Professional Documents
Culture Documents
1. Introduction
When a rectangr la v vessel containing fluid is oscillated at «TnaH amplitudes
about a fixed axis (figure.1V gravity waves appear on the surface of the fluid.
If the width B of the contaraev is large with respect to the depth h0 of the fluid,
the wave height 7} might be uescribed by öie 'linear shallow-water theory'.
In§2itwiUbeshownthataccordi^4fethistiiCOrj the-araplitude of the surface
elevation is proportional to {cos (TIMJZVJA1, where aJ is the angular frequency of
the excitation and , / D w , .A ,..
Hence the linear theory predicts an infinite amplitude at o> = <t>0. The present
paper is concerned with the frequency range where w is near w0.
Experiments carried out with a> a w„ showed the occurrence of a hydraulic r ; 'J,
jump which travelled back and forth between the walls of the container. Ob-
viously the linear theory is invalid in this frequency range and a description must
start from the non-linear shallow-water theory.
The present situation appears to be analogous to that occurring when a
column of gas is oscillated at a resonance frequency, in which case a shock wave
is formed in the gas. This problem has received attention recently in the work of
Betchov (1958), Chu & Ying (1963) and Chester (1964). The most rigorous ac-
count of the travelling shock wave appearing in the gas was given by Chu
& Ying, who used a perturbation method due to Lin (1954). In the present paper
the Chu-Ying—Lin method .is applied to fluid oscillations under resonance
conditions.
47 Fluid Hech. 22
738 J. H. G. Verhagen and L. van Wijngaarden
Following a formulation of the problem in §2, the theory is discussed in §3
and results pertaining to the hydraulic jump are given in §4. Experimental
results are presented in §5 together with the theoretical results.
. S
FIGTJBE 1. Fluid osculating in a container.
2. Theoretical analysis 4
Consider an open rectangular container of width B, filled with fluid to a level ,£
h0 (see figure 1). Let one pair of the side walls be parallel to the {X, r)-plane, 31
where gravity acts in the negative ^-direction. Let the other side walls be located J&
&tx = ± \B. The container is oscillated about the Z-BJOS at small amplitudes S. jijj
We assume that the width of the container in the Z-direction is large enough
for the flow to be two-dimensional. We denote the undisturbed fluid surface by
y = H, the surface elevation with respect to this level by 17, and the angular dis-
placement about O.by SmxoA, a counter-clockwise rotation being considered ||^"f ."K'--fl
positive. Then the bottom is described by S~M * ''-\'_ d v ; §..
y* j/^-ff-Ao + fosinörf, t2)^"^ ^V"^
and the surface of the fluid by y=y = E + t]. ( )Sifc* ' ' , ^ " I :
3
' r . .-H^sk'z •
I t is convenient to consider the surface level relative to the bottom of the co^jJ|p -„*£
caft
tainer. Therefore we introduce *smfr-o-
*"
mii
A = h0 + v — Sxsmojt. (* Yf-Yh) wm^
^rm &£&*,. .,
If hJB -^ 1, the motion of the fluid caused by the oscillation of the container*^ «
can be described by the 'shallow-water theory' (Wehausen & Laitone 1960^^7^
§30, Stoker 1957, ch. 2). In this theory the continuity equation is ' ' H | | R t | § | .^ vH • in the
S&the hy
» :+ "a» ex I ".^.•equatii
, '^sional;
r
' • , '*£• •
'Mi'-
x&. •
«tf-
Non-linear oscillations of fluid in a container 739
'CS-
«rwhere « denotes the velocity in the x-direction, and the momentum equation is
tal • Ür~ -" du du dX _.
-* k..'
(6)
•üf In this formulation the pressure distribution in the vertical direction is assumed
-.P ^ to be hydrostatde. Therefore, the acceleration in the F-direction, introduced by
gf} the excitation, must be small with respect to the acceleration due t o gravity;
i.e. SBaP/g <§ 1. The boundary conditions for u are determined by the velocity
produced in the horizontal direction by the exciting oscillation. In the shallow-
water approximation u does not vary between the bottom and the surface. Taking
the value at the surface, we require that
"~~~ u = — SBtooostut at x = + \B. (7)
We seek a solution of equations (5)-{7) in which u and A vary periodically.
For small enough S, one expects the linearized form of (5) and (6) to be valid, i.e.
BX , du n
2-
&* fa ». , \
¥ + f f - + ^ s m ^ = 0.
The solution of these equations for A, satisfying (7), is
jpiif j , SBo>Q 1+HaPfg . # .
