Académique Documents
Professionnel Documents
Culture Documents
Abstract
The paper deals with a mathematical and numerical model for simulating the
sediment transport within bottom boundary layer. The mathematical model uses
Reynolds equations for the hydrodynamic model and the two-dimensional
convection-diffusion equation for the sediment transport model. Results of the
numerical model are compared with laboratory data of Ribberink et al. (1994) and
Katopodi et al. (1994). Based on the analysis of comparisons, the validity of the
model is discussed.
Introduction
The flow field and sediment transport within the bottom boundary layer (BBL)
has been a long time research subject for coastal engineers who work in the field
of sediment transport. Scientists understood that the sediment entrainment
mechanism is strongly related to the flow field in the area close to the bed and that
flow field calculation is a key in the sediment transport phenomena.
Another major problem facing the researchers dealing with the boundary
layers flow field and sediment transport modelling is the scarcity of experimental,
both field and laboratory. The problem is mainly due to: (1) the limitations of the
present measuring equipment and techniques used to sample the data in the highly
turbulent flow regime and (2) the high concentrations of moving sediment
particles which prevents accurate measurements in the area close to the bottom.
134
Velocity and sand transport characteristics in the area very close to the bed,
known as the bottom boundary layer, are still under analysis. The reasons are both
the lack of accurate, reliable laboratory data that are difficult to obtain and the
theoretical problems involved by the formulation of the numerical models. The
bottom boundary layer has a major role in the whole sand transport process:
entrainment, suspension, deposition and resuspension of particles are largely
dependent upon the correct description of the processes occurring here.
The general layout of the proposed model is outlined in Figure 1.
TIME-DEPENDENT
SEDIMENT FLUX
~(z,t) = u(z,t )xc(z,t )
The numerical model for investigating the flow field and sediment transport inside
Downloaded from ascelibrary.org by KUNGLIGA TEKNISKA HOGSKOLA on 09/13/15. Copyright ASCE. For personal use only; all rights reserved.
Hydrodynamic Model
For the case of the flow field inside the bottom boundary layer,
considering the two-dimensional vertical (2-DV) case in a (x,z)-plane, the scalar
form of the equations governing the flow field can be written as
bu bw
--+--=0 (1)
~x 3z
bu ~u Ou lOr~_~u o Ouo
-~+U~x+WOz p Oz ~t ~u~ C2)
where u and w: the horizontal and respectively vertical components of the
Reynolds averaged velocity vector inside the bottom boundary layer, "r= : the
Reynolds shear stress, p : the water density, and u o : the horizontal free stream
velocity at the upper edge of the bottom boundary layer. The system of
coordinates is chosen so that x is positive in the direction of the wave
propagation and z is positive upward from the bottom,
The right hand side of Equation (2) represents the "forcing factor" of the
outer flow, that is, the oscillatory fluid motion outside of the boundary layer.
Since the Reynolds stresses are dominating in the uniform turbulent flow near a
hydrodynamically rough boundary, the viscous forces are thus neglected in the
present formulation.
The hydrodynamic model for the bottom boundary layer is interconnected
with the upper layer flow. A complex model (Duy and Shibayama, 1997) is used
to compute the time-series of the free stream velocities at the upper edge of the
boundary layer.
The Navier-Stokes equations in their Reynolds averaged form are employed
for the upper layer hydrodynamic model. The time-series of the free stream
velocities at the upper edge of the bottom boundary layer are then used as
boundary condition for the hydrodynamic layer, as it will be shown later. Further
references on the governing equations for the upper layer can be referred in the
listed references (Duy and Shibayama, 1997).
Figure 2 presents schematically the coordinates system for the upper layer and
bottom boundary layer.
