You are on page 1of 9

PHYSICALLY BASED HYDRAULIC JUMP MODEL FOR

DEPTH-AVERAGED COMPUTATIONS

By Abdul A. Khan] and Peter M. Stet1ler,z Associate Member, ASCE

ABSTRACT: Consideration of momentum conservation within a hydraulic jump leads to the conclusion that
both the momentum correction due to the nonuniform mean velocity profile and the depth-averaged turbulent
normal stress are important mechanisms. A model is constructed where the turbulent stresses are approximated
with a simplified algebraic stress model. These stresses are shown to depend primarily on the vertical gradient
of the longitudinal velocity. An estimate for the jump velocity distribution is then obtained from a moment of
momentum equation. A single new term in the St. Venant momentum equation, combining the turbulent stress
and velocity distribution effects, in terms of the depth and depth-averaged velocity is proposed. The new jump
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

momentum flux term is nonlinear and diffusive in character. With an appropriate calibration of a single coeffi-
cient, the model gives good results for the location, length, and profile of hydraulic jumps ranging in Froude
numbers from 2 to 7. The numerical results are obtained from a finite-element model with and without numerical
dissipation.

INTRODUCTION high upstream Froude numbers) are well known to predict


jumps of infinitesimal length, justifying the shock-capturing
The hydraulic jump is a transitional state between an up- approach. The main limitation of these equations appears to
stream supercritical and downstream subcritical flow and, for be the assumption of a uniform flow-type velocity distribution
a given set of flow conditions, has a fixed position and length. which cannot generate the turbulent stresses and total momen-
For design purposes, an open-channel flow model should ide- tum flux necessary to spread the jump over its observed length.
ally be capable of accurately predicting both the location and Any improvement in jump modeling, therefore, must be based
the length of a hydraulic jump as a part of general flow con- on including more physics in the form of velocity, turbulence,
ditions. Most St. Venant equations models which presently al- and possibly pressure distribution information.
low jumps, however, treat the jumps as thin shocks, using a Narayanan (1975) and McCorquodale and Khalifa (1983)
dissipative, shock-capturing scheme adapted from the gas dy- used the two-dimensional Reynolds equations with an integral
namics methods. The main emphasis is on suppressing the approach for solving hydraulic jumps. An empirical relation-
oscillations before and after the jump and maintaining the mo- ship, based on the experimental studies of Rouse et al. (1958)
mentum balance across the jump (Abbott et al. 1969). The and Resch and Leutheusser (1972), was used to model the
jump location is well predicted but the jump length is neces- turbulent stresses. Madsen and Svendsen (1983) also used an
sarily dependent on the spatial discretization, typically spread integral method with a one-equation turbulence model and
over one to three intervals. Methods developed include finite were able to predict some of the turbulence characteristics
difference with an explicit artificial viscosity [e.g. Gharangik within the jump. Although these studies were able to predict
and Chaudhry (1991), Younus and Chaudhry (1994)], or with some of the detailed characteristics of hydraulic jumps, they
flux limiters (lha et al. 1994; Garcia-Navarro et al. 1994) and are too complex to be incorporated into general open-channel
Petrov-Galerkin finite-element schemes (Katopodes 1984; flow models.
Hicks and Steffler 1992). This study begins with an analysis of momentum conser-
Gharangik and Chaudhry (1991) attempted to model the hy- vation within a jump, which identifies that the important mo-
draulic jump using both the St. Venant and Boussinesq equa- mentum flux mechanisms governing the jump profile are the
tions. Second- and fourth-order finite-difference shock-captur- velocity distribution and the difference in longitudinal over
ing techniques with added artificial diffusion were used to vertical turbulent stresses. A simplified algebraic stress tur-
numerically model these equations. The Boussinesq equations bulence model, assuming local equilibrium of production and
did not give an improved solution while the fourth-order dissipation, is used to relate the turbulent stresses to the ve-
scheme was found to be superior to the second-order method. locity distribution. The velocity distribution is then evaluated
The jump length was matched to experimental results but was using a moment of longitudinal momentum equation, coupled
only two or three discretization intervals long. Younus and with a simple linear velocity distribution (Steffler and lin
Chaudhry (1994) incorporated a depth-averaged k-e turbulence 1993). A single new momentum flux term is thereby devel-
model but found that the jump length was primarily governed oped, which includes one empirical parameter, related mainly
by the numerical diffusion coefficients chosen. It appears that, to the turbulent length scale. Comparison of numerical solu-
regardless of the equations and overall numerical scheme, the tions with experimental jump profiles shows that a unique
jump length is still governed by the artificial diffusion and value of this parameter gives acceptable results for jumps
spatial discretization. ranging in Froude number from 2.3 to 7.0. A Petrov-Galerkin
The St. Venant and Boussinesq equations (for sufficiently finite-element scheme, with and without numerical dissipation,
is used to solve these equations.
IGrad. Student, Dept. of Civ. Engrg., Univ. of Alberta, Edmonton,
Alberta, Canada, T6G 2G7. MOMENTUM CONSERVATION WITHIN A HYDRAULIC
2Prof., Dept. of Civ. Engrg., Univ. of Alberta, Edmonton, Alberta, Can- JUMP
ada, T6G 2G7.
Note. Discussion open until March I, 1997. To extend the closing date An analysis of momentum conservation within a hydraulic
one month, a written request must be filed with the ASCE Manager of jump is undertaken to identify the important physical mecha-
Journals. The manuscript for this paper was submitted for review and nisms which determine the overall characteristic of jump
possible publication on February 3, 1995. This paper is part of the Jour-
lUll of Hydraulic Engineering, Vol. 122, No. 10, October, 1996. ©ASCE, length and profile. Consider a simple, classical, hydraulic jump
ISSN 0733-9429/96/0010-0540-0548/$4.00 + $.50 per page. Paper No. in a wide rectangular, horizontal channel as shown in Fig. 1.
10101. The longitudinal coordinate is x and the vertical coordinate is
540/ JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996

