Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Technical Note

Cylindrical Central Baffle Flume for


Flow Measurements in Open Channels
Aniruddha D. Ghare 1; Ankur Kapoor 2; and Avinash M. Badar 3

Abstract: This work experimentally investigated a modified venturi flume formed by placing portable cylinders vertically upright in trap-
ezoidal channels, referred to herein as the cylindrical central baffle flume (CCBF). This mobile device causes flow constriction that creates a
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

critical flow condition. Experiments were conducted in a specially fabricated experimental trapezoidal channel, having the facility to change
the side slopes. The side slope of the channel was varied between 0.50 H:1 V and 2 H:1 V at an interval of 0.25 H. Experimental investigation
of a mobile flume for such a wide range of side slopes has never been reported. Different variables describing the flow through this flume were
selected, and a number of forms of equations relating stage with discharge were developed using Buckingham’s Π theorem and self-similarity
hypothesis. The various forms of the stage–discharge equation were calibrated using the measurements made during the laboratory experi-
ments. The stage–discharge equations were evaluated for fit by statistical measures, root mean square error (RMSE), relative mean error
(RME), predicted residual error sum of squares (PRESS), and predicted R2 . Based on the values of the statistical measures, two forms of
stage–discharge equation were selected. The two selected forms were further compared for their performance, first by using additional ex-
perimental observations recorded for the same contraction ratio and then by using experimental observations collected for the changed flume
dimensions. The performance of the selected forms of equation was also assessed by comparing them with a model available in the literature
for a similar flume in a trapezoidal channel having a 1∶1 side slope. Based on the results of the comparison and distribution of errors, one of
the two forms of stage–discharge equation that had a maximum relative error between predicted and measured discharge of less than 10% was
proposed for flow measurement in such flumes under free flow conditions. The proposed mathematical model proves useful for its versatility,
as it was developed and validated for a range of side slopes and contraction ratios, and is valid for a submergence ratio up to 62%. The
proposed flume facilitates flow measurement at any desired location including small laterals or turnouts in agricultural settings or water
treatment plants conveying water with trapezoidal channels. The flume is mobile, inexpensive, easy to install, and does not require high
maintenance. DOI: 10.1061/(ASCE)IR.1943-4774.0001499. © 2020 American Society of Civil Engineers.
Author keywords: Venturi flume; Cylindrical central baffle flume; Open channel; Free flow; Flow measurement; Regression models.

Introduction accurately measured and regulated at all important points in an irri-


gation system.
With available fresh water becoming scarcer, accounting for water Existing flow-measuring devices for open channels are rela-
use via improved flow rate measurement has become more impor- tively expensive to construct and install, and require proper main-
tant. The demand for water has forced water suppliers and irrigation tenance. There are several approaches to measure discharge in open
departments to improve allocation methods and distribution of water channels such as orifices, weirs, or flumes. Measuring flumes allow
shares to users. One fundamental method to meet conservation goals accurate flow measurement (inducing critical flow conditions) by
is universal metering. Researchers have suggested that water conser- creating either side contraction (Cone 1917; Skogerboe and Hyatt
vation efforts should be primarily directed toward farmers, in light of 1967) or bottom contraction (Parshall 1926), thus effectively alter-
the fact that crop irrigation accounts for 70% of the world’s fresh- ing the channel. In flumes, discharge measurement is based on the
water use. The efficient and economical use of water resources is principle of critical flow. By installation of an appropriate device of
becoming a critical goal if the needs of every water user are to be known geometry, discharge can be computed by recording flow
addressed. To improve water management and to accurately assess depth at only one location.
use and cost, it is strongly recommended that irrigation water be Hager (1985) first proposed the concept of a portable device for
discharge measurement in rectangular, trapezoidal, and U-shaped
channels in which the constriction was achieved along the channel
1
Professor, Dept. of Civil Engineering, Visvesvaraya National Institute axis instead of at its side walls, using a cylinder positioned vertically
of Technology, Nagpur 440010, India. ORCID: https://orcid.org/0000 into the channel. Hager (1986) also investigated the flow pattern of a
-0002-2734-8002 cone immersed in a rectangular channel. Its rating curve was devel-
2
Research Scholar, Dept. of Civil Engineering, Visvesvaraya National oped by accounting for streamline curvature effects. The geometrical
Institute of Technology, Nagpur 440010, India (corresponding author). parameters defining the cross section at various locations of the coni-
Email: ankurkapoor06@yahoo.com cal flume proposed by Hager (1986) was thoroughly studied, and an
3
Vice-Principal, KDK College of Engineering, Nagpur 440009, India. explicit stage–discharge relationship was developed by Kapoor et al.
ORCID: https://orcid.org/0000-0003-4466-3511
(2019). Hager (1988) presented a graphical approach for a cylinder
Note. This manuscript was submitted on October 4, 2019; approved on
May 1, 2020; published online on July 2, 2020. Discussion period open placed in a circular channel that was replaced by a computer model
until December 2, 2020; separate discussions must be submitted for indi- by Samani et al. (1991) to predict discharge with maximum error
vidual papers. This technical note is part of the Journal of Irrigation and of 6%. Another computer model was developed by Samani and
Drainage Engineering, © ASCE, ISSN 0733-9437. Magallanez (1993) for a trapezoidal flume, valid for a 1∶1 side slope