For the hydrodynamic module of present bottom boundary layer, the eddy
viscosity model was introduced by expressing the Reynolds shear stress in the
following form
"rz_~= c)u (3)
P vT ~z
where Vr: the eddy viscosity within the bottom boundary layer. Based on a
simple approach (Kajiura, 1968), the eddy viscosity is considered time-invariant
but space dependent in vertical direction throughout the thickness of the bottom
boundary layer.
v r = tcu.z (4)
where tr : the Karman constant with a value of 0.4, u. : the friction velocity and
z: the vertical elevation taken from the solid bottom. The friction velocity is
assumed to be proportional to the free stream velocity, the roughness
characteristics of the bed and the fluid particle excursion amplitude. It is given as
I
(u,)~ = (5)
where fw: the friction factor and (Uo).~ x : the maximum horizontal free stream
velocity. The friction factor is based on the momentum integral method for the
case of rough beds (FredsCe and Deigaard, 1992),
I
Downloaded from ascelibrary.org by KUNGLIGA TEKNISKA HOGSKOLA on 09/13/15. Copyright ASCE. For personal use only; all rights reserved.
where ab'. the free stream particle amplitude and ku: the bed roughness which
was taken to be equal to mean sand grain diameters times two. At the same time,
the thickness of the bottom boundary layer was assumed constant in the direction
of the wave propagation and equal to
=2 (8)
O)
The physical processes related to the sand particle movement in the flow are
described by the turbulent convection-diffusion equation.
-~--~x 'gz- 3z
/
~-~x(eS,~x)+-~z(es,-~z, (9)
where c: the equi-phase mean sand concentration, u, w: the equi-phase mean
velocity components calculated in the hydrodynamic module for the bottom
boundary layer, w : the mean fall velocity of sand, es~ and es~: the turbulent
diffusion coefficients corresponding to the two axis of coordinates. However, for
the present model, it was assumed that the coefficients of turbulent diffusion, esx
and es~ are taken to be equal.
The difference between the two forms of sand transport - suspended sand and
sheet flow layer conditions - was outlined by various researchers. The present
work defines the two forms as separate but assumes a direct interaction between
them in the form of sand transfer between the two layers (bottom boundary layer
and the upper layer).
Both convection and diffusion are assumed to take place inside the boundary layer.
The last two terms of the right hand side of Equation (9) express the diffusion
process of the sand, which is being suspended inside the boundary layer while the
last two terms in the left hand side express the convection phenomenon. The
importance of the two processes is still being argued at present. However, for the
area very close to the bed, both processes, in connection with several other
phenomena such as the interaction between the sediment particles and the high
density of the fluid-sand mixture, seem important as proved by experimental work
performed in the last few years.
The boundary conditions for the both the hydrodynamic and sand
concentration model are given in Figure 3.
u=u" l at i = N , j = I , . . , M ~ 1
w--w~ I _~ r/ - - [ level'--"'--"--~prior
Static bed
wavacti
e~ to
,cc_at CG'
'u-O............................................................................
atl 1 1 M '''''~ ~'" Bonombc forconcentrationl' .......................
"&{"" ~ i .......
w- 0 ' = ' j = "" field = non-movable bed I
The bottom boundary condition assumes that sand concentration at the sea bed is
equal to the maximum concentration of immobile, immersed sand.
C=Cm~, for Z = 0 (10)
The reference level is assumed to be the level, experimentally observed, below
which there is no sand particle movement. This level is located beneath the initial,
expected, the reference level for sand concentration is not yet clearly established
through some sort of a clear mathematical formulation. However, such a condition
is assumed to be a first step in describing closer to reality the inception of sand
transport inside the bottom boundary layer.
As seen in Figure 3, at the upper edge of the bottom boundary layer, the
concentration gradient is assumed to be negligible, when compared to the middle
and lower part of the same area.
0c
--=0 (11)
0z
The location of the bottom boundary conditions for the hydrodynamic and sand
transport modules is not the same since the convection-diffusionequation requires
values of the vertical and horizontal velocities. Therefore, as noticed in Figure 3,
the location of the bottom boundary condition for the sand concentration model is
located one vertical interval grid, Az, upward with reference to the location of the
bottom boundary condition for the hydrodynamic model. Also, one major
hypothesis for the presented model is that the diffusion coefficient is assumed to
be equal to the eddy viscosity.
esx = esz = e s = v r (12)
With the computed values of the time-dependent velocity flow and time-
dependent concentration field, the time-dependent sand flux, r is calculated
as
t#(Z,t ) = u(z,t )c(z,t )/ p, (13)
where p, : the sand density, u(z,t): equi-phase mean values of horizontal velocity
and c(z,t): equi-phase mean values of sand concentration.
A fully implicit scheme used in the present model allows the determination of the
values of velocity and concentration at the same grid points. The Finite Difference
Method is employed and the node equations are written depending on their
location in the numerical grid.