J. Hydraul. Eng., 1996, 122(10): 540-548


Eq. (5) can be used to eliminate the sequent depth ratios so
that the right-hand side depends on F, only. As a rough esti-
mate for jumps in the 6-8 Froude number range, take MIfF
... I, P,fF ... 0, MzfF ... 0.1, and PzfF ... 0.9.
The next step is to identify and estimate the magnitude of
h,
the terms that contribute to the total specific force within the
jump. Consider the point along the jump where the depth is
half of the final depth, h z. This point will be close to the
FIG. 1. Definition Sketch for a Hydraulic Jump middle of the jump and likely in the vicinity of maximum
nonuniformity and turbulence. The mean momentum flux and
pressure force contributions can be estimated based on mea-
z. At any point along the jump, the mean (time-averaged) flow sured profiles. The remainder gives an estimate of the turbu-
depth is indicated by hand U is the depth-averaged mean lent stress contribution. This estimate will give some idea of
longitudinal velocity. The time-averaged velocity components the turbulent intensity within the jump. At this point
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

at any point are indicated by U and w, respectively. Section 1


is upstream of the jump and section 2 is downstream. h2
The depth-averaged momentum equation can be derived by h-
- -2'' (10, 11)
integrating the Reynolds equation in the longitudinal direction
over the depth of flow. The Liebnitz rule and the kinematic The mean momentum flux contribution can be evaluated
surface condition are used to eliminate terms evaluated at the approximately by means of the momentum flux correction fac-
surface. The resulting equation can be written as follows: tor, 13, as in

a
-
ax lh
0
pu dz
2 a
+-
ax
lh
0
P dz - -
a
ax
lh0
ax dz = -Tb (I) M = lh pu 2 dz = I3pV 2h (12)

where p = water density; p = pressure; ax = Reynolds normal Hager (1992) suggested a function that can be used as a rea-
stress; and Tb = bed shear stress. Denoting the momentum flux sonable approximation for the velocity profile for jumps rang-
integral as M, the pressure integral as P, and the stress integral ing in upstream Froude number from 4 to 9. This can be ex-
as S, (1) can be integrated from section I to any point x along
pressed as
the jump, resulting in
2
U - Us 5'lT Z

1.,r
---= cos ( - -) ] (13)
(M + P - S) = (MI + PI - SI) - Tb dx (2) Ub - Us [
9 h
The subscripts indicate bed and surface velocities. Since an
For the end of the jump, at section 2, (2) becomes the usual estimate of the depth-averaged velocity for the distribution
jump momentum equation from (11) is already available, only one of Ub or Us needs to
be specified. Hager (1992) showed that the maximum back-
ward surface velocity occurs near the middle of the jump and
(3)
is approximately equal in magnitude to the downstream uni-
form velocity. Thus
Assuming uniform velocity and hydrostatic pressure distribu-
tions and neglecting the contribution of the turbulent normal h,V, I
Us = - V 2 = - - - = - - V, (14)
stresses and the total bed shear force results in the elementary h2 /

jump equation
Using (11) with (13) and (14) gives an estimate for the max-
h2 h22 imum forward velocity, for Froude numbers from 4 to 9, at
pV~h, + "'2 1= pVih 2 + "'2 = F (4) the jump midpoint as