© ASCE 06020007-1 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


and a contraction ratio range of 0.90–0.98. The concept of a cutthroat some surcharge weights can be placed over the top of the cylinder
flume (Robinson and Chamberlain 1960) was combined with that of to stabilize it against the strong flow current. Alternatively, the hol-
circular flume by Samani and Magallanez (2000) by attaching two low cylinders can be filled with sand or locally available materials
semicircular cylinders to the side walls of the rectangular channel to such as stones. For the development and calibration of the stage–
achieve the constriction, which was later studied experimentally by discharge relationship for a CCBF, extensive experiments were
Ghare and Badar (2014), who proposed a model that predicted dis- conducted, and the detailed investigation and results are reported
charge with a maximum error of 5%. Samani (2017) offered a single herein.
equation for discharge prediction in open channels in which the cal-
ibration coefficient value depends on the shape of the cross section of
the channel. Later, Kolavani et al. (2018) reanalyzed the formulation Dimensional Analysis of Flow
given by Samani (2017) for circular channels.
The flow passing through the cylindrical central baffle flume theo-
The venturi effect was also studied by Peruginelli and Bonacci
retically depends on the upstream energy head (H), bed width of
(1997) using a pier-shaped prism to propose a discharge prediction
the channel section (B), bottom diameter (D) of the obstruction
model with an error of 6%. Ferro (2016) theoretically developed a
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

being inserted, horizontal projection of the side walls (Lm ) (Fig. 1),
stage–discharge relationship of a flume with a central baffle using
viscosity of water (μ), acceleration due to gravity (g), and density
dimensional analysis and calibrated this relationship using the ex-
(ρ) of water (Badar and Ghare 2012; Kolavani et al. 2019). From a
perimental data of Peruginelli and Bonacci (1997). Later, Kolavani
practical point of view, the flow depth at the cylindrical face (y) is
et al. (2019) and Bijankhan and Ferro (2019) experimentally inves-
used to represent the upstream energy (Samani and Magallanez
tigated the effect of different geometrical parameters of a central
2000). Dimensional analysis along with the self-similarity concept
baffle flume and proposed an explicit stage–discharge relationship
was used to obtain a functional relationship between the variables
using a dimensional analysis approach.
in terms of nondimensional parameters (Ferro 2002), the relation-
It can be concluded after reviewing the literature that, with
ship of which was calibrated using experimental observations re-
proper calibration, the aforementioned portable flumes can be used
corded on the CCBF.
as substitutes for the traditional fixed flow-measuring flumes for
According to Buckingham’s Π method of dimensional analysis,
flow measurement in small open channels. The literature also re-
the functional relationship describing the free flow hydraulics of the
vealed that the flow prediction models proposed by various re-
CCBF can be expressed (Bijankhan and Ferro 2019) by
searchers have not been experimentally investigated for different
values of channel side slope. In the research reported herein, a port- fðQ; y; B; D; Lm ; g; μ; ρÞ ¼ 0 ð1Þ
able device that measures the flow rate in small trapezoidal chan-
nels is experimentally calibrated to provide a universal discharge As per the Π theorem, the functional relationship Eq. (1) can be
model which can be used over a wide range of channel side slopes. expressed by five dimensionless Π terms. Considering D, g, and μ
A cylinder was inserted vertically upright in the trapezoidal chan- as independent variables, and Q as the ultimate independent var-
nel, referred to herein as a cylindrical central baffle flume (CCBF). iable for which to solve, Π terms have been constructed as follows:
The cylinder creates critical flow condition via contraction caused
Q
by the cylinder. The CCBF (Fig. 1) is easy to install and eliminates Π1 ¼ 5=2 1=2
ð2Þ
the problem of incorrect placement. After recognizing that practical D g
use of a stage–discharge relationship requires incorporation of two y
Π2 ¼ ð3Þ
vital parameters—side slope of the trapezoidal channel m and the D
contraction ratio r of the flume—a laboratory investigation pro-
B
gram was designed. The primary objective of this study is to es- Π3 ¼ ð4Þ
tablish a stage–discharge relationship to determine the rate of flow D
in a trapezoidal channel having side slopes from 0.50 H:1 V to Lm
2 H:1 V using a CCBF over a range of contraction ratios (0.63 ≤ Π4 ¼ ð5Þ
r ≤ 0.1). The use of a CCBF has advantages over conical baffles for D
channels where higher flow rates are expected. In such scenarios, ρD3=2 g1=2
Π5 ¼ ð6Þ
μ
D

Lm
Eqs. (4)–(6) can be rewritten as follows:
Water level
Π4 Lm D Lm
1 Π2;4 ¼ ¼ ¼ ¼m ð7Þ
y
m Π2 D y y

Also
B
Π1 Π5 Q ρD3=2 g1=2 D ρQ
Π1;3;5 ¼ ¼ 5=2 1=2 ¼ ¼R ð8Þ
Π3 D g μ B μB
Cylindrical baffle Eq. (1) can then be expressed as
 
Q Q y
y ¼ f1 ; m; R ð9Þ
yc
D5=2 g1=2 D

where Q=D5=2 g1=2 = dimensionless discharge; y=D = dimension-


less head; m = channel side slope; R = Reynolds number; and f 1 =
Fig. 1. Definition sketch of cylindrical central baffle flume.
functional symbol.