The determination of the equi-phase mean values of the velocity and sand
concentration fields is made within the common domain and then, the time-
dependent sand flux is computed. In order to ensure model stability, the time- step
should be small enough (At = T/150...T/IO0) where T is the wave period. The
model requires around 4 wave periods to obtain a stable solution.
Ribberink et al. (1994) have investigated the full-scale reproduction of the wave
flow in the Large Oscillating Water Tunnel (LOWT) of Delft Hydraulics, Holland.
The test section was 14 m long, 1.1 m high and 0.3 m wide. A 30 cm thick sand
bed was arranged into the test section, leaving 0.8 m height for the oscillatory
flow above the bed.
The results of the measurements included also time-series of intra-wave
measurements of sand grain velocities in sheet flow regime using special
measuring techniques and equipment. The grain velocities in the thin sheet flow
layer were measured using detailed High-Speed Video recordings (HSV),
covering a depth range between - 8 mm and + 28 mm with respect to the initial
bed level. In their experiment, the sand grain velocity data were collected and
analyzed so that their equi-phase mean values were finally obtained. For the
analyzed case, termed as El, the oscillation period was T = 7 . 2 s e c while the
mean sand diameter was Dso = 0.21 mm. Maximum horizontal velocity recorded
at 20 cm above the initial bed level was ~ = 1.60 m/s. In Fig. 4, the measured
values of the sediment grain horizontal velocities are shown.
1.2-(ern)
1.o: . . . .
0.8 - + - - A - - t/T=0.55 '
0.6-
\\i
9 x.*
f /" 7
=
'
.
0,- \\ I/(./ /
0.2-~ '
0.0-
-0.2-
-30-25-20-15-10-5 O 5 10 15 2 0 25 3 0 35 4 0 4 5 5 0 5 5
Velocity (cm/s)
Figure 4. Measured horizontal grain velocities (HSV) inside bottom
boundary layer, case E1 (Ribberink et al., 1994).
z (cm)
1.2 . , . , . , - , . , . i - , - , . , . , - , . , . , . , . , . , .
I - - ~ I/T--0.14s ~AX&A
,o
~T~TT --e-- t/T=O.27S
--&~ t/T=O.41S
\~l i\ -'e-- trr=O55 s
0.8- x * q --+-- t/T=O.68s
--x-- I/T=O.81s B
0.6- ~ t/T=0.95s
- §
0,4-
0.2- +%
0.0~
-a0-~s-~0-l's'-;o-; "; "1b'1;'2;"215"30"3~5"4;"415'5()"55
Velocity (era/s)
In the present work, measured grain particle velocity is compared with the
computed fluid particle velocity as seen in Figures 6 and 7.
Z (cm)
1.2 " l ' l ' T ' , ' l ' J ' l ' , ' = ' 1 " , ' l ' l ' J ' , ' l '
1.0-
i \
0,6-
0.4- ~
0.2 -
9 )., --A--Fluid velocity (comput.)
-- Sediment velocity (lab, exper.)
0.0
-0.2 " l ' l ' , ' , ' l " " , , , , ' J ' , ' , ' l ' , ' ~ "
-30-25-20-15-10 -5 5 10 15 20 25 30 35 40 45 50 55
1~ (era) Velocity (cm/s)
1.0
0.8
o.o -/
/I/~' :9 Rt~d~oaty (comet)
-0.2 -- S~liment velocity (lal~ exper.)
-30-25-20-15-10 -5 0 5 10 15 20 25 30 35 40 45 50 55
V e l o c i t y (cm/s)
Figure 6. Comparison between water and sediment particle velocities inside BBL
(laboratory data of Ribberink et al., 1994).
Figure 7. Comparison between water and sediment particle velocities inside BBL
(laboratory data of Ribberink et al., 1994).
The comparisons revealed phase and amplitude differences between the velocities
of water and sand particles during one wave cycle. If at the beginning of the wave
cycle the differences were practically not significant, with an increasing value of
the relative time (t/T), significant phase differences seem to occur. However, by
the end of the wave period, the sand grain velocity seems to "catch u p " with the
water particle velocity.
Measurements of sand flux inside the bottom boundary layer have been rarely
reported in literature. The laboratory data which the authors considered for the
present study were performed by Katopodi et al. (1994) at Delft Hydraulics,
Holland, using a Large Oscillating Water Tunnel with reported dimensions as 15
m • 1.1 m x 0 . 3 m.
