DUET Journal Vol.

1, Issue 1, June 2010

Efficiency of an Expansive Transition in an Open Channel Subcritical Flow

B. C. Basak and M. Alauddin

Department of Civil Engineering

Dhaka University of Engineering & Technology, Gazipur, Bangladesh
E-mail: mauddin_duet@yahoo.com

ABSTRACT

Open channel transitions involving an expansion of width are a common feature of canals and flumes. Subcritical flow
through an expansion transition can result in significant head loss due to separation of flow and subsequent eddy
formation. The body of the hydraulic structure is subjected to lateral vibrations due to intermittent shedding of
eddies, which are dangerous and hence, undesirable. Moreover, uneven distribution of velocity may cause asymmetry
of flow and thus develop scour at places of highly concentrated velocities. This paper presents the results of
experimental investigations on subcritical flow through gradual expansion in rectangular rigid-bed channels. The
velocity distributions of flow through the transition models are made, thus, the efficiencies of the transitions evolved
by different investigators are evaluated.

1. INTRODUCTION

Open channel expansions for subcritical flow are b0

F low
bx bL
encountered in the design of hydraulic structures such as
aqueducts, siphons, barrages, and so on. In these structures
x
the flow tends to separate while subjected to the positive
L
pressure gradient associated with flow deceleration, thus
resulting in a considerable loss of energy. In an expanding PLAN

flow, the distribution of velocity in the cross section can be

extremely uneven, and uneven distribution of velocity may
cause asymmetry of flow and thus develop scour at places
yx yL
of highly concentrated velocities. This study involves the y0

performance-evaluation of transitions evolved by different S 00

investigators in an open channel subcritical flow. To
evaluate the transition profiles, efficiencies of the transition
E L E V A T IO N
models are determined in a laboratory setup flow, defining
this as the ratio of gain of potential energy to loss of kinetic Fig. 1: Definition sketch: Rectangular expansion transition
energy.
surface profile. For this purpose Eq. (1) is taken as a form
2. AVAILABLE METHODS loss equation. The loss due to surface resistance is
neglected, as it is small. The form loss, hL, is assumed to
Because of the importance of knowledge concerning vary uniformly along the transition length and is expressed
expansions in rigid-bed channels, several investigators as:
studied with different aspects of flow in expansion. The V 2 VL2
methods available for the design of expansion transitions hL = K H 0 (1)
were contributed by Hinds [1], Hartley et al [2], Chaturvedi 2g
[3], Nashta and Garde [4], and Swamee and Basak [5]. A where, V0 and VL = the velocity at the inlet and outlet of
brief outline of each method is given as follows (Referred expansion respectively, KH = the loss coefficient lying
to Fig. 1). between 0.3 and 0.75 [6], and g = the acceleration due to
gravity. The equations of the two reverse parabolas
Hinds [1] assumed the water-surface profile in the representing the water depth y at a distance x from the inlet
transition to be composed of two reverse parabolas of equal are given by
length connected at the centre of the transition, and found y = y 0 + 2 ( y L y 0 ) 2 ; 0 0 .5 (2a)
the bed-width profile corresponding to the assumed water-
and
y = y L 2( y L y 0 )(1 ) 2 ; 0.5 1 (2b)

