.047, because the Meyer- incipient motion is shown.
Peter and Muller bed load formula will be used (see Simons and In order to model the volumetric solid discharge for unit width
Senturk, 1992). Qs , we suppose, from (10), a functional relation as
If the bottom is not flat and is described by the equation z =
τ̄cr (φ + zbx )
zb(x, t), the transport conditions have to be modified to account Qs = Qs (ψ), ψ ≡ τidr −  (11)
for the effect of the bottom slope. In the following description nf φ 1 + zbx2
and τf are the unit vectors, respectively, orthogonal and tangent We assume that the solid transport is due mainly to Qs , i.e. the
to the bottom, with nf · k < 0, where k is an upward unit vector suspended load is negligible, consequently the velocity of the
parallel to the z axis, i is the unit vector parallel to the x axis and solid particles is locally parallel to the bottom. The continuity
668 Adduce and Sciortino

flume cross-section. Given the above assumptions it follows that



δ/ν = Ni χ/(Um g), hence

αρUm2
τidr = 1/4 √ . (17)
Ni (χ/ g)1/4
We use the following maximum velocity decaying law (see
Schlichting, 1979; Hogg et al., 1997):
 k
M ν2
Um = C1 (18)
ν M(x + Lp )
which for k ∼
= 0.5 (see Schlichting, 1979; Hogg et al., 1997) is
Figure 9 Sketch defining the stresses acting on a grain.

M
Um = C1 (19)
equation for the solid material can be written as (see Graf, 1971, x + Lp
1998): where
  h0
∂zb 1 ∂ M= u2 dz
+ (Qs τf · i) = 0 (12) (20)
∂t 1−p ∂x z

C1 is a constant, which can be calculated by the measured wall jet


or
velocity profile at the end of the rigid apron, h0 is the undisturbed
Downloaded by [IAHR ] at 07:23 17 January 2012

   
∂zb 1 ∂ Qs water depth on the mobile bed, and u is the local velocity compo-
+  =0 (13) nent along the longitudinal direction of the flume. If we assume
∂t 1−p ∂x 1 + z2x
h + zb ∼= h0 , as observed experimentally, where h is the instanta-
neous water depth and use the approximation M ∼ = q2 /(h0 − zb),
Hence using (11) the following equation holds
where q is the unit discharge, Eq. (17) becomes, with β = (αC12 )
   
∂zb 1 Qs ∂τidr τ̄cr (1 − φzbx ) ∂2 zb ρβq2
+  −  3/2 2 τidr = √ (21)
∂t 1−p 1 + zbx2 ∂x φ 1 + zbx2 ∂x 1/4
Ni (χ/ g)1/4 (x + Lp )(h0 − zb)

Qs ∂zb ∂2 zb This relation reasonably approximates the bed shear stress at a
− 3/2 =0 (14) certain distance from the initial profile of the jet for the case
1 + zb2 ∂x ∂x2
x
of a fixed bottom. During the scouring process the wall-jet like
where  denotes the derivative with respect to the variable ψ and profile on the mobile bed changes and the shear stress related to
p is the porosity of the sediments. The equation is nonlinear and it has been object of recent investigations (see Dey and Westrich,
it has a parabolic structure. 2003). At the beginning of the scouring process, the mobile bed
At a distance δ from the wall, the hydrodynamic bed shear at the downstream edge of the rigid apron starts to be eroded. As
stress, caused by a turbulent 2D horizontal wall jet, can be a consequence the wall jet, moving first onto the rigid apron and
modelled following Schlichting (1979) as then onto the eroded bed, changes into a jet like a free jet, whose
transport action is very weak (see Fig. 7). This free jet resembles
αρUm2 a wall jet after a length scale on the order of one half of the
τidr = 1/4
(15) maximum scour length. This length scale is very short during the
Um (δ/ν)1/4
first period of the scouring process and becomes longer as the time
where α = 0.0283 is a numerical coefficient, Um is the maximum advances. When the aspect ratio of the eroded profile is not small,
velocity of the jet and ν is the kinematic viscosity of the fluid. In as in the last stage of the scouring process, the hypothesis of a
order to estimate the transport action at the moving bottom, δ is shear stress like (21) is a crude approximation. This is because
chosen as the viscous sublayer thickness. We assume the velocity profile is no longer like a wall jet, since the more
the aspect ratio increases the less flat the bed becomes. (Here the
Ni · ν · χ aspect ratio is defined as the ratio of depth to streamwise extent.)
δ= √ (16)
Um · g For this reason we follow a phenomenological approach, like
Hogg et al. (1997) and suppose that the bed shear stress evolves
valid for a uniform and turbulent rough flow. In Eq. (16), following a law obtained by multiplying Eq. (21) by a Gaussian
Ni = 11.5 is the Nikuradse number, assuming δ as a length function G(x, z). In case of absence of scour, this function is
scale, Um is the maximum velocity of the wall jet and is used G(x, 0) = 1, while it has to simulate the transition from a free-