© ASCE 06020007-2 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


The Reynolds number (R) calculated for the experimental data experimental investigation was carried out at the hydraulics labo-
associated with the CCBF lies in the range 1.68 × 104 < R < ratory of K. D. K. College of Engineering, Nagpur, India. The ex-
8.70 × 104 , the fully turbulent flow regime, for which the viscous perimental runs were conducted on a trapezoidal channel fabricated
effects can be neglected (Kolavani et al. 2019). Therefore, Eq. (9) using mild steel angles and transparent acrylic sheets, which cre-
reduces to ated the ability to adjust the side slopes. The trapezoidal channel
  had a base width (B) of 0.20 m and length of 4.90 m. The flume
Q y
¼ f 2 ; m ð10Þ used in the experiment was formed using a cylinder made of
D5=2 g1=2 D galvanized iron sheets, inserted vertically upright in the trapezoidal
channel. For each set (only Qm varied) of experimental trials, one of
where f2 = functional symbol.
six different cylinders were tested in a trapezoidal channel having
The self-similarity principle (Barenblatt 1979, 1987) can be ap-
plied to obtain the mathematical shape of Eq. (10). A phenomenon one of the following side slopes: 0.50 H:1 V; 0.75 H:1 V; 1 H:1 V;
can either be defined as complete self-similar (CSS) or incomplete 1.25 H:1 V; 1.50 H:1 V; 1.75 H:1 V; or 2 H:1 V. The cylinders were
self-similar (ISS) in a given Πn dimensionless group depending fabricated such that the contraction ratio ranged between 0.63 and
upon the behavior of the function φ (Ferro 2016). The self-similar 1. The dimensions of the tested flumes are shown in Table 2. To
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

solutions of a physical phenomenon are searched for boundary con- maintain the stability of the cylinder against the current of flow,
ditions. For Πn → 0 or Πn → ∞, if the function φ tends to a finite surcharge weights were placed at the top. Alternatively, sand filling
limit other than 0, then the functional relationship Π1 ¼ φðΠ2 ; was also done in some of the cases to achieve better cylinder sta-
Π3 ; ···; Πn Þ is considered to be independent of Πn and complete bility. The floor of the trapezoidal channel was horizontal in both
self-similarity occurs. In such case, Πn can be dropped from the the longitudinal and transverse directions. The cylinder was placed
functional relationship Π1 ¼ φ1 ðΠ2 ; Π3 ; ···; Πn−1 Þ, in which φ1 is at a distance of 1.50 m from the outlet of the channel (Fig. 2). A
the functional symbol. However, for Πn → 0 or Πn → ∞, if the horizontally positioned centrifugal pump was used to deliver a
function φ has a limit equal to 0 or ∞, it is regarded as incomplete continuous flow of water with recirculation. At the end of the chan-
self-similarity (Ferro 2002) and the functional relationship can be nel, a gate was installed for variation of downstream flow depth
expressed as Π1 ¼ Πpn φ2 ðΠ2 ; Π3 ; ···; Πn−1 Þ, in which φ2 is the and determination of submergence limit. For a given measured dis-
functional symbol and p is a numerical constant. charge, the downstream gate was closed incrementally to increase
In the present study, for the given value of side slope m, when the tailwater depth until the increase in upstream flow depth was
y=D → 0, then Q=D5=2 g1=2 → 0, and when y=D → ∞, then not more than 1%. The submergence limit was then defined as the
Q=D5=2 g1=2 → ∞. Hence, ISS occurs. Therefore, Eq. (10) can ratio of maximum tailwater depth to the corresponding upstream
be rewritten as flow depth. To stabilize the flow, baffle walls were provided at
 n the upstream end of the channel.
Q y Experimental observations of flow depth (y) were recorded
5=2 1=2
¼ ψðmÞ ð11Þ
D g D at the upstream face of the cylinder using the depth graduations
inscribed on the cylinder. The corresponding flow rates were
where ψ = functional symbol; and n = exponent that is a numerical measured volumetrically using a measuring tank having a capac-
constant. ity of 665 L. The flow rate was varied between 0.0030 and
The choice of independent variable affects the form of the Π 0.0155 m3 =s. For every experimental run, the flow rate was al-
terms, which directly affects the form of stage–discharge equation. lowed to stabilize before upstream flow depth and flow rate read-
Eq. (11) is one such form that should be generalized in terms of ings were recorded under subcritical free flow conditions. The
dimensionless discharge (Y), dimensionless head (X), and side
process was continued for each of the six cylinders and seven side
slope (m) as
slopes of the trapezoidal channel. A total of 561 trials were re-
Y ¼ ψðmÞðXÞn ð12Þ corded (Tables S1–S7) and used to derive the proposed discharge
prediction model. Fig. 3 shows the experimental set up of the
In the present study, the dimensionless discharge terms (Y) were CCBF.
obtained using combinations of bed width (B) and cylinder diam- Discharge prediction models were developed for the CCBF
eter (D) of the CCBF and the possible forms are reported in Table 1. using dimensional analysis (Pairs 1 and 2 in Table 1) and self-
similarity concept including the ones that were developed by alter-
ing the dimensionless discharge term (Pairs 3 and 4). The models
Materials and Methods obtained from the selected pairs of dimensionless terms were de-
veloped using the experimental data recorded on laboratory flumes.
To obtain a calibrated model for the CCBF for accurate prediction
To evaluate the performance of the models, the observed (mea-
of discharge in a free flow subcritical regime, a comprehensive
sured) values were compared with calculated (predicted) discharge
values using two statistical indices: root mean square error (RMSE)
Table 1. Pairs of dimensionless terms
Pair No. Dimensionless discharge, Y Dimensionless head, X
Table 2. Dimensions of the tested cylindrical central baffle flumes
1 Q y=D
D5=2 g1=2 S.No. B (m) D (m) r
Q 1 0.20 0.126 0.63
2 y=D
B5=2 g1=2 2 0.20 0.140 0.70
Q 3 0.20 0.160 0.80
3 y=D
D3=2 Bg1=2 4 0.20 0.166 0.83
Q 5 0.20 0.1823 0.91
4 y=D
B3=2 Dg1=2 6 0.20 0.200 1.00

