Refined Energy Correction for Calibration of Submerged

Radial Gates
Tony L. Wahl, P.E., M.ASCE1

Abstract: The energy-momentum 共E-M兲 method for calibrating submerged radial gates was refined using a large laboratory data set
collected at the Bureau of Reclamation hydraulics laboratory in the 1970s. The original E-M method was accurate in free flow, and when
the gate significantly controls submerged flow, but for large gate openings with low head loss through the gate, discharge prediction errors
were sometimes large 共approaching 70%兲. Several empirical factors were investigated with the laboratory data, including the combined
upstream energy loss and velocity distribution factor and the submerged flow energy correction. The utility of the existing upstream
energy loss and velocity distribution factor relation was extended to larger Reynolds numbers. The relation between the relative energy
correction and the relative submergence of the vena contracta was shown to be sensitive to the relative jet thickness. A refined energy
correction model was developed, which significantly improved the accuracy of submerged flow discharge predictions. Although the focus
of this work was radial gates, the energy correction concept and these refinements potentially have application to all submerged sluice
gates.
DOI: 10.1061/共ASCE兲0733-9429共2005兲131:6共457兲
CE Database subject headings: Gates; Discharge coefficients; Discharge measurement; Submerged discharge; Submerged flow;
Submerged jets; Hydraulic jump.

Introduction channel widths and differing channel invert elevations relative

to the gate sill,
Radial gates are common features of most irrigation projects. The • Potentially better accuracy when structures include multiple
ability to accurately measure discharge through these structures gates that are not operated uniformly, and
would allow project operators to more effectively deliver water to • Accurate determination of free- and submerged-flow condi-
end users in a timely and accurate manner, and would reduce the tions and accurate calibration continuously through free-flow,
need for the construction of separate dedicated flow measurement transition, and submerged-flow conditions.
structures. The calibration of radial gates for flow measurement is The E-M method is theoretically based, but also makes use of
a challenging hydraulic problem due to the number of possible empirical relations for gate contraction coefficients, energy loss,
gate, structure, and channel configurations, and the sensitivity of and velocity distribution factors in free flow, adjustments to the
calibrations to such factors as gate seal type and downstream energy equation in submerged flow, and hydrostatic forces on
channel width. Calibration methods for gates operating in a free- downstream channel boundaries. The initial empirical relations
flow condition are available in standard references and have rea- were developed from a series of experiments performed by the
sonable accuracy and ease of use, but calibrations for submerged Agricultural Research Service 共ARS兲 at the U.S. Water Conser-
gates are often very inaccurate, with errors of up to 50% reported. vation Laboratory, Phoenix, 共Tel 2000兲. These experiments used a
Most available calibration methods rely primarily on the energy single radial gate structure with a sharp-edged gate leaf. In free-
equation, although some incorporate the momentum equation to flow conditions the experiments covered a broad range of gate
distinguish between free- and submerged-flow conditions. openings, but in submerged-flow conditions the tests were per-
A procedure that uses both the energy and momentum equa- formed at only one gate opening with four different flow rates and
tions for flow calibration was recently developed 共Clemmens et a range of tailwater conditions. The submerged flow tests covered
al. 2003兲. The energy-momentum 共E-M兲 method uses an iterative an intermediate range of relative gate openings 共the ratio of gate
solution of the energy and momentum equations and offers sev- opening to upstream head兲, but did not include very small or very
eral potential advantages over previous methods: large relative gate openings. Despite the limited data, perfor-
• Ability to account for differing upstream and downstream mance of the method has thus far been encouraging, but there is a
great need for testing against other data sets, especially over a
1
Hydraulic Engineer, U.S. Dept. of the Interior, Bureau of wider range of submerged-flow conditions. Clemmens et al.
Reclamation, Water Resources Research Laboratory, Denver, CO. E-mail: 共2003兲 speculated that the relative gate opening might be a crucial
twahl@do.usbr.gov parameter affecting submerged-flow calibration.
Note. Discussion open until November 1, 2005. Separate discussions A series of tests performed at the Bureau of Reclamation hy-
must be submitted for individual papers. To extend the closing date by
draulic laboratory in Denver, Colorado 共Buyalski 1983兲 offers an
one month, a written request must be filed with the ASCE Managing
Editor. The manuscript for this paper was submitted for review and pos- opportunity to test the E-M method over a wide range of condi-
sible publication on September 23, 2003; approved on August 24, 2004. tions and possibly refine some of its empirical components. Buy-
This paper is part of the Journal of Hydraulic Engineering, Vol. 131, alski’s data were originally used to develop an energy-based cali-
No. 6, June 1, 2005. ©ASCE, ISSN 0733-9429/2005/6-457–466/$25.00. bration method, which was implemented in the RADGAT

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005 / 457

 ␦ = 1.001 − 0.2349␪ − 0.1843␪2 + 0.1133␪3 共1兲
with ␪ given in radians. This equation closely matches the experi-
mental results of Toch 共1955兲 and others, summarized in Clem-
mens et al. 共2003兲.

Free Flow
Clemmens et al. 共2003兲 wrote the energy equation from Section 1
to 2 using the free-flow jet velocity at the vena contracta

v2j v2j
H1 = H j + ⌬H = y j + ␣ j +␰ 共2兲
2g 2g
Fig. 1. Definition sketch for flow through a radial gate
where H1 = energy head at Section 1; H j = energy head at the vena
contracta 共Section 2兲; ⌬H = head loss between Sections 1 and 2;
y j = flow depth at the vena contracta, v j = average jet velocity; ␣ j
= velocity distribution coefficient for the jet; g = acceleration of
computer program. Buyalski tested nine gate configurations com-
gravity; and ␰ = energy loss coefficient. The velocity distribution
posed from three seal configurations 共sharp-edged, hard rubber
coefficient ␣ j was assumed to be 1.0, with any deviation from
bar, and music note or “J” seal兲, and three different ratios of gate unity accounted for in ␰, making 1 + ␰ a combined energy loss and
radius to trunnion pin height. Seven different gate openings were velocity distribution factor. Noting that the discharge is Q
tested for each configuration, with gate opening to trunnion height = ␦wbcv j, where y j = ␦w and bc is the gate width, one may substi-
ratios varying from 0.1 to 1.2. Nearly 2,650 test runs were made, tute for v j and y j in Eq. 共2兲 and solve for discharge to obtain
in both free and submerged conditions. The availability of both
free- and submerged-flow data for the same gates makes it pos-
sible to analyze the data in several ways to determine contraction
coefficients and other empirical factors in the E-M method. Buy-
Q = ␦wbc 冑 2g共H1 − ␦w兲
1+␰
共3兲

