You are on page 1of 13

FMFP14 – A – 50

SECONDARY MOTIONS IN A STRAIGHT OPEN CHANNEL OVER


ARTIFICIAL BED STRIPS

Sushant K. Biswal

In this study, cellular secondary flows are artificially generated over an alternate smooth and
rough bed strips aligned longitudinally in an open channel. Flow velocities are recorded with a
three-dimensional Acoustic Doppler Velocimeter. Experimental results reveal that the
distribution of the primary velocity and streamwise Reynolds shear stress is altered in the
presence of the secondary flows. This study also displays that the flow field can be linearized so
that a flow quantity can be decomposed into two components, one being related to the average
base flow and the other representing the perturbation caused by the bed configuration. This
perturbation-based approach provides an analytical formulation of the altered Reynolds shear
stress distribution that differs from the linear profile used for two dimensional open channel
flows.
Secondary motions in a straight open channel over artificial
bed strips

Sushant K. Biswal, PhD

Assistant Professor, Civil Engineering Department


National Institute of Technology Agartala, e-mail: drsushantbiswal@gmail.com

Abstract. In this study, cellular secondary flows are artificially generated over an alternate
smooth and rough bed strips aligned longitudinally in an open channel. Flow velocities are
recorded with a three-dimensional Acoustic Doppler Velocimeter. Experimental results reveal
that the distribution of the primary velocity and streamwise Reynolds shear stress is altered in
the presence of the secondary flows. This study also displays that the flow field can be linear-
ized so that a flow quantity can be decomposed into two components, one being related to the
average base flow and the other representing the perturbation caused by the bed configuration.
This perturbation-based approach provides an analytical formulation of the altered Reynolds
shear stress distribution that differs from the linear profile used for two dimensional open chan-
nel flows.

Keywords: Open channel flow; Secondary flow; Velocity profile; Reynolds


shear stress

1. Introduction

Secondary flows are commonly ubiquitous in natural streams and in laboratory flumes
either in the laminar or turbulent conditions. Prandtl (1952) classified steady second-
ary flows into two categories. The secondary flow of Prandtl’s first kind is that in-
duced by skewing of the mean flow in curved channels or meandering rivers.
Prandtl’s second kind of secondary flows are those caused by the cross-sectional non-
homogeneity and anisotropy of turbulence. They are also called shear-or turbulence-
driven secondary flows and are initiated by the sidewall effect, asymmetry of channel
boundaries, free-surface effect or the variation of bed conditions and instabilities in
turbulent flows in straight channels or non-circular ducts. Secondary currents influ-
ence significantly the mean flow and turbulence characteristic of the flow. However,
many researchers have inferred that the existence of secondary flows in straight wide
open channels attributed to periodic lateral variations of streamwise velocity and sus-
pended sediment concentration. The vorticity equation governing the secondary cur-
rents has the following form:

W
d:
dz
V
d:
dy
d 2 '2
dydz
§ d2
d2 ·
v  w  ¨¨ 2  2 ¸¸v cwc  X’ 2 :
'2

© dz dy ¹
(1)

adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
dW dV
: 
dy dz

where Ω is the vorticity, V and W are the mean velocities of secondary currents in
spanwise and vertical directions, respectively v c and wc are the velocity fluctuations
and X is the kinematic viscosity. Generally, in fluvial environment, longitudinal bed
forms appear as periodic and refer to large-scale bed elements that are elongated
parallel with the primary flow and characterized by alternate bed roughness or/and
elevation variations in the spanwise direction. Owing to the recurrence of bed forms
in the lateral direction, the generated secondary flows often appear as pair of counter-
rotating flow cells in the cross stream plane. Therefore, such secondary flows are
usually called cellular secondary flows and are such vortices that can be specified by
channel bed conditions.
The effect of bed form on flow characteristics have been studied experimen-
tally and numerically by several investigators. Tsujimoto (1989) conducted an exper-
iment and found that the non-uniformity of bed particles contributes to generating or
strengthening the cellular secondary currents. Experimental results of Nezu and Nak-
agawa [1993] showed that lateral variations in bed topography and roughness can lead
to the formation of secondary currents, which are independent of the sidewall effect
or the corner induced secondary currents. Authors remarked that “the mechanisms for
initiation and maintenance of cellular secondary currents in wide channel flows are
not yet well understood.” Colombini (1993) conducted a linear stability analysis of
flow in wide channels with an erodible bed and indicated that the initiation of the
secondary flow and sand ridges is intrinsically associated with instability of the erodi-
ble bed rather than the effect of sidewall-induced vortices. Colombini (1993) showed
that the erodible bottom with uniform sediment is responsible for initiating mecha-
nism of cellular secondary currents rather than corner vortices by the sidewalls. Wang
and Cheng (2003, 2005) examined experimentally cellular secondary flows induced
by the lateral roughness variation in a flat-bed channel and suggested that the cellular
secondary flows are such vortices that can be specified by channel bed conditions.
Souliotis and Prinos (2011), made an experiment over number of parallel smooth and
rough patches in a laboratory channel. They examined the effect of these patches on
different aspects of flow behaviour and concluded that such a scenario is not only a
common situation in practical channel but also affects the flow behaviour considera-
bly.
The above brief literature review clearly demonstrates that there are still
knowledge gaps on the formation of secondary currents. Regarding the generation of
the cellular secondary currents, no clear mechanism has been demonstrated so far.
Two factors have been thought to be associated with initiating cellular secondary
currents, i.e., the corner-induced vortices and the bottom sediment. In this paper,
characteristics of cellular secondary flows artificially generated with a bed configura-
tion, which comprised alternate smooth and rough strips aligned longitudinally in an
open channel are explored experimentally. Based on the experimental data collected,
analytical formulae are then proposed for the description of vertical profiles of the
streamwise velocity and Reynolds shear stress, which are modified by the secondary
flows. The main objectives of this study are to: (1) investigate the driving force of the
secondary currents; (2) illustrate mathematically whether the boundary plays a crucial
role in the formation of secondary currents; and (3) demonstrate the interactions of
secondary currents, sediment transport and morpho-dynamics.

2. Experimental setup

Schematic diagram of the experimental setup is shown in Figure 1. In this study, ex-
periments are carried out in a straight rectangular brick masonry flume with a bottom
slope of 0.0008 and are of 12m long, 0.7m wide and 0.6m deep. The flume bed com-
prises smooth and rough longitudinal strips of O 100mm wide each which are jux-
taposed in an alternate manner as shown in Figure 1. The rough strip is prepared by
sediment particles of uniform median diameter of 2.0 mm, and densely packed. The
surface of all the strips is set at the same level across the channel. Water is discharged
from upstream constant head tank which is controlled by valves in the pumping cir-
cuit. The flow depth at the outlet is controlled by an adjustable tailgate. In the channel
inlet section, a set of honeycomb and perforated plates are mounted to reduce turbu-
lence and obtain parallel flow over the channel length. In order to obtain the velocity
distribution in the channel, a three-dimensional downlooking Acoustic Doppler veloc-
imeter (Vectrino, 2009) is employed.

Fig. 1. Schematic diagram of the experimental setup


The shortest distance from the measuring point to the bed was 15mm, and spacing
10mm was adopted between grid points. At each location, the sampling duration is
taken as 240s, and sampling rate is about 50Hz. The data analysis yielded the mean
velocities and other statistical properties. The origin of the coordinate system is the
bed at the centre of the channel (Figure 2). The flow discharge is varied from 0.15-
0.25m3/s during experiment and the average velocity in the central zone of the chan-
nel is in range of 0.45-0.60 m/s. Observations for velocity is recorded without chang-
ing the flow settings.