© ASCE 06020007-3 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


Water supply with
Upstream reservoir
recirculating arrangement

Baffle wall
0m
1.2 Gauge 4.90
. m
Cylinder

0.709 m 0.60 m Flow Adjustable side


1
.y slope
H m

0.2 m Measuring tank


Pump 1.50 m 1.50 m

Supporting jack .
1.0 m 1.50 m
F.L. F.L.
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 2. Experimental setup of cylindrical central baffle flume.

(a) (b)

Fig. 3. Laboratory experiment: (a) view from downstream of channel; and (b) side view of channel.

and relative mean error (RME). The RMSE can be computed using plotted against X for different side slopes m ¼ 0.50, 0.75, 1.00,
the following expression: 1.25, 1.50, 1.75, and 2.00, and power regression equations [Eq. (12)]
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN ffi were obtained. This provided seven equations, one each for each of
2
ðQ p − Q m Þ the seven side slopes. According to the ISS concept, a constant value
RMSE ¼ i¼1
ð13Þ of parameter n should be used; therefore, an average value of n for a
N
dimensionless pair was determined for 0.50 ≤ m ≤ 2.00. The varia-
where Qp and Qm = predicted and measured discharge values; and tion of ðXÞn in terms of dimensionless discharge (Y) for one of the
N = number of observations. pairs (Q=B3=2 Dg1=2 ; ðy=DÞ2.4 ) was investigated graphically (Fig. 4),
An alternative to RMSE is the RME, which is computed as which shows that the stage–discharge relationship can be defined by
N   a linear regression, for which the line slope depends on the side slope
1X  Qp  m of the channel. Consequently, ψðmÞ was obtained from the slope
RME ¼ − 1 ð14Þ
N i¼1  Qm of associated regression lines. It was found that the coefficient ψðmÞ
varies with side slope m, and therefore was expressed as a function of
m, as shown in Fig. 5. A similar procedure was followed for all the
Analysis of Data for Development of Calibration dimensionless pairs simultaneously, and the results are listed in
Equation Table 3.
Thereafter, discharge values were calculated using all forms of
The upstream flow depth was measured at the upstream face of the the equation (Table 3). The performance of all forms of stage–
cylinder, which served as the upstream energy head, as it was as- discharge equation (Table 3) were evaluated by comparing the dis-
sumed that all of the velocity head was converted to the energy charge values calculated by the equations with the corresponding
head. The recorded observations of upstream head (y) and the cor- measured values. The RMSE and the RME were initially used as
responding discharge (Qm ) for CCBFs of different diameters D, parameters for statistical comparison between calculated and mea-
placed vertically upright in trapezoidal channels with different side sured discharge values. It is evident from Table 4 that the values of
slopes m, are reported in Tables S1–S7. For any dimensionless pair RMSE and RME for all the forms of stage–discharge equation are
(Table 1), the dimensionless discharge (Y) and dimensionless head very close to each other, and therefore their performance could not
(X) was calculated using all the observations. Thereafter, Y was be judged solely on the basis of RMSE and RME. Thus, another

© ASCE 06020007-4 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


0.50
Table 3. Forms of stage–discharge equation
0.45
Pair No. Form of equation ψðmÞ n
0.40
Q  y n
0.35 1 ¼ ψðmÞ 0.62m0.69 2.8
5=2
D g 1=2 D
 y n
/(B3⁄ 2 D 1⁄ 2)

0.30 Q
m = 0.50 2 ¼ ψðmÞ 0.27m0.47 2.0
0.25 B5=2 g1=2 D
m = 0.75
Q  y n
0.20 m = 1.00 3 ¼ ψðmÞ 0.45m0.60 2.5
m = 1.25 D3=2 Bg1=2 D
0.15
Q  y n
m = 1.50
4 ¼ ψðmÞ 0.39m0.58 2.4
0.10 m = 1.75 B3=2 Dg1=2 D
0.05 m = 2.00