as velocity scale and χ = 2.49 g ln(13.3Rfr /d50 ), assuming like jet to a wall-like jet in a region between the downstream
a wall log-layer and where d50 is the mean diameter of the edge of the apron and the maximum scour abscissa, as described
sand, R is the hydraulic radius and fr is the shape factor of the before. The Gaussian function varies the transport action, which
Scour due to a horizontal turbulent jet 669

is weak due to a free-like jet and becomes stronger as the free-like The latter gives the abscissa of the instantaneous maximum scour
jet changes into a wall-like jet. We assume depth as a functional of the scour profile and it can be used in
 Eq. (22) to define G(x, z).


1 if x > Xs or in absence of scour
 
 2   Similarly to Hogg et al. (1997), we assume that the shear stress
G(x, z) = Exp − x−X s
ln ς1 if x ≤ Xs (22) is described by


Xs

and in presence of scour ρβq2
τidr = 1/4 √ G(x, z) (29)
and G(0, z) = ζ, where 0 < ζ < 1 is a calibration parameter of Ni (χ/ g)1/4 (x + Lp)(h0 − zb)
the model and considers that the transport action due to a free-
like jet is weaker than the transport action due to a wall-like jet, The bed shear stress asymptotic decaying law (∼ 1/x) for a flat
in the region downstream of the rigid apron. The abrupt change bed is similar to the law proposed by Hogg et al. (1997), while the
between the rigid bed of the apron and the erodible bed implies the definition of the function G(x, z) is different. Equation (14) has
necessity to calibrate ζ in order to obtain an agreement between the structure of a partial integro-differential equation, because
the measured and simulated scoured depths close to the edge of G(x, z) depends on z with an integral relation.
the apron. The Meyer-Peter and Muller transport formula can be used
In the following we define the abscissa of the instantaneous (see Hogg et al., 1997) for the solid transport, using ψ
maximum scour depth as a functional of the scour profile. We as independent variable (instead of τ − τcr for a flat bed),
assume, as experimentally detected (see Fig. 5), that the local given by
scour is self-similar, if Ds (t) and Xs (t) are used as the vertical and 8
horizontal length scale, respectively, where Ds (t) is the maximum Qs (ψ) = √ ψ3/2 η(ψ) (30)
ρ(γs − γ)
Downloaded by [IAHR ] at 07:23 17 January 2012

