The Design- of High'-' -Speed :Free Surface
''Water
by
josEPti
H. PRESTbiki:Proceedings of the NATO Advanced Study Institute
an 'Surface; HydrOdYnnmi
es!
Bressanone, August 26 September 7 1966
".Tedmitche._
The Design of High Speed Free Surface Water
Channels
by JOSEPH H.
PRESTON1. Introduction
The definition of a high speed water channel adopted in this paper is not restricted to supercritical conditions of
flow. For the present purpose a Channel with a maximum velocity greater than 5 ft/sec. say will be called a high speed channel.
- The design of high speed water channels is particularly
difficult, since not only must the velocity distribution through-out the working section (excluding the boundary layer) be
uni-form as in,a wind tunnel, but the free surface must be free from waves and remain level. Air entrainment must be avoided, as must cavitation on the impeller and cascades if the channel
is depressurised. Leakage of water out of joints and seals must be prevented and also the leakage of air into the system when depressurised. High water speeds involve high dynamic 'forces on the structure -- for instance 2D ft/sec with water
introduces forces comparable with those of a wind tunnel of 600 ft/sec -- thus the structure tends to.be expensive. The need to avoid corrosion limits the choice of the materials of construction or precautions must be taken and these raise the
The first circulating water Channel with a free surface appears to have been a small one constructed by L. Prandtl in 1904 for the observation of vortices. Other channels were de signed between 1920 and 1930 for ship model testing with
di-mensions of about 5 ft wide x234 ft deep and low
velocities.-References to these will be found in ref. [2]: Generally, the
velocity distribution and the quality of the free surface of these channels was poor,by modern standards and not much.uae
appears to have been made of them.
In Germany (at the V.W.S. - in Berlin) a free surface
water channel with working section 7 m long x 1.8 m wide x 1.2m
deep and a top speed of about 6.m/s was built in the 1920's and then rebuilt in 1957. This channel had a compact return circuit resembling that of a wind tunnel (see ref.
Ii]
and Fig. 1) and the water emerged into the working section via the closed contraction nozzle. The free surface was held level andfree from major waves by means of a false tilting floor. This device is now an essential feature of all successful recent water channels. Also the upper part of the flow was diverted
at the exit from the working section to prevent wave reflection
and to get rid of air bubbles. This channel must be regarded. as the first successful channel for ship model testing aid has
set. the design pattern for a number of recent channels!:
A very.large channel was designed for the David Taylor
Model Basin and completed in 1944. This is described in ret.Efl:
The circuit was of the wind tunnel type with an enclosed con-traction nozzle and the test section was 60 ft long x 22 ft
wide x 9 ft deep. The maximum speed was about 16 ft/sec
Ob-tained.from an input of 1250 h.p. The overall length was 157 ft_'.
No details of the performance is given in reference [2]; but
it is understood to-have not been very. successful. An outline
sketch of its elevation is shown
in
Fig. 2.: In the U.K.;theBritish Ship Research Association sponsored research at Cambridge under Bihnie on the design
of
high speedEle..1E4.3,
Elbe-Fi gr.
a
De s ign of the V W S
Water channel
Ver.ticar. longitudinal section- through. T
1
I
water,..chaurret (from ref.
,.
.
water channels. Binnie 16J studiedthe discharge from a large stilling tank into a long channel via a nozzle. He found that
at high speeds a trough' water surface developed which he could
not sufficiently improve by skimming. Also at subcritical
speeds, and particularly as the critical speed was approached,
large waves could develop. He then turned to a natural draw
down instead of a closed nozzle and by development of a cascade type of weir at exit from the working section was able to hold a satisfactory level surface up to a Froude number of 0.6 (see
ref. [7]). Following this research, the N.P.L. Ship Division considered the design of a large flume and built a 1/10 scale model (Fig. 3 and ref. [8]). This model operated extremely
satisfactorily when suitable smoothing screens were introduced
into the stilling tank upstream of the working section. The full scale channel was then built and completed in 1966. The
working section is 50 ft long x 12 ft wide x 8 .ft deep and the top speed is 12 ft/sec corresponding to a Froude number of 0.75.
The water flows from a stilling reservoir via a natural draw down through a 9:1 contraction into the working section. This
leaves' via a cascade type weir and falls under gravity into a large sump from 'which entrained air bubbles escape before being
pumped into the upstream reservoir through a 2000 h.p. axial flow pump: The overall length is 350 ft approximately.
A water channel has recently been installed at the Duisburg
Technical College. This was designed and built by Kempf and
Remmers, Hamburg. The working section is approximately 5 m long x.11/2 m wide x 1 m deep. The top speed is 8 ft/sec (Z%
m/s)-This follows, the design of the Berlin channel in having a tilt-ing floor and a closed contraction nozzle (2:1). Entrained air
is removed by diverting the Upper layers of water leaving the working section over a plate and through screens, so that the water is slowed down in a passage of increasing area before being returned to the main circuit between the first two
cas-cades. Flap controls, at the end of the nozzle, at the beginning
DI FF USER UPSTREAM STIU.ING ICoNTRACTIoN SECTION
Fig., 3
-:. WORKING SECTIj WEIR SECTION .-Fig.
3 a- General layout of circuit
,
.
W TER ' LEVEL. IN: /
"fAWOi
WORKING:SECT . . D . . l grRP-.: GUMP_ wxrgit 'T :,, LENEL , SUMP ' cf tilting weir
di:mined(- [rola re 1..
,.-with the main return, help in maintaining a level surface and in preventing gulping of air into the main circuit. We were
priviledged to see this channel in operation and were impressed
by the level water surface and its ease of control by tilting the floor.
At Liverpool we have very recently replaced an old low
speed water channel, installed in 1925, by a new channel. The new channel has a working section 14 ft long x 43 ft wide x
23/i ft deep and a top speed of nearly 21 ft/sec. The channel was
designed in conjunction with Kempf and Remmers who built-it (Fig. 4). It follows the now established German design with a
closed contraction nozzle (4:1) and a false floor which can be
raised and lowered and tilted. Entrained air is removed as in
the Duisburg flume by a diverter plate with controllable flaps,
which also prevent gulping of air into the main circuit. The design is novel, for a channel of this size, in that full
de-pressurisation of the free surface is intended. A model, 1:4.67
scale, was built by Kempf and Remmers to investigate the de-pressurisation aspects and the general flow features. This model was very successful. Results for the model channel and the full scale channel will be presented later in Part II of
this paper.
In the United States there are two very high speed water channels, one at C.I.T. and the other at Hydronautics near
Washington. Both have small working sections. which can be
de-pressurised and have large deaeration chambers through which the whole flow is passed. The C.I.T. channel has a working section 8 ft long x 2.0 ft wide x 14 ft to 2 ft deep. The miximum speed is 30 ft/sec and the free surface can be de-pressurised down to 1/15 atmosphere. The Hydronautics channel has a working section 11 ft long x 2 ft wide x 1.0 2.0 ft
deep-. The maximum speed is 40 ft/sec and it can be depressurised
li/j3 (53.0.Frr)t &es . /3300 _03.6 rt) 5000 (LUTE) . rioxfraN4 Jsulem 7
.1.:iyerpool.,l:Rivers.it.rtwit;er channel'
hia:
5Variablepressure, freesurface, highspeed
channel. Hydronautics Inc. (from ref.
5)Fig. 6
The freesurface water tunnel at the Hydrodynamics
Laboratory of the California Institute of Technology (from
hig. 7
a
area ratioLriiversity water
channel,....
' 20,028 A4Tatranre..Darsra of Swcarcamty dicristrk. -Fig. 7la ,
-chann
_. .Fig. 7 Effect of contraction ratio on size of channel
The design of this.is fairly straight forward. The sides
should be parallel and smooth, with the maximum of window area for viewing consistent with strength. The junction of the steel
and glass (or perspex) surfaces needs to be smooth in the neighbourhood of the water surface. Enough freeboard must be
left to cope with large solitary waves which can be accidentally.
generated when adjusting the false floor or if an emergency shutdown occurs. The top of the sides of the workinE section must be designed to carry a variety of instruments including the model suspensions and resistance dynamometers, and in the case of depressurisation be designed for sealing.
b) The first diffuser
The purpose of this is twofold. .Firstly, to slow up the
flow before it reaches the first two cascades of turning vanes,
for these can contribute very greatly to the power factor of
the circuit if the velocity is high. Secondly, to commence the necessary transition from a rectangular to square section before
finally changing to a circular section before the impeller.
The angle between the walls should be such as to avoid' boundary layer separation and alternate switching of the flow from wall to wall, the effects of which may be felt in the working section. The maximum angle which is permissible here
depends on the boundary thickness at entry to the diffuser and on the area ratio finally reached. References [10]; [11]; [12]'
and [13] and[14] should be consulted and appeal to model tests may be necessary.
c). The
guide vanes in the bendsThe design of the guide vanes system for the first two
bends tends to be the most difficult, since the velocities and hence the forces are large, and, in the-depressurised channel,
is not : far belovWthe...free surface. _
A:variety
of
guide Vanes exists for 'Wind. tunnel use (see references "[10]-and-,[11]). 'Vanes for water =channels andwaterthnneig..haie
sharp leadingedges,andAhp sn11.4ce§
-are formed by .circular arcs A -modificationfhf this fdrin
based '
on
the assumption of a free vortex flow in the passagebetween neighbouring blades..
Bladeg:Offthii'ddsigh:
have -.constantpregatie
on each surface 'which is an advantage from-a cavitation. _
-standpoint : For the guide vanes listed'. in
'refg3,11.61-
and,.[1.1];-the thickest sections have a minimum ' loSg. coefficient of 0. 25
for -a gap/chard. _raft&
of
0 5 whilst. for very,thin
circular :alf&profiles the minimum : loss coefficient is 0 15i for a gap/chord' . ratio of 0.3 strength and vibration requirements in
-profiles
of
some thickness essentialuhless:a..large
number OfcidggeupportS are
:Used,
which 'liutS the cost 40.-::Wilesg theflow is slowed up.. after leaving the Workini. section it is cidar
that the first twd,gets:-Of guide vanes will make' a considerable _
if not major contribution to the total ;losses in the circuit:
d) The Impeller and straightener design
The aim here is to achieve good efficiency and ,cavitation
performance In the case of the depressurised - channel .-_ The _latter
- is linked to thedePth:of the impeller below the free surface,
The blade profiles in .conjunction with the gelected-.radial-.
loading or lift coefficient, must retain as high
a
-value Of the- ,
-.pressure minimum as possible. - The ..camber and fairing of the
:
profiles determines the .chordwise PreSSure'distribution which
should be nearly constant over the 'majorpett:Athe.,guctiOn
_surface_ Low circuit losses (i e. a low powerl factor) are
im--
-. .
portaht in that they reduce the disc loading of the
impellerLo*
lift coefficientsraise the minimum pressure Or reduce ' the
-.maiiinuth
velocity
over the blade. This leads to an .impeller of-high :gol id ity ' which becomes difficult to design and also to
..:SOthe'ireduetioh
of efficiency Low blade velocities clearlyhave
an advantageous effect on cavitation but again increase the
solidity. In appendix II, the influence of the various factors
on the cavitation performance is dealt with in an elementary
way. References [9]; [10] 'and [17] should be consulted. As regards the general design, this could be of the 'free vortex' type (as is usual in wind tunnels) with straighteners which are designed to remove all the whirl; or the design may be such as to make good the loss of total pressure in the boundary layer and thus give a uniform outlet flow. With this design no arrangement of fixed straighteners can completely
remove the rotation; zero rotation only is possible with contra,
rotating impellers and these are impracticable for a water channel. The behaviour of the diffuser which follows and the selection of its angle is undoubtedly linked to these aspects of impeller design. We may note that the presence of whirl downstream of impeller is undesirable since it may find its way through to the working section unless a fine honeycomb is used. On the other hand it is claimed that the presence of
whirl enables a larger diffuser angle to be used_ The available
information on this aspect and on the general problem of the impeller design on diffuser performance is meagre.
As regards the straightener (if one is employed) its de-sign depends on the magnitude and nature of the whirl. If the whirl is of the free vortex type and if the ensuing angle of
flow relative to the axial direction does not exceed 12°, then
the simple untwisted radial straightener can be employed (see ref. [19]). Otherwise a more complicated, twisted version is necessary. The whirl associated with a free vortex design is considered in appendix III.
A favourable point arising from the use of a low lift
coef-ficient at the tip, in order to avoid cavitation, is that the hub diameter comes out quite small and there is little danger of its `shadow' or wake penetrating to the working section as occasionally happens in wind tunnel designs, when a hub/tip
diameter ratio of 0.5 or 0.6 is usual.
e) The main diffuser downstream of the impeller
The purpose of this diffuser is to slow down the mean velocity to the value set by the contraction ratio of the
nozzle. It must do this in as Short a distance as possible
con-sistent with good efficiency and the absence of separation. This implies a fairly small cone angle. As regards minimising
the losses, little is gained by continuing the small angle dif-fusion beyond an area increase of 4:1. The angle selected here
varies a good deal in practice and the cost and length of the structure must be born in mind. Wind tunnels frequently keep
to as low as a 50 cone angle (refs. [1.0], [11]). Angles up to
8° have been used without separation occurring, but we think that this implies that the impeller has been deliberately de-signed to make good the loss of head in the boundary layer or
that swirl has been deliberately used to prevent separation and
also that the area ratio is less than 4:1. The whole problem
of diffuser design is a complex one and at present is not
com-pletely amenable to mathematical analysis, though an attempt is made in refs. [12]and [13]. Inlet conditions have an im-portant effect and scale effect is not understood. If a large
contraction ratio is employed, then, after the small angle
dif-fuser has reduced the velocity to 1/4 of the working section value, a rapid diffuser of large cone (200) angle can be em-ployed, using wire screens at appropriate intervals to avoid separation ref. 110, 111: Alternatively, the constant wall pressure diffuser suggested by Squire and developed by Gibson (ref. [15]) can be used. The constant wall pressure is achieved by correct shaping of the walls, plus the use of one screen whose resistance and position can be calculated. Another form of rapid diffuser is obtained by introducing partitions, so that in effect the diffuser is composed of a number of fine angle diffusers. The pressure recovery is poor and the flow uniformity is much improved by having a screen at the exit.
Settling length
This is placed after the last bend and before the
con-traction. Its purpose is to provide a place for smoothing screens
and honeycomb and a length of low speed flow to allow eddies and turbulence to decay. Its length will partly depend on how uniform the flow is after the last bend and how many screens are required. A long settling length adds to the cost, but it may be necessary for very high speed channels.
wire screens
Tne purpose of screens is to even out spatial variations in velocity and to reduce turbulence. They are also used to
prevent separation in large angle diffusers. Ideally, for recti-fying the velocity distribution, the resistance should be twice
the local dynamic head. Bradshaw [18] has shown recently that such screens can create weak vortex flows of large scale. He
has found that this is prevented by using screens of resistance
about 1.6 local dynamic head (open area ratio 0.57). It is
recommended that screens of this type be employed in the
settling length. It should be noted that screens in high speed
water channels have to support very high loads and they may need to be supported on a coarse honeycomb structure.
Pro-vision must be made for access to them for cleaning. Stainless
steel would seem to be essential from both the standpoint of
corrosion and strength. If a number of screens is to be used,
then a large contraction ratio is essential if the losses are to be kept small. with a contraction ratio of 4:1, only one
screen seems possible.
The honeycomb
The purpose of this is to remove large scale rotation such
as may arise from the impeller, secondary flows in the cascade passages and weak vortex flows from screens. It is essential that a uniform cell structure is maintained and the smaller
the cells. the better. The ratio (length of the cell to its cross
. dimension) should be large -- 6.to 8 seems satisfactory. For
water channels, the cells are usually square and the honeycomb
is built up by halfslitting long strips of brass or stainless steel. With small 'diameter' cells the cost can be high. A honeycomb is essential in all channels!
1) The contraction nozzle
The purpose of this is to accelerate the flow from the settling region into the working section. It plays a very im-portant part in reducing the longitudinal component of turbu-lence and the spatial variations of total head. A monotonic
behaviour of pressure distribution along its walls is desirable
and particularly at the high speed end in the case.of water
channels. To achieve this an exponential approach is necessary
and a slow approach to the working section for water channels
helps in maintaining a wave free surface. There is a
consider-able literature on the design of contraction nozzle for'wind
tunnels (references [9]; [10]; [16]. and Cl?];
Choice of Contraction Ratio
This has such a very important influence on the design,
performance and costs that Some discussion of it is essential.
It is readily shown that spatial variations of total
pressure expressed in terMs of the mean local dynamic head are
reduced
(r
contraction ratio) between entry and exit.1.2
Thus it is seen that the contraction ratio plays a big
part
in obtaining of uniform velocity distribution in. the workingsection.
Also it is shown (ref. [10]) that the longitudinal
components are increased by a factor
A
.-Moreover, if smoothing screens of relatively high total resistance are to be employed, then a large contraction ratio
is necessary if the losses due to these are to be kept small,
1
since these losses vary Thus, hydrodynamic arguments tend rz
to favour large contraction' ratios, but economic considerations
indicate that both the cost of the channel and the building
-in which it is housed rise rapidly with increase in contraction
ratio and a compromise has to be effected. Generally it seems
that where the 'dimensions of the working section are moderately large (1.e..5' x 3' say) a modest contraction ratio between
2 and 4 is chosen. Certain small, but very high speed, channels (working section 2' x 2' approximately) have large contraction
ratios. Provided that very low turbulence levels are not re-quired, then there, is evidence that a fairly satisfactory velocity distribution in the working section can be obtained with a.contraction ratio of about 41 and this might allow the
use of one smoothing. screen. However, significant gaits in the
level of turbulence and uniformity of velocity distribution are obtainable by going up to a contraction ratio Of 9:
Fig. 7 shows the Liverpool channel with a contraction ratio of
4:1 and also shown is a channel with the same working section and return portion up to the diffuser, but with the rest of
the circuit modified to accord with a contraction ratio of 9:1.
j) secondary flows
- These are cross flows associated with streamwise trailing.
vorticity in the boundary layers on the walls. Apart from the
variation of flow direction which they imply, they.have a much more important effect on the distribution of the boundary layer thickness. Generally their effect is to thin the boundary layer
on the side walls of.channels and to thicken it on the bottom
about the centre line of the bottom. It is this accumulation
which is mainly responsible for the gradual departure from
uni-formity of the velocity distribution in depth as the floor is approached. Also cross traverses near the floor show a region
of low velocity or total pressure about the floor centre line.
The origin of the trailing vorticity stems from three
sources: (a) the wall boundary layers in passing through the cascade passages are bent in their own plane and as a conse-quence generate streamwise vorticity which sets up secondary
flows in the cascade passages (see refs. [21 to 23]). On leaving, these vortices, which are all of the same sign for any one wall,
combine to form a large single vortex. Thus there are a pair of vortices of opposite signs associated with the boundary
layers on the side walls, (b) on passing through the contraction
nozzle, which has double curvature on each wall, the boundary layers are again curved in their own plane and the issuing
secondary flows may not always have the same sign as those under (a); see ref. [24] andFig. 17. Finally (c) there is the well known secondary flow associated with turbulent boundary layer
flow in corners whicn distorts the contours of velocity and total pressure, tending to produce a thick boundary layer at
the centre of each wall and hence on the bottom wall of a channel to add to the effect under (a). Appendix III outlines the main
results of secondary flow theory.
Clearly the source of trailing vorticity under (a) is as-eociated particularly with the boundary layer from the diffuser
entering the last two cascades and giving rise to a very poor distribution of total pressure at the entry to the nozzle.
The honeycomb,probably,is effective in killing much of the swirl associated with these secondary flows, but a screen is necessary
to even up the distribution of total pressure and perhaps a screen at the end of the diffuser is the best solution. The
secondary flows under (b) and (c) are not so amenable to control by correct design at present and further research is required.
2.2. Settling Tank - Sump Type of Channel
Fig. 3 shows this arrangement diagrammatically. Water flows
from a long settling length containing smoothing screens and honeycomb through a contracting nozzle with natural drawdown
to the working section. It leaves this via a special weir, falling under gravity into a large sump from which entrained air escapes. From the sump it is.pumped along a long pipe (of
area about equal to that of the working section) to the settling length which it enters via a rapid angle diffuser. This may be of the 'egg box' type i.e. partitioned or it may be controlled by screens. The power factor for this type of circuit is large
since all the K.E. of the working section flow is destroyed.
The return pipe has large skin friction losses and there are
4 bends in this duct (with cascades) for which the losses are high, since the velocity is high. Hence a motor of large power
is required and the impeller may be difficult to design since it is at a relatively low depth below the free surface.
It seems that this type of channel is likely to be employed when the dimensions of the working section are large, because
it is structurally and economically more attractive than the wind tunnel type of circuit. The design of the various indi-vidual features of the circuit e. g. impeller/straightener, guide vanes, screens, rapid diffuser, contraction is already
covered under 2.1 and reference [8] gives details of an actual
design. The pump intake may need special attention to avoid
vortices.
3. The Free surface
Ideally, the free surface should be level and free from long or short standing waves over the whole speed range. If the
speed range covers both subcritical and supercritical ve-locities this presents a severe problem.
The presence of waves and the nature of the surface (Whether -glassy Or rough) depends on (a) the speed of flow, (b) the type
of entry to the working. section (whether via an enclosed nozzle or natural.dram=down), (c) the provision of an adjustable floor,
(d) the nature of the exit (e) the water content.
3.1. The Entry
(a) Enclosed nozzle
Binnie and his associates (refs. [6]. and [7]) in their experiments with nozzles found that the water surface at exit
was 'rough', and this was particularly noticeable at high speeds. Attempts- to skim off the rough surface with a blade gave little
improvement and this is supported by the experience of Dr. Remmers of Hamburg. However, the designers and operators of the Berlin flume have accepted this limitation up to speeds of about 6.m/S, as also has Dr. .Remmers for the lower speeds of the Duisburg flume and we ourselves in conjunction with
Dr. 'Remmers for the new Liverpool flume. At Liverpool we find
that at speeds below the critical (9.4 ft/sec) the roughness is very small and that up to 15 ft/sec it does not exceed
1"
However, at topspeed 21 ft/sec.the surface is very rough
and this together with entrained air constitutes a serious limitation above 18 ft/sec (design top speed). we.think this roughness arises from the boundary layer of the nozzle, which
becomes a highly turbulent wake on the free surface. The
rough-ness is then a manifestation of the fluctuating velocities associated with this turbulence and possibly with the general
stream turbulence. Possibly some form of boundary layer control
on the nozzle surface would reduce the roughness.
Needless to say the nozzle surfaces should be smooth i.e., free from waves and roughness and have exponential approach to the working section sides and the free surface. A design' point to be .noted, if an enclosed nozzle is used, is that suction
to run full at low speeds.
b) Natural drawdown
The large N.P.L. channel (working section 12 ft x8 ft deep)
is based on tests of.
0'
scale model (ref. [8]) and has a natural
1
drawdown. Both the model and the full scale flume operate in . the subcritical speed range. The natural drawdown appears to
be very successful; there is no roughness of water surface, but
there are small wavelets which originate mainly from the con-traction side walls. Some are no doubt due to surface tension effects, but others could arise from contour imperfections on the contraction wall. It appears highly desirable to maintain continuous curvature of the surface of the contraction wall (with absence of even small waves in the contour) if dis-sturbances to the water surface are to be avoided.
so
far as is known, there are no research channelsoper-ating at speeds above the critical which employ natural draw down. There appears to be no fundamental difficulty, though there is a practical problem associated with the variation of head in the settling length when the speed is changed. In Ap-pendix I we have investigated, on a onedimensional basis, the form of the contraction nozzle to give an asigned monotonic shape of drawdown for different Froude numbers. Contractions in both the vertical and horizontal planes are considered. In
both cases it is seen that the contraction shape must vary with Froude number and, in particular, at speeds above the critical (F= 1.0) a bump appears on the nozzle (nonmonotonic behaviour). Provided the contraction is gentle, the calculations should be
correct and therefore, for an assigned shape of drawdown, adjustable walls must be incorporated to cover the required Froude number range. The need for variable wall shape perhaps
in part explains. the success of the adjustable floor which is dealt with below.
3.2. Adjustable Floor
With a fixed horizontal floor, Binnie Ds] found increasing
.difficulty in maintaining a wavefree level surface in the working section of his model channel as the speed was raised.
At speeds approaching the critical large solitary waves appeared. The C.I.T..water channel has a fixed floor and there are limited
ranges of speeds or rather Froude numbers for which wave free
flow is possible. The Hydronautics channel has a fixed inclined floor (to compensate for boundary layer growth) and satisfactory operation is claimed. over a wide speed range. Both these Ameri-can channels are primarily intended for very high speeds in the
supercritical regime. The very high speed channel now under design by Kempf and Remmers has an adjustable floor.
The Berlin flume (described by Schuster [I]) appears to
have been the first to have made use of an adjustable floor.
Both height and inclination of the floor can be varied to give
a wave free surface over the whole speed range. The Duisburg flume and the new Liverpool flume incorporate this device and all these flumes have enclosed contraction
nozzles. The
ex-perience at Liverpool is that there is one angle of inclination
which is practically satisfactory throughout the speed range and this is the angle estimated to be necessary to compensate for boundary layer growth. The large N.P.L..channel and its model operate below the critical speed and both have an
ad-justable floor, but in the large channel it appears to be used at a fixed setting.
It is clear that an adjustable floor must be considered as essential for channels' intended to operate from low sub critical Froude numbers through to supercritical conditions. It also has the. advantage that variable depth can be readily
obtained. The Hydronautics channel is interesting in that vari-able depth is achieved in the working section by incorporating
3.3. Exit from working section
A successful design of exit must prevent waves being re-flected upstream at subcritical speeds. It must prevent draw down and gulping of air into the return circuit and it must allow entrained air bubbles to escape so that they are not circulated around the channel.
(a) Wind tunnel type return circuit
In the German type of design some 5% of the main flow is diverted over a long splitter plate lying above the main exit and over the first bend. Because of the shallow depth of the diverted water, the initial flow over the diverter plate is supercritical and no waves can return upstream.
A number of wire screens intercept the diverted flow and
a 'jump' occurs at the first screen, or ahead of it at low speeds, causing bubbling and foaming. As the flow passes these screens the depth increases and its speed decreases so that bubbles of entrained air have a chance to rise to the surface. The diverted
flow rejoins the main flow behind the first cascade through a
passage the width of which is controlled by an adjustable flap.
Movement of this flap in a direction to further constrict the
main passage (which also has been reduced inarea)andix)increase
the secondary passage tends to prevent drawdown at the end of the working section at high speeds. It is used in conjunction
with an adjustable flap at the beginning of the diverter plate
to prevent air entrainment and splashing. The setting of the adjustable floor and increase of water content are important in this connection at high speeds.
These arrangements, though complicated, are quite satis-factory on the German type of design up to speeds of about 18 ft/sec. Above this speed more and more small air bubbles
are entrained into the main flow. If higher speeds are required
then other arrangements become necessary. A possible solution would be to carry the diverted water into a settling tank so
that the bubbles have time to rise. The water could -then be
pumped back into the main circuit and used as a means of boundary layer control to improve the velocity distribution- at the dif-fuser exit. In the very high speed channels of C. I. T. and
Hydro-nautics the whole of the flow at exit is passed to specially designed de-aeration chambers of large volume. The working sections of these channels are fairly, small and so this is possible but with large working sections this solution would
be very expensive. In the projected design for the Berlin channel
Kempf and Remmers adopt a similar solution of a large
de-aer-ation chamber but the geometry differs from the American type.
b) Sump type 'return circuit
In this design, the water in leaving the working section
falls under gravity over a weir into a large open -sump. The air entrainment is considerable, but, if the sump is large enough,
air bubbles have time to reach the surface ahead
of
the pumpentry.'
The weir needs to be of special design to avoid draw-down of the free 'surface in the working section and also the possible
formation of long waves at 'sub-critical speeds: Orkney [7]. carried out a number of experiments on this aspect of channel design and evolved a cascade type of weir which is employed
with, success 'in the N. P. L. 1/10 scale channel [8] and in the
full scale channel.
3.4. Water Content
In both the wind tunnel and sump types of return circuits the water level at rest is of some importance. In the former, for low speeds the channel is filled so that the level is very
slightly above that of the nozzle exit.. At high speed§ for optimum operating conditions the water content can be slightly
of air into the main circuit.
4. Free surface with Depressurisation
The very high speed channels in the U.S.A. at C.I.T. and
Hydronautics and new German high speed channel can be
de-pressurised down to atmosphere. These have relatively
15 11
snail working sections. The new channel at Liverpool (working
section 4% ft x 23 ft, top speed 20 ft/sec) has been designed for depressurisation down to about 1 ft of water absolute
pressure. This greatly increases the scope of the channel.
Ignoring viscosity and surface tension, the drag D (say
of a partially submerged body depends on the velocity U, a
representative length 1, gravity g, the density p and (p. - pvl where p0 is the pressure at the surface and pv is the vapour
pressure i.e.
D = f[U, /, g, p, -
PO]
There are thus 3 non-dimensional groups involved and hence
U
pn - p
2 2 =
f(
2 f(P, a)½pUt
1 gwhere F Froude number - Cavitation number
For dynamical similarity with the above yariables between model
(1) and full scale (2), F and a must be constant for the two systems. Now both g and p are the same for the two systems, and if
model .(1) , Or 12 (Fib
- Pv/,
(p0
.1)v 1 Pv U 2(Po
12 (PO Pv)2.l2
As an example, suppose we need to inVestigaWthe...7nose-i,
. - - ,,,
bility Of cavitation on the stabilisers, ._ . a passenger
ship--.
,
_.
._.
.or the behaviour of tbe:propeller. Take:
ft U
knots-F =0.274a =
Here we note that in the model test the
aPeedA'a'lWL-and.-ab-iia.
tfie:miriate-pressure.
-
This
test would be not possible if theObtraction:hWto rmitull,' since suction below the free surface pressure has to tei.applie-d. However it is
possible thaty-at-AOW speeds';
.1 .
satisfactory flow can be obtained without the contraction
,
running full.
500 ft 20
As a second example we take the case of a small high speed craft e.g. hydrofoil boat.
-ft U knots
(po -
ftF = 1.33
a =0.0743
Here the surface pressure is higher than in the former example, as also is the speed, so that the contraction problem does not
arise. However, since the speed in the channel is 10 knots
and because 0- is small, it is necessary to consider the
cavi-tation performance of the channel; particularly in regard to
the impeller and the first set of guide vanes.
This has been done in Appendix II where the results have
been presented in non-dimensional form showing
acrit
v-whilst, in theory, it is possible to predict the two constants appearing in the theory, one of these is best
ob-tained by model tests. The design of the Liverpool channel was
based on tests of a complete model. Fig. 16 shows the graph of
a
. vcrit
for this case. and within the speed range of the full scale
channel cavitation trouble seems unlikely, but within the speed
range of the model channel serious impeller cavitation can
occur.
8hiP
(2)
80- 40 345. Selection of Channel Size, Speed, Type of Design, and Materials of Construction
The selection of channel size and speed is governed prima-rily by the money available and by the need of the model under
test to be reasonably large whilst satisfying the full scale
Froude number and cavitation number -- if the latter is likely
to be significant. However, whilst small scale models require
low speeds on a Froude number basis, they require low absolute free surface pressures for the same cavitation number. Also, as a consequence of Froudes law, the Reynolds number of the model is the (scale) 2 of the full scale Reynolds number. This is
a serious limitation in those cases where the drag and boundary
layer is dependent on Reynolds number.
From the present survey it appears that the designs fall into 3 groups. The first relates to the testing of very high
speed surface craft and underwater weapons. These channels tend
to have small working section dimensions, high water speeds (30 50 ft/sec) and have depressurisation. The C.I.T. and
Hydronautics channels in the U.S.A. and the new Berlin University Channel falls into this group and references [s]' and [4] discuss
the design requirements. At the other end of the scale is the
large channel in which ship models are to be tested on a Froude number basis only. Large models are necessary to give'accuracy
in manufacture and to yield boundary layers which are thick enough to allow of their exploration with Pitot tubes for measurement of velocity and, or, skin friction. The size of model is not too different from the ship tank model so that
comparison of tests is possible. The D.T.M.B. channel and the new N.P.L. channel were designed to meet this need. The former has
a working. section 22 i4wide x 9 ft deep and for the latter the dimensions are 12 MOO x 8 ft deep. The complete structure
and the building are large and expensive, as is the supporting
instrumentation. Hence such a channel will only be found in a
national establishment. Finally there is the intermediate type
quantitative information from model tests on small to medium size high speed surface craft. It should be capable of pro-viding useful information relating to pressure distribution
and on flow directions in the boundary layer for models of large ships. However, its Reynolds number will be small so that drag measurements will not have much significance in this case, where skin friction is the major source of drag. The Berlin, Duisburg. and Liverpool channels fall into this category and have much the same size of working section 5 ft wide x 3 ft deep with a maximum speed of up to 20 ft/sec and operate over the whole
speed range. The flexibility and scope of such channels can be greatly increased by depressurisation. It would seem that
gener-al purpose channels which are smgener-aller than this have a re-stricted usefulness because of the small model size. On the other hand increase in dimensions rapidly puts the cost up so that firms and universities cannot afford a larger channel.
It is interesting that, for all three groups, the Reynolds number (based on maximum velocity and working section width)
comes out about the same value..
Type of Design
The small and intermediate sizes of channel
will
be of the enclosed nozzle wind tunnel type circuit because of therelatively high top speed and the need for depressurisation. The material of construction will be steel -- preferably of stainless steel for the parts in contact with water. The
pro-tection of mild steel against corrosion appears to be difficult
for open channels where the supply of oxygenated air to the water is constantly being renewed. However, where cost is im-portant, a suggested formula for protection is two coats of phosphating solution, two coats of aluminium primer paint and a finishing coat of bitumastic paint. Of the two very large channels in existence, the D.T.M.B. channel has a wind tunnel
whilst the N.P.L. type has an open settling tank sump layout
with natural draw-down. The latter probably is capable of pro-ducing a smoother water surface and it is probable that structur-ally and economicstructur-ally it is more attractive. However, we think
that there is a third solution which utilizes a wind tunnel circuit laid out in a horizontal plane. The circuit could be
enclosed partially to give a natural draw-down. The exit from
the working section would utilize the German design of a di-verter plate. The diverted water would be freed from bubbles and returned to a constricted section or to a separate tank from which it could be pumped into the main diffuser. Fig. 8 Shows a layout of this channel. The power factor of such a channel could be as low as 0.5. Taking the N.P.U..dimensions
and top speed of 10 ft/sec it would require only about 100 h.p.. to drive it instead of about 1200 h.p. for N.P.L. design. The material of construction for large channels will be reinforced
of concrete for the bulk of the structure.
Again, in connection with the very large flumes and the very high speed channels, it is worth considering the possi-bility of siting these near to a hydro-electric station. The waste water or compensation water could be used to supply a
large stilling tank or small reservoir and then be led to the
working section and returned to the river. Thus the expense of a return circuit and the problem of aeration and de-pressurisation disappear.
6: Speed Control
The main drive should be designed to give continuously. -variable control of speed down to _-L.the maximum speed in
200
the case of a general purpose channel, which may. be used to
calibrate oceanographical instruments. For model testing a 1 to
-00" I./ :
-riS
SIDE ;ELEVATION'Fig. 8
holding to 1.0 % for any speed in these ranges, which should
be by automatic control of Shaft r.p.m..The WardLeonard system is a well tried one and in our experience is capable of achieving
the above requirements without gearbox on a single pulley.
system with a toothed belt drive. There are alternative systems of obtaining continuous speed control such as hydraulic
trans-mission or by a variable pitch impeller, as is used in the large N.P.L..channel..
7. Basic Instrumentation
7.1. speed measurement
This will normally be obtained from the pressure drop
across the contraction nozzle, which has to be calibrated against
traverses by a Pitotstatic head. At very low speeds of the
order of ft/sec it will be necessary to use an accurate 10
tachometer of the kind which continuously counts impulses or signals which are proportional to impeller.r.p.m..The cali-bration of this can be done by the timing Of a float. The ta-chometer reading can be also very useful at all speeds, but it
should be borne in mind that its calibration may change with settings of the various controls. Deaeration of all water Manometers should be provided..
7.2. False floor position and flap controls
The position of the floor can be obtained by direct measure-ment using a steel scale at.the.entry.and exit positions.
How-ever, it is wellworth the expense to have distant reading counters on the control panel, which record. the revolutions
of. the jacking motors and hence floor positions. The flap
7.3. Balances for measurement of forces
If the channel is not depressurised or when used in the depressurised condition for testing ship models, a drag dyna-mometer combined with trim and displacement meters will be
essential.
In depressurised channels, balances will need to be of
the strain gauge type'because of the limited head room available above the free surface. Serious consideration must be given to
the forces and moments.which are most likely to be required.'
It seems preferable to order a comparatively simple balance to
cover the most likely measurements and to improvise for the
others, rather than to order an expensive multicomponent balance
which may not be satisfactory for all the measurements..
7.4. gavemaker
Provision should be made in the design of medium size and
large channels for the mounting ofa wavemaker at the end of
the contraction nozzle.
7.5. Conversion to an enclosed water tunnel
It is worthwhile to make provision in the design for the mounting of a plate to suppress the free surface waves in the case of a medium size channel. This then converts the channel into a comparatively large water tunnel which, at 20 ft/sec, is equivalent to a wind tunnel of the same size operating at
230 ft/sec on a Reynolds number basis. As has been pointed out the dynamic forces are much larger and it is likely that certain
problems in industrial aerodynamics, involving the elasticity
8. Miscellaneous Problems
8.1. whirl and vibration of the impeller shaft
Generally whirl is not likely to be a problem because of
the low shaft r.p.m. and the necessity of having a fairly stiff
shaft to transmit the torque.
There is the possibility of resonance at the first natural
frequency of bending vibration if this corresponds to blade frequency. This can arise if the peripherial distribution of
velocity is nonuniform, as may happen when the main flap control (which controls the diverted water) is in an extreme position.. Then each blade is subjected to a fluctuating bending moment,
Which will provide an exciting force at blade frequency. The
cure of this, in the design state, is to ensure that the first
natural bending frequency is sufficiently high.
8.2. The fairing of instrument leads and the method of support
The fairing of Pitotstatic head leads is often in the form
of
a fine streamline shape formed by two biconvex circular .arcs cantilevered from a central support above the channel. Thecentre of torsion is at midchord point and the centre of
pressure of the lift forces on the .fairing is at the quarter chord point. Clearly, with this arrangement, slight
misa-lignment in direction calls into play .a torsional moment tending
to increase the misalignment and thus to further increase the
lift and twisting moment. A critical speed is reached when the
elastic restoring moments are exceeded by the hydrodynamic moments and 'divergence' both in torsion and bending
occurs.-This will prevent the use of the instrument in this form at high speeds and large depths. The cure is to introduce a pivot ahead
of the centre of pressure so that the instrument has 'weather
cock' stability and will always align itself with the flow.
Small rubber pivots used in suspensions.are very suitable in this
WENN_
LWstnt ELEVATION;
TILT1i1C _ FLOOR. : ' . 1\\L\
-:,At:sugge-ated- design for:alarge water channelv:-'
DIVERTED. WATER
'holding to 1.0 % for any speed in these ranges which should
be by automatic: control of shaft r.p..m. The Ward-Leonard system is a well tried one and in our experience is capable of achieving
the above requirements Without gearbox on a single pulley
system with a toothed belt drive. There are alternative system of Obtaining continuous speed control
such
as hydraulictrans-mission or by a variable pitch impeller, as is'useclin the. large N.P.L. channel.
7. Basic Instrumentation
7-1, speed measurement
This will normally be obtained from the pressure drop
across the contraction nozzle, which has to-be calibrated against
traverses by a Pitot-static head. At very lo* speeds of the
order of ft/sec it will be necessary to use an accurate 10
tachometer of the kind which continuously counts impulses or signals which are proportional to impeller r.p.m..The cali-bration of this can be done by the timing Of a float. The
ta-chometer reading can be also very useful at all speeds, but it
should be borne in mind that its calibration may change with settings of the various controls.. De-aeration of all water Manometers
Should
be provided_7.2. False floor position and flap controls'
'The position of the floor can be obtained by direct measure-ment using a- steelscale at the entry and exit positions.
How-ever, it is well worth the expense to have distant reading counters on the control panel, which record. the revolutions
of,the jacking. motors and hence floor positions. Theflap
7.3. Balances for measurement of forces
If the channel is not depressurised or when used in the depressurised condition for testing ship models, a drag dyna-mometer combined with trim and displacement meters will be
essential.
In depressurised Channels; balances will need to be of
the strain gauge type because of the limited head room available.
.
above the free -strface serious consideration must be given to
the forces and moments which are most likely to
be
required. It Seems preferable to order' a Comparatively simple balance to.cover. the most likely. measurements and to improvise for the
others, rather than to order, an expensive multi-component balance which may not be .satisfactory for
all
the Measukeinents.7.4.. Wave-maker
,
Provision
should
be made in the-designof
medium size and_
-large channels
for
the mounting of :a wave-Maker at the endof
- the contractionnozzle:
7.5. Conversion-to an enclosed water tunnel
It is worthwhile to make provision in the design. for, the
.
mounting....Ofr,a, plats to suppress the free surface Waves' in the case of a medium Size channel: This then converts the channel
.,
into a comparativelY, large water' tunnel which at 20. ft/sec, .
_
is equivalent to sa wind tunnel of the same size operating
j60 ft/sec on a :Reynolds number basia. As has been Pointed out
-*the. dynamic _forces are much larger:. and it is likely that
certain
problems 1.n.,indUStrial aerodynamics, involving the elasticity
8. Miscellaneous Problems
8.1. Whirl and vibration of the impeller shaft
Generally whirl is not likely to be a problem because of
the low shaft r.p.m. and the necessity of having a fairly stiff shaft to transmit the torque.
There is the possibility of resonance at the first natural
frequency of bending vibration if this corresponds to blade frequency. This can arise if the peripherial distribution of
velocity is non-uniform, as may happen when the main flap control (which controls the diverted water) is in an extreme position.. Then each blade is subjected to a.fluctuating bending moment,
which will provide an exciting force at blade frequency. The
cure of this, in the design state, is to ensurethat the first natural bending frequency is sufficiently high.
8.2. The fairing of instrument leads and the method of support
The fairing of Pitot-static head leads is often in the
form of a fine streamline shape formed by two bi-convex circular .arcs cantilevered from a central support above the channel. The
centre of torsion is at mid-chord point and the centre of pressure of the lift forces on the fairing is at the quarter-chord point. Clearly, with this arrangement, slight
mis-a-lignment in direction calls into plava torsional moment tending to increase the mis-alignment and thus to further increase the lift and twisting moment. A critical speed is reached when the
elastic restoring moments are exceeded by the hydrodynamic moments and 'divergence' both in torsion and bending
occurs.-This will prevent the use of the instrument in this form at high speeds and large depths. The cure is to introduce a pivot ahead
of the centre of pressure so that the instrument has 'weather
cock' stability and will always align itself with the flow.
Small rubber pivots used in suspensions.arevery suitable in this
counteracts any tendency of the P.V.C.. connection tubes to throw
the instrument out of alignment at low speeds.
9. Conclusions
Very high speed open channels can be designed. for satis-factory operation up to 50 ft/sec..In the two examples of these
in the U.S.A..it is probable that the inclusion of a tilting floor 'would widen the range of operation. The problem of de
aeration seems to have been satisfactorily overcome by the use
of specially-designed large deaeration tanks in circuit. It seems likely that improvement to the power factor of these Channels could be made by improved hydrodynamic design of the
circuit.
Water channels operating at these speeds are likely to be small because of the nature: of the testing to be carried out, the large powers required, and the high cost of larger
channels.-The medium speed channels (up to 20 ft/sec) of medium size
can be designed with a very efficient circuit from the stand-point of Dower input and satisfactory deaeration is possible up to 20-ft/sec by diverting some 5% of the main flow at exit
for the working section. Use of a tilting floor enables channels of this type to operate with a wave free surface over the whole
speed range'.
There is need for research to improve the roughness of water surfacewhich occurs with the use of an enclosed nozzle
possibly by some form of-boundary.layer control or by the
use of a sluice gate.
It is thought that. the small drawt-down at exit can be pre-vented or much reduced by suitable shaping of the floor contour at exit. Also the provision of a separate tank to deaerate the
dimensions small; places a limit on the contractionratio of about 4: land makes the use of large diffuser angle (7° - 8°)
essential. This can lead to a slight non-uniform velocity
distri-bution through the depth of the working section and there is
scope for research to find out how compact channels can be de-signed with good velocity distribution. Essentially the problem
is one of obtaining a more uniform flow at the diffuser exit and in reducing secondary flow.'
(c) Water channels for the .testing of orthodox ship models must be large in order.that the Reynolds numbers are as large
as possible and that the boundary layers are sufficiently thick ' for experimental investigations. On economic grouds it is likely.
thatthese will be of the open stilling tank - sump design. with a natural draw-down through the Contraction nozzle. There appears
to be scope for improvement.in the surface flow by reduction of the wavelets generated from the walls of the contraction.. This might be achieved by greater smoothness of the walls and
possibly the merits of contractions in the vertical plane versus those in the horizontal plane should be investigated. A tilting
floor is necessary for wave free operation throughout the speed range. It would also be interesting to know whether a channel of this kind could operate satisfactorily.through the
critical speed.
(d) Blade frequency vibration as well as whirl of the impeller Shaft should be considered in the design
stage.--..(e). Fairing of instrument leads and holders needs special consideration in channels of medium to large size when the instruments have to operate at deep immersions and at high
10.
Part II - The Liverpool Channel
10,1 Introduction
The initial design was based on a simple wind tunnel type of circuit with a contraction ratio of 4:1. This design was put,
into -a practical form by Dr. Remmers of Kempf and Remmers
(Hamburg), who incorporated his experience of dealing with the exit flow.
Initially, the working section dimensions were 15 ft long
x 5 ft wide x 3 ft deep and the top speed was to be about 18 ft/sec for an input of 100 h.p..It was then decided to in-corporate depressurisation and the extra cost of this was
partially met by an 8% reduction in overall dimensions, so that the working section is now 14 ftlong x 434 ft wide x 24 ft deep.
The question of smoothing screen was left open until the
com-pletion of model tests...
Fig.A shows an outline of the 'channel. It is seen. that the overall length (excluding motor) is 53 ft and the maximum height is 18.6 ft and width is 7.7 ft. The decision was made
to have those surfaces in contact with water. made of stainless steel, with the exception of the honeycomb (brass) and the
im-peller (phosphor bronze). The external stiffeners are of mild steel.
The uses to which the channel is to be put- demanded a speed range of (1 - -1-) max, and speed holding to 1% at any. selected
-200
speed. The electrical drive is of a standard Ward-Leonard-type, with automatic electronic speed control. .The impeller is driven
via pulleys and a toothed belt without gear box..
10.2. The model tests
These are described in reference [9]: The impeller was driven by a 6.b.h.p..motor to give a top speed of 6.m/s: The
scale model was 1/4.67 full scale.
The water surface .
The water surface could be controlled by the tilting floor
to be remarkably level and free from waves at all speeds. The surface roughness was slight below the critical speed but be-came pronounced at top speed.
The velocity distributions
The velocity distributions both across the channel and in , depth were remarkably good. These are extracted from ref.' [9J' and are shown in Fig. .9.
Depressurisation tests
Tests were also run at different free surface pressures
and from these it was deduced that the performance of the full scale channel should be satisfactory. Reference [9] gives details
of these tests and we have presented these in a .compact non
dimensional form in Fig. 16, based on analysis given in Appendix
Concluding remarks
The very good performance of the model has led to Kempf
. and Remmers offering it as a standard production for University
and Technical College teaching.
From our standpoint, the results meant that subject to
possible sdale, (Reynolds number). effects on velocity distri-bution we could have confidence in the full Scale performance.
The roughness of water surface is not expected to scale up, but will mainly be a function of
speed.-DEPTH IN CMS.
16 15 13 12 11 -10 9 8
Vinax 6.46 FT /SEC.
53.5 EM' FROM NOZZLE
Vmax 13.10 FT/SEC. .
.
DISTANCES ACROSS I CHANNEL CMS.
' 8 10 12 14 18 18 40 22 24 28 ,26 30
-
1.0 as'
0 4. 2Fig. 9 Kempf and Remti4ii model channel
Variation of velOCity with depth and,iariation aCo-aW at 6 cms , depth for different speeds and positions along working sections
1716 15 -.14 13 '12 ; 11 10
I
6 .1 0-:17
I.
5 CM. FROM DISTANCES NOZZLE . V ma6 ACROSS, Vinal .13.12 CHANNEL 6.48 FI/SEC ki/SEC. CMS -02 0 10 12 . 14 I 16 1820' 4 S '26 30. S. 1.0 09 08 07 as10.3. The full scale channel tests
The water surface
Tests amply confirm the model results that a level wave free surface can be maintained over the whole speed range by.
operation of the adjustable floor. It has been found that there
is one angle of tilt which is approximately correct for all
speeds. This is 1 in 230 and is about that required to
compen-sate for boundary layer growth. At speeds of about 3 ft/sec small standing waves occur for this particular floor setting, but they can be removed by suitable adjustment of the floor. There is also a slight drawdown at exit at high speeds, which
can be corrected by readjustment of the floOr, operation of the flap controls and increasing the water content.
. The roughness of the water surface is very small at speeds
below the critical (9.4 ft/sec) and is barely noticeable ( +1/2)
ins), At 15 ft/sec it is quite small amounting to about .± 1/16
ins. However, at 18 ft/sec it has increased to 1/10 ins and
at top speed -- 21 ft/sec it is quite marked (± 1/8 ins) and is probably unacceptable for some tests...
Air entrainment
Up to 15 ft/sec, air entrainment in the form of small
bubbles
is
slight provided the channel has been run for some time after filling. Above this speed, bubbles of air becomevery numerous and of larger size. Depressurisation is expected to reduce this somewhat.
The velocity distrioution
The velocity distribution in depth (on the centre line) and across the channel (1.0 ft depth) are in Figs. 10 and 11 for stations 1 ft, 6. ft and 12 ft downstream of nozzle and at
'2 rt FROM NOZZLE
ibex 4 136
F7,4EO.-max
rT/SCC:-max!S.çCu
gi 1 .Liverpool Channel
FT. FROM NOZZLE
L
DISTANCES ACROSS i CHANNEL CMS ' ;
70 60 50
-Le
"la': _20 30 -41:1 50 10 . 09. , vi Q8,Y,0?".5 , -4.86F7F/SEC. FT( SEC.,-_ FT/SE.C::11-Fig. 11 Liverpool channel
1_ ft,,,,depth..; Variat.ion of velocity: across_ chafin'el for different..
' ', speeds and positions in working.,7ire'ctiOri
, .. . .,,--J:...:.., _ -=:-.- - ... . ,yri,-4-,..-463 rrtsEc. .: '-' 12 FT, FROM NOZZLE - : : - ." - -: FT/SEC:.' .Vin.ij'i..,° 14.9 0 0 . ,.. -r ,
The distribution across the channel is very uniform (within 170) except in the boundary layer of the side walls (Fig. 11).
The distribution in depth is very uniform up to half the
maximum depth (Fig. 10). After the position the velocity slowly falls, so that at 0 75 of the maximum depth it has fallen by
3%.-There is a slight worsening with increase of speed and also
compared with the model results (Fig. 9). In other words there is some adverse scale effect,
which is
thought to be due to thesecondary flows becoming relatively stronger as the Reynolds
number increases.
-Our view is that the full scale distributions are
satis-factory for the testing of models of surface craft and that for
calibration of current meters and other oceanographic instru-ments the error will be small. However, we think that the
ob-served fall-off of velocity with increase of depth results from the effect of secondary flows on a non-symmetrical distribution
of velocity at the outlet from the diffuser. This gives rise
to a rather poor distribution of total pressure at the entrance to the nozzle. We think this can be rectified by a screen
(possi-bly graded) resting on the cascade immediately after the
dif-fuser_ This will even up the velocity distribution at the dif-fuser exit and reduce the secondary flows. Alternatively, vortex
generators in the diffuser might xtify the velocity
distri-Outiodwithout introducing addition/ losses (see reference [16]).
d) Impeller vibration
At impeller revolutions between 285 300 r.p.m. and with certain flap control settings, a mild vibration occurred. This was traced to resonance between fluctuations in the blade bend7 ing moment (due to non-uniform velocity distribution) and the-first natural bending frequency of the shaft. This is not serious audit can be avoidedtg suitable setting of the various controls.
.Tile metal tubes leading from the Pitot-static head are
streamlined by a fairing formed by hi-convex arcs. The whole is cantilevered from a support on the top of the channel. At medium,
to high speeds, depending on the depth, a critical speed is
reached at which 'divergence' in torsion and bending occurs.. This is because the centre of torsion is behind the centre of
pressure of the 'lift' forces on the As already mentioned
in Part I; this divergence.was avoided by introducing a pivot
axis ahead of the 'quarter-chord point, using weak rubber
sus-pension bearings.
10.4. Conclusions
This design of channel (with a contraction ratio of 4:1).
should be satisfactory, both in quality of free surface and in
velocity distribution up to 13 ft/sec. Also up to this speed
air entrainment is small.-The,adjustable floor is an essential
control. Model results for the
velocity
distribution in depthtend to be optimistic.
11. Acknowledgements
TO Dr.-Norbury, who was chiefly responsible for Appendices I and II, and to Dr. Alexander who superintended the calibration of the channel.
To Dr. K. .Remmers of Kempf and Remmers for making available
his wide experience in the design of water channels for per-mission to use results and photographs from his report on the :modal of the Liverpool channel and for permission to refer to
the-very high speed channel now.under
construction.-TO Mr,.Silverleaf, Superintendent of Ship Division,
Nation-al
Physical Laboratory, for permission to refer to the large National Physical Laboratory Channel which.has just cone intoTo Dr. Benedini of the Istituto di Idraulica, University
of Padoua and Mr. Rossiter of Liverpool University who took most
of the observations of velocity distribution,
Appendix I (Dr. J.F. Norbury)
Design of free surface contractions by onedimensional theory
We use the onedimensional theory to obtain general ideas
about the form of a free surface contraction for a given Froude
number
in
the working section. The notation used is given in Fig. 12, with in additionz = breadth of contraction. Z = breadth of working section
u = local velocity in contraction
A = local crosssectional area = z(H + h +
= u/U
h = h/H working section
z =
2/Z
Froude number,WITT
A=
A/HZ2
The suffix zero refers to conditions at the entry.to the con-traction. Bernoulli' s equation gives 112 = U2 - 2gh, and the
equation of continuity is
= UHZ = Alz(H
4 h + c)
In-non-dimensional form this gives
Also and =
1
so that the contraction ratio,
A0
==---co
1 -
2210
For given values of Ao and F. we can determine co and
E,.
Now if T1 is specified as a function ofxc
we can calculate 71(xc)and A(x). This means that for a given surface profile the
longitudinal variation of area is fixed. The area variation can be obtained in one of three ways.:
i) the constant breadth contraction, with = L Then
_
-
-c =
A -(1 + h)
"ii) the contraction with horizontal floor, so that c = 0. In
this case
A/ ( 1 +
iii) the more general case in which both Z and C vary with
xc.
Calculations have been carried out for values of F = 0.5,