Fig. 2. Cross-sectional view of the channel bed form with two-dimensional sampling meshes
for ADV measurement

3. Results and Discussion

3.1 Velocities

Fig. 3 shows the contour maps of the normalized time-average streamwise and verti-
U W
cal velocities ( and respectively) over the tested section. Here U is the depth
U U
averaged velocity in the central region (y/h = 0.42 ~ -0.42). The figure clearly shows
the downflow occurs over the rough strips while upflow occurs over the smooth
strips. In the central region, the maximum vertical velocity is about 2.15% of the av-
erage primary velocity while the maximum downflow velocity is about 1.75%. The
contours of velocity are gradually skewed while approaching the side wall, especially
y
3.5   2.5 . This indicates that the side wall also plays a role in generating sec-
O
ondary flows. The figure displays that the sidewall effect is strong within a distance of
about 0.67h from the sidewall, and beyond the near-sidewall region, say, 1.25h from
the sidewall, the mean flow field is subjected to the lateral variation of the bed rough-
ness. The central zone is approximately 2.5h wide, which can be considered free from
the sidewall effects. For open channel flows with uniform bed, sidewall effect is usu-
ally restricted within a lateral region (2–3) h. Therefore, it is safe to suppose here that
in the central zone, the sidewall effect is insignificant and the secondary flow ob-
served is mainly related to the bed strips. In this study, only central region is consid-
ered for the analysis, where a complete secondary flow cell occurred (see Fig.4).

Fig. 3. Figure3. Contour plots of: (a) dimensionless mean streamwise velocity (U U ) ; and
(b) dimensionless mean vertical velocity (W U ) u 10 3
in the tested half cross-section.

Due to the effect of the roughness elements on the flow, the near-bed streamwise
velocity above the rough strips is slightly less than that above the smooth strips [Fig.
3 (a)]. In contrast, the flow in the upper portion exhibits higher velocity above the
rough strips and smaller velocity above the smooth strips. The maximum velocity in
the central region approaches to the water surface so that the velocity dip disappears.
The flow vectors across the section are shown in Fig. 4, which clearly shows the pres-
ence of secondary flows. The mean magnitude of the velocity vectors is 0.025 U and
the largest magnitude is 0.085 U . A pair of secondary flow cells can be clearly rec-
ognized. However, the shape of the flow cells is not symmetric with respect to the
circulation center nearly locating over the strip interface. Near the sidewall, the circu-
lar flow comprises two flow cells and the interaction of these cells transfers the higher
momentum fluid from the surface region to the bottom corner. The central region is
characterized by the pairs of counter-rotating secondary flow cells in the scale of the
flow depth. At the bottom, the adjacent flow cells separate over the rough strip and
reattach together over the smooth strip, and vice versa at the free surface. The flow
structures illustrate that cellular secondary flows could be effectively generated and
maintained by the alternate smooth and rough bed strips or the lateral roughness var-
iation. The results indicate that secondary flow cells can be generated independent of
any particular strip widths, but the detail of the cellular structure may vary with the
width ratio. Similar results have been observed for air flows by Nezu and Nakagawa
(1984).

Fig. 4. Vector plot of cross-sectional flow velocities

3.2 Linear approximation of the flow field.

The flow subject to the generated secondary motion is assumed as a result of the per-
turbation to the base flow due to the lateral roughness variation. Colombini (1993)
proposed that the disturbed flow is a linear superimposition of the perturbation on the
base flow in the following form,
f y, z f 0 y  F y, z (2)

Where,
f is the disturbed flow quantity; f 0 is the flow quantity related to the base
flow and F is the corresponding perturbation. First, the log-law is used to represent
the profile of the primary velocity related to the base flow, i.e.

U0 1 y
ln (3)
u* N z0

U
Where, 0 is the streamwise velocity for the base flow; u * is the average shear ve-
locity; κ is the von Karman coefficient (= 0.4); and z 0 is the zero-velocity level or
hydrodynamic roughness length for the base flow. In the presence of the cellular sec-
ondary flows, the distribution of the primary velocity U would deviate from Eq. (3).
Therefore, as a first approximation, modified the log-law is to be

U 1ª y yS º
« a1 ln  a 2 sin 2 (4)
u* N¬ z 2h »¼

Where, z is the velocity level used for the logarithmic term; and a1 and a 2 are the
parameters related to the perturbation-induced variations due to the effect of the sec-

ondary flows. The term 2


a sin 2 >Sy 2h@ is proposed only to empirically account for
the curvature of the velocity profile induced by the secondary flow. In other words;
Eq. (4) would reduce to Eq. (3) if the effect of the secondary flow is not considerable.
Therefore, for this particular case, we can take 1
a 1
and 2
a 0
. To further exam-
ine the velocity profile obtained above with the concept of the linearization given by
Eq. (2), U is decomposed into two components, one representing the base-flow and
the other being related to its lateral undulation, i.e.

