Discharge Formulas of Crump-De Gruyter Gate-Weir For Computer Simulation
Discharge Formulas of Crump-De Gruyter Gate-Weir For Computer Simulation
Discharge Formulas of Crump-De Gruyter Gate-Weir For Computer Simulation
Abstract: One of the problems of interest to professionals in the field of irrigation and drainage is the computer simulation of discharge
or level control structures. Particularly troublesome are structures that display a marked change of behavior when the downstream water
level exceeds a certain limit. The Crump-de Gruyter gate displays several such changes of behavior. Not only does it exhibit a transition
from free to drowned flow when the downstream water level rises, it can also go from weir to gate flow. A series of experiments in a
laboratory flume provided the basic data to test a simple mathematical model of this structure. The model assumes the structure is located
between two reaches with sub-critical flow in the upstream and downstream reach.
DOI: 10.1061/共ASCE兲0733-9437共2003兲129:4共270兲
CE Database subject headings: Water discharge; Computer aided simulation; Models; Weirs; Head loss.
Weir
From Fig. 3, it is clear that the head loss over the drowned weir is
very small. To distinguish between different discharges for a
given downstream water level we need to know the head loss with
an accuracy that is not likely to be attainable in the field. For
example, in our laboratory setup the difference between 27 and 35
L/s is less than 0.5 mm 共Fig. 3兲. It would seem that an accurate
estimate of the water level over the weir is not the central prob-
lem. Therefore, in the case of the drowned weir we use the stan-
dard approximation for expansions, i.e., we take the force on the
sides and bottom of the expansion equal to the static force that a
water level equal to the downstream water level would exert. This
results in the momentum balance given in Eq. 共3a兲
B2 2 Q2 B2 2 Q2
Fig. 2. Location of cross sections h 2⫹ ⫽ h 3⫹ (3a)
2 gB 2 h 2 2 gB 3 d 3
For the free weir we use the critical depth corresponding to the
discharge, given by Eq. 共3b兲 to provide the water level over the
weir
whereas ⫽1 corresponds to a force on the weir based on the
h 2 ⫽h c ⫽ 冑
3 Q2
gB 22
(3b)
water level at Cross Section 2. A discussion of the effects of
different choices of can be found in the section on comparison
of theory and experiment.
Next we will use our results to derive a formula for h 2 that
Gate does not contain Q and as such can be used to determine the
For the gate, we calculated the force on the weir from our experi- discharge.
mental data and found an alternative to Eq. 共3a兲 that results in a
better model of the free gate to drowned gate transition 共cf. the Algorithm to Determine Flow State and Discharge
section on comparison of theory and experiment兲. We will now
describe this alternative. The previous sections have provided us with, on the one hand, a
We combine the small expansion due to the widening of the discharge formula in terms of the upstream depth and the water
canal and the straight drop at the end of the weir and approximate level over the weir, and on the other hand formulas for the water
the forces on the walls and the toe of the weir by the static level over the weir and the downstream water level and the dis-
pressure corresponding to the downstream water level 关Eq. 共4兲兴 charge.
Please note that Eq. 共2兲, the starting point of our algorithm,
F e ⫽g
B 3 d 23
⫺g
冉
B 2 d 3⫺
a
5 冊 2
(4)
contains a coefficient C that, for real-life structures, typically de-
pends both on the flow state and the discharge 共cf. Bos 1990兲. In
2 2 our algorithm we allow for different values of C for different flow
states, but we do not let C depend on the discharge in each sepa-
The calculation of the force on the downstream end of the weir rate flow state. This choice was made because it results in equa-
is based on the following approximation of the water level. As- tions that can be solved without the iteration steps. Equations with
sume the water level follows a quadratic curve running from the a Q dependent C would probably need some form of iteration in
water level h 2 just past the gate 共at Cross Section 2兲 to the level the solution process.
h 3 at the vertical drop at the end of the weir 共Fig. 4兲. The curve Our choice implies that we will need measurements of the
has a horizontal tangent at the latter point. structure in each flow state to determine the actual values for
This results in a horizontal force component as described by these constants. In the section on our experiments we will discuss
Eq. 共5兲 on the slope of the weir. When combined with Eq. 共4兲 this a procedure to determine C for the different flow states. For now,
provides us with Eq. 共6a兲 for h 2 with as specified in Eq. 共6b兲 we assume we have available the following values for C: C weir,free
for free weir flow, C weir,drowned for drowned weir flow, C gate,free for
冉 冊
7/4⫺1/5
F s ⫽B 2 g 冕
x⫽3/4
ah 2 ⫹a 2 x⫺
3
4
free gate flow, and C gate,drowned for drowned gate flow.
The computer program using our algorithm needs to calculate
a discharge for given upstream (h 1 ) and downstream (h 3 ) water
冉 冊
h 23 ⫺h 22 ⫹2a 共 h 3 ⫺h 2 兲 ⫽
2Q 2
gB 22 w
1⫺
B 2w
B 3d 3
(6a)
water level. However, the aim of the algorithm is to supply a
discharge, not to signal the precise moment of change of state.
Question: Can the structure be in free weir flow 共A1兲?