7T0) C08(7T(0/2(i)Q) '
7T6JX
B(t)0'
(8)
du a?3p_ 0
dt+p0dx~~ '
where the undisturbed quantities are indicated with the subscript 0, and a is the
velocity of sound in the gas. If u = 0 at the closed end x = L, and at the piston
u ~ ho cos oit, the solution to the above equations is
Bm{ü)(L~x)la}
U = kü) COS Cut (9)
sin ((oL/a)
Resonance occurs when (oLja is a multiple of ?r. Then a shock wave is generated
in the gas, which travels periodically to and fro through the column, similar to
the hydraulic jump described in this paper. Indeed it is well known that the
equations of shallow-water theory are equivalent to the equations of one-dimen-
sional gas dynamics.
47-2
740 J. E. O. Verhagen and L. van Wijngaarden
The problem of resonance oscillations in a gas column was treated by Betchov
(1958), Chester (1964) and Chu & Ying (1963), starting from the assumption
that the excitation amplitudes were sufficiently small to permit a linearized
solution of type (9) for conditions far from resonance. All these authors succeeded
in obtaining approximate solutions at resonance, including the case of shock
waves travelling periodically up and down through the gas.
Betchov (1958) and Chester (1964) derived such a solution by both physical
and mathematical arguments and also discussed the influence of viscosity.
Chu & Ying used a method of characteristics perturbation developed by Lin
(1954), which, if properly adapted, appears to be.applicable to the hydraulic- EV»
jump problem. We shall give a brief outline of the method, referring for details
to the work of Chu & Ying, henceforth denoted by C. Y.
- :-&
i- ^'Sndjalong 0~ 5 j j « - 2 c - ^ e ^ , e o s ( ö r f - ^ ) | = 0.
(Nofcê/that the Riemann invariants, i.e. the expressions {} in (17) and (IS),
(18)
^ilsume a simple form involving only u and c, owing to the fact that the bottom
<lÜope does not depend on x. A periodic solution of equations (15)-( 18), involving
Tn^draulio jumps and satisfying the boundajy condition (7) has to be found.
0& possible approach to the problem might be to start from the undisturbed il
J f conditions and to construct the development of theflowby the method of charac-
3Jp.terist£cs. This development would lead to the formation of a hydraulic jump.
feUsing the methods given in Courant & Friedrichs (194S) for the gas-dynamic
jf.. case, the procedure might be continued till a quasi-stationary situation is reached,
j | ' in which a jump travels periodically to and fro. Such an approach would be
1
• ^
f. conceivable with the aid of a computer.
f- Another approach is to start from the concept of the quasi-stationary situation
• , <"-y*\^V'. •EJ-
(mentioned above) and to attempt an analytical construction of such a solution.
Consider the (z, *)-plane (figure 3). The paths of the jumps are represented by
h
PQ> QR, RS and so on, and the aforementioned solution has to satisfy equations
(15)-(18) in the different regions I, I I , H I , etc. The solutions for these regions,
which have different energies, must be related to each other by the jump con- If
ditions, which require the conservation of mass and momentum across the
•^"•-•iï; hydraulic jump. (An important difference occurs with shock waves, because
across a shock wave energy is preserved.) From consideration of the periodicity
it follows that the flow in region I must be repeated in region H I , and so on.
.u. The difficulty is that the paths of the jumps are not known at the outset. We
know that for S-+0, these paths cannot be far from dxjdi = + c0. The deviations
of the jumps from these directions are most conveniently expressed as perturba-
tions'in terms of the characteristic co-ordinates (Lin 1954), the appropriate
perturbation parameter e being defined by (12). For, if the difference in level
across the jump is AA, then the rate of loss of energy is given byf
dE (AA)* .
Pit
Ï*-
m
-|B O x \p
FiGTjaE 3. Paths of hydraulic jumps in (x, *)-plane.
t. f^l
*•• da
S - * S + ( « I + « J Ax'
da^ ^ Sij
S.
. a*0
SIB
(25)
,'1
. t
i
I
5 !_.
?.:-i;r.,
Nonlinear oscillations of fluid in a container 743
*••dxda -sr-^si+ci+^s+^+^a.
s 8t
da da
dt
da
*i n
dxt dt, fa
'ay?
A.
dfi (26)
To solve these equations in region I (figure 3), we have first to formulate the con-
ditions at the boundaries of this region in terms of a and /?.