where F = total specific force. An approximate «1 % error for 5.65


Ub=j V, (15)
F, > 2) solution to (4) for the sequent depth ratio, f, is (Hager
1992)
This result corresponds very well with the measured maximum
h • ~ I velocities as shown in Fig. 2.11 of Hager (1992). From this
/=-2 = v2F, - - (5) distribution 13 can be evaluated and is found to be very close
h, 2
to 2.5. Now the total momentum flux of the mean flow can
where F = UlVih = Froude number. For F, = 7.5, / is very be estimated as
close to 10, which is a useful scale in the following analysis.
Manipulation of (4) gives the following relative contribu-
M=l3pVh=2.5p
2 (2)2
fVI
h 5 2 5
2=fPv,h'=fM
2
1 (16)
tions of the individual momentum balance terms:
MI / Thus for a Froude number of 7.5, the total momentum flux of
-=1 (6)
F /3 + 2F~ the mean flow represents about 50% of the total specific force.
The contribution due to nonuniform mean velocity distribution
P, I = / is about 30%. The relative contribution decreases as the
- = 1 _M (7)
/ F /3 + 2F~ Froude number increases. This estimate makes sense only for
Froude numbers greater than about 5. Likely, for smaller
M2 /2
-= 1 (8) Froude numbers, 13 should be smaller than 2.5.
F 1 + 2F~ The pressure force can also be estimated based on some
/2 presumptions about the pressure distribution within the jump.
P2 M2
-= 1 (9) The pressure distribution is governed by the vertical momen-
/ F + 2F~ tum equation, which can be expressed as
JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996/541

J. Hydraul. Eng., 1996, 122(10): 540-548


U -
ow
ox
+ w-
OW
oz
=--p1 -op
oz
OTxz
ox
ocr,
+- +- -
oz
g (17) ( -;'2
U -w (3 I5) I2
~) = - - - -U 2
4 I
(25)

where T x , = Reynolds shear stress; and g = acceleration due to The turbulence must be highly anisotropic for this difference
gravity. The first two terms of (17) represent the vertical ac- to exist. If it is assumed that the vertical intensity is about half
celeration. Since w is an order of magnitude smaller than u, of the long~tudinal intensity (typical for turbulent shear flows),
the first of these terms should be significantly larger than the then an estImate for the typical longitudinal fluctuation can be
second. The first term will be positive near the beginning of obtained in terms of the depth ratio and inflow velocity
the jump as the flow begins to expand. Past the middle of the

U'=~=~(~-7)ju,
jump, this term will become negative as the flow straightens
out into the subcritical parallel flow. It can be expected, there- (26)
fore, that in the vicinity of the middle of the jump, this term
is at a transition and should be close to zero. In terms of
curvature, the flow in the upstream half of the jump is concave For a Froude number of 7.5, the turbulent velocity is about
upwards while in the downstream half the flow is concave 30% of the inflow velocity. In terms of local velocity scales
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

downwards. Near the middle, at the maximum upward angle at the jump midpoint, the depth-averaged turbulent velocity is
of flow, an inflection point exists where the curvature, and approximately 1.5 times the depth-averaged velocity or half of
therefore the acceleration, is zero. the maximum velocity near the bed. These values are some-
Further, assume that the longitudinal variation of shear what larger than those reported in experimental studies. Rouse
stress and the density variation due to air entrainment can be et al. (1958) indicated a value of about 15% of inflow velocity
neglected. Both effects would tend to lead to reduced pres- for a Froude number of 6.0 while the foregoing estimate is
sures, but the bulking effect of air entrainment would result in 25%. For a Froude number of 6.0, Resch and Leutheusser
a slightly higher integrated pressure force. However, Mc- (1972) also showed about 30% for fully developed inflow but
C.orquodale and Khalifa (1983) found that except for the much less (about 10%) for undeveloped inflow. The former
hIgher stage at the end of the jump, the air entrainment plays experiments were performed in a closed conduit, which pre-
a minor role in determining the shape of the hydraulic jump. vented the large-scale surface oscillations that are observed in
As turbulence in the jump is strong, the normal stress is re- free jumps. The latter experiments were performed with a hot
tained. The approximate vertical momentum equation then be- film probe which may not have been accurate in flow regions
comes with a high ratio of turbulence to mean flow velocities, par-
ticularly where the flow direction was constantly changing.
o
- (p - cr,) = -pg (18)
The foregoing estimates, while approximate, indicate that there
oz remains a need for definitive turbulence measurements in a
hydraulic jump.
Integration as usual leads to the pressure distribution The main conclusion of this analysis is that the momentum
transfer effects of the nonuniform mean velocity distribution
p - cr, = -pg(h - z) (19) and the turbulent normal stress difference are both important
The pressure force contribution is therefore components of the momentum balance within a jump. For a
Froude number of 7.5, these two mechanisms account for
!!.- + L
h h
about 55% of the total momentum flux. The key to a distrib-
P =
L o
P dz = pg 2
2 0
cr, dz (20)
uted jump model, therefore, is to express these effects in terms
of the usual mean flow properties and their gradients. An at-
which indicates that the pressure is effectively hydrostatic, but tempt in this direction is introduced in the following section.
reduced by the vertical turbulence. In terms of downstream
pressure force, the pressure force is
DEPTH-AVERAGED JUMP MODEL
h
P2
P ="4 + L0 cr, dz (21) The transient St. Venant equations for plane flow in a wide
rectangular channel, neglecting lateral flows, can be written as
Combining the momentum and pressure contributions re-
sults in ah oq
-+-=0 (27)