scour depth at a time t and Xs (t) is the instantaneous abscissa of


the maximum scour depth. Hence including the no transport condition η(ψ), which is defined by
  the Heaviside step function
x
zb(x, t) = Ds (t)f (23)
Xs (t) 
1 if ψ ≥ 0
It follows that η(ψ) = (31)
0 if ψ < 0
Ls Ls /Xs
A= zb dx = Ds Xs f(η)dη =Ds Xs C2 (24) The potential law expressed by Eq. (30) is a continuous function
0 0
together with its derivatives, because the exponent (3/2) > 1.
where A is the scoured area, Ls is the instantaneous maximum
scoured
 Llength, η = (x/(Xs )) is the dimensionless abscissa and
/X
C2 = 0 s s f(η)dη, is a constant linked to the self-similarity of 5 Numerical integration of the model
the local scour profile.
We define Numerical integration of the mathematical model was per-
L L
formed by applying a MacCormack-like scheme (see Garcia and
A2 = zb2 dx = Ds2 Xs f 2 (η)dη = Ds2 Xs C3 (25)
0 0 Kahawita, 1986) to Eq. (13), the conservative form of Eq. (14).
where C3 is a constant linked to the self-similarity of the scoured The MacCormack scheme is structured in two time steps, i.e.
profile. a predictor and a corrector step, each one of them applied to
From Eqs (24) and (25) it follows that the conservative form (Abbott, 1979) of the differential equation
to be solved. The conservative form assures the correct discrete
C3 A2 A2 form of the conservation laws, generally expressed by integral
Xs = 2
=C (26)
(C2 ) A2 A2 balance to control volume, or through differential equations in
where C = C3 /(C2 )2 is a universal constant, which is a control conservation law form (see Abbott, 1979). The MacCormack
parameter of the mathematical model and can be experimentally scheme is applied to the conservative form of Eq. (14), expressed
detected. by Eq. (13). As a consequence, both in the corrector and pre-
Because all the experiments were conducted without sediment dictor step, the conservation law of solid discharge is correctly
supply from the upstream into the scour zone, i.e. clear water imposed on each computational cell. A more detailed discretiza-
scour, and if we assume a 2D bed profile, the scoured area is tion should be required to obtain the same precision, applying
equal to the dune area, hence the same numerical scheme to Eq. (14). Two difference opera-
Ls tors, denoted respectively by B and F, backward and forward,
1 L were defined as
A= zb dx = |zb|dx (27)
0 2 0
zbin − zbi−1
n
and Bzbin ≡ (32)
  2 x
1 L
2 0 |zb|dx n
zbi+1 − zbin
Xs = C L (28) Fzbin ≡ (33)
zb2 dx x
0
670 Adduce and Sciortino

where
zbin ≡ zb|x=ix,t=nt (34)
x = L/Mint is the spatial step of integration, Mint is the number
of intervals of discretization and t is the temporal step of inte-
gration, in this study Mint = 100 and t = 0.02 s. The following
predictor–corrector numerical scheme is defined according to the
following integration scheme
   
1 Qs
zbi = zbi − t
P n
B  (35)
1−p 1 + Fzbin
 
 
1  Qs 
zbiC = zbiP − t F   (36)
1−p 1 + BzbiP

1 n 
zbin+1 = zbi + zbiC (37)
2
where C and P denote the corrector and predictor step,
respectively.
This algorithm is completely defined when suitable boundary
conditions at x = 0 and x = L are assigned. These boundary
Downloaded by [IAHR ] at 07:23 17 January 2012

conditions are the clear water scour condition at the inlet of the
mobile bed and the absence of solid transport at the end of it.
From Eq. (10) the condition of absence of solid transport at
the end of the mobile bed can be written as
τidr (z)
=λ<1 (38)
τeff (zx )
where τeff = τcr − τg (see Eq. (10)) and λ is a parameter of the
model, whose value has to be suitably chosen.
The condition of clear water scour at the inlet of the mobile
bed can be discretized as
τidr (zb0n+1 )
= λ0 (39)
τeff ((zb1n+1 − zb0n+1 )/x)
here zb1n+1 is known, because it is calculated at an internal grid
point, hence in Eq. (39) the only unknown is zb0n+1 , and (39) can
be numerically solved for the assigned λ0 values, by a trial and
error process.
The condition of absence of solid transport at the end of the
mobile bed can be discretized as
n+1
τidr (zbM )
= λM (40)
τeff ((zb1n+1 − zb0n+1 )/x) Figure 10 Predicted (line) and laboratory measured (dots) scour profiles
Equation (40) is numerically solved as Eq. (39), to find the for test 4, test 5, test 6, test 7, test 8, and test 9, respectively.
n+1
unknown zbM .