In defining a and y? we follow C.Y. The ^--characteristics (a = const.) are
defined by the value of a: at the intersection with the hydraulic jump PQ. PQ is
thus determined by a = x. The ^-characteristics (fi = const.) are defined in the
following manner: through each point of the wall PR passes a <7+-characteristic.
The value of fi along this characteristic is equal to the value of a along the
(^"-characteristic through the same point of PR. By this definition of the
C+'lines, the wall PR is described by a = fi. J •
Along a = fi we have two conditions
« = - H(o3 cos (art - <f>), (27)
'and x = -iB. (28)
The hydraulic jump PQ tends to fi = — Ji? as e -* 0. Therefore, along PQ,
^ fi = -$B + e/t1(a) + e*M*) + - , (29)
where fi^-^B) = /i 2 (-.S£) = 0, since at P,fi = a = ~\B. Along (29) we have
the condition x = a.
The second condition that is needed is provided by stipulating that along
x==a
c = e0 + ey1(a) + e*ya(<z) + .... (30)
The functions y1 and y 8 are determined later.
Finally, we fix the origin of time t at a = fi = — $B, so that at t = 0 the jump
is at x — — £i?. The boundary conditions are also determined in terms of e.
Mftlring use of Taylor expansions, we obtain
e°: at .fi *= a,
Xo = a at
*o = 0 at y ? - a « - ^ ;
«1 = 0, «! = 0 at ^ = a,
ftr0
e^ft at ^ = - £ 5 ;
at £ = * = - $ £ ;
« 2 = 0, «2 = -7TiShc0B-tcos((o0t0^^) at ft = a,
dx* . *dh;a dxa . 3c,
at fi=-\B
*B = 0 at >? = <* = -$.#.
Following C.Y., the procedure was as follows.
744 J. S. G. Verhagen and L. van Wijngaarden
First, the solution in region I (figure 3) was obtained by solving the differential
equations (24)-(26) with the above conditions. From this solution the flow in
region I along QR, denned similarly to that along PQ (cf. equation (29)) as
a « }2?+ed 1 (i) + ..., was determined. The flow along QR in region I I could be
obtained subsequently from the conditions of conservation of mass and momen-
tum across a jump. If the quantities in front of the jump are denoted by the
subscript ƒ, and behind the jump by b, these conditions are as follows:
(«,-©<*-(«»-©.*
4. Results
Using the method of the preceding sections, we obtained the following results
(reference being made tofigure3):
a1 = éeCoAsinÜuot-t-Wsm^+^ + Oie*), (33)
m7 '*r-U,
?~-
_ # , • "
~--~J?
~*~> • <*. r t"L - ;-
r..-^*-
-r^.
-.=¥*.
•*; J ' .
Non-linear oscillations of fluid in a container 745
!
3. •£.-•
• ^ v
^ : : ; ; : _ rf^+. I + _—H—sm(^
+ _i + j7r)
w(a: + iJï)
+ 2CQ^ cos i(&>0f — $ — iff) cos
25
?] + 0(e«)f (34)
, '~J. Q = (üj-a> 0 )/e, (35)
where
¥)r-
s
r 2 /
and 1 + 77
In region II,
•t( - (36)
-» t n
«n = 46c 0 ^ cos i ( a i , i - ^ -.fr) oos g ( a ! + p * g ) + Ote2),
|\BQ 2c
\BQ. 2coJ04_ . , , . , . ,
2B (37)
c = c n +e
^
+ 2c0 4 sin \{o)Qt — <f> — \n) sin.-^>]+0(e), (38)
The path of the jump PQ, travelling from left to right, appears to be
Kj? ? rfe"
x = -\B + c0t + e — t +—(c0t-B) sin ($<f> + \TT)
- 2 sin {\oiQt -\4>- in) cos &ü0t\ + 0(e2), (39)
while the equation'for the jump QR, travelling from right to left is
z =~ c 0 t - e ^ t + ^(c0t-2B)sm(W + fr)
2 77
2BA 2
+ cos ($<o0t — 1$ — JTT) sin |wsl0$+ Ote ). (40)
77
The phase difference <f> between the jump and the container is determined
—from the condition that a t all times the total water volume per unit length in the
z-direction should be Bh0f i.e.
Xdx = BKQ,
J-is
or, by use of (10) and (11),
•*»
-IB
f.
c*dx = Bc%.
The integral can be evaluated with the help of (34) and (38). To obtain the con-
tribution óf order e, the relation between x and t at the jump is needed only to
zeroth order. If we consider a time f a t which the jump travels from the left to
the right, it follows from (39) to this order that at the jump,
z = -$B+c0f.
