Вы находитесь на странице: 1из 21

ARTICLE IN PRESS

JID: APM [m3Gsc;March 9, 2017;4:24]

Applied Mathematical Modelling 000 (2017) 1–17

Contents lists available at ScienceDirect

Applied Mathematical Modelling


journal homepage: www.elsevier.com/locate/apm

Hydrodynamic interaction in suspended sediment distribution of open


channel turbulent flow

Debasish Pal , Koeli Ghoshal
Department of Mathematics, Indian Institute of Technology Kharagpur, Kharagpur 721302, India

article info abstract

Article history: In the present study, a theoretical model is proposed to predict the vertical distribution of suspended sediment
Received 29 October 2015 in open channel turbulent flow. We derive our mathematical model based on the six prominent hydrodynamic
Revised 8 February 2017 mechanisms, which are upward sediment flux due to the turbulent diffusion, downward gravitational settling
Accepted 22 February 2017
of the sediment, hindered settling phenomenon, secondary current, fluid induced lift force on the suspended
Available online xxx
particles and the gradient of Reynolds normal stress of the suspended sediment. The importance of the
Keywords: hydrodynamics is described by the changes of suspended sediment concentration pro-file in terms of the
Open channel particle-flow interaction caused by those mechanisms. We also address the significance of the co-existence of
Turbulent flow those mechanisms for estimating the suspended sediment concentration profile. The present model agrees
Particle concentration satisfactorily with a wide spec-trum of experimental data in available literature. Unlike the previous
Particle-flow interaction researchers, we select a broad range of previously published models of vertical sediment concentration
Secondary current distribu-tion for comparison analysis, and the proposed model shows better prediction accuracy which is
confirmed by error analysis.

© 2017 Elsevier Inc. All rights reserved.

1. Introduction

Due to the highly turbulent nature of flow in natural streams, shear stress is generated which results in erosion of sediment. The eroded
particles are conveyed by those natural streams and play an important role in water resources man-agement, formation of various landforms,
economic development and many other aspects that are integral to our daily life. The prediction of sediment dynamics in open channel turbulent
flow is a topic of interest to the community of researchers [1–7] due to its many applications. Sediment particles are transported mainly by bed
load and suspended load modes in which the first one is a thin layer zone near the channel bed whereas the second one covers the remaining
portion of the flow depth. The objective of this research is to achieve a comprehensive understanding of suspended load dynamics by uncovering
the underlying physics in the flow region.

During the prediction of suspended load in a flow field, the complexity stems from the coupling of particle suspen-sion and the turbulent
nature of the fluid flow. Rouse [8] was the pioneer in solving the intricacy of suspended sediments and proposed a concentration equation based
on the upward and downward fluxes of sediment due to turbulent diffusion and gravitational force, respectively. This fundamental theoretical
work did not take into account particle-flow interaction,


Corresponding author. Present address: Pillar of Engineering Systems and Design, Singapore University of Technology and Design, 8 Somapah Road, Singapore 487372.

E-mail address: bestdebasish@gmail.com (D. Pal).

http://dx.doi.org/10.1016/j.apm.2017.02.045
0307-904X/© 2017 Elsevier Inc. All rights reserved.

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
2 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Nomenclature

a reference level
b width of channel
C volumetric concentration of sediment
c time-averaged volumetric concentration of sediment
ca time-averaged reference concentration of sediment
cd depth averaged concentration of sediment
cm maximum volumetric concentration of sediment
c fluctuation of c
D particle diameter
D normalized particle diameter
∗ damping of turbulence caused by sediment
df damping of sediment turbulence relative to fluid
dp mean relative error
Er fluid induced lift force on particle
FL gradient of Reynolds normal stress of particle
FS
g gravitational acceleration
hflow depth
J parameter
N total number of data points in an experimental run
nh reduction exponent of settling velocity
Rep particle Reynolds number
uf time-averaged streamwise velocity of fluid
ur relative streamwise velocity between fluid and particle
us time-averaged streamwise velocity of sediment
u
s fluctuation of us
u∗ shear velocity
U velocity vector of sediment
V computed value of ith data point
ci
V observed values of ith data point
oi
v exponent represents the effect of particle crowding
vf time-averaged vertical velocity of fluid
v vertical velocity of clear fluid in particle-laden flow
f1
v additional upward vertical velocity of fluid due to sediment
f2
vs time-averaged vertical velocity of sediment
vs fluctuation of vs
ws time-averaged lateral velocity of sediment
ws fluctuation of ws
x streamwise direction
y vertical height from the channel bed
yo vertical height from the channel bed at which u f = 0
z lateral direction
αs 1.3 exp(−b/2h)
β s/ m
p
(ρ p − ρ f ) / ρ f
m turbulent diffusion coefficient
s sediment diffusion coefficient
κ von Karman constant in sediment-fluid mixture
κc von Karman constant in clear fluid
νf kinematic viscosity of fluid
ωp settling velocity of particle in clear fluid
ωm settling velocity of particle in sediment-fluid mixture
ρf mass density of fluid
ρp mass density of particle
τp particle relaxation time
ξ y/h
ξ aa/h

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 3

stratification, secondary current and several other turbulence effects; still this research continuously motivates investigators for further study on
suspended load. Through the inclusion of volume occupied by suspended sediment, Hunt [9] pro-posed another analytical model of concentration
distribution. Apart from the traditional models [8,9], a large number of concentration models [10–12] were formed by using the advection–
diffusion equation. In the last few decades, investigators [13–16] developed mathematical models from the viewpoint of a two-phase flow
approach. Researchers also included the effect of stratification [17,18] in concentration profiles. Some published studies [19,20] are found
regarding the significance of secondary current in terms of the vertical velocity of fluid to predict the concentration distribution; however these
inves-tigations neglect the effect of suspended particles in the secondary current as mentioned by Yang [21]. Fu et al. [22] showed the importance
of fluid induced lift force on the suspended particles and the gradient of Reynolds normal stress of the sus-pended sediment in estimating
concentration distribution, though the effect of vertical velocity of fluid is ignored in the research of Fu et al. [22]. The aforementioned models
did not include the effect of the hindered settling phenomenon which reduces the settling velocity of particles in a sediment-fluid mixture in
comparison to that in clear fluid.

We conclude from the above discussions that upward sediment flux due to the turbulent mixing, downward gravita-tional settling of the
sediment, hindered settling phenomenon, secondary current, fluid induced lift force on the suspended particles and the gradient of Reynolds
normal stress of the suspended sediment have a significant role in predicting the suspended sediment concentration. No previous investigation
has been done on the importance of co-existence of these mechanisms together with particle-flow interaction in calculating concentration profile
for dilute and non-dilute open chan-nel turbulent flow. This study endeavors to address the aforesaid overlooked research problem and is
organized as follows. Section 6.2 describes the detailed mathematical formulation of the concentration equation through a theoretical approach
together with the inclusion of the aforementioned six hydrodynamic mechanisms. For the verification of the proposed model, Section 3 discusses
the selection of a wide range of experimental data existing in the literature. Section 4 exhibits the comparison between our model and
experimental data considered. Section 5 assesses the prediction accuracy of the present model in comparison to other existing models. Section 6
demonstrates the significance of the six hydrodynamic mechanisms considered in our study and the importance of their co-existence in predicting
the concentration profile.