alski also collected more than 450 data points from 13 prototype The energy loss and velocity distribution factor 1 + ␰ was related
gates for use in a field verification program. to the Reynolds number of the flow at the upstream face of the
Unfortunately, all but one prototype site operated only in sub- gate. In the laboratory tests 共Tel 2000兲, values of 1 + ␰ varied from
merged flow, and the other operated in free flow only. Thus, it about 1.04 to 1.12 in tests covering a Reynolds number range of
would be much more difficult to isolate the various empirical about 0.5⫻ 105 to 2.7⫻ 105. Clemmens et al. 共2003兲 developed a
parameters in the same way as can be done with the laboratory relation between 1 + ␰ and the Reynolds number,
data. Also, at most of the prototype sites, current metering meth-
ods were used for independent discharge measurement, so uncer- −6R
1 + ␰ = 1 + 0.15e−5⫻10 共4兲
tainties in the measured data are much greater than in the labora-
tory data. where e = base of natural logarithms. The Reynolds number is R
This paper uses the Buyalski laboratory data set to test the E = VRh / ␯, where ␯ is the kinematic viscosity, V is the characteristic
-M method as proposed by Clemmens et al. 共2003兲, and then velocity determined at the gate opening, V = Q / 共bcw兲, and Rh is
presents modifications to incorporate the relative gate opening the hydraulic radius just upstream from the gate, between the gate
and its influence on the submerged-flow energy correction term. piers, Rh = b1y 1 / 共b + 2y 1兲. The upstream channel width is b1, and
y 1 is the upstream flow depth. Using this model, the value of 1
+ ␰ approaches 1.0 at large Reynolds numbers.
Methods
Submerged Flow
The E-M calibration method is described in detail by Clemmens For submerged flow, the energy equation is applied from Section
et al. 共2003兲. To provide a basis for the analysis and discussion 1 to 2 and the momentum equation is applied from Section 2 to 3.
that follow, the method is briefly reviewed here as originally pro- Application of the momentum equation requires estimates of flow
posed. Later, modifications to improve the model are presented in forces on the boundaries of the downstream channel, which are
the section entitled, “Analysis and Results.” given by empirical relations developed from the ARS experi-
The E-M equations are applied to the flow situation shown in ments.
Fig. 1. The energy equation is applied to the flow from Section 1 Clemmens et al. 共2003兲 described how the transition into sub-
to Section 2, and the momentum equation is applied to the flow merged flow causes a thickening of the jet issuing from beneath
from Section 2 to Section 3. For free flow, only the energy equa- the gate, accompanied by a velocity reduction. These changes
tion is needed to determine the flow calibration of the gate. Key take place as a result of the incomplete hydraulic jump against the
parameters are those shown in Fig. 1, and the contraction coeffi- downstream side of the gate and the associated adverse pressure
cient, ␦, for the flow beneath the gate, ␦ = y j / w, where y j is the jet gradient. Rather than model the actual changes in jet thickness
thickness at the vena contracta; and w is the vertical gate opening. and velocity, they proposed an alternative approach, modifying
In the experiments conducted by ARS, Tel 共2000兲 found that the the energy equation to include an energy correction term Ecorr that
contraction coefficient from a sharp-edged gate was a function of accounts for the reduced velocity head of the jet. The energy
the gate leaf angle, ␪, at the edge of the opening equation is then written as

458 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005

 v2j v2j Table 1. Radial Gate Configurations Tested by Buyalski 共1983兲
H1 = y 2 + ␣ j + ␰ − Ecorr 共5兲
2g 2g Trunnion
Gate pin height
where y 2 = flow depth at the now submerged vena contracta loca- designation 共mm兲 Seal design
tion. The value of v j is held constant at its free-flow value. Ecorr is a
1 461 Hard rubber bar
zero under free-flow and fully submerged conditions, but varies
2 511 Hard rubber bar
with the relative submergence of the jet in the transition zone, as
shown by Clemmens et al. 共2003兲. Eq. 共5兲 can be solved for 3 409 Hard rubber bar
discharge, as described previously for free flow, again assuming 4 409 Music note
that ␣ j = 1 5 461 Music note

冑
6 511 Music note
2g共H1 − y 2 + Ecorr兲 7 409 Sharp edge
Q = ␦wbc 共6兲
1+␰ 8 461 Sharp edge
9 511 Sharp edge
Clemmens et al. 共2003兲 performed submerged-flow tests at a a
single gate opening and a range of flow rates and tailwater levels Gate designations are correct. The tests with the hard rubber bar seal
were performed in a different order than the later tests.
to determine values of the energy correction. They found it help-
ful to examine the energy correction relative to the increase in
flow depth at the vena contracta, y 2 − y j, and the jet thickness y j. required because of the Reynolds number dependence of 1 + ␰.
They related the relative energy correction, Ecorr / 共y 2 − y j兲, to the After a free-flow discharge is computed, the momentum equation
relative depth increase 共y 2 − y j兲 / y j and developed the predictive is used to determine whether the gate is submerged. If so, the
equation momentum equation and submerged-flow energy equation are