Dhaka University of Engineering & Technology, Gazipur 27

DUET Journal Vol. 1, Issue 1, June 2010

in which y0 and yL = the depth of flow at inlet and outlet of width 9.525 cm and length 2.50 m was constructed and
x placed at upstream portion of the flume, the walls of which
channels respectively; and = . were of varnished wood. Then the gradual expansion of
L length 55.56 cm was provided in the reach to have the
Equating total energies at the inlet and at a section normal channel width 25.4 cm and then continued for
0 0.5 , and using Eq. (1) and (2a), the bed width remaining 1.50 m length. The transition models were made
profile is obtained as of wooden bed and the side walls of Perspex sheet. The
Q 1 KH Q2( 1 K ) length of transition governed by side splay of 7:1 has been
H
b= + 2 g S 0 L used in the present study, which is claimed to be the
y0 + 2( y L y0 ) 2 b0 2 y0 2 optimal value [7], and all the experiments were conducted
4 g ( y L y 0 ) 2 4 g ( y L y 0 ) 2 ]
0.5
(3a)
in a rectangular channel with an expansion ratio of 2.67. A
tail gate was provided at the downstream end of the flume
in which Q = the discharge, and S0 = the channel-bed slope. for depth regulation. Water was circulated through the
Similarly, applying energy equation at the outlet and at a channel by one electrically driven centrifugal pump with
section 0.5 1 and using Eq. (1) and (2b), the constant speed closed impellers. The rails were provided
along the entire working length of the flume, which
corresponding bed-width profile is
supports a moving carriage, and a continuous scale
Q 1 KH Q 2 (1 K ) calibrated in millimeters is provided along the length of one
H
b= of the rails. The carriage with pointer gauge and Pitot tube
2 2 2
y L 2( y L y 0 )(1 ) b L y L is used for depth of flow and velocity measurements
2 g S 00 L(1 ) + 4 g ( y L y0 )(1 ) 2 ]0.5
(3b) respectively (Fig. 2).

Hartley et al [2] assumed the following linear variation of

the velocity:
V = V0 + (VL V0 ) (4)
in which V = the velocity of flow at a distance x and further
assumed constant depth throughout the transition,
b0V0 = bLVL = bV (5)
Combining Eqs. (4) and (5), the bed-width profile was
obtained as:
[ (
b = b01 + bL1 b01 )] 1
(6)
Chaturvedi [3] generalized Eq. (6) in designing the
rectangular expansion transition in the following manner:

[ (
b = b0n + bLn b0n )] 1
n
where from experimental investigation the best value of n
was claimed to be 1.5.
Nashta and Garde [4], based on minimization of the form
loss and friction loss recommended the following equation
for the transition:
b = b0 + (bL b0 ) [1 (1 ) 0.55 ] (8)
Based on optimal control theory, a methodology has been
presented for optimal design of a rectangular subcritical
expansive transition by Swamee and Basak [5]. Analyzing
a large number of optimal profiles, an equation for the
design of rectangular transitions was presented as
0.775
L
1.35
b = b0 + (bL b0 )2.52 1 + 1
x of sections across the width at 1/6, , and 5/6 times the
However, no experimental evidence is available for the
latest development.
3. EXPERIMENTAL SETUP AND PROCEDURE
The experiments were carried out in a flume of Water
Resources Engineering Laboratory, DUET, Gazipur, which
consisted of steel frame and bed, side walls of Perspex
sheet. To test the transition models a contracted reach of
(7)
Fig. 2: Water surface profile and setup for velocity
measurements
To know the velocity distribution of flow through gradual
expansion and hence to evaluate the efficiency of
transition, the (i) velocity at various points, and (ii) depth of
flow at various sections were to be known. Pitot tube with
its setup was to be held at various points to know the
velocities for those points. The velocity was evaluated from
the differential head (known as velocity head, ( V / 2 g )
2
in the Pitot tube over the water surface in the channel.
Water level along the centre line of the channel was
measured with a pointer gauge. Velocities were measured
at near surface, 0.20y, 0.40y, 0.60y, 0.80y and near bottom
(9) in the vertical. These measurements were made at a number
width, (Fig. 3) and along the length of transition at inlet,
mid-length, and outlet sections. Each transition was tested
for five different discharges, Q as 0.0015, 0.0030, 0.0045,
0.0060, and 0.0075 cumec. For each discharge, experiments
were conducted at five different depths so that Froude
number F1, at entry were 0.25, 0.32, 0.40, 0.47, and 0.55
respectively. A typical table (Table 1) is given below which
reveals the data organizing.