2. Mathematical model

2.1. Governing equation

In a steady sediment-laden open channel turbulent flow where no mass is transferred from the solid phase to the fluid phase or vice versa, the
continuity equation for the solid phase is written as:
(1)
∇ .(U C) = 0,
where C is the volumetric concentration of sediment and U is the velocity vector of suspended sediment. Using the Reynolds decomposition
theory of turbulent flow in Eq. (1) and then taking the time average, we obtain
+ ∂ (vsc + + 0, (2)
∂ (usc usc ) + vsc ) +∂ (wsc wsc ) =
∂x ∂y ∂z
where x, y and z denote a three dimensional coordinate system along streamwise, vertical and lateral directions respectively, c and c are the time-
averaged and fluctuation parts of C, respectively; us, vs and ws are the time-averaged values of U along x, y and z directions respectively; us, vs and
ws are the fluctuation parts of us, vs and ws respectively and the over bar represents the time-averaged value. For a uniform flow along streamwise
direction, we can write Eq. (2) as:
∂ (vsc + + 0. (3)
vsc ) +∂ (wsc w sc ) =
∂y ∂z
At the central region of open channel through its cross-sectional area, the magnitude of second term in the left hand side is negligible in
comparison to that of the first term at the central region of cross-section of the open channel [19,21], therefore Eq. (3) turns into

∂ (vsc + 0. (4)
vsc ) =
∂y
Integrating Eq. (4) together with the boundary condition that the mass flux is zero at the free surface of the open channel, we get

vsc + = 0. (5)
vsc
2.2. Diffusion coefficients

We connect the turbulent diffusion in Eq. (5) with concentration gradient by


vs
c
− =
∂y
vs
s ∂c
, (6)
c
Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
4 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

where s is the sediment diffusion coefficient which can be written in terms of the turbulent diffusion coe fficient m by

s =β m,
(7)

where β is the inverse of Schmidt number. The well known and widely used form of m in terms of the streamwise velocity gradient of fluid is
given by [17,20,23]
∂ u f = u∗2 (1 − ξ ),
(8)
m ∂y
where u∗ is the shear velocity, uf is the time-averaged streamwise fluid velocity and ξ (= y/h) is normalized vertical height from channel bed in
which h is the flow depth. Using Eqs. (6)–(10), the following expression is obtained from Eq. (5):
β u ∗ ( 1 − ξ ) ∂ c = c vs . (9)
2
∂u
f ∂y
∂y

2.3. Secondary current

We need a mathematical formula for the time-averaged vertical velocity vs of particle in the sediment-fluid mixture to get the model of
concentration profile from Eq. (9). Hunt [9] wrote vs as the summation of the time-averaged vertical velocity vf of fluid in a sediment-fluid
mixture and the settling velocity ωp of particle in clear quiescent fluid. For sediment-laden turbulent flow, we replace ωp by ωm which
represents the settling velocity of a particle in that sediment-fluid mixture and write vs as:

vs = v f − ωm. (10)

Due to the presence of suspended particles in the sediment-fluid mixture, the settling velocity of a particle is hindered in comparison to that in
clear fluid and this hindered settling phenomenon can be modeled by the extensively used Richardson and Zaki [24] equation given by

ωm = ωp(1 − c)n h , (11)

where the exponent nh measures the amount of reduction of settling velocity in the sediment-fluid mixture. Using Eqs. (10) and (11), we express
Eq. (9) as:
β u (1 − ξ ) ∂ c + ωp c(1 − n = c v f .

(12)
c) h
2
∂u
f ∂y
∂y

All open channel flows are three-dimensional due to the appearance of secondary current [21,25–27] which is the com-bination of vertical
and lateral velocities of the flow. At the commencement of our modeling, the magnitudes of lateral variation terms are ignored as we predict the
sediment concentration at the middle region of the cross-sectional area; therefore we consider the effect of secondary current through the vertical
velocity vf of fluid in Eq. (12). Previous researchers [21,22] already pointed out the non-zero value of vf in sediment-laden flow and argued that
its magnitude and direction vary with particle concentration. It was also shown that the mass flux cvf in Eq. (12) can alter velocity, concentration,
stress and other several turbulent features in the flow field.

To investigate vf, Yang [21] examined the profile of Reynolds shear stress in sediment-laden flow which was a function of the momentum
2
flux and vertical height from channel bed as −u f v f = u ∗ (1 − ξ ) + u f v f . On the other hand, it is observed from the experimental data of
sediment-laden flow [28,29] that the value of Reynolds shear stress increases after adding sediment in clear fluid. Yang [21] gave the argument
that the increased Reynolds shear stress is obtained either due to the decreased downward value of vf, which implies increased magnitude along
upward y direction, or due to the suspended particles which contribute an additional upward vertical velocity as a compensation of downward vf
caused by the secondary current. From this argument, vf has been expressed in the literature [21,30] as:
v =v +v , (13)
f f1 f2

where v f1 represents the vertical velocity of clear fluid in particle-laden flow and v f2 describes the additional upward verti-
cal velocity of fluid due to the suspended sediment.
To derive the concentration distribution from Eq. (12) together with Eq. (13), we need to find suitable expressions for v f1 and v f2 . Due to
the irregular behavior of the secondary current, theoretical knowledge on these features is lacking in the literature. For particle-free flow, Yang
[31] suggested a semi-theoretical equation for v f1 based on its boundary condition i.e. zero value at channel bed and free surface of the open
channel. Ghoshal and Kundu [19] considered the formula for
v f1 by Yang [31] as a reference and suggested a modified form for sediment-laden flow by taking into account the effect of suspended sediment
concentration. We consider the following expression of v f1 by Ghoshal and Kundu [19] as it was verified with a wide range of experimental data [32,33]
and good agreement was observed.
v = αsκ ξ (1 − ξ )(1 − c). (14)
f1

u∗
Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 5

In Eq. (14), κ is the von Karman coefficient in the sediment-fluid mixture and the unknown parameter αs needs to be determined.

For the remaining component, v f2 , investigators [21,30] expressed it simply as a function of settling velocity of parti-cle and sediment
concentration; however the expression does not include the effect of particle-flow interaction. Fu et al. [22] proposed v f2 by taking into account
the particle-fluid interaction in terms of the fluid induced lift force FL acting per unit mass on particle and the gradient, FS, of the Reynolds
normal stress of the suspended sediment. Moreover, they delin-eated the contributions of FL and FS in secondary current as well as in entraining
the suspended sediment near the channel bed. Therefore, we select the expression of v f2 by Fu et al. [22] given as:
v f2 = τp(FL + FS ), (15)

where τ p is the particle relaxation time. Following the previous researchers [13,22], we write τ p by
v v
ωp(1 − c) ρp ωp(1 − c) ( p + 1) , (16)
τp = =
g(ρp − ρ f ) g p

where g is the gravitational acceleration and p = (ρp − ρ f )/ρ f in which ρf and ρp are the mass densities of fluid and sediment, respectively. The
effect of particle crowding is considered in Eq. (16) by the exponent v (Di Felice [34]). Using Eqs. (13)–(16), we express Eq. (12) as:
β u∗ (1 − ξ ) ∂ c = −ωp c(1 cu (1 − ξ )(1 c) v (F F ). (17)
n ∗ κ ωpc(1 − c) ( p + 1)
2 c) h

∂uf ∂y − +
αs ξ − + g p L
+ S
∂y

We need two appropriate expressions of FL and FS to obtain the suspended concentration profile from Eq. (17).

2.3.1. Fluid induced lift force on particle


The shear of flow and particle rotation have important role in the fluid induced lift force FL on particle [35]. Kallio and Reeks [36] showed
that the first one is very important in comparison to the second one in the turbulent boundary layer. In this context, McLaughlin [37] expressed
FL for shear flow by
FL =
2π 2 ρ p D νf ∂y 1 ur J,
(18)
27 ρfνf 1 ∂uf 2

where νf is the kinematic viscosity of fluid, D is the particle diameter, ur is the streamwise velocity lag between particle and fluid and the
parameter J depends on ur, gradient of uf and the distance between center of particle and channel bed. It follows that a suitable expression for uf
is required to calculate Eq. (18). In spite of a wide literature available on uf, we select the commonly used log-law to keep our model simple and
it is expressed by
u 1 y
f
= κ ln , (19)
u yo


where yo represents the dimensional vertical height from channel bed at which u f = 0.
2.3.2. Gradient of Reynolds normal stress of the suspended sediment

The gradient FS of Reynolds normal stress of the suspended sediment is expressed as:
FS ∂ vs 2
. (20)
=−
∂y
The relation between Reynolds normal stresses of particle and fluid, being denoted by vs 2 and v 2 respectively, is still
f
2 2
inconclusive in existing literature. Some researchers [28,38] found lower value of vs than v far away from the channel
f
2 2
bed where others [39,40] discovered that vs is larger than v near the bed. Greimann and Holly [14] resolved this issue
f
by a simple relation which is
vs2 d pv
(21)
2 ,