= F =f
5 P2 (h r at ax
M +P- S M, + 4 - Jo crx dz + Jo cr, dz (22)
-oq + -0
ot ax
(l)- + -
h
a (gh
-
ax 2
2
)
+ -aJ = gh(S -
ax [)
S)
f
(28)
from which the difference in magnitude of turbulent stresses
can be deduced as follows: where t = time; q(=hU) = discharge per unit width; and So and
. .:.h(.:. ;cr.::.,'_---=-cr~x) = 1 _ ~ M , _ .!. P 2
Sf = usual bed and friction slopes, respectively. The transient
(23) equations are used for generality and to facilitate an orderly
F IF 4F development of steady-state solutions. J represents the mo-
where the angle brackets indicate a depth average. For a mentum fluxes due to the nonuniform velocity distribution and
Froude number of 7.5, the stress difference appears to account the turbulent normal stresses. For generality, J should auto-
for about 25% of the total specific force. This magnitude may matically become active in the presence of a jump and be
be more easily appreciated by using M I == P2 == F and sub- negligible elsewhere. Essentially, the purpose of this paper is
stituting the velocity fluctuation correlations for the Reynolds to develop and test an expression for J. The present develop-
stresses. ment is for one-dimensional flow, but the intent is to eventu-
ally apply the method for general two-dimensional depth-av-
-;'2
h( pu ~)
- pw (3 5)
= :4 - f phI U 2I (24)
eraged problems.
From the proceeding section J can be written as

which can be written as (29)


542/ JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996

J. Hydraul. Eng., 1996, 122(10): 540-548


Velocity Distribution term will be neglected, but an allowance will be made by
allowing adjustment of a model constant.
To proceed, the velocity and turbulence distributions must The turbulent stresses can then be evaluated as
be specified. For simplicity, a crude linear velocity profile
2
(2 *- 1)
"""'
u i2 =
_ - 4 A k"'(l - A)
2 k(l - A ) + - (au)2
U ~V+ UI (30) 3 3 £2
-
az
(39)

-
is assumed. The parameter Ul is the velocity at the surface in W,2 ~ -2 k( I - A) (40)
excess of the depth-averaged velocity. The corresponding mo- 3
mentum correction coefficient is then and

(31) -,-, 2 eA(l - A) au


U w ~ -- - (41)
3 £ az
The second part of the jump flux expression then becomes The assumption of production equal to dissipation, as well as
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

using only the largest of the production terms, allows us to


(13 - 1)hV 2 =.!3 hui (32) write
__ au e 12
Turbulence Model -u'w'-=c
iJz
-L J)
(42)

Leaving the estimation of UI for the moment, consider the where L = a turbulent length scale; and Cn = a model constant.
turbulent stress component of J. Since the anisotropy of the Substitution of (41) allows evaluation of k and E, which may
stresses is important, an algebraic stress closure, albeit highly then be reintroduced into (39)-( 41). The final result can be
simplified, will be used. From the ASCE Task Committee on expressed as
Turbulence Models (1988), assuming that the production of
turbulence is approximately equal to the dissipation, the tur- -U,2 ~ ae (-au 2. -W,2 == be 2 (-au )2
(43,44)
bulent stresses can be written as az ) '
m m az
-u; u; = [23
k 8 jj + A (7J>ij
E - 328ij)] (33) and

(45)
where i and j = tensor indices for the three coordinate direc-
tions; k = turbulent kinetic energy; 8jj = Kronecker delta (=1
for i = j, = 0 for i ~ j); A = modeling constant; and E = rate where
of dissipation of turbulent energy. The components of produc- 312

tion of turbulent energy, 7J>jj' are given by em = -32 A(l


2

[
- A)
]
L
2"
CD
(46)

fTh aUj -,-, aUj


- ,- ,
OTij = -UjUj -aXj - UjUj -
ax/
(34) is the effective mixing length. The constants a and b are, in
principle, functions of A only. Using A ~ 0.25, for example,
From the measurements of Rouse et al. (1958), turbulence gives a ~ 2.8 and b ~ 1.4. In view of the gross simplifications
production is greater than dissipation in the upstream part of made, however, these constants will be allowed to be adjust-
the jump while dissipation is greater than production in the able. In particular, it is expected that a should be somewhat
downstream part. Production equal to dissipation appears to larger to account for the neglected longitudinal gradient pro-
be a reasonable approximation near the middle of the jump duction term. The key point of the preceding analysis is the
and leads to considerable simplification. Further simplification final form of (43)-(45), where the turbulent stresses are sep-
can be obtained by consideration of the order of magnitude of arately related to the principal velocity gradient.
the velocity gradients in (34). If the lateral mean velocity and An estimate for the average velocity gradient can be ob-
lateral gradients are zero, estimates for the remaining gradients tained directly from the assumed distribution, (30), as
near the middle of jump are as follows:
au Ul
-=2- (47)
(35a,b)
az h
In addition, it will be assumed that the average mixing length
aw _ !!... _ .! VI. aw _ 0
(35c,d)
is a proportion, ex, of the flow depth
az 5h f h 2' ax (48)
Clearly the vertical gradient of the longitudinal velocity is by Thus, simple expressions for the depth-averaged turbulent
far the largest for any Froude number. For turbulent shear stresses can be obtained
flows estimating that U,2 - 2W,2 - 3u'w' (Resch and
Leutheusser 1972), (34) can be simplified to (49,50)