0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
( / ) 2.4
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

Table 4. Statistical measure of fit for selected regression models


3=2 1=2
Fig. 4. Dimensionless discharge (Q=B Dg ) in terms of dimen- Pair Form of Predicted
sionless head ðy=DÞ2.4 associated with different m values. No. equation RMSE RME PRESS R2 (%)
Q  
0.69 y 2.8
1 ¼ 0.62m 0.00169 0.1641 23.27 90.60
D5=2 g1=2 D
measure of error is used, the predicted residual error sum of squares Q  y 2
2 ¼ 0.27m0.47 0.00133 0.1323 14.14 89.21
(PRESS), and consequently predicted R2 for statistical comparison B5=2 g1=2 D
between calculated and measured discharge values. Q  
0.60 y 2.5
3 ¼ 0.45m 0.00059 0.0564 03.36 98.15
The PRESS statistic is a measure of the deviation between the D3=2 Bg1=2 D
fitted values and the observed values, which can be considered as Q  y 2.4
4 ¼ 0.39m0.58 0.00046 0.0448 01.55 99.03
an adjusted sum of squared error (SSE). It is a well-known “leave- B3=2 Dg1=2 D
one-out” statistic, commonly used in regression analysis for cross-
validation (Hu 2016). To calculate PRESS, an observation (e.g., i),
is selected from the N observations and a regression model is fitted R-squared value is also calculated to capture the predictive power
to the remaining N − 1 observations. This equation is used to pre- of the model. Predicted R2 indicates how well a regression model
dict the withheld observation zi . This predicted value is denoted by predicts responses for new observations, and is calculated by
zbi , by which the prediction error for the ith observation is calculated
as εi ¼ zi –b zi . This prediction error is called the ith PRESS residual predicted R2 ¼ 1 −
PRESS
ð16Þ
(Montgomery 2017). This procedure is repeated for each observa- SST
tion i ¼ 1; 2; : : : ; N, producing a set of N PRESS residuals ε1 ; ε2 ;
ε3 ; : : : ; εN . The PRESS statistic is then defined as the sum of where SST = total sum of squares of the dependent variable. The
squares of the N PRESS residuals, as larger the predicted R2 value, the better the model (Hu 2016).
It can be concluded from Table 4 that, on the basis of lowest
X
N X
N RMSE, RME, and PRESS values and highest predicted R2 value,
PRESS ¼ ε2i ¼ ðzi − zbi Þ2 ð15Þ the form of equation with dimensionless discharge term Q=B3=2
i¼1 i¼1 Dg1=2 (i.e., Pair 4) more accurately predicted discharge than
Pairs 1, 2, and 3. However, the small difference in PRESS (3.36 and
The smaller the PRESS statistic, the better the model’s predictive 1.55) and predicted R2 (98.15 and 99.03) values between Pairs 3
capability. However, in addition to PRESS statistics, the predicted and 4 suggest that Pair 3 should also be taken into consideration.

1.25

1.00
Pair 3
m) = 0.45 m 0.60
0.75 Pair 1
R2 = 0.99
m) = 0.62 m 0.69
R2 = 0.99 Pair 4
0.50 m) = 0.39 m 0.58
R2 = 0.99

0.25 Pair 2
m) = 0.27 m 0.47
R2 = 0.99
0.00
0.00 0.25 0.50 0.75 1.00 1.25 1.50 1.75 2.00 2.25
Side slope, m
Pair 1 Pair 2 Pair 3 Pair 4

Fig. 5. Relationship between function ψðmÞ and side slope m.

© ASCE 06020007-5 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


70 60
Present Study Eq. (17) Present Study Eq. (17)
60 50
Present Study Eq. (18) Present Study Eq. (18)
50
40

Frequency (%)
Frequency (%)

40
30
30
20
20
10
10

0
0 <5 5-10 10-15 >15
<5 5-10 10-15 >15
Relative Absolute Error |( )/ |%
Relative Absolute Error |( − )/ |%
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Frequency distribution of relative absolute error for additional


Fig. 6. Frequency distribution of relative absolute error associated with
experimental observations.
proposed Eqs. (17) and (18).

100 Present study Eq. (17) Present study Eq. (18)


30
85 %
80 20
Submergence Ratio (%)

+ 10 %

Relative Error (%)


62 % 10
60
0%
r = 0.63 0

40 r = 0.7 - 10 %
-10
r = 0.8
r = 0.83 -20
20
r = 0.91
r=1 -30
0.0 0.3 0.6 0.9 1.2 1.5
0 Dimensional Head (y/D)
0.0 0.5 1.0 1.5
Dimensionless Head ( y/D) Fig. 9. Relative error values associated with Eqs. (17) and (18) for
additional experimental observations.
Fig. 7. Submergence limit conditions in experiments.