= f

where dp describes the damping of particle turbulence relative to the fluid. Nezu and Rodi [41] provided an expression of
2
turbulence intensity v f for clear fluid in open channel flow. Greimann and Holly [14] modified the expression of Nezu and Rodi [41] for
particle-laden flow and suggested the Reynolds normal stress of fluid by

v f2
= d 1.51 exp(−1.34 ξ ), (22)
u∗2 f
0.5
where df represents the damping of turbulence caused by the suspended sediment. Generally, researchers [13,14,16] calcu-lated df by (κ /κ c)
in which κ c represents the von Karman constant of clear fluid whose value is 0.41. In most of the
Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
6 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

available literature, the value of κ for sediment-fluid mixture is not mentioned clearly and Yang [21] suggested it as a non-linear function of
particle concentration c and κ c by analyzing experimental data. We observe that the suggested formula for df by Yang [21] creates a large
difficulty in getting a numerical solution of Eq. (17) through Eq. (20) as c is a function of y. To avoid this problem, a constant value of df is
needed. As a result, c is replaced by the depth averaged value cd of concentration in df proposed by Yang [21]. The modified expression is given
by
1 1

= cm3 − c d3

0.14 c 3 2

df 1 1
d
1
, (23)

where the value for maximum volumetric concentration cm is taken as 0.67. We select this value as the diameter of particle is determined from an
equivalent spherical volume and Cheng [42] evaluated cm by a packing arrangement of uniform spheres in a cylinder of unit length with diameter
same as the spheres. Using Eqs. (18)–(23), we convert Eq. (17) into:

∂ c
βκ ξ (1 − ξ ) ∂ ξ ωp n
=− c(1 − c) h + cκ αsξ (1 − ξ )(1 − c) u∗
1
+ v 27 ρ f ν f u∗
1
u∗
ωpc(1 − c) ( p + 1) 2
1 2 ur J + 2.02d κ
g p
2
2π ρp D νfh ξ u∗ pd f h exp(−1.34 ξ ) . (24)

To find the concentration distribution from Eq. (24), we need expressions for streamwise velocity lag ur between fluid and suspended particle,
settling velocity ωp of particle in clear still fluid and the reduction exponent nh of settling velocity in sediment-fluid mixture.

2.4. Relative velocity between fluid and particle

In the available literature, a scarcity of theoretical research for ur have been observed due to the tangled dynamics of suspended particles and
fluid in turbulent flow region. Zhong et al. [16] used a constant value of ur throughout the flow depth which is not reasonable since the relative
velocity varies with vertical height shown by previous researchers [13,15]. Cheng [43] evaluated an analytical expression for ur based on the
hindered drag force in sediment-laden flow and we select it for this study as it can explain real phenomena of streamwise velocity lag and has
reasonable accuracy in estimating experimental data. The model of Cheng [43] is given by

1.5

u∗
ur
= ( 2 − 2 ξ ) 1.5 1
+
14 32 ν
u∗D
f
2 −
12 32 ν f
u∗ D
1 . (25)

1.5 1.5

2.5. Settling velocity of particle in clear fluid

The settling velocity ωp of a particle in clear fluid plays a significant role in determining the downward flux of sediment and a large number
of literature studies are available on this turbulent feature. From this wide literature, we consider an advanced formula proposed by Zhiyao et al.
[44] as it provides better prediction accuracy in comparison to the other frequently used existing models in particle-laden flow [45,46]. They
suggested ωp as:
12 7

νf
8

ωp = D D∗3 38.1 + 0.93D∗7 . (26)

The formula of normalized particle diameter D∗ in Eq. (26) is given by [44–46]


D∗ = ν2f
1
D. (27)
3
pg

2.6. Reduction exponent of settling velocity

The reduction exponent nh plays a notable role in reducing the settling velocity in a sediment-fluid mixture and it gener-ally depends on
particle Reynolds number Rep(= ωpD/ν f ) and sediment concentration [24,42,47]. The interaction between suspended particles and flow plays a
significant role in nh and this effect is taken into account by Pal and Ghoshal [48]. We consider their expression for nh given as:
− −
×
7
= ln(1c) − h cm 38.1 5.74Rep7
12 4 8
− +
3
n 1 ln(1 c) 3 f ln 1 c 38. 1 + 5.74 fh Rep , (28)

4 7 7
h 4

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 7

where the expression for fh in Eq. (28) is



f
h = (1
+ p )(1 − c ) p − 1
p
1
cm c 2 (1
− c)

1
1
. (29)

2.7. Final expression of the proposed model

Using Eqs. (25)–(29), we can write the normalized expression for the concentration gradient given by Eq. (24) in the final form as:

∂c
, (30)
fu ∂ ξ = fdh + fsc + fl f + fnsg
where the functions fu, fdh, fsc, flf and fgns are given by

1
0.14 c 3 (31)
fu = βκc 1− 1
d
1 ξ (1 − ξ ),
cm3 − cd3
ωp n
fdh = − c(1 − c) h ,
(32)
u∗
1
0.14 c 3
fsc = αsc 1− 1
d
1 ξ (1 − ξ )(1 − c), (33)
3 3
cm − cd 1
f −
= 3
c(1 − c) 1 − 0.14 cd1 ur
1
lf 2π 2 p gDh
ν f ωp
νf 2
κcξ 1
2

27 J u∗ h 1 v

2
1 1 (34)
cm3 − c3
d
u∗
1 1 ,

= p gh − −
cm3 − cd3

f nsg 2.02d p p +1 ωpu∗ v 0.14 cd3


c(1 c) 1 (35)
exp(−1.34 ξ ).
11

We can observe that the final expression for the concentration profile given by Eq. (30) together with Eqs. (31)–(35) is derived based on six
significant hydrodynamic mechanisms. In Eq. (30), fu and the concentration gradient simultaneously represent the upward flux of sediment due
to the turbulent diffusion, fdh expresses the downward gravitational settlement together with the hindrance of settling velocity of the suspended
sediment, fsc describes the vertical velocity of clear fluid, flf denotes the fluid induced lift force on suspended particles and fnsg explains the
gradient of Reynolds normal stress of the suspended sediment.

2.7.1. Boundary condition


To solve Eq. (30), the required boundary condition is c(ξ = ξa ) = ca where ca is the reference concentration at the nor-malized reference
level ξa (= a/h) from the channel bed in which a is the dimensional reference level from the channel bed or the bed load layer thickness.
Sediment particles move by rolling, sliding and saltation through bed load mode; therefore a is generally defined by the saltation height of
particles and it is of the same order as the particle size [49,50]. In the available literature, the selection process of a is controversial as we can see
that some investigators considered it to be 0.1 h [13,14,22], some other took it to be 0.05 h [15,16] and some used mathematical expressions
[49,50]. Regarding the reference concentration ca also, we can find a few explicit expressions [51–53]; however these formulae either
underestimate or over-estimate the measured data. To avoid the difficulty of choosing a and ca, some researchers took a to be the lowest
available height near the channel bed in experimental data and ca as the concentration measured thereat [2,54]. Here also we follow the same
assumption as the measured values of a and ca are quite reasonable. Finally, we solve Eq. (30) numerically through Eqs. (31)–(35) together with
the selected boundary condition.

3. Experimental data

After deriving Eq. (30) in Section 6.2 for predicting vertical concentration distribution of particles, we need to verify it with experimental
observation. We collect experimental data from published literature where the measured concentration profiles are influenced by the effects
considered in this present work. Apart from the two dominant influences on concen-tration modeling, namely upward and downward sediment
fluxes, secondary current plays a leading role in our work and it has significant impact on particle concentration [19–21]. Researchers [25,31,55]
emphasized that the secondary current shows prominent behavior in narrow open channels where b/h < 5, in which b denotes the width of the
channel. Coleman
[56] conducted experiment in narrow open channel and gradually increased the amount of sand in the flow region until

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
8 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Table 1

Summary of experimental data considered for the verification of our model.