with increasing discharges (see Fig. 10a–f ). For smaller dis-


6 Application of the model and discussions charges a small dune with a sharp dune crest follows the scoured
hole. By increasing the discharge the dune crest becomes flatter
In Fig. 10(a–f) the comparison between the measured laboratory and longer.
bed profiles (dots) and the numerically simulated profiles (lines) Figure 10 shows a good agreement between the predicted and
is presented. All the bed profile measurements refer to the centre the measured scour profiles for all the simulated tests. The math-
of the flume at the end of the run. ematical model, as a consequence of the boundary conditions,
Six different tests are simulated: test 4 (Fig. 10a), test 5 predicts at any instant a bed profile in which the volume of the
(Fig. 10b), test 6 (Fig. 10c), test 7 (Fig. 10d), test 8 (Fig. 10e) and scoured bed is strictly equal to the volume of the dune. When
test 9 (Fig. 10f ). The maximum scour depth and length increase the model simulates the dune profile downstream of the scour
Scour due to a horizontal turbulent jet 671

hole some differences arise between the simulated and measured


scour profiles. For large discharges (tests 7–9) the dune pro-
files predicted by the model are similar to the measured profiles,
while when the discharges are smaller (tests 4–6), the dunes pre-
dicted by the model are always longer than those measured. This
behaviour can be explained by the fact that the simulated scour
profiles are 2D, while the measured scour profiles are quasi-2D
when the discharges are large (tests 7–9) and symmetrical with
respect to the centre line of the flume but no longer quasi-2D
for smaller discharges (tests 4–6), as experimentally detected
(see Figs 3 and 4). As the discharge increases, the scour profile Figure 13 Time evolution of dimensionless predicted (line) and labo-
becomes more quasi-2D. ratory measured (dots) maximum scour depths for test 4.
In Fig. 11 the ratio τidr /τcr is plotted versus x for a flat bed and
for four different tests (test 4, test 6, test 7 and test 9, respectively).
Velocity profiles measured at the end of the apron were used to
calculate the values of constant C1 for each test using Eq. (19).
These values of C1 were used in relation (21) to determine τidr .
Nowadays a knowledge of the wall jet velocity distribution close
to the edge of the apron is necessary to run the model. A more
detailed experimental study could furnish a functional relation
Downloaded by [IAHR ] at 07:23 17 January 2012

between the constant C1 and some characteristics of the sill and


of the flow, avoiding a direct measurement of the velocity field.
The intersection of each curve, τidr /τcr , with the horizontal unit-
line (τcr /τcr = 1) lies near the end of the corresponding dune (see Figure 14 Time evolution of the dimensionless predicted (line) and
Fig. 10a, c–e) and confirms that the erosive action of the jet, at laboratory measured (dots) maximum scour depths for test 8.
least for a flat bed, stops at a distance close to the end of the dune.
In Fig. 12 the temporal evolution of the scour profile (test 8) 830 min, respectively. As experimentally detected, the time evo-
simulated by the developed mathematical-numerical model is lution of the scour hole as its length and depth increase during
shown. Twelve different instantaneous simulated scour profiles the first minutes of the simulated test is very fast. As the time
are shown, at 3, 8, 17, 25, 42, 58, 75, 92, 175, 258, 341 and advances, the scour hole develops more slowly until it reaches
its maximum scour depth; while the dune continues to develop
together with the scour length, which increases very slowly.
In Figs 13 and 14 a comparison of the temporal evolution of
the experimental (filled circles) and numerical maximum scour
depths is presented. Measurements of maximum scour depth
temporal evolution were performed by an image analysis tech-
nique. Figures 13 and 14 confirm the fact observed by Rajaratnam
(1981) about how erosion increases linearly with the logarithm
of time up to a certain time. Both the experimental and numer-
ical instantaneous maximum scour depths (Ds ) are divided by
the maximum scour depth at the end of the test (Dsmax ). The
numerical model agrees with the measured time development of
the maximum scoured depth. Figures 13 and 14 refer to tests 4
and 8, respectively.
Figure 11 Dimensionless bed shear stress due to a 2D turbulent wall
jet over a flat bed for test 4, test 6, test 7 and test 9, respectively.
7 Conclusions