冋 冉 冊册
solved iteratively until y 2 and Q converge. The energy correction
Ecorr y2 − y j term and effective velocity are incorporated into the solution pro-
= 0.52 − 0.34 arctan 7.89 − 0.83 共7兲
共y 2 − y j兲 yj cess.
For the analysis discussed in this paper, a Visual Basic com-
关It should be noted that a bracket was misplaced in this equation
puter program was written to process the data from the Buyalski
in the original publication of Clemmens et al. 共2003兲.兴 Eq. 共7兲
laboratory tests. The program is able to perform the E-M method
does not include any influence of the relative gate opening, w / H1,
calculations to compute flow rate, and is also able to solve in
but Clemmens et al. 共2003兲 speculated that with additional data a
reverse for several different parameters. First, for free-flow situ-
family of curves might be defined relating the relative energy
ations, the program can solve for the gate contraction coefficient
correction and the relative depth increase for different values of
that would produce perfect agreement between the observed dis-
w / H 1.
charge and the computed discharge from Eq. 共3兲, assuming that
In the ARS tests, y 2 was determined from pressure measure-
Eq. 共4兲 for 1 + ␰ is valid. Second, also in free flow, assuming that
ments made in the jet with a Prandtl tube. However, y 2 is very
the gate contraction coefficient is known, the program can solve
difficult to measure in the field due to large velocities and turbu-
for the value of 1 + ␰ that would produce the perfect agreement of
lence. To address this, the downstream depth, y 3, is measured, and
observed and computed discharge. Finally, in submerged-flow
the momentum equation is used to relate y 2 and y 3. The momen-
cases, the program can solve for the value of Ecorr 共and the ac-
tum equation can be written as
companying values of y 2, y j, and ve兲 that produces perfect agree-
y 22 Fw F3 ment of discharge, assuming that relations for the contraction co-
Q v e + b cg + = Qv3 + 共8兲 efficient and 1 + ␰ are given. In submerged flow, the program
2 ␳ ␳
assumes that the empirical weighting factors for the y w depth
where ve = effective velocity in the jet; ␳ = fluid density; F3 calculation are those obtained by Clemmens et al. 共2003兲. These
= hydrostatic pressure force exerted by the downstream water weighting factors appeared to be correct in that they properly
depth; and Fw = streamwise component of the force of water on all identified free versus submerged flow in almost all cases. This is
surfaces between Sections 2 and 3, including hydrostatic forces discussed in more detail later in the paper.
on all walls. Clemmens et al. 共2003兲 discuss the effective velocity
and other application details. Briefly, the effective velocity ac-
counts for the increased thickness and reduced velocity of the Buyalski Data
submerged jet discussed earlier, which is accounted for by the Buyalski 共1983兲 tested one radial gate configured in nine different
Ecorr term in the energy equation and must also be accounted for ways through a combination of three trunnion height settings and
in the momentum equation. The hydrostatic force Fw is computed three different types of gate seals. He identified these tests as
from an effective water depth, y w, computed as a weighted aver- shown in Table 1. Sketches of the laboratory model gate seals are
age of y 2 and y 3 shown in Fig. 2. Detailed drawings of Buyalski’s test flume and
the prototype and laboratory model gate seals can be found in
y w = py 3 + 共1 − p兲y 2 共9兲
Figs. 7, 9, and 10 of his report, which is available online 关see
The ARS tests were used to determine the empirical weighting Buyalski 共1983兲兴.
factors for y 2 and y 3, and a value of p = 0.643 was obtained for For all tests, the gate arm radius was 702 mm 共2.302 ft兲 and
their specific gate and downstream channel configuration. the gate width was 711 mm 共2.333 ft兲. The upstream and down-
stream channel widths 共at the locations where y 1 and y 3 were
measured兲 were 762 mm 共2.50 ft兲. The 51 mm 共2 in.兲 difference
Solution Procedure
between the gate width and the channel width was due to half-pier
The equations presented above must be solved iteratively to de- installed on one side of the laboratory flume. The floor of the test
termine the flow rate through a gate. In free flow, iteration is flume was level throughout its length. The channel approaching

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005 / 459

 Fig. 2. Radial gate seals 关after Buyalski 1983, Fig. 9共b兲兴

the gate was 3.12 m long 共10.25 ft兲 and the tailwater channel was
8.46 m long 共27.75 ft兲. An adjustable picket fence tailgate was
used to regulate downstream water levels. Data were collected at
six different gate openings 共seven gate openings for Gates 1, 2,
and 3兲, corresponding to nominal gate opening to trunnion height
ratios of 0.1, 0.2, 0.4, 0.6, 0.8, 1.0, and 1.2 共only for Tests 1–3兲. Fig. 3. Errors in prediction of submerged flow through sharp-edged
The greatest volume of data 共about 1,200 of the 2,650 runs兲 was radial gates
collected from Gate 1.
Data collection consisted of discharge, water level, and pres-
sure measurements. For each test, the flow condition was de-
scribed as FREE, SUBMERGED, or JUMP 共assumed to mean the within ±2% of one another, and in 99% of the cases, the predicted
flow was in the transition zone兲. The upstream water level, y 1, and observed discharge were within ±4% of one another. The
was measured 1.22 m 共4 ft兲 upstream from the gate seal position, mean relative error was +0.22%, and the standard deviation of the
at a section where the channel width was 762 mm 共2.50 ft兲. The relative errors was 1.48%. Errors were biased slightly positive
downstream water level, y 3, was measured 3.05 m 共10 ft兲 down- 共predicted flow greater than observed flow兲 for larger discharges.
stream from the gate seal position in free flow and 5.49 m 共18 ft兲 Discharge prediction for the submerged flow cases was good at
downstream in submerged flow, again in sections where the chan- small gate openings, but poor for low flows at larger gate open-
nel width was 762 mm 共2.5 ft兲. Pressures were measured at 25 ings where submergence is slight and the gate exerts little control
locations along the centerline of the channel floor between 0.3 m on the flow. Errors ranged from −13 to +70%. Despite the fact
共1 ft兲 upstream from the gate and 0.9 m 共3 ft兲 downstream from that there were some large prediction errors, 25% of the sub-
the gate. Pressures also were measured at the four locations on the merged flow cases were modeled with an error in the range of
sidewalls of the test section just downstream from the gate leaf, ±2%, 66% had errors in the range of ±5%, and 80% had errors in
and at three locations on the gate leaf itself. The floor pressure the range of ±10%. The mean relative error was +4.80%, but this
measurements would be useful for determining values of y 2 and was strongly influenced by a few large positive errors; the median
the contraction coefficient ␦, but unfortunately they could not be error was −1.48%. The standard deviation of the relative errors
located. Only the upstream and downstream water level measure- was 15.3%, again heavily influenced by a few large errors. One
ments were included in the final report 共Buyalski 1983兲. test failed to converge numerically 共Gate model 9, Test 113, for
The downstream water levels, y 3, reported by Buyalski were which w / H1 = 0.88兲.
adjusted “to an equivalent depth for a rectangular channel having Fig. 3 shows the submerged flow errors as a function of the
a width equal to the model gate width.” Buyalski reported that relative gate opening, w / H1, the relative depth increase at the jet,
this adjustment was necessary to eliminate the effect of the half 共y 2 − y j兲 / y j, and the relative energy correction, Ecorr / 共y 2 − y j兲. The
pier. The upstream depths were not adjusted. Because the E-M largest flow measurement errors occurred for w / H1 ⬎ 0.3, and for
method applies the momentum equation between Sections 2 and 3 共y 2 − y j兲 / y j = 0 to 1.5. Thus, the flow condition producing large
and thus accounts for wall forces applied by the pier, it was nec- errors is relatively low submergence 共i.e., transition zone flow兲 at
essary to adjust Buyalski’s reported downstream depths back to gate openings that are a large fraction of the upstream head. This
the original values that would have been measured downstream is consistent with the findings of Clemmens et al. 共2003兲 who
from the half pier. This allows accurate application of the momen- noted that data were lacking for large relative gate openings, and
tum equation to the actual measured data. This adjustment was that the largest errors occur when the flow is in the transition
made using the energy equation, assuming no head loss in the zone, where Ecorr / 共y 2 − y j兲 is in the range of ⬃0.2– 0.8 and is
expansion from the 711 mm 共2.333 ft兲 width to the 762 mm changing rapidly as a function of the relative increase in jet thick-
共2.50 ft兲 width 共believed to be the reverse of the adjustment made ness. There is a slight, but noticeable negative bias 共predicted
by Buyalski, although his report does not give details兲. flows too low兲 for highly submerged-flow conditions, reflected in
the difference between the mean and median errors discussed
above.
Evaluation of Original Energy-Momentum Model It should be noted that low submergence in this context means
only that the downstream flow depth is not dramatically greater
Before making modifications to the E-M model, the free- and than the theoretical free-jet thickness for a given gate opening.
submerged-flow data from the sharp-edged gates 共Gates 7, 8, and For a large relative gate opening, the submergence by this defini-
9兲 were used to test the E-M model as originally proposed by tion can be low at the same time that the downstream depth is
Clemmens et al. 共2003兲. almost equal to the upstream depth, a condition that would be
Discharge predictions for the free flow cases were very good. described as large submergence by those familiar with the sub-
In 79% of the cases, the predicted and observed discharge were mergence definition used for flumes and weirs.