Literature D (mm) h (cm) u∗ (cm/s) p 100 × ca β


Coleman [56] 0.105 16.90–17.30 4.10 1.65 0.170–2.300 0.44–0.64
0.210 16.80–17.20 4.10 1.65 0.098–1.200 1.62–2.08
Lyn [57] 0.150 6.45 3.58 1.65 0.377 0.61
0.190 5.68–5.84 3.69–4.34 1.65 0.050–0.334 0.66–0.89
Bennett et al. [58] 0.230 11.00 4.58 1.60 7.301–7.899 0.60–0.62
Wren et al. [59] 0.550 14.00 8.40 1.65 0.827 1.35
Muste et al. [40] 0.230 2.10 4.20 1.65 0.575–0.973 1.36–1.88
Einstein and Chien [60] 0.274 13.20–13.40 10.09–10.61 1.65 1.185–13.283 0.26–0.54
0.940 12.80–14.30 11.53–12.60 1.65 1.057–9.906 0.75–0.92
1.300 10.90–12.00 12.85–14.50 1.65 4.566–12.377 0.91–1.08
Xingkui and Ning [61] 0.150 8.00 7.37–7.40 1.64 2.170–9.640 0.49–0.75
Zheng et al. [62] 0.100 100.00 1.81–6.08 1.65 0.014–0.090 1.70–4.40

there was a continuously moving sheet of sand at the channel bed with no occurrence of sediment deposition. From the collected data of Coleman
[56], we choose twenty-eight experimental runs for two diameters of sediment.
The shape of the sediment bed at the bottom of the channel varies for the entrainment and deposition of particles. To verify the present model
with the data of varying channel bed, we select two experimental runs over an equilibrium bed and six experimental runs over a non-equilibrium
bed from Lyn [57], two experimental runs over a mobile upper stage plane bed from Bennett et al. [58] and one experimental run over low-relief
antidunes from Wren et al. [59]. The fluid induced lift force on suspended particles and the gradient of Reynolds normal stress of the suspended
sediment are prominent in very small flow depth which is considered in the experimental runs of Muste et al. [40]; consequently, we select their
data obtained using natural sand. In this context, we admit that near the channel bed a few concentration data of runs NS2 and NS3 of Muste et
al. [40] have not been digitized since a consistent nature of concentration profile is not observed there and the reason might be due to high
turbulence near the bed or the limitation of equipment which were used for the measurement of turbulent features.

Apart from the aforesaid perspectives on data selection, we would also like to verify our model with non-dilute data of narrow sediment-
laden open channel flow where the six hydrodynamic mechanisms considered in our study are inher-ently observed. Einstein and Chien [60]
experimental observations for three sediment diameters are the most widely cited non-dilute data where measurements were done up to 40% of
the flow depth in narrow open channel. We select twelve experimental runs of three sediment diameters from their collected data. In addition,
Xingkui and Ning [61] collected three experimental runs for sand up to 90% of the flow depth in non-dilute narrow open channel flow with a
smooth bed surface and these runs are considered for verification. Apart from this, we pick eight experimental runs for nearshore data taken at
Oujiang estuary and Zhoushan available in Zheng et al. [62] to verify our model.

As a whole we select sixty-five experimental runs and it should be noted that this is the widest spectrum of experimental data chosen so far
for the verification of a theoretical model of suspended concentration profile. We show a brief summary on the ranges of flow parameters used in
those experimental runs in Table 1. We refer all the published papers [40,56–62] to the readers for detailed description of the experimental data
considered.

4. Comparison of present model with experimental data

Eq. (30) is a non-linear differential equation and its analytical solution is not possible. To solve it numerically, we use the well known fourth
order explicit Runge–Kutta method due to its higher prediction accuracy in comparison to other methods. Moreover, the values of the unknown
parameters J, dp, v, αs, cd and β are required for the numerical solution of Eq. (30). As J depends on various flow constituents and its
mathematical expression is very complex as discussed in McLaughlin [63], we simply pick the constant value of this parameter as 2.255 which is
available in Fu et al. [22]. A cumbersome task is to determine an appropriate value of dp due to its physical significance mentioned after Eq. (21).
We consider dp as 1.2 evaluated by Greimann and Holly [14] through analyzing experimental data and it is applicable for a broad range of flow
conditions as shown by Zhong et al. [16]. In spite of having the non-dilute characteristic in selected experimental runs of Einstein and Chien [60],
Xingkui and Ning [61] and Bennett et al. [58], these runs lie within the demarcation of non-dense flow and this phenomenon leads us to accept v
in Eq. (16) as 1 suggested by Greimann et al. [13] for non-dense nature of flow. Therefore, the values of parameters are chosen as [J, dp, v] =
[2.255, 1.2, 1] from existing literature to compute the concentration profile from Eq. (30).

Any explicit expression or specific range on αs in Eq. (14) is not found in literature. Researchers [19,31] mentioned that αs depends on
experimental data. A proper approximation of αs is required to solve Eq. (30) and we approximate it as follows. Yang et al. [25] suggested v f1
2
through uv f1 /u ∗ ∝ −ξ and proposed the value of the proportionality parameter as 1.3 exp(−z/h) where z represents the lateral length from
sidewall to central region along the cross-sectional area of the open channel. In our study, we predict the concentration profile at the central
region, therefore we select z = b/2 where b is the width of the open channel and consider αs = 1.3 exp(−b/2h). Moreover, we examine the effect
of αs on the characteristics of

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 9

0
10

-1
10

10-2
-1 0
10 10

Fig. 1. Effect of the step length (SL) of independent variable ξ in the numerical solution of Eq. (30).

Fig. 2. Comparison between computed concentration (solid line) by Eq. (30) and observed data (circle) for D = 0.105 mm (6 and 12) and 0.210 mm (24 and 30) of Coleman [56].

Eq. (30) and an appropriate nature is observed. Due to the unavailability of depth averaged concentration cd in experimental data, we calculate it
−1 1
by (1 − ξa ) ξ a cR dξ where cR denotes the well known analytical concentration profile given by Rouse [8].

Though several investigations have been done on the inverse of Schmidt number β due to its ubiquitous importance in the flow region, still
no model can claim universality for all flow conditions. Considering a wide spectrum of reported data, β is estimated as an empirical parameter
from experimental data. Before the estimation of β, we do need to check the robustness of the numerical solution of Eq. (30) in terms of the step
length of the independent variable ξ . Taking β to be 1 and with the aforementioned values of other parameters, we solve Eq. (30) for different
step lengths of ξ and show the solutions in Fig. 1. We notice that the solution is independent of the step length except when its value is taken as
0.1. The reason behind this behavior is that the large step length ξ is not compatible for the small range of ξ which is ξ a <

ξ < 1. From observation of Fig. 1, the step length of ξ is taken as 0.01 for the solution of Eq. (30). After the selection of step length, the
estimation of β is done by arranging the best fit between measured data and the computed concentration profile by Eq. (30). We show the
estimated ranges of the value of β for different sediment diameters in Table 1 and observe that the obtained ranges agree with previous
researchers [21,59,64,65]. For analysis of the agreement between computed profiles and experimental data, it would not be possible to show the
comparison graphs for sixty-five experimental runs considered for verification. To avoid that, we arbitrarily choose a limited number of
experimental runs from the data to show the determination accuracy of Eq. (30). For two selected experimental runs of each particle diameter
0.105 and 0.210 mm from the Coleman [56] data, Fig. 2 shows the comparison between computed concentration profiles and corresponding
measured data for narrow open channel flow and good agreement between them can be observed.

For experimental data of Lyn [57], we take one run over equilibrium bed and two runs over non-equilibrium bed. Fig. 3 illustrates that
computed concentration profiles assess well the experimental runs over equilibrium and non-equilibrium beds. Fig. 4 depicts the comparison
between calculated and measured concentration for chosen runs of Muste et al. [40], Bennett et al. [58] and Wren et al. [59] in which good
estimation accuracy of our model is observed for the data of Muste et al. [40]. Apart from this, though the computed profile of c gives appropriate
agreement with the data of Wren et al. [59] over low-relief antidunes, but a little deviation of calculated profile from measured data is observed
for Bennett et al.
[58] data over mobile upper stage plane bed. The reason for the disagreement is the highly fluctuating behavior of turbu-lence over the mobile
bed.
From non-dilute runs of Einstein and Chien [60], we pick up three experimental runs S4, S8 and S12 for three particle diameters 1.3 mm,
0.94 mm and 0.274 mm, respectively. Also, we collect one experimental run SQ2 from the non-dilute data of Xingkui and Ning [61]. We
compare these experimental runs with the corresponding concentration profiles computed

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
10 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Fig. 3. Comparison between computed concentration (solid line) by Eq. (30) and observed data (circle) for equilibrium (EQ) and starved beds (ST) of Lyn [57].

Fig. 4. Comparison between computed concentration (solid line) by Eq. (30) and observed data (circle) of Muste et al. [40] (NS2), Bennett et al. [58] (Set 2) and Wren et al. [59]
(WAD).