7J> ~ -2 (U'w' au + U,2 au). 7J> ~ -W,2 au (36,37)


and
xx az ax' Xl az (u'w') == 4ex21ullul (51)
and Substituting (32), (49), and (50) into (29) gives a relation for
(38) the jump momentum flux in terms of the excess surface ve-
locity
The second term of (36) should be about half the size of
the first term, as the size of the stress partially compensates
for the smaller gradient. In the following development this
J= [~+ 4(a - b)a
2
] hui (52)

JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996/543

J. Hydraul. Eng., 1996, 122(10): 540-548


0.30 r----------------------.
=
Manning's n 0.0064

0.20
~
'V 'V

=
g
0.10
........• Jump Flux (Eq. 57)
- - Jump Flux (Eq. 58)
'V Measured
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

0.00 .....- - -......- _ _..&. ........ .L-. ...

o 2 3 4 5

Distance along channel (m)


FIG. 2. Surface Profile for F = 2.30

0.30
Manning's n = 0.0074

0.20

0.10
........• Jump Flux (Eq. 57)
- - Jump Flux (Eq. 58)
'V Measured
0.00 ..... ...... ..&. ........ .L-._ _- - I

o 1 2 3 4 5

Distance along channel (m)


FIG. 3. Surface Profile for F= 4.23

To complete the development, a model for the excess surface Two further roots are also possible; one positive and one
velocity is required. negative. There may be some physical justification for a pos-
itive root, if it is interpreted as indicating a separated jump or
Surface Velocity Model a jump with an undeveloped roller (Hager 1992). Such jumps
A general equation for Uh derived from the moment of mo- seem to occur with fully developed upstream flow conditions.
mentum principle, is provided by Steffler and Jin (1993). As- Again, this is consistent with (53), since the incoming flow
suming that the vertical turbulence is the only significant effect would enter the jump with some positive u" which would tend
on the pressure distribution and neglecting the effect of bed to grow. While qualitatively plausible, quantitative prediction
stress, the equation reads would be unreliable because the flow field is more complex
than that of the classical jump and many of the assumptions
made would no longer be valid.
- - = -3 (-;'2)
aUt+ -auu, w - 3 (12)
ah - - u - 6 (-'-')
ah+ - uw (53)
at ax h ax h ax h The negative root of (54) is the one corresponding to the
classical jump situation. This solution is
Using the turbulent stress relations (49)-(51), assuming steady-
state conditions and that u, reaches a maximum near the mid-
point of the jump, gives
h au
t (55)
U = [ 12 ] ax
u 1 -au
2
+ -12 (a
=-ah [ - 241 u'15 - b)u ] u (54)
a
2
24 - 5 (a - b)
ax l'

Three solutions are possible for u,. The first is simply Ut = O. The velocity predicted by (55) is a better indication of the
Consideration of (53) shows that this solution is stable for (aU/ overall velocity distribution than it is of the actual surface
ax) > 0, but unstable for (aU/ax) < 0, which is the jump con- velocity. In particular, the surface velocity estimate will tend
dition. This equation appears to capture, in general form at to be higher than the actual due to the difference between the
least, the significant difference between accelerating and de- assumed and actual velocity distribution shapes. Substituting
celerating flows. (55) into (52) yields
544/ JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996

J. Hydraul. Eng., 1996, 122(10): 540-548


0.40
Manning's n = 0.0072

0.30

g
oS 0.20
g
0.10 ......... Jump Flux (Eq. 57)
- - Jump Flux (Eq. 58)
'\1 Measured
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

0.00
o 2 3 4 5

Distance along channel (m)


FIG. 4. Surface Profile for F= 5.74

0.30 _ - - - - - - - - - - - - - - - - - - - - - - ,
Manning's n =0.0076

0.20

0.10
......... Jump Flux (Eq. 57)
- - Jump Flux (Eq. 58)
'\1 Measured
0.00 L... ...... ....... .....I oL-_ _----I

o 2 3 4 5

Distance along channel (m)


FIG. 5. Surface Profile for F= 7.0

fication is physically reasonable since it reduces the jump flux


at the toe and end of the jump. The modified model is

J = K2 (:~r h (~~r
3
(58)