Therefore, these two forms (Pair 3 and Pair 4, respectively) of the different dimensions (Table S9). The models were also compared
stage–discharge equation were retained as discharge prediction models with Samani’s (2017) proposed model.
1. Comparison using additional independent set of observations:
Qp1 ¼ 0.45m0.60 D3=2 Bg1=2 ðy=DÞ2.5 ð17Þ The discharge prediction models, Eqs. (17) and (18), were
compared using independent observations (Table S8), recorded
Qp2 ¼ 0.39m0.58 B3=2 Dg1=2 ðy=DÞ2.4 ð18Þ for the same contraction ratio (0.63 ≤ r ≤ 1) but a different dis-
charge range of 0.0007–0.0341 m3 =s. A histogram of the relative
The error in discharge prediction using the proposed Eqs. (17) and absolute error associated with proposed Eqs. (17) and (18) was
(18) for the experimental observations (Tables S1–S7) were calculated. plotted for the experimental data (Table S8) as shown in Fig. 8. It
The frequency distribution of these errors are shown in Fig. 6. Free was found that 67.86% and 41.67% of the calculated discharge
flow conditions were ensured in the laboratory while performing values obtained using Eq. (17) show less than 10% and 5%
experiments—that is, the stage–discharge relationship can only be ap- deviation, respectively, from the measured discharge, whereas
plied when the tailwater submergence is not too high (Hager 1985). It all predicted discharge values using Eq. (18) were within the
was found that the submergence limit was reached when the ratio of 10% zone (Fig. 9), out of which 51.19% values were within
tailwater depth to the corresponding upstream flow depth was greater a 5% relative absolute error. The maximum relative error using
than 0.62. The maximum permitted tailwater depth (yd ) was recorded Eq. (17) was found to be 21.73%, with a mean absolute relative
to allow the free flow condition for all the experimental observations value of 7.73%. For the discharge prediction model given by
(Tables S1–S7). Submergence ratio versus the dimensionless head was Eq. (18), the maximum relative error was 9.96%, with a mean
plotted (Fig. 7) which indicates the submergence limits. absolute relative error value of 4.92%. Therefore, it was con-
cluded that Eq. (18) is more accurate in predicting discharge than
Eq. (17) for 0.63 ≤ r ≤ 1 and 0.0007 ≤ Q ≤ 0.0341 m3 =s.
Comparison of the Discharge Prediction Models
2. Comparison using observations on flumes with changed
The selected discharge prediction models Eqs. (17) and (18) were dimensions:
compared initially using additional experimental observations The laboratory flume was modified by increasing the bed
(Table S8), then using observations recorded on flumes with width of the channel from 0.2 to 0.314 m. Additional laboratory

© ASCE 06020007-6 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


experiments were conducted on this modified channel having (Table S3) on flumes having side slope 1∶1 and contraction ratios of
side slope of 1∶1. The diameters of the cylinders being inserted 0.91 and 1, so that the contraction ratio fell within the range 0.88–1
in the channel were also changed such that the contraction ratio of applicability of the Samani (2017) model.
now varied between 0.53 and 0.85 as against the earlier range The performance of the three models was compared using rel-
of 0.63–1. The observations recorded on the modified flumes ative error between measured and calculated values of discharge
(Table 5) are reported in Table S9. by the three discharge Eqs. (17)–(19), and a histogram of error is
The performance of Eqs. (17) and (18) were compared using plotted in Fig. 12. The plot shows that 28.12% and 25% of the ob-
relative error associated with the discharge values calculated servations gave more than 10% error, with maximum error values
using both the equations. Fig. 10 shows a histogram of relative of 17.05% and 16.75% when calculated using Eqs. (17) and (19),
absolute error associated with Eqs. (17)–(18), in which 70% of respectively. Another plot of predicted versus measured discharge
the experimental observations gave more than 10% error in dis- (Fig. 13) shows that all the data points were found to be within
charge prediction when calculated using Eq. (17), with a maxi- 10% error when obtained using Eq. (18) of the present study, with
mum relative error of 22.34% and mean absolute relative error maximum relative error value of 8.40%.
value of 12.58%. The predicted discharge using Eq. (18) lies From the three comparisons reported herein, it can be concluded
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

within the 10% error zone (Fig. 11) when compared with the that the free flow stage–discharge relationship given by Eq. (18)
actual measured discharge having maximum relative error of can predict discharge for CCBFs having different contraction ratios
9.89% with mean absolute relative error equal to 6.37% for the and side slopes more accurately than Eq. (17), as also evidenced
data points reported in Table S9. This indicated that Eq. (18) by the frequency distribution of errors. Hence, Eq. (18) is proposed
predicted discharge more accurately than Eq. (17) for contrac- as the discharge prediction model, applicable for a contraction ratio
tion ratios ranging between 0.53 and 0.85. range of 0.53–1 and side slope varying from 0.50 H:1 V to 2 H:1 V
3. Comparison with Samani’s (2017) model: for the CCBF. The stage–discharge relationship was characterized
Samani (2017) developed a single model for flow measure- by relative absolute error E which, for 59% of the measured dis-
ment in open channels having different cross sections using a charge values, when compared with those calculated by Eq. (18),
vertical column to create the contracted section. The equation was less than 5% with maximum error equal to 9.78% for data
proposed by Samani (2017) to measure flow in channels of trap-
ezoidal cross sections was
Present study Eq. 17 Present study Eq. 18
30
QSamani ¼ 0.226Bc 5=2 g1=2 ðH=Bc Þ1.51 ð19Þ
20
where Bc ¼ B − D þ 2mH. + 10 %
Relative Error (%)