Fig. 5. Comparison between computed concentration (solid line) by Eq. (30) and observed data (circle) for D = 1.3 mm (S4), 0.94 mm (S8) and 0.274 mm (S12) of Einstein and
Chien [60] and SQ2 of Xingkui and Ning [61].

using Eq. (30) in Fig. 5. We observe from Fig. 5 that our model displays reasonable agreement with the non-dilute data of Einstein and Chien
[60] and Xingkui and Ning [61]. Our model also provides good determination accuracy in Fig. 6 for nearshore data of Oujiang estuary and
Zhoushan from Zheng et al. [62]. Each graph in Figs. 2–6 contains the corresponding name of experimental run mentioned in the associated
literature except Wren et al. [59] and Zheng et al. [62] where we ourselves give run names due to their unavailability in literature [59,62]. We
finish this section by concluding that the present model accurately estimates a wide spectrum of reported experimental data. It confirms the
applicability of proposed model for a broad range of turbulent flow conditions in sediment-laden open channel flow.

5. Comparison of present model with other existing models

Given the comparison analysis is accomplished between our model and considered experimental data, it remains to examine its determination
accuracy to predict experimental data in comparison to other existing models. Hence we choose appropriate models of suspended sediment
concentration profile from published literature. First and foremost, we consider the widely cited concentration equation of Rouse [8] which is
inevitably used for comparison in the literature of suspended load. In addition, we pick up another accepted traditional concentration model given
by Hunt [9] for comparison. We choose four existing concentration profiles of Greimann et al. [13], Greimann and Holly [14], Jiang et al. [15]
and Zhong et al. [16] to compare our model with those models which were derived from the viewpoint of two-phase flow theory in sediment-
laden

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 11

Fig. 6. Comparison between computed concentration (solid line) by Eq. (30) and observed nearshore data (circle) of Zheng et al. [62].

Table 2
Calculated errors from different models selected for comparison and present model.

Er (%)
Literature D (mm) M1 M2 M3 M4 M5 M6 M7 M8 PM
Coleman [56] 0.105 169.9 172.8 166.4 168.1 170.5 164.1 46.2 21.3 ∗
9.9
0.210 70.6 70.4 75.0 67.8 70.5 68.2 76.5 53.0 ∗

11.7
Lyn [57] 0.150 166.3 167.1 146.0 177.7 168.0 178.9 52.9 19.7 ∗

17.4
0.190 51.3 51.4 38.6 68.7 52.7 70.9 27.0 22.6 ∗

12.2
Bennett et al. [58] 0.230 392.8 431.2 355.0 651.7 399.5 703.7 255.8 76.4 ∗

68.6
Wren et al. [59] 0.550 81.3 81.2 52.1 62.5 80.7 102.5 50.7 104.4 ∗

48.2
Muste et al. [40] 0.230 71.8 71.6 53.7 35.7 71.0 65.8 29.1 50.6 ∗

14.2
Einstein and Chien [60] 0.274 62.3 72.8 34.8 26.5 40.9 118.9 22.0 29.1 ∗
4.8
0.940 40.0 38.2 24.1 63.7 19.4 13.5 22.2 21.3 ∗

12.4
1.300 59.3 57.6 ∗ 40.5 40.4 30.9 54.0 15.5 13.3
12.0
Xingkui and Ning [61] 0.150 33.0 36.1 38.0 63.7 33.6 50.0 57.2 38.8 ∗
7.7
Zheng et al. [62] 0.100 44.4 44.4 44.6 44.3 44.4 44.4 56.8 37.0 ∗
3.1

open channel flow. To predict the particle concentration profile, the kinetic model of Fu et al. [22] explained the importance of FL and FS which
are inserted in our study; therefore we take this model also for comparison. The secondary current plays a significant role in our model and this
phenomenon prompts us to select the secondary current based concentration model for comparison analysis. Among the existing very few models
based on secondary current approach, we select the recent model of Kundu and Ghoshal [20]. We rename the models of Rouse [8] by M1, Hunt
[9] by M2, Greimann et al. [13] by M3, Greimann and Holly [14] by M4, Jiang et al. [15] by M5, Zhong et al. [16] by M6, Fu et al. [22] by M7
and Kundu and Ghoshal [20] by M8.

A huge space is required to show the comparison analysis through figures between all experimental runs and all models mentioned in the
previous paragraph. Because of space limitations, we perform the comparison analysis between the present model and other existing models
considered through an error analysis. For the calculation of error, we choose the mean value Er of absolute relative errors and its
percentage value is given by
V
r= N
i 1 oi
V −V
E (%) 100 N ci oi , (36)
=

where Vci and Voi denote calculated and observed values of the ith data point for a particular experimental run and N repre-sents the number of
measured data in that run. Due to the lack of space, it is not possible to provide the calculated values of Er(%) for all models for each of the
sixty-five experimental runs selected. Therefore, for all models, we evaluate the average value of Er(%) for all experimental runs corresponding
to different particle diameters D in each literature selected [40,56– 62]. We show the calculated values of Er(%) in Table 2 where PM denotes Eq.
(30). It is observed that for all experimental data,
the range of Er(%) obtained from PM (2.8–68.7) is very small in comparison to other models M1 (33.0–392.8), M2 (36.1– 431.2), M3 (12.0–
355.0), M4 (26.5–651.7), M5 (19.4–399.5), M6 (13.5–703.4), M7 (22.2–255.8) and M8 (15.5–104.4). It is clear from Table 2 that all the models
give large value of Er(%) for the data of Bennett et al. [58] except PM and M8. We envisage that the inconsistent nature of the Bennett et al. [58]
data which can be inferred from Fig. 4 is responsible for the large value of Er(%). Here we want to highlight also that in spite of having an
inconsistent behavior of experimental data, our model approximates it with a reasonable estimation accuracy compared with other models. Apart

from this, for each row in Table 2, we denote the least error among all models by a superscript star ( ). It can be seen that excepting D = 1.3 mm
of Einstein and Chien [60], the present model provides lower error than other models. The aforementioned attributes of the proposed model
confirm its better determination accuracy with respect to others in estimating measured concentration profile for a wide region of open channel
turbulent flow.

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
12 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Fig. 7. Effect of (a) hindered settling, (b) vertical velocity of fluid, (c) fluid induced lift force on suspended particles and (d) gradient of Reynolds normal stress of the suspended
sediment by ignoring nh (in fdh ), fsc , flf and fnsg respectively in our model given by Eq. (30).

6. Discussions

Here we discuss the significance of including the hydrodynamic mechanisms considered in our model and explain the importance of the co-
existence of those mechanisms in determining the concentration profile of suspended particles.

6.1. Significance of hydrodynamic mechanisms considered in the model

The mathematical expression of our model is constructed on the basis of six hydrodynamic mechanisms, namely the upward sediment flux,
downward sediment settlement, hindered settling, secondary current, fluid induced lift force on the suspended particles and the gradient of
Reynolds normal stress of the suspended sediment. The first two are fundamental mechanisms required for particle concentration modeling;
therefore we justify the importance of the remaining turbulent features through their physical importance. The justification has been given
through describing real phenomena of sus-pended sediment concentration in terms of particle-flow interaction. The effect of particle-flow
interaction is illustrated by comparing concentration computed using Eq. (30) and the same equation with nh = 0, fsc = 0, fl f = 0 and fnsg = 0
assumed in turn. During the calculation of the numerical solution, the values of parameters included in Eq. (30) are taken from the ar-bitrarily
chosen experimental run SQ3 of Xingkui and Ning [61]. In each case, we observe decreased concentration when the turbulent feature terms are
turned off one by one. Fig. 7 exhibits the percentage value of concentration difference obtained.

Fig. 7(a) shows the profile of concentration difference when hindered settling is not considered i.e. nh = 0 in Eq. (30). We observe that up to a
very small distance from the channel bed which is approximately 5%, the concentration difference increases up to 25% and then decreases with
increasing vertical height ξ from the channel bed. However, the decreasing concentration phenomenon is very rapid within the turbulent
logarithmic-layer which is 20% of the flow depth. We explain the nature of concentration difference observed as follows. At the channel bed, the
hindered settling dynamics is not very active as sediment are not in suspension there; however, near the channel bed, a large number of particles
entrain into suspension due to fluid stress. The higher entrainment rate of particles near the channel bed hinders the deposition of particles which
are already in suspension; therefore, the concentration difference increases within a very thin layer near the channel bed. On the other hand, the
entrainment of sediment decreases with increasing ξ which reduces the impact of nh and as a result, the profile of concentration difference
decreases.