As also shown in the following section, K 2 = 441 gives good


The turbulence constants can be collapsed into a single coef- results for both length and profile shapes for all jumps con-
ficient, K h which will be determined from numerical experi- sidered. The values of the constants appear to be physically
ments. The jump flux model is then simply reasonable; for example, a = 3, b = 1.5, and 0: = 0.12 give the
foregoing K!. These constants give a reverse surface velocity
J= K)h
3
(~~r (57)
about 30% larger than measured, but that is expected due to
the crude linear profile assumed. The value of K 1 also approx-
imates the average of 441 [(ahlax)f over the length of the
The diffusive character of (57) is apparent when substituted jump. Either of the jump flux formulations is highly nonlinear
into the St. Venant momentum equation, (28). A second de- and very sensitive to the longitudinal gradients. Included in
rivative of U term arises, which has a negative sign set by the the general equation, they will have negligible effect except in
longitudinal gradient of U. the vicinity of rapidly varying flows. Care should be taken,
Eq. (57) is developed to apply near the middle of the jump however, in regions of rapid acceleration, to set the jump flux
where J is maximum. As shown in the following section, if to zero. In addition, the sign of J should change with the flow
(57) is used over the entire jump, a value of K 1 = 17.4 gives direction.
good jump lengths but tends to predict an overly smooth,
rounded-off jump profile. Slightly better results are obtained NUMERICAL MODEL
by a modification which allows K 1 to vary with the square of To test the proposed jump flux model, a Petrov-Galerkin
the water surface slope. While somewhat ad hoc, this modi- finite-element numerical formulation is used. Although any
JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996/545

J. Hydraul. Eng., 1996, 122(10): 540-548


0.30
Modified Jump Flux Scheme ~ ~ ~ ~
Manning's n =0.0076
Froude Number =7.0
LU is spatial discretization
0.20
.-.
5
~
g --AX =0.305 m
0.10
········-AX =0.152 m
- - - Ax =0.061 m
- • - - •Ax =0.030 m
~ Measured
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

0.00
0 2 3 4 5

Distance along channel (m)


FIG. 6. Variation of Spatial Discretization (COG Scheme)

0.30
Modified Jump Flux Scheme
'Sl
Manning's n =0.0076 lo" •
Froude Number =7.0 : ,"
ll·
Ax is spatial discretization 1/'\10'
/,/ ,
0.20 : ,
1,: I ,
e
'-" ,
J..': "
~ '
,/
B' ,/q
: ,
Q 1,: , I
--Ax =0.305 m
0.10 ,i
: ,
I
·······--AX =0.152 m
/.: I
- - - Ax =0.061 m
t/ \(
: , -----AX =0.030 m
_2.' ~ Measured
0.00
0 2 3 4 5

Distance along channel (m)


FIG. 7. Variation of Spatial Discretization (B-G Scheme)

conservative numerical scheme may be used to solve the equa- experiments, also reported in Gharangik and Chaudhry (1991),
tions, the code developed by Khan and Steffler (1996) is used. were performed in a rectangular, horizontal, metal flume 14-
The code is based on the characteristic dissipative Galerkin m long, 0.915-m high, and 0.46-m wide. Jump profiles for
one-dimensional (CDG-ID) finite-element scheme; Hicks and inflow Froude numbers ranging from 2.3 to 7.0 were measured
Steffler (1992) provide details of the scheme. The method re- using a point gauge with an accuracy of 0.3 mm. In modeling
duces to a standard Bubnov-Galerkin without any numerical these experiments, the element length is set to 0.061 m (0.2
dissipation when the upwinding parameter is set to zero. ft), which is one to two times the supercritical flow depth.
A fully implicit time-stepping scheme is used to rapidly Manning's n varies from 0.0063 to 0.0077, essentially to prop-
solve for the final steady-state jump profiles from an arbitrary erly locate the jump for profile comparison. For all the tests,
initial condition. The implicit set of nonlinear algebraic equa- calibrated values of 17.4 and 441 for K 1 and K 2 , respectively,
tions at each time step is solved by a Newton-Raphson itera- are used to evaluate J.
tive procedure with analytical evaluation of the Jacobian de- Figs. 2-5 show the measured and calculated jump surface
rivatives. Linear interpolation functions for depth and velocity profiles for inflow Froude numbers of 2.3,4.23,5.74, and 7.0.
are used. In running the model, it was found that some care Results obtained using both jump flux formulations, from (57)
was necessary while the jump was forming and moving. The and (58), are shown in Figs. 2-5. In general, both methods
Courant number during this stage is limited to no greater than give good estimates of the jump length and overall profile.
about five. Once the jump is properly located, however, con- Underprediction of the downstream depth can be explained by
vergence is very rapid and very large time steps can be used. bulking due to air entrainment in the experiments, which was
For boundary conditions, both depth and velocity are specified not accounted for in the model. In all cases the depths matched
at the upstream (supercritical inflow) section and depth is spec- well a short distance further downstream. The values of the
ified at the downstream (subcritical outflow) section. constants in the computational models give very close results
for the intermediate Froude numbers. The simple model, how-
EXPERIMENTAL VERIFICATION ever, gives a slightly shorter jump at the high Froude number
The experimental data from Gharangik (1988) is selected and a slightly longer jump at the low Froude number. The
for calibration and verification of the jump flux model. These modified model gives uniformly close jump lengths for all
546/ JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996

J. Hydraul. Eng., 1996, 122(10): 540-548


0.30

- - dx = 0.305 m
·········dx = 0.152 m
=
- - • dx 0.061 m
0.20 - - - -' dx = 0.030 m
\l Measured