A laboratory and a numerical model were presented to study


local scour downstream of a sill, followed by a rigid apron, in
clear water condition. Velocity measurements by UDV were per-
formed both on the rigid apron and in the scour hole at the end of
the run. The flow field on the apron shows a wall-jet like veloc-
ity distribution, while immediately downstream of the apron it
has a free jet-like velocity distribution. A 1D numerical model,
Figure 12 Predicted time development of scour profile for test 8. simulating the evolution of the scour hole was developed. The
672 Adduce and Sciortino

hypothesis, presented in the mathematical model, of the tempo- T = Duration of experiments


ral self-similarity of dimensionless scour profiles is confirmed u = Horizontal velocity component
by laboratory measurements. Agreement between the measured Um = Maximum mean horizontal velocity com-
and the predicted scour profiles at the end of the run is observed. ponent of the wall jet
The agreement between the model predictions and the labora- w = Vertical velocity component
tory measurements is better for higher discharges, because in Wg = Submerged weight of the particle divided
this regime the measured scour profiles tend to be quasi-2D. The by a characteristic particle area
predicted temporal evolution of the maximum scour depth agrees x = Horizontal axis
with the measured one. Knowledge of the wall jet velocity dis- Xs = Abscissa of the instantaneous maximum
tribution close to the edge of the apron is necessary to run the scour depth
proposed model. For further development of this study, a more z = Vertical axis
detailed experimental investigation could furnish a functional zb(x, t) = Instantaneous elevation of the mobile bed
relation between the properties of the jet flow and the charac- zbx = Space derivative of zb(x, t)
teristics of the sill and of the flow, avoiding a direct measurement β = Coefficient of the mathematical model
of the velocity field. γ = Specific weight of water
γs = Specific weight of the sediment
δ = Viscous sublayer thikness
ζ = Calibration parameter of the mathematical
Notation model
η = Dimensionless abscissa
λ0 = Parameter of the boundary condition at the
Downloaded by [IAHR ] at 07:23 17 January 2012

A = Scoured area
A2 = Area of the scour profile squared inlet of the mobile bed
B = Backward difference operator λM = Parameter of the boundary condition at the
C1 = Coefficient of the mathematical model outlet of the mobile bed
C2 = Coefficient of the mathematical model ν = Kinematic viscosity
C3 = Coefficient of the mathematical model ρ = Water density
d50 = Mean sediment size ρs = Sediment density
Dp = Sill height σg = Geometric standard deviation of the sand
Ds = Instantaneous maximum scour depth τcr = Critical bed shear-stress
Dsmax = Maximum scour depth at the end of the test τ̄cr = Critical bed shear-stress for a flat bottom
f = Functional relation τeff = Difference between the critical bed shear-
fs = Shape factor of the flume cross-section stress and the stress due to the gravity
F = Forward difference operator τf = Unit vector tangent to the bottom
Frs = Froude number at the sill τg = Stress due to the gravity
Fr0 = Downstream Froude number τidr = Hydrodynamic bed shear-stress
g = Gravitational acceleration φ = Friction factor for a grain
G(x, z) = Gaussian function ϕ = Angle of repose of the sand
hs = Water depth on the sill χ = Resistance coefficient of Chezy
h0 = Downstream water depth x = Spatial step of the numerical integration of
i = Unit vector parallel to the x axis the mathematical model
k = Constant t = Temporal step of the numerical integration
k = Upward unit vector parallel to the z axis of the mathematical model
L = Length of the sediment recess section
Lp = Length of the rigid apron
Ls = Instantaneous maximum scour length
nf = Unit vector orthogonal to the bottom References
Ni = Nikuradse number
M = Momentum flux 1. Abbott, M.B. (1979). Computational Hydraulics, Elements
Mint = Number of interval of discretization in the of the Theory of Free Surface Flows, Chapter 5. Pitman,
model London, pp. 224–227.
p = porosity of the sediments 2. Adduce, C. (2004). “Local Scour Downstream of a Tur-
q = Water unit discharge bulent Jet”. PhD Thesis, University of Rome “RomaTRE”.
Q = Water discharge Roma, Italy, 20 October, 167 pp.
Qs = Volumetric solid discharge for unit width 3. Altinbilek, H.D. and Basmaci, Y. (1973). “Localized
at the bottom Scour at the Downstream of Outlet Structures”. Proceedings
R = Hydraulic radius of the 11th Congress on Large Dams, Madrid, pp. 105–121.
Scour due to a horizontal turbulent jet 673