The profile of concentration difference obtained when not considering vertical velocity of clear fluid by making fsc = 0 in Eq. (30) is
displayed in Fig. 7(b). We find that the concentration difference increases up to 7% in the turbulent logarithmic-layer and then gradually
decreases with increasing ξ . The vertical velocity v f1 of clear fluid plays a notable role in conveying suspended particles from the channel bed
up to the suspension region. It can be observed from Eq. (14) that except near the channel bed and free surface of the open channel, v f1
significantly affects the entire vertical flow region. The secondary current which is a combination of lateral and vertical velocities of fluid follows
an intricate dynamics of clockwise and anti-clockwise circular path along the cross-sectional area of the open channel. During the circular
motions, according to the strength of upward vertical velocity v f1 , it carries sediment from the channel bed to a certain suspension region in the
turbulent inner layer; as such we observe increase in concentration difference. Since after a certain region the magnitude of

v f1 decreases with increasing ξ , less sediment is carried there which results in decreasing concentration difference.
Fig. 7(c) illustrates the profile of concentration difference with ξ owing to the neglect of fluid induced lift force FL on particles which is
achieved by setting fl f = 0 in Eq. (30). Like the impact of hindered settling exponent nh in Fig. 7(a), we follow similar impact of FL; however, the
concentration difference is observed higher for FL near the channel bed which is 34% in magnitude. Near the channel bed, the magnitude of the
fluid induced lift force FL can maintain a large amount of particles in suspension. Hence the difference in concentration is very high near the
channel bed in the absence of FL. With increasing ξ from the channel bed, the magnitude of FL decreases; therefore the amount of lifted particles
decreases with increasing ξ and this results in decreased profile of concentration difference.
Using fnsg = 0 in Eq. (30), the influence of gradient FS of Reynolds normal stress of the suspended sediment is shown in Fig. 7(d). The
magnitude of concentration difference is observed as 3% within 10% of the flow depth and after that the profile

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 13

of concentration difference gradually decreases with increasing ξ . Since the amount of suspended sediment decreases with increasing ξ , the
effect of FS decreases and decreasing profile of concentration difference is obtained. It is worth mentioning here that the magnitude of
concentration difference is less than in the absence of nh, v f1 and FL. It implies that the influence of FS is less on the concentration profile in
comparison to other turbulent features.
From the above discussion, we can say that nh and FL have the most influential role near the channel bed whereas v f1 and FS dominate up to
a considerable distance from the channel bed. In addition, all the additional turbulent features play important role on the concentration profile in
the turbulent logarithmic layer. Since the sediment concentration decreases with increasing vertical height from the channel bed, the importance
of these turbulent features becomes insignificant as the suspension region moves near to free surface of the open channel. Overall, the dominant
order of these turbulent features in the suspension region can be arranged as FL > nh > v f1 > FS. It can be claimed from the observed magnitude
of concentration difference that the additional hydrodynamics mechanisms need to be included in the modeling of suspended sediment
concentration.

6.2. Importance of the co-existence of hydrodynamic mechanisms

Apart from giving the justification of including different hydrodynamic mechanisms in our model other than upward and downward sediment
fluxes commonly used in other models, we need to explain the importance of their co-existence in determining particle concentration profile. We
explain this issue by two steps. At first, in the suspension region, we delineate the importance of simultaneous interaction of the additional
hydrodynamics included in the present model. After that, we provide an insight into the limitations of previous models due to not considering
these hydrodynamic mechanisms.

6.2.1. Significance of the simultaneous interaction of additional hydrodynamics


To convey sediment from the channel bed to the suspension region, the simultaneous interaction of additional hydrody-namic mechanisms
plays a significant role. This is discussed as follows. At the channel bed, the sediment experience high value of fluid induced lift force FL as
shown in Section 6.1. Therefore, a huge amount of particles are lifted into the sus-pension and higher concentration is obtained immediately near
the channel bed. Due to the presence of a large number of suspended sediment near the channel bed, the downward settlement of a particle in
suspension is hindered due to the other suspended particles and the hindered settling mechanism in terms of nh starts to play a crucial role there.
Then the suspended particles are conveyed near the channel bed to the upper suspension region by the upward vertical velocity v f1 of fluid. With
the increasing amount of particles in suspension, the Reynolds normal stress of the suspended sediment in-creases which contributes an
additional upward vertical velocity in the clear fluid. Therefore, the strength of v f1 increases to carry the suspended sediment in upper region.
Any sediment-laden flow also wants to maintain an equilibrium state in the suspension region by the downward settlement of suspended
sediment. When the amount of suspended particles increase in a water column along the vertical direction, then they start to diffuse into the
adjacent columns where the quantity of suspended sediment is less. These diffused particles settle along the vertical downward direction as they
are not influenced by the simultaneous interaction of FL, nh, v f1 and FS as the strength of these turbulent features is smaller there due to the less
amount of sediment. The deposited particles again enter in the mechanism of upward movement as discussed and the cycling process is
continuing during the flow. An understanding of the simultaneous interaction of the hydrodynamic effects can be perceived clearly only by a
theoretical description, as is done in this section.

6.2.2. Limitations of previous models


The limitations of previous models [8,9,13–16,20,22] due to not considering the additional hydrodynamics is addressed by explaining the
physics of deviations between experimental data and the concentration profiles computed from them. In this regard, we select the non-dilute
experimental run SQ3 from Xingkui and Ning [61] as the measured data is available throughout the flow depth and the turbulent hydrodynamics
show prominent role in high concentration. We show the comparison between this run with other existing models selected and our model in Fig.
8.
It can be seen from Fig. 8(a) that the traditional model M1 of Rouse [8] deviates from experimental run SQ3 throughout the flow depth. Fig.
8(b) depicts that older model M2 of Hunt [9] shows the same nature of deviation as observed for M1. Though the models were constructed based
on upward and downward sediment fluxes, the models did not include fluid induced lift force FL, hindered settling and vertical velocity of fluid;
as a result there is disagreement between experimental data and computed concentration profiles.

The comparison between SQ3 and model M3 of Greimann et al. [13] is shown in Fig. 8(c) where the model underes-timates the experimental
data measured near the channel bed and overestimates the data far from the channel bed. The main limitation of model M3 is the assumption of
constant value of Reynolds normal stress of the suspended sediment. This assumption gives zero value of the gradient FS which results in a
concentration difference near the channel bed. The reason for the deviation far from the channel bed is due to not including vertical velocity v f1
of fluid which carries sediment into the upper suspension region.

Near the channel bed, good agreement is observed in Fig. 8(d) between SQ3 and model M4 by Greimann and Holly [14]; however the
disagreement is prominent far from the channel bed. The model included particle–particle interaction which plays a significant role near the
channel bed and leads to good agreement there. On the other hand, the model did not

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
14 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Fig. 8. Comparison between experimental run SQ3 of Xingkui and Ning [61] and computed concentration by (a) M1 (Rouse, [8]), (b) M2 (Hunt, [9]), (c) M3 (Greimann et al.,
[13]) (d) M4 (Greimann and Holly, [14]), (e) M5 (Jiang et al., [15]), (f) M6 (Zhong et al., [16]), (g) M7 (Fu et al., [22]), (h) M8 (Kundu and Ghoshal, [20]) and (i) PM (our model
given by Eq. (30)).

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 15

include the effects of secondary currents which results in a disagreement far from the channel bed as discussed for model M3.

Fig. 8(e) depicts the comparison between SQ3 and model M5 derived by Jiang et al. [15]. It is observed that model M5 cannot approximate
the experimental data throughout the flow depth. Basically the model M5 is a modified version of M3 where turbulent diffusion coe fficient was
developed together with varying magnitude of fluid and Reynolds normal stresses. This modification results in decreased overestimation of
computed profiles far from the channel bed compared with model M3; however the model M5 is likely to be constrained due to considering
several assumptions which cause the deviation of experimental data from calculated concentration throughout the flow depth.