0.10
Manning's n = 0.0076
Froude Number = 7.0
dx is spatial discretization
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

0.00
o 2 3 4 5

Distance along channel (m)


FIG. 8. Variation of Spatial Discretization (COG Scheme, J = 0)

1.00
Modified Jump Flux Scheme
CDGscheme
--TermsP+M
......... Terms P+M+J
I
1
0.80 ~

0.60 ~

0040 ------ .,. -------.,


0.20 -

0.00 L- .....I .L.... .....IL.- .....


.-----'

o 2 3 4 5

Distance along channel (m)


FIG. 9. Longitudinal Momentum Balance (F= 7.0)

Froude numbers. The modified model also shows a somewhat need for artificial dissipation mechanisms. Depending on the
sharper transition at the toe of the jump. resolution desired by the modeler, the jump model mayor may
Figs. 6 and 7 show the effect of spatial discretization on the not be effective. Combining both a jump model and artificial
computed results for the modified model. For the CDG scheme dissipation would work well as long as it was understood that
(upwinding parameter = 0.25), the overall profile is not sig- accurate jump lengths would only be obtained for sufficiently
nificantly affected until the element size exceeds the down- fine discretizations.
stream depth. The location of the jump toe shifts to the nearest Fig. 8 shows the results of varying discretization for the St.
node and the profile becomes very sharply varied. Further in- Venant equations (J = 0) using the CDG scheme. The results
creasing the element length past the jump length gives some show that, although the position of the start of the jump is
oscillations with a slight overshoot at the end of the jump. The accurately modeled, the length of the jump is a function of
Bubnov-Galerkin scheme (upwinding parameter = 0) is more spatial discretization.
sensitive to the discretization. The finest discretization (about
A profile of the total specific force along the channel com-
one-tenth of the downstream depth) gives very good results
puted using the modified jump flux model for the 7.0 Froude
with very small oscillations downstream of the jump. As the
discretization becomes coarser, these oscillations become number jump is shown in Fig. 9. The St. Venant line includes
stronger and the jump location shifts upstream. At the coarsest the momentum flux based on the depth-averaged velocity (i.e.,
discretization, oscillations upstream of the jump become evi- 13 = 1) and the pressure force. The jump flux is therefore the
dent. The solution within the jump remains very consistent, difference between the total and the St. Venant lines. The slight
albeit shifted in position. This is consistent with the modified oscillation in the total line is attributable to numerical error in
jump flux model which dramatically reduces the flux at the evaluating the jump flux from the nodal values of the varia-
beginning and end of the jump. The unmodified model (not bles. Interelement fluxes are automatically conserved by the
shown) displayed significantly less difference in profile be- finite-element formulation used. The magnitude and distribu-
tween the dissipative and nondissipative schemes. tion of the jump flux is approximately as indicated by the
The proposed jump model does not, in general, obviate the initial order of magnitude estimate. Away from the jump, the
JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996/547

J. Hydraul. Eng., 1996, 122(10): 540-548


jump flux is entirely negligible, despite being included in the Katopodes, N. D. (1984). "A dissipative Galerkin scheme for open-chan-
calculation. nel flow." J. Hydr. Engrg., ASCE, 118(2), 337 -352.
Khan, A. A., and Steffler, P. M. (1996). "Vertically averaged and moment
equations model for flow over curved beds." J. Hydr. Engrg., ASCE,
CONCLUSIONS 122(1), 3-9.
A new method for simulating hydraulic jumps in depth-av- Leutheusser, H. J., and Kartha, V. C. (1972). "Effects of inflow conditions
on hydraulic jump." J. Hydr. Diy., ASCE, 98(8),1367-1385.
eraged computations was developed and presented in this Madsen, P. A., and Svendsen, I. A. (1983). "Turbulent bores and hy-
study. The model is physically based, using a simplified al- draulic jumps." J. Fluid Mech., 129, 1-25.
gebraic stress turbulence closure and a moment of longitudinal McCorquodale, J. A., and Khalifa, A. (1983). "Internal flow in hydraulic
momentum equation for estimating the vertical distribution of jumps." J. Hydr. Engrg., ASCE, 109(5),684-701.
longitudinal velocity. The model takes the form of a jump Narayanan, R. (1975). "Wall jet analogy to hydraulic jump." J. Hydr.
Diy., ASCE, 101(3), 347 -359.
momentum flux term, which may be added to the usual St.
Resch, F. J., and Leutheusser, H. J. (1972). "Reynolds stress measure-
Venant momentum equation. The model is based on conditions ments in hydraulic jumps." J. Hydr. Res., 10(4), 409-429.
at the middle of the jump but is easily modified to apply over Rouse, H., Siao, T. T., and Nagaratnam, S. (1958). "Turbulence charac-
the full jump length. A single constant, of physically reason- teristics of the hydraulic jump." J. Hydr. Div., ASCE, 84(1), 1528-1-
Downloaded from ascelibrary.org by Dot Lib Information, LLC on 09/29/16. Copyright ASCE. For personal use only; all rights reserved.

able magnitude, is adequate to simulate all measured jumps. 1528-30.