460 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005

Fig. 4. Contraction coefficients of gates tested by Buyalski compared to relation for sharp-edged gates determined in Agricultural Research
Service tests

Analysis and Results ␦music note = 0.1292 + 共0.7884兲␦sharp 共12兲

The analysis objective was to use portions of the Buyalski labo- These relationships are based on free-flow tests conducted at
ratory data set to verify and/or improve the empirical relations angles of ␪ ⬍ 1.35 radians 共77°兲. Buyalski tested larger gate open-
previously developed by Clemmens et al. 共2003兲, especially the ings 共up to 98°兲 in submerged flow, but could not produce free
submerged-flow energy correction term. At each step, the existing flow at gate openings greater than 77°. The ARS relation for ␦sharp
empirical relations were used within previously established predicts increasing contraction coefficients when ␪ is greater than
ranges and attempts were made to verify, extend, or improve the 90° 共1.571 radians兲. This is contrary to the physics of the flow
relations for use in other ranges 共e.g., at larger Reynolds num- situation 共more flow contraction would be expected when the flow
bers兲. The sequence of steps taken was as follows: must turn a corner greater than 90°兲. For this reason, and because
1. Use free-flow data at low Reynolds numbers to compute con- submerged operation at such large gate openings is not a common
traction coefficients for the three gate seal types tested by field practice, Buyalski’s data for ␪ ⬎ 90° were discarded in the
Buyalski 共sharp edge, hard rubber bar, and music note seal兲. remainder of this analysis 共35 tests兲, but his submerged-flow data
2. Use contraction coefficients from Step 1 and free-flow data up to ␪ = 90° were used, making use of Eqs. 共10兲–共12兲 to compute
to examine behavior of 1 + ␰ at larger Reynolds numbers. contraction coefficients. For the sharp-edged gates, the data fit the
3. Confirm value of p, the empirical weighting factor for com- ARS relation 关Eq. 共1兲兴 closely enough that it was used without
puting hydrostatic forces on downstream channel boundaries, modification as Eq. 共10兲
by determining whether the E-M method correctly predicts The general trend in Fig. 4 is ␦music note ⬍ ␦sharp ⬍ ␦hard rubber.
whether flows in Buyalski’s tests were free or submerged. This makes sense physically, since the music note seal is rela-
4. Use submerged-flow data for gates with sharp edges and hard tively thick compared to the other two cases 共see gate seal
rubber bar seals to compute values of Ecorr, making use of the sketches in Fig. 2兲. The bulb is thickest very near the controlling
results from Steps 1 through 3. Examine the relationship be- edge and forces the streamlines away from the gate face just
tween Ecorr and the relative gate opening, and refine the before they begin turning the corner around the seal. Thus, a
model for computing Ecorr. larger change in flow direction must be accomplished, which
5. Test the refinements by attempting to predict discharge for yields more contraction and a smaller contraction coefficient. This
the Buyalski tests of gates with music note seals. effect seems to decrease as the gate angle increases, which makes
sense because at larger gate angles the flow approaching the gate
lip is less aligned with the face of the gate, so the additional
Contraction Coefficients contraction caused by the thick bulb is reduced. In contrast, the
The analysis began with the assumption that the upstream head hard rubber bar seal is much thinner than the bulb of the music
loss and velocity distribution factor, 1 + ␰, could be computed note seal, and the seal is held in place by a clamp bar that is set
from Clemmens’ relation 关Eq. 共4兲兴 for tests in which the gate back slightly from the controlling edge of the seal. The flow can
entrance Reynolds number was less than 2.7⫻ 105 关the upper align itself with the face of this clamp bar, and then the flow can
limit of Tel’s data, used to develop Eq. 共4兲兴. This made it possible actually begin to make its turn around the clamp bar before reach-
to use the Buyalski free-flow data to solve for the contraction ing the edge of the seal itself. This, combined with the fact that
coefficients in the free-flow tests conducted at R ⬍ 270,000. Con- the edge of the rubber bar is likely to be somewhat rounded itself,
traction coefficients for the hard rubber bar and music note seals causes less contraction 共a larger coefficient兲 than for the sharp-
were compared to the contraction coefficients for the sharp-edged edged gate.
gate, as given by Eq. 共1兲, and regression relations for the ratios of
the contraction coefficients were developed as a function of the
Energy Loss and Velocity Distribution Factor
gate angle, ␪. This yielded the following relationships 共Fig. 4兲:
The second step of the analysis was to use the newly developed
␦sharp = 1.001 − 0.2349␪ − 0.1843␪2 + 0.1133␪3 共10兲 contraction coefficient relations and the free-flow data for R
⬎ 270,000 to examine the relation for the upstream energy loss
and velocity distribution factor, 1 + ␰ 关Eq. 共4兲兴, at large Reynolds
␦hard rubber bar = 0.0138 + 共1.0209兲␦sharp 共11兲 numbers. Data from all three gate seal types were used. This may