4. Bormann, N.E. and Julien, P.Y. (1991). “Scour Down- Shedding and Empirical Eigenfunctional Analysis”. Exp.
stream of Grade-control Structures”. J. Hydraul. Engng. Fluids 31, 324–335.
ASCE 117(5), 579–594. 23. Jia, Y., Kitamura, T. and Wang. S.S.Y. (2001). “Simu-
5. Breusers, H.N.C. and Raudkivi, A.J. (1991). Scouring. lation of Scour Process in Plunging Pool of Loose Bed-
Balkema, Rotterdam. material”. J. Hydraul. Engng. 127(3), 219–229.
6. Chatterjee, S.S. and Gosh, S.N. (1980). “Submerged Hor- 24. Karim, O.A. and Ali, K.H.M. (2001). “Prediction of Flow
izontal Jet over Erodible Bed”. J. Hydraul. Div. ASCE 92(3), Patterns in Local Scour Holes Caused by Turbulent Water
1765–1782. Jets”. J. Hydraul. Res. 38(4), 279–287.
7. Chatterjee, S.S., Ghosh, S.N. and Chatterjee, M. 25. Kotulas, D. (1967). Das Kolkproblem im Rahmen
(1994). “Local Scour due to Submerged Horizontal Jet”. der Wildbachverbauung. Mitteil., Schweizer Anstalf f.
J. Hydraul. Engng. 120(8), 973–992. forsliches Versuchswesen, 43/1, Birmensdorf, CH.
8. Dey, S. and Westrich, B. (2003). “Hydraulics of Sub- 26. Longo, S., Losada, I.J., Petti, M., Pasotti, N. and
merged Jet Subject to Change in Cohesive Bed Geometry”. Lara, J.L. (2001). Measurements of Breaking Waves and
J. Hydraul. Engng. 129(1), 44–53. Bores through a USD Velocity Profiler. Technical Report
9. Eckert, S. and Gerbeth, G. (2002). “Velocity Measure- UPR/UCa_01_2001. University of Parma, E.T.S.I.C.C. y P.,
ments in Liquid Sodium by Means of Ultrasound Doppler Ocean & Coastal Research Group Laboratory, Universidad
Velocimetry”. Exp. Fluids 32(5), 542–546. de Cantabria, Spain, 2001.
10. Farhoudi, J. and Smith, K.V.H. (1985). “Local Scour Pro- 27. Nakagawa, and Tsujimono, T. (1980). “Sand Bed Insta-
files Downstream of Hydraulic Jump”. J. Hydraul. Res. bility due to Bed Load Motion”. J. Hydraul. Div. ASCE
23(4), 343–358. 106(12), 2029–2051.
Downloaded by [IAHR ] at 07:23 17 January 2012