Since experiments were carried out in a tunnel, the hydrodynamic
conditions were horizontally uniform and therefore, effects associated with
vertical wave velocities were absent. The measurements were carried out on a
sandy bed with a mean sand particle diameter of ds0 = 0.21 mm. The oscillation
period for the analyzed case was T = 7.2 sec while the thickness of the boundary
layer was found to be ~ = 20 mm. The time-dependent sand flux was calculated
by multiplying the measured sand particle velocity with the sand concentration.
The sand grain velocity was measured using HSV (High Speed Video
recordings) while the time-dependent sand concentration inside the BBL was
measured using a CCM (Conductivity Concentration Meter). The results of the
phase-averaged sand flux (as measured) are presented in Figure 8.
2,5![
25.0 Laboratory
I ..........
data
:-.'-0
~ - - T/8
22,5 ~ - "1"/4
--x-- 3T/8
20.0 4 - - T/2
j ~ - 5T/8
17.5 ~ 3T/4"
12,5
vE
10.0
N
5.0 I
7.5
2.5
0.0 j t
-2.5
-5.0 . J . i , i , , i 9 i , i , i . i ,
-0,20 -0.15 -O.~O -0.05 O.CO 0.O5 O.10 0.t5 0.20 0.25 0.30
grain is the same. Therefore, the sand flux (numerically calculated) naturally
presents some differences from measured results. Figures 9 and 10 present the
comparison between the numerical and laboratory results for each of the two
halves of the wave period.
27.5 . ~ .
25.0
4-- 0 (expedmem) I
22.5
Computation
200
- -
I
17.5
15.0
" 125
E 10.0
v
N 73
5.0
2.5
0.0
-2.5
-5.0 , .0115 , _I 9 * - - I 9 i . i . I , I ,
.020 .0 10 4]05 0.00 0.05 0.10 015 0.20 025 0.30
175
15.0
12.5
v
E 10.0
N
5.O
, ! , , . , . . , . i . J . i . I .
.0.20 .0_15 .0.10 .0.C6 0.00 0.C6 0.10 0.15 0.20 0.25 0.30
25.0
22.5
20.0
--x--
- -
(ER)edrnenl)
3T/8
Computalion
i
17.5
15.0
12.5
100
7.5
N
5.0
2,5
oo
-25
-5.0 ' I , Z , I , , I 9 ' 9 i . , . i .
.0,30 -0.15 -0.10 .O.O5 0.00 0.O5 0.10 0.15 0.20 0.25 0.30
27.5
Downloaded from ascelibrary.org by KUNGLIGA TEKNISKA HOGSKOLA on 09/13/15. Copyright ASCE. For personal use only; all rights reserved.
25.0 [ ~--5T/8(ExperimeNt) I
22.5 - - Compulation
20.0
17,5 ~
150
"~ IZ5
E 10.0
N 75
5.0
Z5
0.0 ~ ,
-2.5
. . . ..0.20 ''.0 15 '.0'1o' ~)I~ . 000
. . 0,06
. . . 0.10
. . 0.15
. . .0.2O
. 025 0.30
Sedimentflux (m/s)
272
15.0
~ 125
E i 0.0
N 75
50
22
00
-2s
-0.2O .0.15 .0.10 .0.(~6 000 0.~ 0.10 0.15 0.20 02S 0.30
Sediment flux (m/s)
275
2S.O ~ 71"/8(Experiment) j
225 --
20.0
Computation
I
175
~ 12.5 !
E 100
N 7.5
5.0
2.5
-2 5
-50 " ' " ' " ' " . I . , . , . t . i .
"O.2O .0.15 41.10 .O.O5 000 O.O6 010 0.1S O2O 0.~ O3O
Sedimentflux (m/s)
Figure 10. Comparison between computed and laboratory determined sand flux
inside BBL, T/2<t<7T/8 (laboratory data from Katopodi et al., 1994).
One can observe the evolution of phase differences between the computed sand
flux values and the measured ones. If for the beginning of the wave period there is
good agreement between the computed and measured values, with an increased
value of the relative time (t/T), a larger phase lag occur towards the middle of the
wave period. As seen in the last figure, the phase lag seems to diminish towards
the end of the wave cycle.
Conclusions
References