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005 / 461

 Downstream Wall Depth Weighting Factor
The next stage of the analysis was to verify the value of the
parameter p used to estimate wall forces in the momentum equa-
tion. A value of p = 0.643 was obtained from the ARS tests 共Clem-
mens et al. 2003兲, by solution of the momentum equation from
measured values of y 2 and y 3. The Buyalski data set does not offer
the same opportunity to solve for p, since y 2 was not measured
independently. The only test that is possible is to determine
whether the momentum equation correctly predicts free- and
submerged-flow conditions for the Buyalski tests. If there are runs
that are not correctly modeled, then one could hypothesize that
the run in question was exactly at the transition from free to
submerged flow and solve the momentum equation in reverse to
determine the value of p that balances the momentum equation.
Fig. 5. Combined energy loss and velocity distribution factor The Buyalski free- and submerged-flow data sets for all gate
configurations and Reynolds numbers 共except gate openings
greater than 90°兲 were processed to determine whether the E-M
at first seem to be a circular argument, since we are using the method would accurately identify the flow as free or submerged.
contraction coefficients we just determined by assuming that Eq. The momentum force of free flow through the gate is compared to
共4兲 was applicable. However, it should be carefully noted that the momentum force associated with the measured tailwater. If
different data sets segregated by Reynolds number are being used. the tailwater force is greater, the flow is submerged; if the tailwa-
We only applied Eq. 共4兲 in the previous section to a range of data ter force is smaller, the flow is free and a hydraulic jump occurs
for which its validity is already well established 共Clemmens et al. downstream from the gate.
2003兲. The contraction coefficients determined from that analysis A total of 458 runs were examined that had been described by
can now be used here, since they do not vary with Reynolds Buyalski as FREE, and 454 were correctly identified as free flow
number. using the E-M method. The four cases that were identified as
The free-flow equations were solved in reverse to determine submerged flow were initially perplexing because the y 3 depth
the values of 1 + ␰, and the results are shown in Fig. 5. The spread was less than the gate opening. A closer investigation showed that
of the Buyalski data is much greater than that of the ARS data; in the momentum equation was unbalanced because the reported y 3
fact, values less than 1 were computed for many cases, which do depth was too low, causing the Qv3 term in the momentum equa-
not make theoretical sense. Despite the scatter in the Buyalski tion to become extremely large. A very slight increase in y 3
data, Fig. 5 appears to confirm that Eq. 共4兲 is applicable to larger caused the flow to be correctly modeled as free.
Reynolds numbers. The submerged-flow data were tested in a similar manner. A
After careful investigation, it was found that the errors visible total of 2,101 runs that were described by Buyalski as SUB-
in Fig. 5 are related to some degree to gate setting. Buyalski’s MERGED were analyzed, and 40 of these runs were identified by
gate openings were set by lowering the gate onto a pair of ma- the E-M method as free flow. For these cases, an attempt was
chined blocks and then clamping the gate in place for a complete made to compute values of p that would balance the momentum
series of tests. Despite this care, a series of tests run on a given equation and cause the E-M method to predict submerged flow.
gate at the same gate opening often plot somewhat below or For all 40 cases, the computed value of p was greater than 1.
above the curve defined by Eq. 共4兲, but with a similar shape. This In addition to free and submerged flows, Buyalski reported 63
suggests an error in the gate opening, or perhaps another param- tests having a flow condition described as JUMP. This is assumed
eter common to a series of tests at a given gate opening. The to mean that the flow is in the transition zone, but it is not clear
problem appears to be greater on the gates with seals, so seal whether we should expect such flows to be modeled as free or
compression on the positioning blocks may have been a factor, submerged when examined with the momentum equation. It was
although the variability appears more random than biased. An- hoped that these tests might provide conditions in which the mo-
other possibility is the variation in the accuracy of the venturi mentum equation would be almost perfectly balanced and might
meters used to measure discharge. Venturi meters ranging from give an opportunity to solve for p at the balanced condition.
75-mm 共3-in.兲 to 356-mm 共14-in.兲 diameter were used, and tests These 63 tests were analyzed, with 44 initially identified as free
at low gate openings probably used smaller venturi meters, while and 19 identified as submerged. The momentum equation was
those at larger gate openings would have used the larger venturi then solved to determine the balancing value of p for each case.
meters 共specific meters were not identified for each test兲. Bias In 39 cases, the balancing value of p was a complex number, in
errors in individual venturi meter calibration could thus show up 21 cases, p was negative, in two cases p was greater than 1, and
as systematic errors related to gate opening. Buyalski’s venturi in one case p was equal to 0.46.
meters were calibrated using a volumetric tank, with an uncer- In summary, in almost every case, physically unreasonable
tainty of about ±0.5%. For comparison, discharge measurements values of p were needed to balance the momentum equation. This
in the ARS tests had an uncertainty of about ±0.1%, using a suggests that the tests were not actually close enough to the tran-
weigh tank. Thus, both gate setting and discharge measurement sition flow condition, or that other random experimental errors
are possible sources of the additional scatter in Buyalski’s data. were unbalancing the momentum equation to a greater degree
For tests at one stated gate opening, the errors are primarily sys- than could be accounted for by changing p within a reasonable
tematic, but over the course of thousands of tests the errors are range.
random, as shown in Fig. 5, so they should not dramatically affect One might have expected that a lower value of p would be
the conclusions drawn from these analyses. appropriate for Buyalski’s tests, since the channel was much nar-