zS
U y, z U 0 ( y)  U 1 cos (5)
O
U
where 1 is the amplitude of the lateral velocity undulation. Substituting Eq. (5)
together with Eq. (3) into Eq. (4) yields
NU 1 ( y) zS y z yS
cos (a1  1) ln  a1 ln 0  a 2 sin 2 (6)
u* O z0 z 2h

Finally, for the primary velocity, in the form of Eq. (2), Eq. (5) can be re-written as

U y, z 1 y 1§ y yS · § zS ·
ln  ¨ G 1 ln  D  G 2 sin 2 ¸ cos¨ ¸ (7)
u* N z0 N © z 2h ¹ © O ¹

Eq. (7) shows the modification of the logarithmic distribution of the streamwise ve-
locity due to the presence of the cellular secondary flow.

3.3 Reynolds shear stress.

Reynolds stresses allowing an accurate prediction of the turbulence anisotropy that


controls the generation of secondary flows. Figure 5 shows the contours of the dimen-
sionless Reynolds shear stress  U u cv c . It is normalized with the average bed shear
stress computed as W b = ρghS in the central region. It can be seen that the shear stress
varies with the bed configuration. In this study, the eddy viscosity concept is assumed
to be applicable to the streamwise Reynolds shear stress  U u cvc and can be related
to the gradient of the streamwise velocity as

dU
 U u cv c UXt (8)
dy
in which ρ is the density of fluid; and νt is the eddy viscosity and can be further ap-
proximated to be

Sz
Xt X t 0  X t1 cos (9)
O

X X
where t 0 is the eddy viscosity related to the base flow; and t1 is the amplitude of
the viscosity perturbation. Substituting Eqs. (5) and (9) into Eq. (8) and ignoring the
second order perturbation yields

dU 0 § dU 1 dU 0 · Sz
 u cvc X t 0  ¨¨X t 0  X t1 ¸¸ cos (10)
dy © dy dy ¹ O
Eq. (10) indicates that the lateral variation of  U u cv c depends on the perturbations
induced by the primary velocity and the eddy viscosity, respectively. Furthermore,
consider that the eddy viscosity can also be modelled using the mixing length, l,

dU
Xt l2 (11)
dy

The mixing length l is decomposed into two components as

Sz
l l 0  l1 cos (12)
O

l
where 0 is the mixing length for the base flow; and l1 is the amplitude of the mixing
length perturbation. Substituting Eqs. (5), (9), and (12) into Eq. (11) and ignoring
perturbation terms higher than the first-order yields

Xt 0 dU 1 dy l
2 1 (13)
Xt1 dU 0 dy l0

where
vt 0 is taken as
l 2 dU 0 dy for the base flow. Noting that
 u cvc u*2 1  y h for base flows, with equations (3), (7) and (13), Eq. (10) can
be re-written as

 u cvc § y ·ª § y Sy · Sz º
¨1  ¸ «1  ¨ G 3  G 4 sin ¸ cos » (14)
u*2 © h ¹¬ © h h¹ O¼

where 3
G 1 2G  2 l l
1 0 and 4 G SG
2 . Obviously, these two coefficients are

associated with the amplitude of the shear stress undulation. To compare with the
measured shear stress profiles, Eq. (14) is rearranged in the following form, which is
similar to Eq. (4) proposed for the primary velocity,

 ucvc § y· § y· y Sy
a3 ¨1  ¸  a4 ¨1  ¸ sin (15)
u 2
*
© h¹ © h¹h h

a3 1  G 3 cos Sz O G 4 cos Sz O
where and a4
Fig. 5. Contour plot of Reynolds shear stress

Figure 6 demonstrates that the measured shear stress profiles can be well represented
by Eq. (15). The deviation of  u cv c from the linear distribution that is associated
with the base flow appears differently over the different bed strips. Higher shear
stresses occur in the upper flow portion above the smooth strip, whereas the shear
stress becomes smaller over the rough strip for z/h > 0.2. Near the bed, the values of
 u cv c over the smooth strip tend to be smaller than those over the rough strip.