It can be seen from Fig. 8(f) that model M6 of Zhong et al. [16] shows discrepancy with SQ3 far from the channel bed whereas good
agreement is observed near the channel bed. Model M6 was developed by taking into account the fluid induced lift force FL on suspended
particles and its importance has been understood from the good agreement near the channel bed. Though the model explicitly considered several
forces of sediment-laden flow, still the limitation of this model is the assumption of constant value for ur in the expression of FL. The assumption
is not logical as ur varies with vertical height from the channel bed [13,15,43]. Due to this assumption, an error occurs in the value of FL which
could be responsible for the less entrainment of particles in suspension compared with the expected value. The smaller entrainment rate near the
channel bed results in less transport of sediment into the upper region and this may be the cause for the overestimation of computed values far
from the channel bed.

It is clear from Fig. 8(g) that the kinetic theory based model M7 of Fu et al. [22] approximates the measured data; however a disagreement is
observed near the channel bed. In spite of considering FL and FS, the model avoided the effect of hindered settling which plays a significant role
in predicting concentration profile near the channel bed. Moreover, the model did not include v f1 . The neglect of nh and v f1 is responsible for
the deviation near the channel bed.
Kundu and Ghoshal [20] included the effect of stratification and v f1 in their model M8 which follows the trend of SQ3 in Fig. 8(h); however
a deviation occurs a small distance from the channel bed. The effect of stratification causes good agree-ment near the channel bed as its impact is
insignificant far from the channel bed due to the rapidly decreased magnitude of sediment concentration. The characteristic of M8 in following
the trend of experimental data indicates the importance of including v f1 for modeling concentration profile.

In comparison to the models M1 and M8, the present model shows better agreement with SQ3 which can be ob-served from Fig. 8(i). We
2
quantify also the coefficient of determination R for each model and the computed values are shown in the corresponding graph of Fig. 8. We
2
observe that our model gives the highest value of R with respect to other models. Good agreement between measured and computed values for
2
the present model together with highest value of R demonstrates the requirement of nh, v f1 , FL and FS and their co-existence in the modeling
of suspended sediment concentration.
Apart from the models considered for comparison in the present study, there are other works done in the last few decades to predict fluid
velocity and particle concentration profile simultaneously [2,21,54,66]. Here difficulty arises when agreement analysis is performed with
experimental data because the computed profiles hardly show good approximation with data of both fluid velocity and particle concentration. For
the sake of good agreement, several empirical parameters are introduced by the researchers and the accurate estimation of these parameters
becomes another tough task. Recently, the velocity and concentration profiles are predicted by Pal and Ghoshal [67] with fewer empirical
parameters; however, they were unable to verify their model with a large range of experimental data as done in the present study. Therefore, the
proposed model may be considered as an advanced and generalized version of the previous models in terms of considering additional
hydrodynamic mechanisms and predicting experimental data of suspended sediment concentration for a wide spectrum of turbulent conditions.

7. Conclusions

We have proposed a theoretical model for the determination of particle concentration distribution in sediment-laden open channel turbulent
flow. The model is developed based on the six hydrodynamic mechanisms which are upward sedi-ment flux due to the turbulent mixing,
gravitational settling of the sediment along the downward direction, hindered settling phenomenon, secondary current, fluid induced lift force on
the suspended particles and the gradient of Reynolds normal stress of the suspended sediment. The consideration of several physical mechanisms
of sediment-laden flow imply that the present model is a more general one than previous models. A detailed explanation is also given about the
importance of the co-existence of additionally considered hydrodynamic mechanisms. The model takes the form of a first order ordinary
differential equation; therefore its numerical solution is obtained easily. Another major advantage of applying this model for assessing
concentration distribution is that only the inverse of Schmidt number β needs to be estimated from experimental data.

Our model shows satisfactory agreement with a wide region of existing experimental data in open channel turbulent flow laden with
sediment. An error analysis confirms the better predictive accuracy of our model in comparison to many other existing models which are
constrained by limited consideration of underlying physics of sediment-laden flow. Finally, we can conclude that because of the ability to fit
laboratory and field observations closely, we can apply the present model in computing the concentration profile for a wide spectrum of particle-
laden open channel flow.

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
ARTICLE IN PRESS
JID: APM [m3Gsc;March 9, 2017;4:24]
16 D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17

Acknowledgments

The authors express sincere thanks to five reviewers for giving their valuable comments and suggestions to improve the manuscript.

References

[1] T.J. Hsu, J.T. Jenkins, P.L.F. Liu, On two-phase sediment transport: sheet flow of massive particles, Proc. R. Soc. Lond. A: Math. Phys. Eng. Sci. 460 (2048) (2004) 2223–
2250.
[2] B.S. Mazumder, K. Ghoshal, Velocity and concentration profiles in uniform sediment-laden flow, Appl. Math. Model. 30 (2) (2006) 164–176.
[3] B. Camenen, M. Larson, A general formula for noncohesive suspended sediment transport, J. Coast. Res. 24 (3) (2008) 615–627.
[4] S.K. Jha, F.A. Bombardelli, Toward two-phase flow modeling of nondilute sediment transport in open channels, J. Geophys. Res. Earth Surf. 115 (F3) (2010) F03015(1–27).
[5] J. Bai, H. Fang, T. Stoesser, Transport and deposition of fine sediment in open channels with different aspect ratios, Earth Surf. Process. Landf. 38 (6) (2013) 591–600.

[6] R. Bravo, P. Ortiz, J.L. Pérez-Aparicio, Incipient sediment transport for non-cohesive landforms by the discrete element method (DEM), Appl. Math. Model. 38 (4) (2014)
1326–1337.
[7] W.B. Chen, W.C. Liu, M.H. Hsu, C.C. Hwang, Modeling investigation of suspended sediment transport in a tidal estuary using a three-dimensional model, Appl. Math.
Model. 39 (9) (2015) 2570–2586.
[8] H. Rouse, Modern conceptions of the mechanics or fluid turbulence, Trans. Am. Soc. Civ. Eng. 102 (1) (1937) 463–505.
[9] J.N. Hunt, The turbulent transport of suspended sediment in open channels, Proc. R. Soc. Lond. Ser. A. Math. Phys. Sci. 224 (1158) (1954) 322–335.
[10] M. Umeyama, Vertical distribution of suspended sediment in uniform open-channel flow, J. Hydraul. Eng. 118 (6) (1992) 936–941.
[11] S.H. Huang, Z.L. Sun, D. Xu, S.S. Xia, Vertical distribution of sediment concentration, J. Zhejiang Univ. Sci. A 9 (11) (2008) 1560–1566.
[12] S.K. Bose, S. Dey, Suspended load in flows on erodible bed, Int. J. Sedim. Res. 24 (3) (2009) 315–324.
[13] B.P. Greimann, M. Muste, J.F.M. Holly, Two-phase formulation of suspended sediment transport, J. Hydraul. Res. 37 (4) (1999) 479–500.
[14] B.P. Greimann, J.F.M. Holly, Two-phase flow analysis of concentration profiles, J. Hydraul. Eng. 127 (9) (2001) 753–762.
[15] J. Jiang, A.W.K. Law, N.S. Cheng, Two-phase modeling of suspended sediment distribution in open channel flows, J. Hydraul. Res. 42 (3) (2004) 273–281.
[16] D. Zhong, G. Wang, Q. Sun, Transport equation for suspended sediment based on two-fluid model of solid/liquid two-phase flows, J. Hydraul. Eng. 137
(5) (2011) 530–542.
[17] S.R. McLean, On the calculation of suspended load for noncohesive sediments, J. Geophys. Res. Oceans 97 (C4) (1992) 5759–5770.
[18] S. Wright, G. Parker, Flow resistance and suspended load in sand-bed rivers: simplified stratification model, J. Hydraul. Eng. 130 (8) (2004) 796–805.
[19] K. Ghoshal, S. Kundu, Influence of secondary current on vertical concentration distribution in an open channel flow, ISH J. Hydraul. Eng. 19 (2) (2013) 88–96.

[20] S. Kundu, K. Ghoshal, Effects of secondary current and stratification on suspension concentration in an open channel flow, Environ. Fluid Mech. 14 (6) (2014) 1357–1380.

[21] S.Q. Yang, Turbulent transfer mechanism in sediment-laden flow, J. Geophys. Res. Earth Surf. 112 (F1) (2007) F01005(1–14).
[22] X. Fu, G. Wang, X. Shao, Vertical dispersion of fine and coarse sediments in turbulent open-channel flows, J. Hydraul. Eng. 131 (10) (2005) 877–888.
[23] M.J. Herrmann, O.S. Madsen, Effect of stratification due to suspended sand on velocity and concentration distribution in unidirectional flows, J. Geo- phys. Res. Oceans 112
(C2) (2007) C02006(1–13).
[24] J. Richardson, W. Zaki, Sedimentation and fluidisation: Part I, Trans. Inst. Chem. Eng. 32 (1954) 35–53.
[25] S.Q. Yang, S.K. Tan, S.Y. Lim, Velocity distribution and dip-phenomenon in smooth uniform open channel flows, J. Hydraul. Eng. 130 (12) (2004a) 1179–1186.