462 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005

 Fig. 6. 共Color兲 Relative energy correction versus relative depth increase, subdivided by y j / H1 ranges

rower than that used by Tel 共2000兲. Buyalski’s downstream chan- relative jet thickness if the contraction coefficients are also differ-
nel was only 7% wider than the gate, whereas Tel’s was 2.7 times ent. The important point is that the flow downstream from the
the width of the gate. In a narrower channel, one would expect the gate should behave the same, regardless of how the final jet thick-
downstream depth, y 3, to have less influence on the forces exerted ness is produced. Thus, we will use y j / H1 in the analysis that
on and by the transitional sidewalls surrounding the gate, since follows, and some implications of this will be discussed later.
there is less room for eddies and recirculation. This expectation Fig. 6 shows computed values of the relative energy correc-
was fulfilled by the one run in which p = 0.46 balanced the mo- tion, Ecorr / 共y 2 − y j兲 versus the relative depth increase at the vena
mentum equation. However, one data point was felt to be insuf- contracta, 共y 2 − y j兲 / y j, subdivided by ranges of y j / H1 values. Re-
ficient evidence to warrant changing the value of p. lationships between the energy correction and a jet Froude num-
ber and jet Reynolds number were also investigated, but were not
Energy Correction Factor in Submerged Flow consistent over the full range of the data. The Ecorr and y 2 values
are those obtained by iterative solution of the momentum and
The next stage of the analysis was to determine values of Ecorr for energy equations to obtain perfect prediction of the discharges
the submerged-flow tests, and to compare those results to the observed in the Buyalski tests. The figure also shows the ARS
previously developed ARS curve relating Ecorr / 共y 2 − y j兲 and 共y 2
fitted curve 关Eq. 共7兲兴.
− y j兲 / y j. Only the data from Gates 1, 2, and 3 共hard rubber bar
The general trend is for the transition zone of the energy cor-
seal兲 and Gates 7, 8, and 9 共sharp edged兲 were used. The
rection curve to become steeper and shift to the left 共toward lower
submerged-flow data for Gates 4, 5, and 6 共music note seals兲 were
values of the relative depth increase兲 as the y j / H1 ratio increases.
saved for verification testing.
Thus, when the gate exerts less control on the flow, transition
The relative energy correction relation proposed by Clemmens
et al. 共2003兲 关left side of Eq. 共7兲兴 varies from 1.0 at 共y 2 − y j兲 / y j occurs much more rapidly. There is some scatter in the data, in-
= 0 to zero at 共y 2 − y j兲 / y j ⬎ 3.89. At each extreme, Ecorr itself ap- cluding a small fraction of computed values of Ecorr / 共y 2 − y j兲 that
proaches zero, since the relative energy correction computed from are greater than 1.0 or less than zero. These data are inconsistent
Eq. 共7兲 must be multiplied by the depth increase, y 2 − y j, to obtain with the physical meaning of the energy correction term and are
Ecorr. The maximum effect of the energy correction actually oc- attributed to experimental errors or anomalies. These data were
curs at intermediate values of the relative depth increase, in the excluded from later curve-fitting efforts.
region where the relative energy correction is changing rapidly. The initial objective was to fit the ARS curve or an equation of
As mentioned earlier, Clemmens et al. 共2003兲 speculated that the a similar form to the Buyalski data for different y j / H1 ratios by
relative gate opening, w / H1, might affect the energy correction changing the values of one or more of the four empirical factors
relationship. in Eq. 共7兲. This proved to be a difficult task. The objective in
Although relative gate opening, w / H1, is a convenient physical development of the ARS curve had been to obtain a function that
ratio, from a hydraulic standpoint a more appropriate dimensional passed through the point Ecorr / 共y 2 − y j兲 = 1 at 共y 2 − y j兲 / y j = 0, and
ratio is the relative jet thickness at the vena contracta, y j / H1. approached a limit of Ecorr / 共y 2 − y j兲 = 0 at larger values of 共y 2
Since different gate seals produce varying contraction behavior, − y j兲 / y j. Eq. 共7兲 accomplished this with a relatively complex
two different gates set to the same relative gate opening might curve using an inverse tangent function and having an inflection
produce a different relative jet thickness, or alternately, two gates point near Ecorr / 共y 2 − y j兲 = 0.5. This functional form appeared to be
set to different openings might produce a flow with the same somewhat compatible with the Buyalski data at large values of