Fig. 6. Distributions of streamwise Reynolds shear stress. Scattered points denote experimental
data; solid lines are computed using Eq. (15).
The value of  u cv c gives an indication of the strength of the momentum exchange at
the interface between the rough and smooth strips due to the sudden change in the bed
roughness. Since it is the velocity gradient dU dy which is associated with  u cv c .
Colombini and Parker (1995) have shown that these flows could be examined in quite
similar approaches, which were developed based on the concept of mixing length.
They also reported that the vertical structure of the mixing length played an important
role in deciding the direction of flow cell rotation.

4. Conclusions
Turbulent open channel flows over alternate smooth and rough strips aligned longitu-
dinally was experimentally conducted. A three dimensional downlooking Acoustic
Doppler Velocimeter was employed to measure the instantaneous flow velocities. The
measurements depicted that secondary flow induced comprises corner vortices near
the sidewall and the cellular motions in the central region of the channel. Lateral dis-
tributions of the primary velocity and Reynolds shear stress are not uniform but vary
with the secondary flow structure.
The vertical profile of the primary velocity deviates from the log-law due to
presence of secondary flow. The deviation is considerable, in particular, for the region
where the upflow or downflow prevails. The results shows that the flow field is line-
arized so that a flow quantity can be decomposed into two components, one being
related to the average base flow and the other symbolizing the perturbation caused by
the bed configuration. This perturbation-based approach leads to an analytical formu-
lation of the altered Reynolds shear stress distribution significantly differs from the
linear profile applicable for 2D open channel flows. It is noted that the formulae pro-
posed for describing the distributions of the streamwise velocity and Reynolds shear
stress may be only applicable for limited bed and flow conditions. This study suggests
that secondary flows could be described reasonably by simple analytical expressions
in the sinusoidal form, which are applicable for various secondary flow structures
with different patterns. Further studies need to be done to generalize the analysis and
also to extend it to the case of secondary flows that are induced by lateral variations in
the bed elevation.

Reference
1. Colombini, M., 1993. Turbulence driven secondary flows and the formation of sand ridges.
J. Fluid Mech. 254, 701-719.
2. Colombini, M. and Parker, G., 1995. Longitudinal streaks. J. Fluid Mech. 304, 161-183.
Ikeda, S., 1981. Self-formed straight channels in sandy bed. ASCE, Journal of Hydraulic
Division 107, 389-406.
3. Nezu I, Nakagawa H. 1984. Cellular secondary currents in straight conduit. ASCE, J Hy-
draul Eng;110,173–93.
4. Nezu, I. and Nakagawa, H., 1993. Turbulence in open channels. IAHR/AIRH Monograph,
Balkema, Rotterdam, The Netherlands.
5. Prandtl, L., 1952. Essentials of fluid dynamics. Hafner Publishing Company, New York,
U.S.
6. Souliotis, D., and Prinos, P., 2011. Effect of a Vegetation Patch on Turbulent Channel
Flow, Journal of Hydraulic Research, Volume 49, No. 2, pp. 157-167.
7. Tsjimoto, T. 1989. “Longitudinal stripes of sorting due to cellular secondary currents.”
Journal of Hydroscience and Hydraulic Engineering, 7(1), 14-18.
8. Vectrino 2009. “Vectrino Velocimeter, User guide.” Nortek AS, Vangkroken 2, No-1351,
Norway.
9. Wang, Z. Q., Cheng, N. S., Chiew, Y. M. and Chen, X. W., 2003. Secondary flows in open
channel with smooth and rough bed strips. Proceeding of 30th IAHR Congress, Thessalo-
niki, Greece, Theme C, Vol. 1, 111-118.
10. Wang, Z.-Q. and Cheng, N.-S. 2005. “Secondary flows over artificial bed strips.” Advanc-
es in Water Resources, 28, 441-450.
11. Wang, Z.Q. and Cheng, N.S. 2006. Time-mean structure of secondary flows in open chan-
nel with longitudinal bed forms. Adv. Water Resources, pp. 1-16.

You might also like