11. Fritz, H.M. and Hager, W.H. (1998). “Hydraulics of 28. Qayoum, A. (1960). “Die Gesetzmäβigkeit der Kolkbil-
Embankment Weirs”. J. Hydraul. Engng. 124(9), 963–971. dung hinter unterströmten Wehren unter spezieller Beruck-
12. Garcia, R. and Kahawita, R.A. (1986) “Numerical Solu- sichtigung der Gestaltung eines beweglichen Sturzbettes”.
tion of the De SaintVenant Equations with the Mac-Cormack Dissertation, Technischen Hochschule carolo-Wilhelmina,
Finite-Diffrence Scheme”. Int. J. Numer. Methods Fluids 6, Braunschweig, Germany.
259–274. 29. Rajaratnam, N. (1981). “Erosion by Plane Turbulent Jets”.
13. Gaudio, R. and Marion, A. (2003). “Time Evolution of J. Hydraul. Res. 19, 339–358.
Scouring Downstream of Bed Sills”. J. Hydraul. Res. 41(3), 30. Salehi Neyshabouri, A.A., Ferreira Da Silva, A.M. and
271–284. Barron, R. (2003). “Numerical Simulation of Scour by a
14. Graf, W.H. (1971). Hydraulics of Sediment Transport. Free Falling Jet”. J. Hydraul. Res. 41(5), 533–539.
Water Resources Publication, Littleton, CO, USA. 31. Schlichting, H. (1979). Boundary Layer Theory, Chapter
15. Graf, W.H. (1998). Fluvial Hydraulics: Flow and Trans- XXIV. McGraw Hill, New York.
port Processes in Channels of Simple Geometry. In collab- 32. Schoklitsch, A. (1932). Kolkbildung unter Überfall-
oration with M.S. Altinakar, John Wiley and Sons, UK. strahlen. Die Wasserwirtschaft.
16. Hassan, N.M.K.N. and Narayanan, R. (1985). “Local 33. Shields, A. (1936). “Anwendunk der anlichkeits mechanik
Scour Downstream of an Apron”. J. Hydraul. Engng. ASCE und der turbulenzforschung auf die geschiebebewegung”.
111(11), 1371–1385. Mitt. Preuss. Versuchsanst. Wasser Schiffs. 26.
17. Hoffmans, G.J.C.M. (1997). “Jet Scour in Equilibrium 34. Simons, D.B. and Senturk, F. (1992). Sediment Trans-
Phase”. J. Hydraul. Engng. 124(4), 430–437. port Technology, Water and Sediment Dynamics. Water
18. Hoffmans, G.J.C.M. and Verheij, H.J. (1997). Scour Resources Publications, Littleton (Colo.).
Manual. A.A. Balkema, Brookfield, Rotterdam, The 35. Takeda, Y. (1995). “Velocity Profile Measurements by
Netherlands. Ultrasonic Doppler Method”. Exp. Thermal Fluid Sci. 10,
19. Hogg, A.J., Huppert, H.E. and Dade, B. (1997). “Erosion 444–453.
by Planar Turbulent Jets”. J. Fluid Mech. 338, 317–340. 36. Takeda, Y. (1999). “Ultrasonic Doppler Method for Veloc-
20. Hurther, D. (2001). “3-D Acoustic Doppler Velocimetry ity Profile Measurements in Fluid Dynamics”. Exp. Fluids
and Turbulence in Open-channel Flow”. These No 2395, 26, 177–178.
EPFL Lausanne, Switzerland. 37. Willemetz, J.C. (1992). “Etude Quantitative de l’Hémody-
21. Istiarto, I. (2001). “Flow Around a Cylinder on a Mobile namique de Vaisseaux Profonds par Echographie Doppler
Channel Bed”. PhD Thesis, No 2368, EPFL, Lausanne, Ultrasonore”. PhD Thesis, No 893, EPFL, Lausanne,
Switzerland. Switzerland.
22. Ito, T., Tsuji, Y., Nakamura, H. and Kukita, Y. (2001).
“Application of Ultrasonic Velocity Profile Meter to Vortex

View publication stats

Vous aimerez peut-être aussi