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005 / 463

 Fig. 7. Regression relation for parameter ␧ in Eq. 共12兲 Fig. 9. Energy correction versus depth increase relative to upstream
head
y j / H1, but for small values of y j / H1, it could not fit the data at
intermediate values of 共y 2 − y j兲 / y j and still pass through the point
共y 2 − y j兲 / y j = 0 and Ecorr / 共y 2 − y j兲 = 1. It is difficult to conclude in- contraction coefficient, ␦, would have remained embedded in the
dependently from the Buyalski data that passage through the final equation, indicating that the energy correction factor was
共0, 1兲 point is necessary because the Buyalski data set contains dependent on the contraction characteristics of the gate. Such an
few points at very low values of 共y 2 − y j兲 / y j, but this requirement analysis was performed previously by Wahl et al. 共2003兲. The
is compelling from a physical viewpoint. form of Eq. 共14兲 suggests that if we plot data for widely varying
After exploring several alternative equation forms, the most y j / H1 ranges as a function of 共y 2 − y j兲 / H1, rather than versus 共y 2
satisfactory model found for the energy correction relationship − y j兲 / y j as was done in Fig. 6, we should obtain a single curve
was a simple exponential power function of the form rather than a family of curves dependent on the relative jet thick-
ness. Fig. 8 shows that this is generally the case, with just a few
Ecorr distinct bands of y j / H1 included on the plot for clarity. This also
= e␧关共y2−y j兲/y j兴 共13兲
y2 − y j suggests that a single exponential function could be fit to all the
data, rather than analyzing distinct bands of y j / H1 values. Doing
where ␧ = empirically determined coefficient; and e = base of natu- so yields a slightly different coefficient 共−6.75兲 in Eq. 共14兲. By
ral logarithms. To fit these data to this relation, curve fitting was analyzing the data in bands, each range of y j / H1 values carries
performed manually with the objective of minimizing the sum of similar weight in the final result, even though the bands did not all
the Pearson residuals, ln关共1 + 兩residual兩2兲0.5兴. This approach mini- contain the same number of data.
mizes the influence of outliers. This procedure produced good We can use Eq. 共14兲 to examine the physical significance of
curve fits for most of the y j / H1 ranges, although a few were less the submerged-flow energy correction, Ecorr, by computing values
than satisfying. Future research may better define the shape of the of Ecorr / 共y 2 − y j兲 as 共y 2 − y j兲 / H1 varies from zero to 1, and then
curve for low values of 共y 2 − y j兲 / y j. multiplying by the associated value of 共y 2 − y j兲 / H1 to obtain the
A total of 13 bands of y j / H1 values were analyzed, and expo- value of Ecorr / H1 as a function of 共y 2 − y j兲 / H1 共Fig. 9兲. The maxi-
nential functions were fitted for each range. Fig. 7 shows ␧ plot- mum energy correction is about 5.5% of the upstream head and
ted versus the average y j / H1 values for each band. A linear re- occurs when the depth increase at the vena contracta is about 15%
gression yields the following final equation for Ecorr / 共y 2 − y j兲, of the upstream head.
which replaces Eq. 共7兲: One question that arises in evaluating the outcome of this
Ecorr analysis is the influence of the other empirical factors whose val-
= e−6.78共y j/H1兲关共y2−y j兲/y j兴 = e−6.78关共y2−y j兲/H1兴 共14兲 ues may have affected the determination of Ecorr. To evaluate this,
y2 − y j
some simple sensitivity analyses were performed. The values of
Note the cancellation of the y j terms which takes place because the upstream energy loss factor, ␰, the contraction coefficient, ␦,
we chose to analyze the energy correction as a function of the and the momentum equation weighting factor, p, were each
relative jet thickness, y j / H1, rather than the relative gate opening, changed individually by 5%, and Ecorr values were then recom-
w / H1. If the analysis had been based on w / H1, some simplifica- puted. The percentage change in the value of Ecorr increases with
tion would have been possible by recognizing that y j = ␦w, but the the level of submergence, but at the point of maximum influence

Fig. 8. Relative energy correction versus depth increase relative to: 共a兲 jet thickness and 共b兲 upstream energy head, for selected bands of y j / H1
values

464 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005

 Table 2. Comparison of Submerged-Flow Errors
Percentage of observations

Original E-M Modified E-M

model applied to model applied to
sharp-edged Gates Gates 4, 5, and 6
7, 8, and 9 共music note seals兲
Description 共%兲 共%兲
±2% error 25 46
±5% error 66 78
±10% error 80 93
±20% error 86 99.6
+20 to + 70% error 14 0.4
Percentage error statistics
Mean error +4.80 −1.35
Median error −1.48 −0.75
Fig. 10. Submerged-flow discharge prediction errors in verification Standard deviation 15.3 4.70
test of the energy-momentum method with the refined energy Note: E-M = energy-momentum.
correction model

nal E-M method for which Clemmens et al. 共2003兲 reported that
of Ecorr 关approximately 共y 2 − y j兲 / H1 = 0.15, as explained in the pre- larger prediction errors occurred when the relative energy correc-
vious paragraph兴, the changes in Ecorr due to ␰, ␦, and p were tion was in the range from 0.2 to 0.8 共the range in which the curve
approximately 1.9, 44, and 1.6%, respectively. Clearly, it is im- was relatively steep兲, as Fig. 3 confirmed.
portant to know the contraction coefficient accurately, while un- To test the sensitivity of the E-M method to the influence of
certainties in the other two parameters have a much smaller, but the weighting parameter, p, in the momentum equation, a sensi-
still significant, influence on our determination of Ecorr. tivity analysis was performed on the verification data set. For a
change in p of 0.1, the resulting change in computed discharge
共average of 236 runs from the verification data set discussed later
Verification Testing
in the paper兲 was about 0.4 to 0.5%. Thus, the influence on cali-
To test the effect of refining the submerged-flow energy correc- bration accuracy of a reasonable variation of p was relatively
tion model, the E-M method was used to predict discharges for slight.
the Buyalski data from Gates 4, 5, and 6 共music note seals兲, It should be noted that the significant random scatter visible in
computing the energy correction with Eq. 共14兲 rather than Eq. 共7兲. Buyalski’s free-flow data 共Fig. 5兲 compared to that in the ARS
Both free-flow and submerged-flow cases were tested, although data may carry over to the submerged-flow analysis as well. Thus,
no changes had been made to the free-flow model. In free flow, some of the scatter in Fig. 10 may be due to the quality of the
174 test runs were analyzed. Discharge was predicted within ±2% Buyalski data set, and the E-M method may perform better with
for 64% of the cases, and within ±5% for 99.4% of the cases. An higher-quality data. The important point to remember is that a
error of +9% was obtained in one case; this test was reported to dramatic improvement was realized when the refined energy cor-
be submerged flow by Buyalski, but was modeled by the E-M rection model was used.
method as free flow, so it is possible that there is an error in the
data. The mean relative error was +0.40%, and the standard de-
viation of the relative errors was 1.97% 共ignoring the one run just Conclusions
mentioned兲. These results are not quite as good as the initial test
made against the free-flow data for the sharp-edged gates 共Gates With the refined energy correction model described here, the E-
7, 8, and 9兲, and the difference is most likely caused by additional M method accurately predicts both free- and submerged-flow dis-
uncertainty in the contraction coefficients of the music note seals. charges through the model radial gates tested by Buyalski. Flow
We should thus expect slightly larger errors in submerged flow as measurement uncertainties in free-flow conditions are about ±2%
well. to ±5%. In submerged-flow conditions, the relative gate opening,
The submerged-flow discharge prediction errors for 236 tests w / H1, plays an important role in determining the energy correc-
of gates with music note seals are shown in Fig. 10. To help tion term needed for the energy equation. With the refined energy
illustrate the flow conditions that are the most error prone, the correction model, discharge in submerged-flow conditions is pre-
errors are plotted versus the relative gate opening, the relative dicted with an uncertainty on the order of ±5% to ±10% in most
depth increase, and the value of the relative energy correction, cases. The largest errors still occur during conditions of slight
similar to the plots shown in Fig. 3 for the original E-M method submergence at large relative gate openings, but the magnitude of
applied to the sharp-edged gates. The error distribution is also such errors has been reduced dramatically from as much as 70%
compared in Table 2 to the results obtained with the original E to about 10%.
-M model applied to the sharp-edged gates. There was significant Several empirical parameters and relations are important to the
improvement on all levels, and dramatic improvement in elimi- performance of the E-M method. In free-flow conditions, the con-
nating the very large errors that occurred when the flow is in the traction coefficient and the energy loss and velocity distribution
transition zone. The figure shows that there is no significant trend factor are the primary sources of uncertainty in the model, with
in the magnitude of errors as a function of the relative energy the contraction coefficient being the dominant source of error. The
correction. This indicates significant improvement over the origi- Buyalski data set provided enough information to create useful

JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005 / 465

models for the contraction coefficients of the hard rubber bar bc ⫽ gate width;
seals and music note seals used on many prototype radial gates, Ecorr ⫽ submerged-flow energy correction;
but further improvement is certainly possible, and variability in e ⫽ base of natural logarithms, 2.718282;
the characteristics of these types of seals on prototype gates is F3 ⫽ hydrostatic force on downstream control volume
likely to be significant. The Buyalski free-flow data were not able boundary;
to shed much additional light on the energy loss and velocity Fw ⫽ force exerted by upstream channel boundaries
distribution factor, although many of the tests were carried out at surrounding gate;
larger Reynolds numbers than those attained in previous work by g ⫽ acceleration of gravity;
ARS. The equation proposed by Clemmens et al. 共2003兲 appears H ⫽ total energy head;
to be the best available relation. This is a question of importance p ⫽ weighting factor used to compute y w;
for field applications, since most prototype gates will have much Q ⫽ discharge;
larger Reynolds numbers than those used to develop the model for R ⫽ Reynolds number;
the energy loss and velocity distribution factor, but fortunately Rh ⫽ hydraulic radius immediately upstream from gate;
1 + ␰ approaches a limiting value of 1.0 at large Reynolds num- r ⫽ gate radius;
bers. V ⫽ characteristic velocity at entrance to gate opening;
In submerged flow, the additional factor that is of great impor- v ⫽ flow velocity;
tance is the energy correction term. The model proposed by ve ⫽ effective velocity in vena contracta jet in submerged
Clemmens et al. 共2003兲 was refined, yielding a relation that in- flow;
corporates the significant effects of the relative gate opening. The w ⫽ vertical gate opening;
form of the energy correction relation proposed here differs sig- y ⫽ flow depth;
nificantly from that proposed by Clemmens et al. 共2003兲, and yj ⫽ jet thickness at the vena contracta;
additional experimental data are needed to verify the behavior of yw ⫽ effective water depth used to estimate Fw;
the energy correction at the beginning of the transition into sub- ␣ ⫽ velocity distribution coefficient;
merged flow 关low values of 共y 2 − y j兲 / y j兴. ⌬H ⫽ energy loss;
The analysis performed here did not produce any change in the ␦ ⫽ contraction coefficient;
empirical weighting factor, p, used to estimate forces on the ␧ ⫽ empirically determined exponent;
downstream channel walls needed for the momentum equation. ␪ ⫽ gate lip angle from horizontal;
Buyalski’s test channel was relatively narrow compared to that ␯ ⫽ kinematic viscosity;
used in the ARS tests, but was otherwise similar, so it is still ␰ ⫽ energy loss coefficient; and
possible that other geometries will require significant adjustment ␳ ⫽ fluid density.
of the value of p. Features such as bed drops downstream from
Subscripts
the gate or flared transitional walls may prove to be important.
1 ⫽ sections upstream from the gate;
2 ⫽ sections at the vena contracta;
Acknowledgments 3 ⫽ sections downstream from the gate; and
j ⫽ conditions in the vena contracta jet.
The writer thanks Dr. Albert Clemmens and the anonymous re-
viewers of this manuscript for their insightful suggestions on the
analysis of these data and the preparation of this paper. The work References
performed here would never have been possible without the ex-
tensive set of laboratory data collected by Mr. Clark Buyalski. His Buyalski, C. P. 共1983兲. “Discharge algorithms for canal radial gates.”
perseverance and thoroughness in the collection and publication Research Report REC-ERC-83-9, Bureau of Reclamation, Denver,
of these data are exemplary. The work reported here was funded Available on the web at 具http://www.usbr.gov/pmts/hydraulicsគlab/
by Reclamation’s Science and Technology Program as part of the pubs/REC/REC-ERC-83-09.pdf典.
Flow Measurement with Canal Gates project. Clemmens, A. J., Strelkoff, T. S., and Replogle, J. A. 共2003兲. “Calibration
of submerged radial gates.” J. Hydraul. Eng., 129共9兲, 680–687.
Tel, J. 共2000兲. “Discharge relations for radial gates.” MSc. thesis, Delft
Technical Univ., Delft, The Netherlands.
Notation Toch, A. 共1955兲. “Discharge characteristics of tainter gates.” Trans. Am.
Soc. Civ. Eng., 120, 290–300.
The following symbols are used in this paper: Wahl, T., Clemmens, A. J., and Replogle, J. A. 共2003兲. “The energy
a ⫽ gate trunnion pin height above base of gate chamber; correction for calibration of submerged radial gates.” Proc. 2nd Int.
b1 ⫽ upstream channel width; Conf. Irrigation and Drainage, USCID, Phoenix.

466 / JOURNAL OF HYDRAULIC ENGINEERING © ASCE / JUNE 2005