Therefore we have to use (34) from x = — $i?tox = —\B + <%?. For the remainder
of the integration interval, i.e. the region in front of the jump, we have to use
(38), with (see figure 3) f + 2Bfc0 substituted'for f. Then we obtain, requiring
the cancellation of terms of order e,
sin(i0 + i7r) = -BQ{6AcQ + 0{e). (41)
746 J. B. O. Verhagen and L. van Wijngaarden
For o) - a>0, (41) with (35) yields <f> = — è77"- ^ ^ means that when the container
starts an oscillation cycle (in the counter-clockwise direction) with a> = w0, the
jump is, on its way from x = $B to x = — $B, just passing the centre of the con-
tainer and is therefore exactly 90° out of phase with the excitation-
As a check on the expressions for the paths of the jump, we calculated the
mean speed _, ,
TTJQ at
from (39). The result obtained is cQ+eQB/n or, by virtue of (35), (oBjn as it
should be.
We now determine the strength of the hydraulic jump, i.e. the difference
[ij] between the surface elevations behind and in front of the jump, divided by
V From (4), (10) and (11) we infer that
c = c0(l + 7}/2h0) + O(e*). (42)
From (42) we obtain, using (34), (35), (38) and (41), for the wave elevations in
region I and region I I
fi . * ^ + ^ e c o s ( K , _ è $ s _ è 3 r ) ^ ^ ^ ^ (43)
3 w„ 2B
^4^^o+^e^(K/_^_^)^^^a+0(e2)_ (44)
3 w„ 2B
Now consider again a time ?. at which the jump travels from the left to the right.
The position of the jump is given by (39), and the strength of the jump follows
from - , . 'r
_^_£jj+Cor+O(e),^H+0(e)j]
= 4Ae cos (£$J + Jff).
Using (41), we obtain
[tfft, = *Ae{l - (BQ/6Ac0}*}*+0{e*).
For a given frequency the strength of the jump is independent of time to
order in e.
5. Measurements
mm v . n#t"'<
The theoretical results given in the preceding section were verified e x p e n g ^ - ;^>--* --*-*
mentally. For this purpose a rectangular container with B = l*20m, was ffllear ;Jr Jvl : '
with tap water to a level h0 = 9 x 10 _a m, and oscillated about an axis paiSulel* -'-J.; m
to the Z-axis as in figure 1 and coinciding with the bottom of the oootgUMpp : <?.*
The experimental value of B was therefore equal to AQ and hence theJs^gond>""
term in (36) could be omitted, being negligible with respect to unity. ; - ^ ^ -'-;'" *•
Three different kinds of measurements were carried out: ^3fc
(i) At an oscillation amplitude <* = ;r/90rad., "the surface elevation^w^
measured at four different values of x, the frequency being o>Q =.2^6.sec~i.
.-. ,r
ȕ
_.;-_. v : m ^ £ S ^ p ë p
=?!•*.
m
O^S-I-0
vn« 0)
: * & •
it
i
» i — fi
W
-40-1 -40-
F I G U B E 4. Surface elevation as a function of wt—$ for various values of
x, a t a) = WQ and A = 2°. , Theory; , experiment.
•
mm mm nun
30 -\ «/«0-0-89 30-i « / « o - l 30-1 « / « o = H 4
*-l
ml ;t ^ TU^lf.^
4 3 S '•-?>
n
:
'i
•<•;•
--e
0-10 -
>l
FIGURE: 6. Amplitude of the moment exerted b y the fluid on the container as a function of
G>/&\, fox various values of S. , Theory (for low values of w/^b» the curve obtained
from linearized theory is drawn).
o) = 0'8Qü)0, (o = o)0, and o) = l*14w0. The results are given in figure 5 in broken
lines and the corresponding theoretical results are represented by solid lines, / .
(iii) The moment about 0 exerted by the fluid was measured as a function of s
o) for several values of the oscillation amplitude o\ A counter-clockwise moment
was considered positive.
The measured moment was subjected to a Fourier analysis
mw=Ji *>•*«•*-«-«• m
mm Non-linear oscillations offiuid in a container
We recall in this connexion that <t)Qt — <j> is the phase of the container (since in
the present analysis &» — w0 = 0(e)), so that ^ „ measures the phase difference
749
between the nth harmonic in M and the nth harmonic of the container. The
experimental values of Mlt for several values of 8, are given infigure6 as a function
ofw.