Dhaka University of Engineering & Technology, Gazipur 28

DUET Journal Vol. 1, Issue 1, June 2010

Inlet tank
Perspex sheet sidewall Transition model Control weir

9.525 cm 25.4 cm Outflow

Inlet transition Wooden channel 1 2

65 cm 35 cm 250 cm 55.56 cm 150 cm

a) Top view of flume

b/6 b/2 5b/6

Near surface
0.2 y

Flow depth, y 0.4 y

0.6 y
0.8 y
Near bottom

Bed width, b

b) Cross-section at inlet of transition model with various points of measurements

Fig. 3: Top view of flume with the section showing the various points of velocity measurements
Table 1: Local velocity (Model IV, Q = 0.0045 cumec) kinetic energy per unit time passing over the entire section
1 3
y 1 at
y 1,
y 2 at
y 2, y 3 at y 4 at
y 5 at 1-1, is obtained by integrating the term v dA , i.e.,
Q m3/s F 1 inlet outlet 3L /4 2
mm mm L /4 mm L /2 mm
mm mm mm 1
2 A1
0.25 153.0 30.6 155.3 31.1 153.8 154.3 154.8 v 3 dA , where A1 indicates integration over the cross-
0.32 129.0 25.8 131.5 26.3 130.0 130.8 131.4
0.0045 0.4 110.0 22.0 114.2 22.8 111.3 112.3 113.2
0.47 101.0 20.2 105.4 21.1 102.8 104.3 105.0 section 1-1. In a similar manner, the kinetic energy per unit
0.55 89.0 17.8 94.8 19.0 91.0 92.8 94.0 time for section 2-2, at the outlet of transition is
1
2 A2
At At b /6 At b /2 At 5b /6 V. at mid-length
v 3 dA . As the fluid passes through the transition, the
No., F 1
Section
Froude

Velocity, Velocity, Velocity, Reading, h

Depth Reading, Reading, Reading, Avg. Vel,
v in v in v in at 0.6y 4,
s h in mm m/sec h in mm m/sec h in mm m/sec mm
v in m/sec
actual reduction in kinetic energy per unit time, or the
0.1 y 10.0 0.443 10.5 0.454 10.0 0.443
0.2 y 10.5 0.454 11.0 0.465 10.5 0.454 power available for transformation in the transition, is the
0.4 y 10.0 0.443 10.5 0.454 10.0 0.443 1 1
Inlet
0.6 y 9.0 0.420 9.5 0.432 9.0 0.420 3.0 0.243 difference of the above two, i.e., v 3 dA v 3 dA .
2 A1 2 A2
0.8 y 8.0 0.396 8.5 0.408 7.5 0.384 5.5 0.328
0.4
0.9 y 6.0 0.343 7.0 0.371 6.0 0.343 5.5 0.328 All of this energy is not transformed into useful work. It
0.1 y 5.0 0.313 6.0 0.343 2.5 0.221 4.5 0.297
may be helpful to imagine the transition as a pump. The
0.2 y 6.5 0.357 7.5 0.384 3.0 0.243 2.5 0.221
0.4 y 6.0 0.343 6.5 0.357 2.5 0.221 at b /3, and 2b /3 at I/O pump raises the pressure of the fluid entering. The term
Outlet
0.6 y 3.0 0.243 4.0 0.280 1.5 0.172 4.5 0.297 1 1
0.8 y 1.0 0.140 2.0 0.198 1.0 0.140 4.0 0.280 v 3 dA v 3 dA might be regarded as the power
0.9 y 0.0 0.000 0.0 0.000 0.0 0.000 2 A1 2 A2
supplied or input to the pump. The head gained by the fluid
4. BACKGROUND OF EFFICIENCY ( )
flowing through the transition is y 2 y1 , where y1 and
For finding the efficiency, it is necessary to consider the y2 refer to the depths of flow at sections 1-1 and 2-2
velocity variation across the channel. Referring to section respectively. If Q be the rate of flow per sec, potential
1-1 at the inlet, Fig. 3, let v be the velocity of flow through energy gained by the fluid per sec will be Qg ( y 2 y1 )
an infinitesimal area dA. Assuming that the flow is which gives the actual power, the pump adds to the fluid.
essentially in the direction of the axis of the transition, the The purpose of transition is to convert kinetic energy into
volume of fluid per unit time passing through the useful pressure energy. Hence, the efficiency of the
elementary area is vdA, and then the mass rate of flow transition is defined by
through this area is vdA. The kinetic energy of this fluid Qg ( y 2 y1 )
= (10)
mass per unit time is (vdA)v , or v dA . The total
1 2 1 3 1 1
2 2 v 3 dA v 3 dA
2 A1 2 A2