Samani (2017) tested the proposed model [Eq. (19)] in a trap- 10


ezoidal field channel with side slope 1∶1 and contraction ratio rang- 0%
ing between 0.88 and 1, using cylinders of diameter 0.3–0.91 m 0
and flows ranging from 0.0195 to 0.528 m3 =s. In the present study,
- 10 %
discharge prediction models Eqs. (17) and (18) were compared -10
with Samani’s (2017) model [Eq. (19)] using observations recorded
-20

Table 5. Dimensions of the modified flumes


-30
S.No. B (m) D (m) r 0.0 0.3 0.5 0.8 1.0
Dimensionless Head (y/D)
1 0.314 0.166 0.53
2 0.314 0.200 0.64 Fig. 11. Relative error values associated with Eqs. (17) and (18) for
3 0.314 0.238 0.76
observations on modified flume.
4 0.314 0.268 0.85

70 80

Present Study Eq. (17) Present Study Eq. (17)


60 70
Present Study Eq. (18)
Present Study Eq. (18) 60
50 Samani 2017
Frequency (%)

Frequency (%)

50
40
40
30
30
20
20
10 10

0 0
<5 5-10 10-15 >15 <5 5-10 10-15 >15
Relative Absolute Error |( − )/ |% Relative Absolute Error |( − )/ |%

Fig. 10. Frequency distribution of relative absolute error associated Fig. 12. Frequency distribution of relative absolute error associated
with Eqs. (17) and (18) for observations on modified flume. with Eqs. (17) and (18) and the Samani (2017) model.

© ASCE 06020007-7 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


0.020
+ 10 % D = diameter of cylindrical column;
E = Relative absolute error;
- 10 % f1 , f2 = functional symbols;
Predicted Discharge (m3/s)

0.015
g = acceleration due to gravity;
H = upstream energy head;
0.010 i = any observation;
Lm = horizontal projection of channel side walls;
m = side slope of channel;
Present study Eq. (17)
0.005
Present study Eq. (18)
N = total number of observations;
Samani 2017 n = exponent in head–discharge relationship;
Line of perfect agreement ρ = density of water;
0.000 Q = flow rate;
0.000 0.005 0.010 0.015 0.020
Measured Discharge (m3/s) Qm = measured discharge;
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

Qp = predicted discharge;
Fig. 13. Comparison of selected discharge prediction models with the QSamani = discharge by Samani’s (2017) model;
Samani (2017) model. R = Reynolds number;
r = contraction ratio;
X = dimensionless head;
associated with CCBF (Tables S1–S7) having 0.63 ≤ r ≤ 1. For on-
Y = dimensionless discharge;
field flow measurements, easily available PVC or GI pipes of ap-
y = upstream flow depth;
propriate diameter can be used as obstruction. Locally available
materials, such as sand or stones, can also be used to further yc = critical flow depth;
strengthen its stability against the current of flow. This makes yd = maximum permissible tailwater depth to allow free
the CCBF an “on-the-go” measuring device for temporary flow flow condition;
measurements in open channels. zi = withheld observation;
zbi = predicted value of the withheld observation;
εi = ith PRESS residual;
Conclusions μ = viscosity of water;
П = dimensionless term; and
A simple flow measuring device for open channels, the cylindrical φ, ψ = functional symbols.
central baffle flume, was experimentally calibrated in this work.
The CCBF can be used to measure flow in trapezoidal channels
having side slopes ranging from 0.50 H:1 V to 2 H:1 V. A Supplemental Materials
stage–discharge relationship, Eq. (18), was developed for a CCBF
having a contraction ratio between 0.63 and 1. The proposed dis- Tables S1–S9 are available online in the ASCE Library (www
charge equation was validated using additional experimental obser- .ascelibrary.org).
vations recorded for the same range of contraction ratio (0.63 ≤
r ≤ 1). The statistical measure (RMSE, RME, PRESS, and pre-
dicted R2 ) values and the maximum relative error (within 10%) va- References
lidated the applicability of Eq. (18) for discharge prediction. The
Badar, A. M., and A. D. Ghare. 2012. “Development of discharge predic-
deviation of the measured and predicted discharge values using tion model for trapezoidal canals using simple portable flume.” Int. J.
the deduced stage–discharge equation for observations recorded Hydraul. Eng. 1 (5): 37–42. https://doi.org/10.5923/j.ijhe.20120105.02.
on CCBF with different bed width and contraction ratio range Barenblatt, G. I. 1979. Similarity, self-similarity and intermediate asymp-
(0.53–0.85) was within acceptable limits (<10%), as evident from totic. New York: Consultants Bureau.
the frequency distribution of errors. The presented approach for Barenblatt, G. I. 1987. Dimensional analysis. Amsterdam, Netherlands:
flow measurement in open channels using the CCBF is applicable Gordon & Breach.
for flow rates corresponding to experimental model discharge range Bijankhan, M., and V. Ferro. 2019. “Experimental study on triangular cen-
of 0.0007–0.035 m3 =s, contraction ratio range of 0.53–1, and sub- tral baffle flume.” Flow Meas. Instrum. 70 (Dec): 101641. https://doi
mergence ratio up to 62%. The model is versatile, as it is applicable .org/10.1016/j.flowmeasinst.2019.101641.
Cone, V. M. 1917. “The venturi flume.” J. Agric. Res. 9 (4): 15–129.
for any commonly found side slope of the channel.
Ferro, V. 2002. “Discussion of ‘Simple flume for flow measurement in open
channel’ by Zohrab Samani and Henry Magallanez.” J. Irrig. Drain.
Eng. 128 (2): 129–131. https://doi.org/10.1061/(ASCE)0733-9437
Data Availability Statement (2002)128:2(129).
Ferro, V. 2016. “Simple flume with a central baffle.” Flow Meas. Instrum.
All data, models, and code generated or used during the study 52 (Dec): 53–56. https://doi.org/10.1016/j.flowmeasinst.2016.09.006.
appear in the published article. Ghare, A. D., and A. M. Badar. 2014. “Experimental studies on the use
of mobile cylinders for measurement of flow through rectangular
channels.” Int. J. Civ. Eng. 12 (4): 504–511.
Hager, W. H. 1985. “Modified venturi channel.” J. Irrig. Drain. Eng.
Notation 111 (1): 19–35. https://doi.org/10.1061/(ASCE)0733-9437(1985)
111:1(19).
The following symbols are used in this paper: Hager, W. H. 1986. “Modified trapezoidal venturi channel.” J. Irrig. Drain.
B = channel bed width; Eng. 112 (3): 225–241. https://doi.org/10.1061/(ASCE)0733-9437
Bc = contracted channel bed width; (1986)112:3(225).

© ASCE 06020007-8 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007


Hager, W. H. 1988. “Mobile flume for circular channel.” J. Irrig. Drain. Peruginelli, A., and F. Bonacci. 1997. “Mobile prisms for flow measure-
Eng. 114 (3): 520–534. https://doi.org/10.1061/(ASCE)0733-9437 ment in rectangular channels.” J. Irrig. Drain. Eng. 123 (3): 170–174.
(1988)114:3(520). https://doi.org/10.1061/(ASCE)0733-9437(1997)123:3(170).
Hu, S. 2016. “Develop PRESS for nonlinear equations.” In Proc., ICEAA Robinson, A. R., and A. R. Chamberlain. 1960. “Trapezoidal flumes for
Professional Development and Training Workshop. Arlington, VA: open channel flow measurement.” Trans. Am. Soc. Agric. Eng.
Tecolote Research. 3 (2): 120–0124. https://doi.org/10.13031/2013.41138.
Kapoor, A., A. D. Ghare, A. D. Vasudeo, and A. M. Badar. 2019. “Channel Samani, Z. 2017. “Three simple flumes for flow measurement in open
flow measurement using portable conical central baffle.” J. Irrig. Drain. channels.” J. Irrig. Drain. Eng. 143 (6): 04017010. https://doi.org/10
Eng. 145 (11): 06019010. https://doi.org/10.1061/(ASCE)IR.1943 .1061/(ASCE)IR.1943-4774.0001168.
-4774.0001427.
Samani, Z., S. Jorat, and M. Yousaf. 1991. “Hydraulic characteristics of a
Kolavani, F. L., M. Bijankhan, and A. M. Mazdeh. 2018. “Discussion
circular flume.” J. Irrig. Drain. Eng. 117 (4): 558–566. https://doi.org
of ‘Three simple flumes for flow measurement in open channels’ by
/10.1061/(ASCE)0733-9437(1991)117:4(558).
Zohrab Samani.” J. Irrig. Drain. Eng. 144 (9): 07018029. https://doi
.org/10.1061/(ASCE)IR.1943-4774.0001324. Samani, Z., and H. Magallanez. 1993. “Measuring water in trapezoidal
Kolavani, F. L., M. Bijankhan, C. Stefano, V. Di Ferro, and A. M. Mazdeh. canals.” J. Irrig. Drain. Eng. 119 (1): 181–186. https://doi.org/10
2019. “Experimental study of central baffle flume.” J. Irrig. Drain. Eng. .1061/(ASCE)0733-9437(1993)119:1(181).
Downloaded from ascelibrary.org by University of Exeter on 07/02/20. Copyright ASCE. For personal use only; all rights reserved.

145 (3): 62. https://doi.org/10.1061/(ASCE)IR.1943-4774.0001370. Samani, Z., and H. Magallanez. 2000. “Simple flume for flow measurement
Montgomery, D. C. 2017. Design and analysis of experiments. Hoboken, in open channel.” J. Irrig. Drain. Eng. 126 (2): 127–129. https://doi.org
NJ: Wiley. /10.1061/(ASCE)0733-9437(2000)126:2(127).
Parshall, R. L. (1926). “The improved venturi flume.” Transportation Skogerboe, G. V., and M. L. Hyatt. 1967. “Rectangular cut throat flumes.”
89 (Mar): 841–880. J. Irrig. Drain. Eng. Div. 98 (4): 569–583.

© ASCE 06020007-9 J. Irrig. Drain. Eng.

J. Irrig. Drain Eng., 2020, 146(9): 06020007

You might also like