The jump flux term can be used as a shock-capturing mech- Steffler, P. M., and Jin, Y. (1993). "Depth averaged and moment
anism, since it becomes active only in the presence of a jump equations for moderately shallow free surface flow." J. Hydr. Res.,
31(1),5-17.
and is entirely negligible elsewhere. Younus, M., and Chaudhry, M. (1994). "A depth averaged k-E model for
In developing this model, some interesting insights into the the computation of free surface flow." J. Hydr. Res., 32(3),415-444.
behavior of the fluid flow within the jump have been gained.
First, half or more of the total specific force within the jump APPENDIX II. NOTATION
is carried by the difference in depth-averaged turbulent normal
stresses and nonuniform momentum flux. Second, a flow in- The following symbols are used in this paper:
stability, with a classical or separated solution, is indicated by
the velocity distribution analysis. As observed, small differ- A = turbulence model constant;
ences in inflow conditions may change the jump conditions. a = turbulence model constant;
Third, rough estimates of the relative magnitudes of the tur- b = turbulence model constant;
bulent stresses and mixing length have been obtained from the CD = turbulence model constant;
F Froude number;
condition that the jump flux constant give the measured jump F total specific force per unit width;
length. These estimates await detailed turbulence measure- f = sequent depth ratio;
ments for independent verification. g = acceleration due to gravity;
The proposed jump model is incorporated into a finite-ele- h = depth of flow in vertical direction;
ment model for illustration and verification, but may be used I, } = tensor indices;
with any conservative numerical scheme. Since the jump flux J = jump momentum flux;
term becomes active only at jump length scales, it does not K1 = jump flux model constant;
remove the need for dissipative mechanisms in general models K2 = jump flux model constant;
where the spatial discretization used may be significantly k turbulent kinetic energy;
larger. The jump flux model is easily adapted to two-dimen- L = turbulent length scale;
sional models, moving surges, and jumps on slopes. However, em turbulent mixing length;
further research and verification in these areas are warranted. M = momentum flux integral;
P = pressure force per unit width;
ACKNOWLEDGMENT W'jj = turbulence production components;
q discharge per unit width of the channel;
This research was supported by a grant from the National Science and
Engineering Research Council of Canada.
S = normal force per unit width;
Sf = friction slope;
So = bed slope;
APPENDIX I. REFERENCES t time;
Abbott, M. B., Marshall, G., and Rodenhius, G. S. (1969). "Amplitude- U = depth-averaged longitudinal velocity;
dissipative and phase-dissipative scheme for hydraulic jump simula- U' = depth-averaged mean-square longitudinal velocity fluctua-
tion." Proc.. 13th Congr. 1AHR, Int. Assn. for Hydr. Res. (IAHR), tion;
Delft, The Netherlands, Vol. 1,313-329. u = time-averaged longitudinal velocity;
ASCE Task Committee on Turbulence Models in Hydraulic Computa-
tions. (1988). "Turbulence modeling of surface water flow and trans-
u' = longitudinal velocity fluctuation;
UI surface velocity in excess of the depth-averaged velocity;
port: Part 11." J. Hydr. Engrg., ASCE, 114(9), 992-1014.
Garcia-Navarro, P., Priestley, A., and Alcrudo, F. (1994). "An implicit
w = time-averaged vertical velocity;
method for water flow modeling in channels and pipes." J. Hydr. Res., w' = vertical velocity fluctuation;
32(5), 721-742. x = longitudinal coordinate;
Gharangik, A. M. (1988). "Numerical simulation of hydraulic jump," z vertical coordinate;
MSc thesis, Washington State Univ., Pullman, Wash. IX = mixing length proportionality constant;
Gharangik, A. M., and Chaudhry, M. H. (1991). "Numerical simulation 13 momentum flux correction factor;
of hydraulic jump." J. Hydr. Engrg., ASCE, 117(9), 1195-1211. 'Y = specific weight of water;
Hager, W. H. (1992). Energy dissipators and hydraulic jump. Kluwer Bij Kronecker delta;
Academic Publishers, Dordrecht, The Netherlands. e dissipation rate of turbulent kinetic energy;
Hicks, F. E., and Steffler, P. M. (1992). "Characteristic dissipative Gal-
erkin scheme for open-channel flow." J. Hydr. Engrg., ASCE, 118(2),
p density of water;
337-352. crx Reynolds normal stress in longitudinal direction;
Jha, A. K., Akiyama, J., and Ura, M. (1994). "Modeling unsteady open- cr, Reynolds normal stress in vertical direction;
channel flows-modification to beam and warming scheme." J. Hydr. Tb bed shear stress; and
Engrg., ASCE, 120(4),461-476. Txz Reynolds shear stress.

548/ JOURNAL OF HYDRAULIC ENGINEERING / OCTOBER 1996

J. Hydraul. Eng., 1996, 122(10): 540-548

You might also like