Dhaka University of Engineering & Technology, Gazipur 29

 DUET Journal Vol. 1, Issue 1, June 2010

Suppose, V1 be the average velocity at section 1-1 and V2 The percentage efficiencies for different discharges and
Froude numbers, and also overall efficiencies of the
be the average velocity at section 2-2. Then one can write
transition models are presented in Table 3a. The
1 1
v 3 dA = QV12 1 (11) efficiencies of the transition models for average of
2 A1
2 discharges and average of Froude numbers are also
and summarized in Table 3b, where the comparative feature of
1 1 performance of the models is observed. The efficiency
v 3 dA = QV12 2 (12)
2 A2
2 Table 3a: Overall hydraulic efficiencies of the transition
where, 1 and 2 are numerical constants, known as models
energy coefficients for non-uniformity of velocity
distribution. Model I Model II Model III Model IV

Froude No.,
Discharge,
Q , cumec
Overall Overall Overall Overall
Now Eq. (10) takes the form

F1
Efficiency Efficiency, Efficiency, Efficiency
Efficiency Efficiency, Efficiency, Efficiency
Qg ( y 2 y1 ) ( y 2 y1 ) ,% , % , % ,%
,% , % , % ,%
(13)
= = 0.25 87.8 84.2 80 89.2
1 2 1 2 V12 V 22 0.32 82.8 70.3 83.8 89
QV1 1 QV2 2 1
2g 2 2g
0.002 0.4 78.4 75.5 80.9 80.9
2 2 0.47 73.4 79.1 68.6 76.7
0.55 76.4 72.7 71.9 79
The energy coefficients, 1 and 2 are obtained 0.25
0.32
88.1
83
77.8
80
78.1
72.8
87.1
77.7
0.003 0.4 79.3 76.5 70.8 78.7
numerically from Eqs. (11) and (12). 0.47 70.1 76.6 74 76.4
0.55 71.4 68.9 70.4 71.7
5. ANALYSIS OF DATA 0.25 85.6 84.7 80.7 92.7
0.32 86.6 81.3 75.9 78.1
0.005 0.4 77.6 75.8 74.4 74.7 68.9 71.4 76.2 78.7
The velocities measured at different depths in the vertical 0.47 77.3 69.8 65.7 72.4
0.55 68.2 65.3 61.2 72.9
indicated the usual turbulent boundary layer profile 0.25 80 81.2 76.6 86.7
(mentioned in data table); as such, only the average 0.32 76 75.3 74.3 80.4
0.006 0.4 70.2 72.1 67 77.6
velocity over a vertical was used in further analysis. Using 0.47 64.6 72.2 59 71.1
the data for depth of flow and local velocity, the flow area 0.55 59.1 61.3 59.8 69.3
0.25 78.8 81.7 77.8 88.1
and average velocity at the sections were known, and thus 0.32 74.8 82.4 71.7 85.5
energy correction factors were calculated. After then these 0.008 0.4 68.3 70.7 66.2 73.8
0.47 69.5 70.9 67.2 67.9
data were used to evaluate the efficiency of transition from 0.55 68.7 63.8 62.5 68.9
the expression shown above. The table (Table 2) given
below shows this in brief. Four transition profiles used Table 3b: Hydraulic efficiencies of the transition models
commonly in the field [2, 3, 4 and 5], are tested for (for avg. of Q and avg. of F1)
performance and compared the efficiency of the models,
i.e. Model I, II, III, and IV respectively. Model I Model II Model III Model IV
Q , cumec
Discharge,
No., F 1
Froude