[26] S.Q. Yang, S.K. Tan, S.Y. Lim, Velocity and sediment concentration profiles in sediment-laden flows, Chin. Ocean Eng. 18 (2) (2004b) 229–244.
[27] Z.Q. Wang, N.S. Cheng, Secondary flows over artificial bed strips, Adv. Water Resour. 28 (5) (2005) 441–450.
[28] M. Muste, V.C. Patel, Velocity profiles for particles and liquid in open-channel flow with suspended sediment, J. Hydraul. Eng. 123 (9) (1997) 742–751.
[29] M. Cellino, W.H. Graf, Sediment-laden flow in open-channels under noncapacity and capacity conditions, J. Hydraul. Eng. 125 (5) (1999) 455–462.
[30] S.Q. Yang, Influence of sediment and secondary currents on velocity, Water Manag. 162 (WM5) (2009) 299–307.
[31] S.Q. Yang, Interactions of boundary shear stress, secondary currents and velocity, Fluid Dyn. Res. 36 (3) (2005) 121–136.
[32] T. Ohmoto, Z. Cui, R. Hirakawa, Effects of secondary currents on suspended sediment transport in an open channel flow, in: Proceedings of the International Symposium on
Shallow Flows, Taylor & Francis, Delft, Netherlands, 2004, pp. 511–516.
[33] K. Ghoshal, R. Mazumder, C. Chakraborty, B.S. Mazumder, Turbulence, suspension and downstream fining over a sand-gravel mixture bed, Int. J. Sedim. Res. 28 (2) (2013)
194–209.
[34] R. Di Felice, The voidage function for fluid–particle interaction systems, Int. J. Multiph. Flow 20 (1) (1994) 153–159.
[35] S. Dey, Sediment threshold, Appl. Math. Model. 23 (5) (1999) 399–417.
[36] G.A. Kallio, M.W. Reeks, A numerical simulation of particle deposition in turbulent boundary layers, Int. J. Multiph. Flow 15 (3) (1989) 433–446.
[37] J.B. McLaughlin, The lift on a small sphere in wall-bounded linear shear flows, J. Fluid Mech. 246 (1993) 249–265.
[38] M. Cellino, W.H. Graf, Measurements of suspension flow in open channels, in: Proceedings of the of Twenty-Seventh IAHR Congress, San Francisco, 1997, pp. 179–184.

[39] M. Garcia, Y. Nino, F. Lopez, Laboratory observations of particle entrainment into suspension by turbulent bursting, in: Coherent Flow Structures in Open Channels, Wiley,
1996, pp. 63–86.
[40] M. Muste, K. Yu, I. Fujita, R. Ettema, Two-phase versus mixed-flow perspective on suspended sediment transport in turbulent channel flows, Water Resour. Res. 41 (10)
(2005) W10402(1–22).
[41] I. Nezu, W. Rodi, Open-channel flow measurements with a laser doppler anemometer, J. Hydraul. Eng. 112 (5) (1986) 335–355.
[42] N.S. Cheng, Effect of concentration on settling velocity of sediment particles, J. Hydraul. Eng. 123 (8) (1997) 728–731.
[43] N.S. Cheng, Analysis of velocity lag in sediment-laden open channel flows, J. Hydraul. Eng. 130 (7) (2004) 657–666.
[44] S. Zhiyao, W. Tingting, X. Fumin, L. Ruijie, A simple formula for predicting settling velocity of sediment particles, Water Sci. Eng. 1 (1) (2008) 37–43.
[45] N.S. Cheng, Simplified settling velocity formula for sediment particle, J. Hydraul. Eng. 123 (2) (1997) 149–152.
[46] J.P. Ahrens, A fall-velocity equation, J. Waterw. Port Coast. Ocean Eng. 126 (2) (2000) 99–102.
[47] T.E. Baldock, M.R. Tomkins, P. Nielsen, M.G. Hughes, Settling velocity of sediments at high concentrations, Coast. Eng. 51 (1) (2004) 91–100.
[48] D. Pal, K. Ghoshal, Hindered settling with an apparent particle diameter concept, Adv. Water. Resour. 60 (2013) 178–187.
[49] N.S. Cheng, A diffusive model for evaluating thickness of bedload layer, Adv. Water Resour. 26 (8) (2003) 875–882.
[50] K. Ghoshal, D. Pal, An analytical model for bedload layer thickness, Acta Mech. 225 (3) (2014) 701–714.
[51] M. Garcia, G. Parker, Entrainment of bed sediment into suspension, J. Hydraul. Eng. 117 (4) (1991) 414–435.
[52] J.A. Zyserman, J. Fredsøe, Data analysis of bed concentration of suspended sediment, J. Hydraul. Eng. 120 (9) (1994) 1021–1042.
[53] Z. Cao, Equilibrium near-bed concentration of suspended sediment, J. Hydraul. Eng. 125 (12) (1999) 1270–1278.
[54] B.S. Mazumder, K. Ghoshal, Velocity and suspension concentration in sediment-mixed fluid, Int. J. Sedim. Res. 17 (3) (2002) 220–232.
[55] R. Absi, An ordinary differential equation for velocity distribution and dip-phenomenon in open channel flows, J. Hydraul. Res. 49 (1) (2011) 82–89.
[56] N.L. Coleman, Effects of suspended sediment on the open-channel velocity distribution, Water Resour. Res. 22 (10) (1986) 1377–1384.

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045
JID: APM
ARTICLE IN PRESS [m3Gsc;March 9, 2017;4:24]

D. Pal, K. Ghoshal / Applied Mathematical Modelling 000 (2017) 1–17 17

[57] D.A. Lyn, A similarity approach to turbulent sediment-laden flows in open channels, J. Fluid. Mech. 193 (1) (1988) 1–26.
[58] S.J. Bennett, J.S. Bridge, J.L. Best, Fluid and sediment dynamics of upper stage plane beds, J. Geophys. Res. Oceans 103 (C1) (1998) 1239–1274.
[59] D.G. Wren, S.J. Bennett, B.D. Barkdoll, R.A. Kuhnle, Distributions of velocity, turbulence, and suspended sediment over low-relief antidunes, J. Hydraul. Res. 43 (1) (2005)
3–11.
[60] H.A. Einstein, N. Chien, Effects of heavy sediment concentration near the bed on velocity and sediment distribution, Missoury Riv. Div., Corps of Eng. US Army 33 (2)
(1995) 78.
[61] W. Xingkui, Q. Ning, Turbulence characteristics of sediment-laden flow, J. Hydraul. Eng. 115 (6) (1989) 781–800.
[62] J. Zheng, R.J. Li, Q. Fend, S.S. Lu, Vertical profiles of fluid velocity and suspended sediment concentration in nearshore, Int. J. Sedim. Res. 28 (3) (2013) 406–412.

[63] J.B. McLaughlin, Inertial migration of a small sphere in linear shear flows, J. Fluid Mech. 224 (1991) 261–274.
[64] L.C.v. Rijn, Sediment transport, part ii: suspended load transport, J. Hydraul. Eng. 110 (11) (1984) 1613–1641.
[65] W.H. Graf, M. Cellino, Suspension flows in open channels; experimental study, J. Hydraul. Res. 40 (4) (2002) 435–447.
[66] C.H. Tsai, C.T. Tsai, Velocity and concentration distributions of sediment-laden open channel flow, J. Am. Water Resour. Assoc. 36 (5) (2000) 1075–1086.
[67] D. Pal, K. Ghoshal, Vertical distribution of fluid velocity and suspended sediment in open channel turbulent flow, Fluid Dyn. Res. 48 (3) (2016) 035501.

Please cite this article as: D. Pal, K. Ghoshal, Hydrodynamic interaction in suspended sediment distribution of open channel turbulent flow,
Applied Mathematical Modelling (2017), http://dx.doi.org/10.1016/j.apm.2017.02.045

Вам также может понравиться