⫽
4
5 冉 冊
1⫺
3,179
3,675
⬇0.108 (6b)
Answer: Only if two conditions are met, the gate must not touch
the water surface and the balance of forces between Cross Sec-
tions 2 and 3 must be possible. The first condition is easily
Remark: ⫽0 corresponds to the conventional approach that we checked. By combining Eq. 共2a兲 with C⫽C weir,free and Eq. 共3兲 we
used for the weir 共force based on downstream water level兲 obtain a cubic polynomial 关Eq. 共7兲兴
冉 冊
B2
B 1d 1
2
h 42 ⫺4C 2
B2 3
B 3d 3 冉
h 2 ⫹ 4C 2
B 2h 1
B 3d 3
⫹4C 2 ⫺
B 2h 3
B 1d 1 冉 冊 冊 2
⫺1 h 22
冉 B 2w
冊 h 1 ⫺h 2
冋 冉 冊册
h 23 ⫺h 22 ⫹2a 共 h 3 ⫺h 2 兲 ⫽4C 2 w 1⫺ 2
B 3d 3 B 2w
1⫺
B 1d 1
(9)
Assuming drowned flow, i.e., C⫽C gate,drowned , we check whether
Eq. 共9兲 has a real root greater than w. If such a root exists then we
are in State B2 or C, if it does not we are in State B1.
We now know the flow state. For a given flow state we calcu-
late the discharge as follows:
• Flow state A1: use Eq. 共2a兲 with C⫽C weir,free and h 2 given by
Eq. 共3b兲,
• Flow state A2: use Eq. 共2a兲 with C⫽C weir,drowned and h 2 equal
to the root of Eq. 共8兲 that lies between h c , and h 3 ,
• Flow state B1: use Eq. 共2a兲 with C⫽C gate,free and h 2 ⫽w, and
• Flow state B2 or C: use Eq. 共2a兲 with C⫽C gate,drowned and h 2
given by the root of Eq. 共9兲 that lies between w and h 3 . The
procedure is summarized in the algorithm shown in Fig. 5.
Laboratory Setup
Laboratory Results Compared with Model Predictions C drowned weir⫽0.80, C free gate⫽0.882, C drowned gate⫽0.85. Measure-
The discharge predicted by our algorithm was compared with 351 ments where the relative measurement error in the head loss ex-
measurements. Five different discharges 共19, 27, 35, 43, and 51 ceeded 20% were excluded. The reason for this is the following:
L/s兲 were used. The following series of steady flow measurements If the measurement accuracy did not allow for accurate measure-
were performed, in Cases 1, 2, 4, and 5 the downstream water ment of the head loss during the experiment, then comparison of
level is the parameter that is varied, for cases 3 and 6 the gate results between experiment and a formula indirectly based on that
position was adjusted. head loss is unlikely to be of use.
1. A series where the first measurement corresponded to free We classified the error in the predicted discharge as follows. If
weir flow and the last to drowned weir flow; the error was less than 5% or less than two times the relative error
2. A series where the first measurement corresponded to free in the head loss 关 ⫽0.001/(d 1 ⫺d 3 ) 兴 , where we assume an abso-
gate flow and the last to drowned gate flow; lute error of 1 mm, then the error was considered acceptable, else
3. A series where the first measurement corresponded to free it was considered ‘‘large.’’ If the error was greater than 20% and
weir flow and the last to free gate flow; more than two times the relative error in the head loss it was
4. A series where the first measurement corresponded to considered ‘‘very large.’’ We found 12.8% ‘‘large’’ errors and
drowned weir flow and the last to drowned gate flow; 0.7% ‘‘very large’’ errors in 282 measurements. ‘‘Large’’ errors
5. A series where the first measurement corresponded to free occur for the free to drowned gate transition and for the sub-
weir flow and the last to drowned gate flow; and merged weir to free orifice transition.
6. A series where the first measurement corresponded to The excluded measurements—with the exception of the peak
drowned weir flow and the last to free gate flow. for free weir flow to drowned gate flow—correspond to drowned
As discussed earlier, the values for the reduction coefficient C weir flows. Tables 1 and 2 contain a comparison of predicted and
were obtained from the 35 L/s experiment, C free weir⫽0.93, actual flows for 27 L/s.
Note the peak in the error for free weir flow to drowned gate Gruyter gate in terms of upstream and downstream water levels.
flow. This is the result of the same kind of problem we have with We found that free weir flow, free gate flow, and drowned gate
drowned weir flow, the head loss becomes very small. This is flow are dealt with correctly. The proposed model consists of a
probably due to the local minimum in head loss found for the free closed set of formulas and is suitable for use in computer simu-
to drowned gate transition 共Fig. 7兲. This in turn is probably re- lations. Further research into the approximation of the force on
lated to the reduction of head loss for increasing downstream the back of the weir is likely to improve the predictions of the
water level seen for weir flow 共Fig. 3兲. transition points.
From both experiments and theory it is clear that for a fully
drowned weir the head loss over the structure is too small to be
Conclusions used as a means of discriminating between different discharges.
For flow conditions where sufficient head loss over the structure In effect, the structure no longer functions as a discharge regula-
is present it is possible to model the discharge of the Crump-de tor or measurement device 共cf. Bos 1985, p. 66兲. Even for a
Parshall flume the crucial depth over the weir, normally used in
combination with upstream and downstream depth to determine
the discharge is simply not available if we see the structure as a
unit connecting two sections of one-dimensional channel flow
simulation. We recommend switching to a formula simulating a
head loss based on the contraction and expansion at the structure
instead of a gate or weir as soon as fully drowned weir flow
occurs. The choice of switching criterion would depend on the
purpose of the model and the relative importance of accurate
simulation in the case of fully drowned weir flow.
Notation
The following symbols are used in this paper:
a ⫽ height of crest of weir relative to channel
Fig. 7. Head loss versus downstream water level for gate flow
bottom 共0.10 m兲;