Average Efficiency Average EfficiencyAverage EfficiencyAverage Efficiency

for for Avg. for for Avg. for Avg. for for Avg. for Avg.
Table 2: Average velocity, energy correction factor, and Avg. Q F1 Avg. Q F1 Q Avg. Q F1
efficiency (Model IV, Q = 0.0045 cumec) 0.25 84.1 81.9 78.6 88.8
0.32 80.6 77.9 75.7 82.1
0.40 74.8 73.8 70.8 77.4
0.47 71.0 73.7 66.9 72.9
Avg. Vel., Avg. V / Avg. V / / / Efficiency,
Section y v y v y
3
0.55 68.8 66.4 65.2 72.4
v strip Section Strip Section ,%
0.002 79.8 76.4 77.0 83.0
0.448 0.0220 0.010 0.0020 0.003 78. 4 76.0 73.2 78.3
0.448 0.0220 0.010 0.0020
Strip I 0.432 0.0220 0.009 0.421 0.0018 1.02 0.005 79.1 75.1 70.5 78.5
0.408 0.0220 0.009 0.0015 0.006 70.0 72.4 67.3 77.0
0.370 0.0220 0.008 0.0011 0.008 72.0 73.9 69.1 76.8
0.459 0.0220 0.010 0.0021
0.459 0.0220 0.010 0.0021
Inlet Strip II 0.443 0.0220 0.010 0.434 0.425 0.0019 1.01 1.01 curves for average of discharges and for average of Froude
0.420 0.0220 0.009 0.0016
0.389 0.0220 0.009 0.0013 numbers for the four models have been shown in Figs. 4
0.448 0.0220 0.010 0.0020 and 5, respectively.
0.448 0.0220 0.010 0.0020
Strip III 0.432 0.0220 0.009 0.419 0.0018 1.02
0.402 0.0220 0.009 0.0014
0.363 0.0220 0.008 0.0011 6. EXPERIMENTAL OBSERVATION AND
76.2
0.335 0.0228 0.008 0.0009 RESULTS
0.350 0.0228 0.008 0.0010
Strip I 0.293 0.0228 0.007 0.248 0.0006 1.49
0.191 0.0228 0.004 0.0002 The surface configuration in the transitions I, II, and III
0.070 0.0228 0.002 0.0000
0.363 0.0228 0.008 0.0011 exhibited local disturbances leading to diagonal waves,
0.370 0.0228 0.008 0.0012
Outlet Strip II 0.319 0.0228 0.007 0.278 0.234 0.0007 1.36 1.39 starting in the inlet section and persisting towards the entire
0.239 0.0228 0.005 0.0003 length of the transition. The separation of flow took place
0.099 0.0228 0.002 0.0000
0.232 0.0228 0.005 0.0003 at the boundaries near the exit of the transition. The points
0.232 0.0228 0.005 0.0003
Strip III 0.197 0.0228 0.004 0.177 0.0002 1.32 of separation were not symmetrical on either side; this was
0.156 0.0228 0.004 0.0001 also observed by several investigators earlier. The
0.070 0.0228 0.002 0.0000
separation points were found to move forward and

Dhaka University of Engineering & Technology, Gazipur 30

DUET Journal Vol. 1, Issue 1, June 2010

95.0

90.0 Model I
Line Model I
Model II Velocity
85.0 Maximum
Q = 0.0015 cumec
Model III
80.0 F1 = 0.40
Efficiency, e, %

Model IV
75.0

70.0

65.0

60.0

55.0 Line
Velocity Model II
Maximum
50.0 Q = 0.0015 cumec
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60
F1 = 0.40

Froude Number, F1

Fig. 4: Efficiency (average of discharges) of transition

90.0
Model I
85.0 Model II Model III
Model III Q = 0.0015 cumec
80.0
Efficiency, e, %

Model IV Maximum
F1 = 0.40
Velocity
Line
75.0

70.0

65.0

60.0
Model IV
Line
0.0010 0.0020 0.0030 0.0040 0.0050 0.0060 0.0070 0.0080 Maximum Velocity
Q = 0.0015 cumec

Discharge, Q, cumec F1 = 0.40

Fig. 5: Efficiency (average of Froude numbers) of

transition

backward on either side. The flow in the downstream All dimensions are in mm
channel after expansion was found to be unstable and Scale :
Velocity-
0 50 100 cm/s
Other dimensions-0 5 10 cm
oscillating in nature. At times, flow was found to swing
completely from one side to the other, thereby reversing the Fig.6: Velocity distributions across the width and along the
picture of separation and velocity distribution. Excepting a length of transition models (Q = 0.0015 cumec, F1 = 0.40)
few cases, flow was never symmetrical with respect to the
centre line of the channel. The maximum-velocity line 1.0
coincided with the centre line of the channel, for a short
length after entry. Thereafter, it shifted to the side to which
the main flow attached. The percentage efficiency, based Swamee & Basak (1993)
0.8
on the difference of kinetic energy at inlet and outlet, and (Profile IV)
head recovery, can be examined from Tables 3a and 3b,
and Figs. 4 and 5. The superiority of Transition IV is
apparent. Provision of smooth outlet in the optimal 0.6 Hartley et al (1940)
(b x-b 0 )/(b L -b 0 )

transition (Profile IV) eliminates the chances of eddy (Profile I)

formation and separation to a significant amount. The eddy
with reverse flow is observed strong for Transition I to III. Chaturvedi (1963)
The velocity profiles of Model IV are close to flat and near 0.4
(Profile II)
ideal; those of Model I, II, and III depict central
deformations indicating one sidedness of maximum
velocity thread (Fig. 6). 0.2
Nashta and Garde
Except Profile IV, all other transitions have abrupt ending (1988) (Profile III)
at downstream part of transition (Fig. 7), which may be
responsible for the separation noticed at the exit section in 0.0
these transitions. The overall hydraulic efficiency of the 0.0 0.2 0.4 0.6 0.8 1.0
transition models decreases from Model I to III, and these x /L
are 75.8%, 74.7%, and 71.4% respectively. Efficiency of
the Model IV is the highest among the models, and it is Fig.7: Comparison of bed-width profiles

Dhaka University of Engineering & Technology, Gazipur 31

DUET Journal Vol. 1, Issue 1, June 2010

78.7%. Abrupt ending in the transition profiles evolved by flow decreases, the local velocity remains
Hartley et al [2], Chaturvedi [3], and Nashta and Garde [4], considerably high, which if allowed to persist in
shifts the efficiency curves (Figs. 4 and 5) downward, erodible channels will scour away bed and sides of
whereas smooth ending in the optimal transition (Profile channel.
IV) makes it efficient over others, and keeps the efficiencyiii.
curve at top. Figure 8 shows the overall hydraulic R EF ER ENC ES
efficiency of the various transition profiles, where
dominance of the Profile IV is observed over others. [1] J. Hinds, The hydraulic design of flume and siphon
transitions, Trans., ASCE, 92, pp. 1423-1459, 1928.
80.0
78.0
76.0
[2] G. E. Hartley, J. P. Jain and A. P. Bhattacharya, Report
Efficiency, , %

74.0 on the model experiments of fluming of bridges on

72.0 Purwa branch, Technical Memorandum. 9, United
70.0
68.0 Provinces Irrigation Res. Inst., Lucknow (now at
66.0 Roorkee), India, pp. 94-110, 1940.
64.0
62.0
60.0 [3] R. S. Chaturvedi, Expansive subcritical flow in open
Model I Model II Model III Model IV channel transitions, J. Inst. of Engrs., India, 43(9),
pp. 447-487, 1963.
Fig.8: Efficiency of the Transition Models
[4] C. F. Nashta, and R. J. Garde, Subcritical flow in
rigid-bed open channel expansions, J. Hydr. Res.,
7. CONCLUSIONS 26(1), pp. 49-65, 1988.

In the light of the present study the following conclusions
could be drawn:
i. The Transition Profile IV yields a design that
produces the highest efficiency among the existing
profiles. So, this can be suggested for field use over
others.
ii. The velocity distribution after expansion in case of
Model I, II, and III becomes highly non-uniform
with the result that although the average velocity of
[5] P. K. Swamee and B. C. Basak, A comprehensive
open-channel expansion transition design, J. of Irrg.
and Drainage Div., ASCE, 119(1), pp. 1-17, 1993.
[6] H. M. Morris and J. M. Wiggert, Applied Hydraulics in
Engineering. Second Ed., The Ronald Press Co., New
York, N. Y., 1972, pp. 184-188.
[7] S. K. Mazumder, Optimum length of transition in
open channel expansive subcritical flow, J. Inst. of
Engrs., India, 48(3), pp. 463-478, 1967.

Dhaka University of Engineering & Technology, Gazipur 32