180'M
Af 150°-
120°-
90°-
* i
60°-
30°-
1 1 1 1 1— -T —r~
- 1 0 - 0 8 -0-6 - 0 4 -0-2 0-2 04 0-6 0-8 1-0
(W-Wo)/ÊH
FIÖTJEE 7. The phase ^ of the moment relative to the phase of the container, as a function
of (W-MO)/Ê&V — , Theory; x : 6 = 1°; A: 8 = 2°; O: * = 3°; D : 8 = 4°.
The moment M can also be calculated. Using the results of §4, we obtain
for the frequency range around w, with p as the density of water,
l28eA A.
= n2^ [ ( - l + c o s ^ 0 t + sin^> 0 f)oos(^ + ^r)
+ ( - M ' + i ? r + s m J w 0 f - c o s J ü ) 0 0 B m ( ^ + Jff)] + O(e2) (48)
for 0 < t < 7TJa>, and
where use has been made of (12), (35) and (36). For ^ we obtain
(53)
The theoretical values for $rlt obtained from (52) and (53) are given in figure 7
as a function of (<y-ü>0)/ew0w (w — a>0)l(g8)i, together with the experimental
values. For a) <^<o0 the moment is in phase with the oscillation of-the container
as shown by (50).
6. Discussion
Figures 4-7 show a good agreement between theory and experiment. In
fact, the agreement is better than could be expected, if the following is considered.
We recall that the theory is a first approximation in terms of e = ($Bf7Th0)i
(see equation (12)), so that differences of the order of e2 may be expected. A
representative value of S for the experiments is 7r/60. Then e is as large as 0-5.
Hence the agreement is, in general, surprisingly good and suggests that the
coefficients of the terms in e2 are small.
Part of the discrepancy between theory and experiment is of course due to the
neglect of viscosity in the theory. In particular, part of the difference between
the measured phase of moment and the calculated phase, must be attributed to
boundary-layer effects (cf. Chester (1964) where it is shown for the gas-dynamic :-£
case how viscosity changes the phase of the shock).
The results as given in §4 are valid for frequencies which do not differ very .-4
much from the resonance frequency <o0. In fact, it follows from consideration of
the expression (53), for the phase of the hydraulic jump, that solutions including
a jump exist only for (a> — d>0)2 < 24gSjB. This inequality can be written as Ay^-^Ly • 1
-*
y.'^ilh
i
(o I /24\* ï
< — e. 0||v
• 'j'^f '-"•
Q)Q\ \irj "ï.:Séf> f
A similar result was obtained by Chester for the gas-dynamic case, using a quite -vr v- ïüi- ^. ;: -- *-
5.
.4 • * ? * • • * • '
different method. During the experiments we observed that at values of & well ' ;(
beyond o>0) the hydraulic jump disappeared, which is in agreement with thé above „.^zy-- rt
result. ' ' "'"^'-TT;'*."::-" •
I t may be of interest to note that at the disappearance of the jump a solitary: -
wave was observed, which travelled back and forth between the walls of the'.
TH
Non-linear oscillations offiuid in a container 751
container. This corresponds with the well-known fact (see Wehausen & Laitone
1960, §31) that the solitary wave travels at a speed slightly higher than the
critical speed corresponding to a Proude number of unity. This solitary wave
cannot be obtained trom the present theory, since a higher-order shallow-water
theory is needed, where in contrast to the first-order approximations (equations
1
J
(5) and (6)) vertical accelerations are taken into account.
At high frequencies, o> ^> w„, a wave pattern represented by a solution of the
type (8) reappeared.
REFERENCES
\
BETCHOV, R. 1958 Nonlinear oscillations of a column of gas. Phys. Fluids, 1, 3.
B I N N I E , A. H . 1941 Waves in an open oscillating tank. Engineering, 151, 227.
If
NTHSSTEB, W. 1964 Resonant osculations in closed tubes. J. Fluid Mech. 18, 1. "
O H D , BOA-TEH & S. J . YING 1963 Thermally driven nonlinear osculations in a pipe with
travelling shock waves. Phys. Fluids, 6, 11.
1
Si
CotTBAKT, R. & FBIEDBICHS, K . O. 1948 Supersonic Flow and Shock Waves. New York:
Interscience.
L m , C. C. 1954 On a perturbation theory based on the method of characteristics. J. Math. \
Phys. 33, 117.
STOKES, J . J . 1957 Water Waves. New York: Interscience.
"WEHAUSEN, J. V. & LAITONE, E . V. 1960 Surface waves. Article in Encyclopedia of
•• Physics (ed. S. Flugge), volume i x , p . 446. Berlin: Springer-Verlag.
-II
M
i • Si
-w
'k: